diff --git a/daemons/based/based_callbacks.c b/daemons/based/based_callbacks.c
index 9985ec4551..d2ff9503d1 100644
--- a/daemons/based/based_callbacks.c
+++ b/daemons/based/based_callbacks.c
@@ -1,1761 +1,1761 @@
/*
* Copyright 2004-2024 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include <crm_internal.h>
#include <sys/param.h>
#include <stdio.h>
#include <sys/types.h>
#include <unistd.h>
#include <stdlib.h>
#include <stdint.h> // uint32_t, uint64_t, UINT64_C()
#include <errno.h>
#include <fcntl.h>
#include <inttypes.h> // PRIu64
#include <glib.h>
#include <libxml/tree.h>
#include <crm/crm.h>
#include <crm/cib.h>
#include <crm/cluster/internal.h>
#include <crm/common/xml.h>
#include <crm/common/remote_internal.h>
#include <pacemaker-based.h>
#define EXIT_ESCALATION_MS 10000
static unsigned long cib_local_bcast_num = 0;
typedef struct cib_local_notify_s {
xmlNode *notify_src;
char *client_id;
gboolean from_peer;
gboolean sync_reply;
} cib_local_notify_t;
int next_client_id = 0;
gboolean legacy_mode = FALSE;
qb_ipcs_service_t *ipcs_ro = NULL;
qb_ipcs_service_t *ipcs_rw = NULL;
qb_ipcs_service_t *ipcs_shm = NULL;
static int cib_process_command(xmlNode *request,
const cib__operation_t *operation,
cib__op_fn_t op_function, xmlNode **reply,
xmlNode **cib_diff, bool privileged);
static gboolean cib_common_callback(qb_ipcs_connection_t *c, void *data,
size_t size, gboolean privileged);
gboolean
cib_legacy_mode(void)
{
return legacy_mode;
}
static int32_t
cib_ipc_accept(qb_ipcs_connection_t * c, uid_t uid, gid_t gid)
{
if (cib_shutdown_flag) {
crm_info("Ignoring new IPC client [%d] during shutdown",
pcmk__client_pid(c));
return -ECONNREFUSED;
}
if (pcmk__new_client(c, uid, gid) == NULL) {
return -ENOMEM;
}
return 0;
}
static int32_t
cib_ipc_dispatch_rw(qb_ipcs_connection_t * c, void *data, size_t size)
{
pcmk__client_t *client = pcmk__find_client(c);
crm_trace("%p message from %s", c, client->id);
return cib_common_callback(c, data, size, TRUE);
}
static int32_t
cib_ipc_dispatch_ro(qb_ipcs_connection_t * c, void *data, size_t size)
{
pcmk__client_t *client = pcmk__find_client(c);
crm_trace("%p message from %s", c, client->id);
return cib_common_callback(c, data, size, FALSE);
}
/* Error code means? */
static int32_t
cib_ipc_closed(qb_ipcs_connection_t * c)
{
pcmk__client_t *client = pcmk__find_client(c);
if (client == NULL) {
return 0;
}
crm_trace("Connection %p", c);
pcmk__free_client(client);
return 0;
}
static void
cib_ipc_destroy(qb_ipcs_connection_t * c)
{
crm_trace("Connection %p", c);
cib_ipc_closed(c);
if (cib_shutdown_flag) {
cib_shutdown(0);
}
}
struct qb_ipcs_service_handlers ipc_ro_callbacks = {
.connection_accept = cib_ipc_accept,
.connection_created = NULL,
.msg_process = cib_ipc_dispatch_ro,
.connection_closed = cib_ipc_closed,
.connection_destroyed = cib_ipc_destroy
};
struct qb_ipcs_service_handlers ipc_rw_callbacks = {
.connection_accept = cib_ipc_accept,
.connection_created = NULL,
.msg_process = cib_ipc_dispatch_rw,
.connection_closed = cib_ipc_closed,
.connection_destroyed = cib_ipc_destroy
};
/*!
* \internal
* \brief Create reply XML for a CIB request
*
* \param[in] op CIB operation type
* \param[in] call_id CIB call ID
* \param[in] client_id CIB client ID
* \param[in] call_options Group of <tt>enum cib_call_options</tt> flags
* \param[in] rc Request return code
* \param[in] call_data Request output data
*
* \return Reply XML
*
* \note The caller is responsible for freeing the return value using
* \p free_xml().
*/
static xmlNode *
create_cib_reply(const char *op, const char *call_id, const char *client_id,
int call_options, int rc, xmlNode *call_data)
{
xmlNode *reply = pcmk__xe_create(NULL, PCMK__XE_CIB_REPLY);
crm_xml_add(reply, PCMK__XA_T, PCMK__VALUE_CIB);
crm_xml_add(reply, PCMK__XA_CIB_OP, op);
crm_xml_add(reply, PCMK__XA_CIB_CALLID, call_id);
crm_xml_add(reply, PCMK__XA_CIB_CLIENTID, client_id);
crm_xml_add_int(reply, PCMK__XA_CIB_CALLOPT, call_options);
crm_xml_add_int(reply, PCMK__XA_CIB_RC, rc);
if (call_data != NULL) {
xmlNode *wrapper = pcmk__xe_create(reply, PCMK__XE_CIB_CALLDATA);
crm_trace("Attaching reply output");
pcmk__xml_copy(wrapper, call_data);
}
crm_log_xml_explicit(reply, "cib:reply");
return reply;
}
static void
do_local_notify(const xmlNode *notify_src, const char *client_id,
bool sync_reply, bool from_peer)
{
int rid = 0;
int call_id = 0;
pcmk__client_t *client_obj = NULL;
CRM_ASSERT(notify_src && client_id);
crm_element_value_int(notify_src, PCMK__XA_CIB_CALLID, &call_id);
client_obj = pcmk__find_client_by_id(client_id);
if (client_obj == NULL) {
crm_debug("Could not send response %d: client %s not found",
call_id, client_id);
return;
}
if (sync_reply) {
if (client_obj->ipcs) {
CRM_LOG_ASSERT(client_obj->request_id);
rid = client_obj->request_id;
client_obj->request_id = 0;
crm_trace("Sending response %d to client %s%s",
rid, pcmk__client_name(client_obj),
(from_peer? " (originator of delegated request)" : ""));
} else {
crm_trace("Sending response (call %d) to client %s%s",
call_id, pcmk__client_name(client_obj),
(from_peer? " (originator of delegated request)" : ""));
}
} else {
crm_trace("Sending event %d to client %s%s",
call_id, pcmk__client_name(client_obj),
(from_peer? " (originator of delegated request)" : ""));
}
switch (PCMK__CLIENT_TYPE(client_obj)) {
case pcmk__client_ipc:
{
int rc = pcmk__ipc_send_xml(client_obj, rid, notify_src,
(sync_reply? crm_ipc_flags_none
: crm_ipc_server_event));
if (rc != pcmk_rc_ok) {
crm_warn("%s reply to client %s failed: %s " CRM_XS " rc=%d",
(sync_reply? "Synchronous" : "Asynchronous"),
pcmk__client_name(client_obj), pcmk_rc_str(rc),
rc);
}
}
break;
#ifdef HAVE_GNUTLS_GNUTLS_H
case pcmk__client_tls:
#endif
case pcmk__client_tcp:
pcmk__remote_send_xml(client_obj->remote, notify_src);
break;
default:
crm_err("Unknown transport for client %s "
CRM_XS " flags=%#016" PRIx64,
pcmk__client_name(client_obj), client_obj->flags);
}
}
void
cib_common_callback_worker(uint32_t id, uint32_t flags, xmlNode * op_request,
pcmk__client_t *cib_client, gboolean privileged)
{
const char *op = crm_element_value(op_request, PCMK__XA_CIB_OP);
int call_options = cib_none;
crm_element_value_int(op_request, PCMK__XA_CIB_CALLOPT, &call_options);
/* Requests with cib_transaction set should not be sent to based directly
* (outside of a commit-transaction request)
*/
if (pcmk_is_set(call_options, cib_transaction)) {
return;
}
if (pcmk__str_eq(op, CRM_OP_REGISTER, pcmk__str_none)) {
if (flags & crm_ipc_client_response) {
xmlNode *ack = pcmk__xe_create(NULL, __func__);
crm_xml_add(ack, PCMK__XA_CIB_OP, CRM_OP_REGISTER);
crm_xml_add(ack, PCMK__XA_CIB_CLIENTID, cib_client->id);
pcmk__ipc_send_xml(cib_client, id, ack, flags);
cib_client->request_id = 0;
free_xml(ack);
}
return;
} else if (pcmk__str_eq(op, PCMK__VALUE_CIB_NOTIFY, pcmk__str_none)) {
/* Update the notify filters for this client */
int on_off = 0;
crm_exit_t status = CRM_EX_OK;
uint64_t bit = UINT64_C(0);
const char *type = crm_element_value(op_request,
PCMK__XA_CIB_NOTIFY_TYPE);
crm_element_value_int(op_request, PCMK__XA_CIB_NOTIFY_ACTIVATE,
&on_off);
crm_debug("Setting %s callbacks %s for client %s",
type, (on_off? "on" : "off"), pcmk__client_name(cib_client));
if (pcmk__str_eq(type, PCMK__VALUE_CIB_POST_NOTIFY, pcmk__str_none)) {
bit = cib_notify_post;
} else if (pcmk__str_eq(type, PCMK__VALUE_CIB_PRE_NOTIFY,
pcmk__str_none)) {
bit = cib_notify_pre;
} else if (pcmk__str_eq(type, PCMK__VALUE_CIB_UPDATE_CONFIRMATION,
pcmk__str_none)) {
bit = cib_notify_confirm;
} else if (pcmk__str_eq(type, PCMK__VALUE_CIB_DIFF_NOTIFY,
pcmk__str_none)) {
bit = cib_notify_diff;
} else {
status = CRM_EX_INVALID_PARAM;
}
if (bit != 0) {
if (on_off) {
pcmk__set_client_flags(cib_client, bit);
} else {
pcmk__clear_client_flags(cib_client, bit);
}
}
pcmk__ipc_send_ack(cib_client, id, flags, PCMK__XE_ACK, NULL, status);
return;
}
cib_process_request(op_request, privileged, cib_client);
}
int32_t
cib_common_callback(qb_ipcs_connection_t * c, void *data, size_t size, gboolean privileged)
{
uint32_t id = 0;
uint32_t flags = 0;
int call_options = 0;
pcmk__client_t *cib_client = pcmk__find_client(c);
xmlNode *op_request = pcmk__client_data2xml(cib_client, data, &id, &flags);
if (op_request) {
crm_element_value_int(op_request, PCMK__XA_CIB_CALLOPT, &call_options);
}
if (op_request == NULL) {
crm_trace("Invalid message from %p", c);
pcmk__ipc_send_ack(cib_client, id, flags, PCMK__XE_NACK, NULL,
CRM_EX_PROTOCOL);
return 0;
} else if(cib_client == NULL) {
crm_trace("Invalid client %p", c);
return 0;
}
if (pcmk_is_set(call_options, cib_sync_call)) {
CRM_LOG_ASSERT(flags & crm_ipc_client_response);
CRM_LOG_ASSERT(cib_client->request_id == 0); /* This means the client has two synchronous events in-flight */
cib_client->request_id = id; /* Reply only to the last one */
}
if (cib_client->name == NULL) {
const char *value = crm_element_value(op_request,
PCMK__XA_CIB_CLIENTNAME);
if (value == NULL) {
cib_client->name = pcmk__itoa(cib_client->pid);
} else {
cib_client->name = pcmk__str_copy(value);
if (crm_is_daemon_name(value)) {
pcmk__set_client_flags(cib_client, cib_is_daemon);
}
}
}
/* Allow cluster daemons more leeway before being evicted */
if (pcmk_is_set(cib_client->flags, cib_is_daemon)) {
const char *qmax = cib_config_lookup(PCMK_OPT_CLUSTER_IPC_LIMIT);
if (pcmk__set_client_queue_max(cib_client, qmax)) {
crm_trace("IPC threshold for client %s[%u] is now %u",
pcmk__client_name(cib_client), cib_client->pid,
cib_client->queue_max);
}
}
crm_xml_add(op_request, PCMK__XA_CIB_CLIENTID, cib_client->id);
crm_xml_add(op_request, PCMK__XA_CIB_CLIENTNAME, cib_client->name);
CRM_LOG_ASSERT(cib_client->user != NULL);
pcmk__update_acl_user(op_request, PCMK__XA_CIB_USER, cib_client->user);
cib_common_callback_worker(id, flags, op_request, cib_client, privileged);
free_xml(op_request);
return 0;
}
static uint64_t ping_seq = 0;
static char *ping_digest = NULL;
static bool ping_modified_since = FALSE;
static gboolean
cib_digester_cb(gpointer data)
{
if (based_is_primary) {
char buffer[32];
xmlNode *ping = pcmk__xe_create(NULL, PCMK__XE_PING);
ping_seq++;
free(ping_digest);
ping_digest = NULL;
ping_modified_since = FALSE;
snprintf(buffer, 32, "%" PRIu64, ping_seq);
crm_trace("Requesting peer digests (%s)", buffer);
crm_xml_add(ping, PCMK__XA_T, PCMK__VALUE_CIB);
crm_xml_add(ping, PCMK__XA_CIB_OP, CRM_OP_PING);
crm_xml_add(ping, PCMK__XA_CIB_PING_ID, buffer);
crm_xml_add(ping, PCMK_XA_CRM_FEATURE_SET, CRM_FEATURE_SET);
send_cluster_message(NULL, crm_msg_cib, ping, TRUE);
free_xml(ping);
}
return FALSE;
}
static void
process_ping_reply(xmlNode *reply)
{
uint64_t seq = 0;
const char *host = crm_element_value(reply, PCMK__XA_SRC);
xmlNode *wrapper = pcmk__xe_first_child(reply, PCMK__XE_CIB_CALLDATA, NULL,
NULL);
xmlNode *pong = pcmk__xe_first_child(wrapper, NULL, NULL, NULL);
const char *seq_s = crm_element_value(pong, PCMK__XA_CIB_PING_ID);
const char *digest = crm_element_value(pong, PCMK__XA_DIGEST);
if (seq_s == NULL) {
crm_debug("Ignoring ping reply with no " PCMK__XA_CIB_PING_ID);
return;
} else {
long long seq_ll;
if (pcmk__scan_ll(seq_s, &seq_ll, 0LL) != pcmk_rc_ok) {
return;
}
seq = (uint64_t) seq_ll;
}
if(digest == NULL) {
crm_trace("Ignoring ping reply %s from %s with no digest", seq_s, host);
} else if(seq != ping_seq) {
crm_trace("Ignoring out of sequence ping reply %s from %s", seq_s, host);
} else if(ping_modified_since) {
crm_trace("Ignoring ping reply %s from %s: cib updated since", seq_s, host);
} else {
const char *version = crm_element_value(pong, PCMK_XA_CRM_FEATURE_SET);
if(ping_digest == NULL) {
crm_trace("Calculating new digest");
ping_digest = calculate_xml_versioned_digest(the_cib, FALSE, TRUE, version);
}
crm_trace("Processing ping reply %s from %s (%s)", seq_s, host, digest);
if (!pcmk__str_eq(ping_digest, digest, pcmk__str_casei)) {
xmlNode *wrapper = pcmk__xe_first_child(pong, PCMK__XE_CIB_CALLDATA,
NULL, NULL);
xmlNode *remote_cib = pcmk__xe_first_child(wrapper, NULL, NULL, NULL);
const char *admin_epoch_s = NULL;
const char *epoch_s = NULL;
const char *num_updates_s = NULL;
if (remote_cib != NULL) {
admin_epoch_s = crm_element_value(remote_cib,
PCMK_XA_ADMIN_EPOCH);
epoch_s = crm_element_value(remote_cib, PCMK_XA_EPOCH);
num_updates_s = crm_element_value(remote_cib,
PCMK_XA_NUM_UPDATES);
}
crm_notice("Local CIB %s.%s.%s.%s differs from %s: %s.%s.%s.%s %p",
crm_element_value(the_cib, PCMK_XA_ADMIN_EPOCH),
crm_element_value(the_cib, PCMK_XA_EPOCH),
crm_element_value(the_cib, PCMK_XA_NUM_UPDATES),
ping_digest, host,
pcmk__s(admin_epoch_s, "_"),
pcmk__s(epoch_s, "_"),
pcmk__s(num_updates_s, "_"),
digest, remote_cib);
if(remote_cib && remote_cib->children) {
// Additional debug
xml_calculate_changes(the_cib, remote_cib);
pcmk__log_xml_changes(LOG_INFO, remote_cib);
crm_trace("End of differences");
}
free_xml(remote_cib);
sync_our_cib(reply, FALSE);
}
}
}
static void
local_notify_destroy_callback(gpointer data)
{
cib_local_notify_t *notify = data;
free_xml(notify->notify_src);
free(notify->client_id);
free(notify);
}
static void
check_local_notify(int bcast_id)
{
const cib_local_notify_t *notify = NULL;
if (!local_notify_queue) {
return;
}
notify = pcmk__intkey_table_lookup(local_notify_queue, bcast_id);
if (notify) {
do_local_notify(notify->notify_src, notify->client_id, notify->sync_reply,
notify->from_peer);
pcmk__intkey_table_remove(local_notify_queue, bcast_id);
}
}
static void
queue_local_notify(xmlNode * notify_src, const char *client_id, gboolean sync_reply,
gboolean from_peer)
{
cib_local_notify_t *notify = pcmk__assert_alloc(1,
sizeof(cib_local_notify_t));
notify->notify_src = notify_src;
notify->client_id = pcmk__str_copy(client_id);
notify->sync_reply = sync_reply;
notify->from_peer = from_peer;
if (!local_notify_queue) {
local_notify_queue = pcmk__intkey_table(local_notify_destroy_callback);
}
pcmk__intkey_table_insert(local_notify_queue, cib_local_bcast_num, notify);
// cppcheck doesn't know notify will get freed when hash table is destroyed
// cppcheck-suppress memleak
}
static void
parse_local_options_v1(const pcmk__client_t *cib_client,
const cib__operation_t *operation, int call_options,
const char *host, const char *op, gboolean *local_notify,
gboolean *needs_reply, gboolean *process,
gboolean *needs_forward)
{
if (pcmk_is_set(operation->flags, cib__op_attr_modifies)
&& !pcmk_is_set(call_options, cib_inhibit_bcast)) {
/* we need to send an update anyway */
*needs_reply = TRUE;
} else {
*needs_reply = FALSE;
}
if (host == NULL && (call_options & cib_scope_local)) {
crm_trace("Processing locally scoped %s op from client %s",
op, pcmk__client_name(cib_client));
*local_notify = TRUE;
} else if ((host == NULL) && based_is_primary) {
crm_trace("Processing %s op locally from client %s as primary",
op, pcmk__client_name(cib_client));
*local_notify = TRUE;
} else if (pcmk__str_eq(host, OUR_NODENAME, pcmk__str_casei)) {
crm_trace("Processing locally addressed %s op from client %s",
op, pcmk__client_name(cib_client));
*local_notify = TRUE;
} else if (stand_alone) {
*needs_forward = FALSE;
*local_notify = TRUE;
*process = TRUE;
} else {
crm_trace("%s op from %s needs to be forwarded to client %s",
op, pcmk__client_name(cib_client),
pcmk__s(host, "the primary instance"));
*needs_forward = TRUE;
*process = FALSE;
}
}
static void
parse_local_options_v2(const pcmk__client_t *cib_client,
const cib__operation_t *operation, int call_options,
const char *host, const char *op, gboolean *local_notify,
gboolean *needs_reply, gboolean *process,
gboolean *needs_forward)
{
// Process locally and notify local client
*process = TRUE;
*needs_reply = FALSE;
*local_notify = TRUE;
*needs_forward = FALSE;
if (pcmk_is_set(operation->flags, cib__op_attr_local)) {
/* Always process locally if cib__op_attr_local is set.
*
* @COMPAT: Currently host is ignored. At a compatibility break, throw
* an error (from cib_process_request() or earlier) if host is not NULL or
* OUR_NODENAME.
*/
crm_trace("Processing always-local %s op from client %s",
op, pcmk__client_name(cib_client));
if (!pcmk__str_eq(host, OUR_NODENAME,
pcmk__str_casei|pcmk__str_null_matches)) {
crm_warn("Operation '%s' is always local but its target host is "
"set to '%s'",
op, host);
}
return;
}
if (pcmk_is_set(operation->flags, cib__op_attr_modifies)
|| !pcmk__str_eq(host, OUR_NODENAME,
pcmk__str_casei|pcmk__str_null_matches)) {
// Forward modifying and non-local requests via cluster
*process = FALSE;
*needs_reply = FALSE;
*local_notify = FALSE;
*needs_forward = TRUE;
crm_trace("%s op from %s needs to be forwarded to %s",
op, pcmk__client_name(cib_client),
pcmk__s(host, "all nodes"));
return;
}
if (stand_alone) {
crm_trace("Processing %s op from client %s (stand-alone)",
op, pcmk__client_name(cib_client));
} else {
crm_trace("Processing %saddressed %s op from client %s",
((host != NULL)? "locally " : "un"),
op, pcmk__client_name(cib_client));
}
}
static void
parse_local_options(const pcmk__client_t *cib_client,
const cib__operation_t *operation, int call_options,
const char *host, const char *op, gboolean *local_notify,
gboolean *needs_reply, gboolean *process,
gboolean *needs_forward)
{
if(cib_legacy_mode()) {
parse_local_options_v1(cib_client, operation, call_options, host,
op, local_notify, needs_reply, process,
needs_forward);
} else {
parse_local_options_v2(cib_client, operation, call_options, host,
op, local_notify, needs_reply, process,
needs_forward);
}
}
static gboolean
parse_peer_options_v1(const cib__operation_t *operation, xmlNode *request,
gboolean *local_notify, gboolean *needs_reply,
gboolean *process)
{
const char *op = NULL;
const char *host = NULL;
const char *delegated = NULL;
const char *originator = crm_element_value(request, PCMK__XA_SRC);
const char *reply_to = crm_element_value(request, PCMK__XA_CIB_ISREPLYTO);
gboolean is_reply = pcmk__str_eq(reply_to, OUR_NODENAME, pcmk__str_casei);
if (pcmk__xe_attr_is_true(request, PCMK__XA_CIB_UPDATE)) {
*needs_reply = FALSE;
if (is_reply) {
*local_notify = TRUE;
crm_trace("Processing global/peer update from %s"
" that originated from us", originator);
} else {
crm_trace("Processing global/peer update from %s", originator);
}
return TRUE;
}
op = crm_element_value(request, PCMK__XA_CIB_OP);
crm_trace("Processing legacy %s request sent by %s", op, originator);
if (pcmk__str_eq(op, PCMK__CIB_REQUEST_SHUTDOWN, pcmk__str_none)) {
/* Always process these */
*local_notify = FALSE;
if (reply_to == NULL || is_reply) {
*process = TRUE;
}
if (is_reply) {
*needs_reply = FALSE;
}
return *process;
}
if (is_reply && pcmk__str_eq(op, CRM_OP_PING, pcmk__str_casei)) {
process_ping_reply(request);
return FALSE;
}
if (is_reply) {
crm_trace("Forward reply sent from %s to local clients", originator);
*process = FALSE;
*needs_reply = FALSE;
*local_notify = TRUE;
return TRUE;
}
host = crm_element_value(request, PCMK__XA_CIB_HOST);
if (pcmk__str_eq(host, OUR_NODENAME, pcmk__str_casei)) {
crm_trace("Processing %s request sent to us from %s", op, originator);
return TRUE;
} else if(is_reply == FALSE && pcmk__str_eq(op, CRM_OP_PING, pcmk__str_casei)) {
crm_trace("Processing %s request sent to %s by %s", op, host?host:"everyone", originator);
*needs_reply = TRUE;
return TRUE;
} else if ((host == NULL) && based_is_primary) {
crm_trace("Processing %s request sent to primary instance from %s",
op, originator);
return TRUE;
}
delegated = crm_element_value(request, PCMK__XA_CIB_DELEGATED_FROM);
if (delegated != NULL) {
crm_trace("Ignoring message for primary instance");
} else if (host != NULL) {
/* this is for a specific instance and we're not it */
crm_trace("Ignoring msg for instance on %s", host);
} else if ((reply_to == NULL) && !based_is_primary) {
// This is for the primary instance, and we're not it
crm_trace("Ignoring reply for primary instance");
} else if (pcmk__str_eq(op, PCMK__CIB_REQUEST_SHUTDOWN, pcmk__str_none)) {
if (reply_to != NULL) {
crm_debug("Processing %s from %s", op, originator);
*needs_reply = FALSE;
} else {
crm_debug("Processing %s reply from %s", op, originator);
}
return TRUE;
} else {
crm_err("Nothing for us to do?");
crm_log_xml_err(request, "Peer[inbound]");
}
return FALSE;
}
static gboolean
parse_peer_options_v2(const cib__operation_t *operation, xmlNode *request,
gboolean *local_notify, gboolean *needs_reply,
gboolean *process)
{
const char *host = NULL;
const char *delegated = crm_element_value(request,
PCMK__XA_CIB_DELEGATED_FROM);
const char *op = crm_element_value(request, PCMK__XA_CIB_OP);
const char *originator = crm_element_value(request, PCMK__XA_SRC);
const char *reply_to = crm_element_value(request, PCMK__XA_CIB_ISREPLYTO);
gboolean is_reply = pcmk__str_eq(reply_to, OUR_NODENAME, pcmk__str_casei);
if (originator == NULL) { // Shouldn't be possible
originator = "peer";
}
if (pcmk__str_eq(op, PCMK__CIB_REQUEST_REPLACE, pcmk__str_none)) {
// sync_our_cib() sets PCMK__XA_CIB_ISREPLYTO
if (reply_to) {
delegated = reply_to;
}
goto skip_is_reply;
} else if (pcmk__str_eq(op, PCMK__CIB_REQUEST_SYNC_TO_ALL,
pcmk__str_none)) {
// Nothing to do
} else if (is_reply && pcmk__str_eq(op, CRM_OP_PING, pcmk__str_casei)) {
process_ping_reply(request);
return FALSE;
} else if (pcmk__str_eq(op, PCMK__CIB_REQUEST_UPGRADE, pcmk__str_none)) {
/* Only the DC (node with the oldest software) should process
* this operation if PCMK__XA_CIB_SCHEMA_MAX is unset.
*
* If the DC is happy it will then send out another
* PCMK__CIB_REQUEST_UPGRADE which will tell all nodes to do the actual
* upgrade.
*
* Except this time PCMK__XA_CIB_SCHEMA_MAX will be set which puts a
* limit on how far newer nodes will go
*/
const char *max = crm_element_value(request, PCMK__XA_CIB_SCHEMA_MAX);
const char *upgrade_rc = crm_element_value(request,
PCMK__XA_CIB_UPGRADE_RC);
crm_trace("Parsing upgrade %s for %s with max=%s and upgrade_rc=%s",
(is_reply? "reply" : "request"),
(based_is_primary? "primary" : "secondary"),
pcmk__s(max, "none"), pcmk__s(upgrade_rc, "none"));
if (upgrade_rc != NULL) {
// Our upgrade request was rejected by DC, notify clients of result
crm_xml_add(request, PCMK__XA_CIB_RC, upgrade_rc);
} else if ((max == NULL) && based_is_primary) {
/* We are the DC, check if this upgrade is allowed */
goto skip_is_reply;
} else if(max) {
/* Ok, go ahead and upgrade to 'max' */
goto skip_is_reply;
} else {
// Ignore broadcast client requests when we're not primary
return FALSE;
}
} else if (pcmk__xe_attr_is_true(request, PCMK__XA_CIB_UPDATE)) {
crm_info("Detected legacy %s global update from %s", op, originator);
send_sync_request(NULL);
legacy_mode = TRUE;
return FALSE;
} else if (is_reply
&& pcmk_is_set(operation->flags, cib__op_attr_modifies)) {
crm_trace("Ignoring legacy %s reply sent from %s to local clients", op, originator);
return FALSE;
} else if (pcmk__str_eq(op, PCMK__CIB_REQUEST_SHUTDOWN, pcmk__str_none)) {
*local_notify = FALSE;
if (reply_to == NULL) {
*process = TRUE;
} else { // Not possible?
crm_debug("Ignoring shutdown request from %s because reply_to=%s",
originator, reply_to);
}
return *process;
}
if (is_reply) {
crm_trace("Will notify local clients for %s reply from %s",
op, originator);
*process = FALSE;
*needs_reply = FALSE;
*local_notify = TRUE;
return TRUE;
}
skip_is_reply:
*process = TRUE;
*needs_reply = FALSE;
*local_notify = pcmk__str_eq(delegated, OUR_NODENAME, pcmk__str_casei);
host = crm_element_value(request, PCMK__XA_CIB_HOST);
if (pcmk__str_eq(host, OUR_NODENAME, pcmk__str_casei)) {
crm_trace("Processing %s request sent to us from %s", op, originator);
*needs_reply = TRUE;
return TRUE;
} else if (host != NULL) {
crm_trace("Ignoring %s request intended for CIB manager on %s",
op, host);
return FALSE;
} else if(is_reply == FALSE && pcmk__str_eq(op, CRM_OP_PING, pcmk__str_casei)) {
*needs_reply = TRUE;
}
crm_trace("Processing %s request broadcast by %s call %s on %s "
"(local clients will%s be notified)", op,
pcmk__s(crm_element_value(request, PCMK__XA_CIB_CLIENTNAME),
"client"),
pcmk__s(crm_element_value(request, PCMK__XA_CIB_CALLID),
"without ID"),
originator, (*local_notify? "" : "not"));
return TRUE;
}
static gboolean
parse_peer_options(const cib__operation_t *operation, xmlNode *request,
gboolean *local_notify, gboolean *needs_reply,
gboolean *process)
{
/* TODO: What happens when an update comes in after node A
* requests the CIB from node B, but before it gets the reply (and
* sends out the replace operation)
*/
if(cib_legacy_mode()) {
return parse_peer_options_v1(operation, request, local_notify,
needs_reply, process);
} else {
return parse_peer_options_v2(operation, request, local_notify,
needs_reply, process);
}
}
/*!
* \internal
* \brief Forward a CIB request to the appropriate target host(s)
*
* \param[in] request CIB request to forward
*/
static void
forward_request(xmlNode *request)
{
const char *op = crm_element_value(request, PCMK__XA_CIB_OP);
const char *section = crm_element_value(request, PCMK__XA_CIB_SECTION);
const char *host = crm_element_value(request, PCMK__XA_CIB_HOST);
const char *originator = crm_element_value(request, PCMK__XA_SRC);
const char *client_name = crm_element_value(request,
PCMK__XA_CIB_CLIENTNAME);
const char *call_id = crm_element_value(request, PCMK__XA_CIB_CALLID);
crm_node_t *peer = NULL;
int log_level = LOG_INFO;
if (pcmk__str_eq(op, PCMK__CIB_REQUEST_NOOP, pcmk__str_none)) {
log_level = LOG_DEBUG;
}
do_crm_log(log_level,
"Forwarding %s operation for section %s to %s (origin=%s/%s/%s)",
pcmk__s(op, "invalid"),
pcmk__s(section, "all"),
pcmk__s(host, (cib_legacy_mode()? "primary" : "all")),
pcmk__s(originator, "local"),
pcmk__s(client_name, "unspecified"),
pcmk__s(call_id, "unspecified"));
crm_xml_add(request, PCMK__XA_CIB_DELEGATED_FROM, OUR_NODENAME);
if (host != NULL) {
peer = pcmk__get_node(0, host, NULL, pcmk__node_search_cluster);
}
send_cluster_message(peer, crm_msg_cib, request, FALSE);
// Return the request to its original state
pcmk__xe_remove_attr(request, PCMK__XA_CIB_DELEGATED_FROM);
}
static gboolean
send_peer_reply(xmlNode * msg, xmlNode * result_diff, const char *originator, gboolean broadcast)
{
CRM_ASSERT(msg != NULL);
if (broadcast) {
/* @COMPAT: Legacy code
*
* This successful call modified the CIB, and the change needs to be
* broadcast (sent via cluster to all nodes).
*/
int diff_add_updates = 0;
int diff_add_epoch = 0;
int diff_add_admin_epoch = 0;
int diff_del_updates = 0;
int diff_del_epoch = 0;
int diff_del_admin_epoch = 0;
const char *digest = NULL;
int format = 1;
xmlNode *wrapper = NULL;
CRM_LOG_ASSERT(result_diff != NULL);
digest = crm_element_value(result_diff, PCMK__XA_DIGEST);
crm_element_value_int(result_diff, PCMK_XA_FORMAT, &format);
cib_diff_version_details(result_diff,
&diff_add_admin_epoch, &diff_add_epoch, &diff_add_updates,
&diff_del_admin_epoch, &diff_del_epoch, &diff_del_updates);
crm_trace("Sending update diff %d.%d.%d -> %d.%d.%d %s",
diff_del_admin_epoch, diff_del_epoch, diff_del_updates,
diff_add_admin_epoch, diff_add_epoch, diff_add_updates, digest);
crm_xml_add(msg, PCMK__XA_CIB_ISREPLYTO, originator);
pcmk__xe_set_bool_attr(msg, PCMK__XA_CIB_UPDATE, true);
crm_xml_add(msg, PCMK__XA_CIB_OP, PCMK__CIB_REQUEST_APPLY_PATCH);
crm_xml_add(msg, PCMK__XA_CIB_USER, CRM_DAEMON_USER);
if (format == 1) {
CRM_ASSERT(digest != NULL);
}
wrapper = pcmk__xe_create(msg, PCMK__XA_CIB_UPDATE_DIFF);
pcmk__xml_copy(wrapper, result_diff);
crm_log_xml_explicit(msg, "copy");
return send_cluster_message(NULL, crm_msg_cib, msg, TRUE);
} else if (originator != NULL) {
/* send reply via HA to originating node */
crm_trace("Sending request result to %s only", originator);
crm_xml_add(msg, PCMK__XA_CIB_ISREPLYTO, originator);
return send_cluster_message(pcmk__get_node(0, originator, NULL,
pcmk__node_search_cluster),
crm_msg_cib, msg, FALSE);
}
return FALSE;
}
/*!
* \internal
* \brief Handle an IPC or CPG message containing a request
*
* \param[in,out] request Request XML
* \param[in] privileged Whether privileged commands may be run
* (see cib_server_ops[] definition)
* \param[in] cib_client IPC client that sent request (or NULL if CPG)
*
* \return Legacy Pacemaker return code
*/
int
cib_process_request(xmlNode *request, gboolean privileged,
const pcmk__client_t *cib_client)
{
// @TODO: Break into multiple smaller functions
int call_options = 0;
gboolean process = TRUE; // Whether to process request locally now
gboolean is_update = TRUE; // Whether request would modify CIB
gboolean needs_reply = TRUE; // Whether to build a reply
gboolean local_notify = FALSE; // Whether to notify (local) requester
gboolean needs_forward = FALSE; // Whether to forward request somewhere else
xmlNode *op_reply = NULL;
xmlNode *result_diff = NULL;
int rc = pcmk_ok;
const char *op = crm_element_value(request, PCMK__XA_CIB_OP);
const char *originator = crm_element_value(request, PCMK__XA_SRC);
const char *host = crm_element_value(request, PCMK__XA_CIB_HOST);
const char *target = NULL;
const char *call_id = crm_element_value(request, PCMK__XA_CIB_CALLID);
const char *client_id = crm_element_value(request, PCMK__XA_CIB_CLIENTID);
const char *client_name = crm_element_value(request,
PCMK__XA_CIB_CLIENTNAME);
const char *reply_to = crm_element_value(request, PCMK__XA_CIB_ISREPLYTO);
const cib__operation_t *operation = NULL;
cib__op_fn_t op_function = NULL;
crm_element_value_int(request, PCMK__XA_CIB_CALLOPT, &call_options);
if ((host != NULL) && (*host == '\0')) {
host = NULL;
}
// @TODO: Improve trace messages. Target is accurate only for legacy mode.
if (host) {
target = host;
} else if (call_options & cib_scope_local) {
target = "local host";
} else {
target = "primary";
}
if (cib_client == NULL) {
crm_trace("Processing peer %s operation from %s/%s on %s intended for %s (reply=%s)",
op, client_name, call_id, originator, target, reply_to);
} else {
crm_xml_add(request, PCMK__XA_SRC, OUR_NODENAME);
crm_trace("Processing local %s operation from %s/%s intended for %s", op, client_name, call_id, target);
}
rc = cib__get_operation(op, &operation);
rc = pcmk_rc2legacy(rc);
if (rc != pcmk_ok) {
/* TODO: construct error reply? */
crm_err("Pre-processing of command failed: %s", pcmk_strerror(rc));
return rc;
}
op_function = based_get_op_function(operation);
if (op_function == NULL) {
crm_err("Operation %s not supported by CIB manager", op);
return -EOPNOTSUPP;
}
if (cib_client != NULL) {
parse_local_options(cib_client, operation, call_options, host, op,
&local_notify, &needs_reply, &process,
&needs_forward);
} else if (!parse_peer_options(operation, request, &local_notify,
&needs_reply, &process)) {
return rc;
}
if (pcmk_is_set(call_options, cib_transaction)) {
/* All requests in a transaction are processed locally against a working
* CIB copy, and we don't notify for individual requests because the
* entire transaction is atomic.
*
* We still call the option parser functions above, for the sake of log
* messages and checking whether we're the target for peer requests.
*/
process = TRUE;
needs_reply = FALSE;
local_notify = FALSE;
needs_forward = FALSE;
}
is_update = pcmk_is_set(operation->flags, cib__op_attr_modifies);
if (pcmk_is_set(call_options, cib_discard_reply)) {
/* If the request will modify the CIB, and we are in legacy mode, we
* need to build a reply so we can broadcast a diff, even if the
* requester doesn't want one.
*/
needs_reply = is_update && cib_legacy_mode();
local_notify = FALSE;
crm_trace("Client is not interested in the reply");
}
if (needs_forward) {
forward_request(request);
return rc;
}
if (cib_status != pcmk_ok) {
rc = cib_status;
crm_err("Operation ignored, cluster configuration is invalid."
" Please repair and restart: %s", pcmk_strerror(cib_status));
op_reply = create_cib_reply(op, call_id, client_id, call_options, rc,
the_cib);
} else if (process) {
time_t finished = 0;
time_t now = time(NULL);
int level = LOG_INFO;
const char *section = crm_element_value(request, PCMK__XA_CIB_SECTION);
const char *admin_epoch_s = NULL;
const char *epoch_s = NULL;
const char *num_updates_s = NULL;
rc = cib_process_command(request, operation, op_function, &op_reply,
&result_diff, privileged);
if (!is_update) {
level = LOG_TRACE;
} else if (pcmk__xe_attr_is_true(request, PCMK__XA_CIB_UPDATE)) {
switch (rc) {
case pcmk_ok:
level = LOG_INFO;
break;
case -pcmk_err_old_data:
case -pcmk_err_diff_resync:
case -pcmk_err_diff_failed:
level = LOG_TRACE;
break;
default:
level = LOG_ERR;
}
} else if (rc != pcmk_ok) {
level = LOG_WARNING;
}
if (the_cib != NULL) {
admin_epoch_s = crm_element_value(the_cib, PCMK_XA_ADMIN_EPOCH);
epoch_s = crm_element_value(the_cib, PCMK_XA_EPOCH);
num_updates_s = crm_element_value(the_cib, PCMK_XA_NUM_UPDATES);
}
do_crm_log(level,
"Completed %s operation for section %s: %s (rc=%d, origin=%s/%s/%s, version=%s.%s.%s)",
op, section ? section : "'all'", pcmk_strerror(rc), rc,
originator ? originator : "local", client_name, call_id,
pcmk__s(admin_epoch_s, "0"),
pcmk__s(epoch_s, "0"),
pcmk__s(num_updates_s, "0"));
finished = time(NULL);
if ((finished - now) > 3) {
crm_trace("%s operation took %lds to complete", op, (long)(finished - now));
crm_write_blackbox(0, NULL);
}
if (op_reply == NULL && (needs_reply || local_notify)) {
crm_err("Unexpected NULL reply to message");
crm_log_xml_err(request, "null reply");
needs_reply = FALSE;
local_notify = FALSE;
}
}
if (is_update && !cib_legacy_mode()) {
crm_trace("Completed pre-sync update from %s/%s/%s%s",
originator ? originator : "local", client_name, call_id,
local_notify?" with local notification":"");
} else if (!needs_reply || stand_alone) {
// This was a non-originating secondary update
crm_trace("Completed update as secondary");
} else if (cib_legacy_mode() &&
rc == pcmk_ok && result_diff != NULL && !(call_options & cib_inhibit_bcast)) {
gboolean broadcast = FALSE;
cib_local_bcast_num++;
crm_xml_add_int(request, PCMK__XA_CIB_LOCAL_NOTIFY_ID,
cib_local_bcast_num);
broadcast = send_peer_reply(request, result_diff, originator, TRUE);
if (broadcast && client_id && local_notify && op_reply) {
/* If we have been asked to sync the reply,
* and a bcast msg has gone out, we queue the local notify
* until we know the bcast message has been received */
local_notify = FALSE;
crm_trace("Queuing local %ssync notification for %s",
(call_options & cib_sync_call) ? "" : "a-", client_id);
queue_local_notify(op_reply, client_id,
pcmk_is_set(call_options, cib_sync_call),
(cib_client == NULL));
op_reply = NULL; /* the reply is queued, so don't free here */
}
} else if ((cib_client == NULL)
&& !pcmk_is_set(call_options, cib_discard_reply)) {
if (is_update == FALSE || result_diff == NULL) {
crm_trace("Request not broadcast: R/O call");
} else if (call_options & cib_inhibit_bcast) {
crm_trace("Request not broadcast: inhibited");
} else if (rc != pcmk_ok) {
crm_trace("Request not broadcast: call failed: %s", pcmk_strerror(rc));
} else {
crm_trace("Directing reply to %s", originator);
}
send_peer_reply(op_reply, result_diff, originator, FALSE);
}
if (local_notify && client_id) {
crm_trace("Performing local %ssync notification for %s",
(pcmk_is_set(call_options, cib_sync_call)? "" : "a"),
client_id);
if (process == FALSE) {
do_local_notify(request, client_id,
pcmk_is_set(call_options, cib_sync_call),
(cib_client == NULL));
} else {
do_local_notify(op_reply, client_id,
pcmk_is_set(call_options, cib_sync_call),
(cib_client == NULL));
}
}
free_xml(op_reply);
free_xml(result_diff);
return rc;
}
/*!
* \internal
* \brief Get a CIB operation's input from the request XML
*
* \param[in] request CIB request XML
* \param[in] type CIB operation type
* \param[out] section Where to store CIB section name
*
* \return Input XML for CIB operation
*
* \note If not \c NULL, the return value is a non-const pointer to part of
* \p request. The caller should not free it directly.
*/
static xmlNode *
prepare_input(const xmlNode *request, enum cib__op_type type,
const char **section)
{
xmlNode *wrapper = NULL;
xmlNode *input = NULL;
*section = NULL;
switch (type) {
case cib__op_apply_patch:
{
const char *wrapper_name = PCMK__XE_CIB_CALLDATA;
if (pcmk__xe_attr_is_true(request, PCMK__XA_CIB_UPDATE)) {
wrapper_name = PCMK__XA_CIB_UPDATE_DIFF;
}
wrapper = pcmk__xe_first_child(request, wrapper_name, NULL,
NULL);
input = pcmk__xe_first_child(wrapper, NULL, NULL, NULL);
}
break;
default:
wrapper = pcmk__xe_first_child(request, PCMK__XE_CIB_CALLDATA, NULL,
NULL);
input = pcmk__xe_first_child(wrapper, NULL, NULL, NULL);
*section = crm_element_value(request, PCMK__XA_CIB_SECTION);
break;
}
// Grab the specified section
if ((*section != NULL) && pcmk__xe_is(input, PCMK_XE_CIB)) {
input = pcmk_find_cib_element(input, *section);
}
return input;
}
// v1 and v2 patch formats
#define XPATH_CONFIG_CHANGE \
"//" PCMK_XE_CRM_CONFIG " | " \
"//" PCMK_XE_CHANGE \
"[contains(@" PCMK_XA_PATH ",'/" PCMK_XE_CRM_CONFIG "/')]"
static bool
contains_config_change(xmlNode *diff)
{
bool changed = false;
if (diff) {
xmlXPathObject *xpathObj = xpath_search(diff, XPATH_CONFIG_CHANGE);
if (numXpathResults(xpathObj) > 0) {
changed = true;
}
freeXpathObject(xpathObj);
}
return changed;
}
static int
cib_process_command(xmlNode *request, const cib__operation_t *operation,
cib__op_fn_t op_function, xmlNode **reply,
xmlNode **cib_diff, bool privileged)
{
xmlNode *input = NULL;
xmlNode *output = NULL;
xmlNode *result_cib = NULL;
int call_options = 0;
const char *op = NULL;
const char *section = NULL;
const char *call_id = crm_element_value(request, PCMK__XA_CIB_CALLID);
const char *client_id = crm_element_value(request, PCMK__XA_CIB_CLIENTID);
const char *client_name = crm_element_value(request,
PCMK__XA_CIB_CLIENTNAME);
const char *originator = crm_element_value(request, PCMK__XA_SRC);
int rc = pcmk_ok;
bool config_changed = false;
bool manage_counters = true;
static mainloop_timer_t *digest_timer = NULL;
CRM_ASSERT(cib_status == pcmk_ok);
if(digest_timer == NULL) {
digest_timer = mainloop_timer_add("digester", 5000, FALSE, cib_digester_cb, NULL);
}
*reply = NULL;
*cib_diff = NULL;
/* Start processing the request... */
op = crm_element_value(request, PCMK__XA_CIB_OP);
crm_element_value_int(request, PCMK__XA_CIB_CALLOPT, &call_options);
if (!privileged && pcmk_is_set(operation->flags, cib__op_attr_privileged)) {
rc = -EACCES;
crm_trace("Failed due to lack of privileges: %s", pcmk_strerror(rc));
goto done;
}
input = prepare_input(request, operation->type, §ion);
if (!pcmk_is_set(operation->flags, cib__op_attr_modifies)) {
rc = cib_perform_op(NULL, op, call_options, op_function, true, section,
request, input, false, &config_changed, &the_cib,
&result_cib, NULL, &output);
CRM_CHECK(result_cib == NULL, free_xml(result_cib));
goto done;
}
/* @COMPAT: Handle a valid write action (legacy)
*
* @TODO: Re-evaluate whether this is all truly legacy. The cib_force_diff
* portion is. However, PCMK__XA_CIB_UPDATE may be set by a sync operation
* even in non-legacy mode, and manage_counters tells xml_create_patchset()
* whether to update version/epoch info.
*/
if (pcmk__xe_attr_is_true(request, PCMK__XA_CIB_UPDATE)) {
manage_counters = false;
cib__set_call_options(call_options, "call", cib_force_diff);
crm_trace("Global update detected");
CRM_LOG_ASSERT(pcmk__str_any_of(op,
PCMK__CIB_REQUEST_APPLY_PATCH,
PCMK__CIB_REQUEST_REPLACE,
NULL));
}
ping_modified_since = TRUE;
if (pcmk_is_set(call_options, cib_inhibit_bcast)) {
crm_trace("Skipping update: inhibit broadcast");
manage_counters = false;
}
// result_cib must not be modified after cib_perform_op() returns
rc = cib_perform_op(NULL, op, call_options, op_function, false, section,
request, input, manage_counters, &config_changed,
&the_cib, &result_cib, cib_diff, &output);
// @COMPAT: Legacy code
if (!manage_counters) {
int format = 1;
// If the diff is NULL at this point, it's because nothing changed
if (*cib_diff != NULL) {
crm_element_value_int(*cib_diff, PCMK_XA_FORMAT, &format);
}
if (format == 1) {
config_changed = cib__config_changed_v1(NULL, NULL, cib_diff);
}
}
/* Always write to disk for successful ops with the flag set. This also
* negates the need to detect ordering changes.
*/
if ((rc == pcmk_ok)
&& pcmk_is_set(operation->flags, cib__op_attr_writes_through)) {
config_changed = true;
}
if ((rc == pcmk_ok)
&& !pcmk_any_flags_set(call_options, cib_dryrun|cib_transaction)) {
if (result_cib != the_cib) {
if (pcmk_is_set(operation->flags, cib__op_attr_writes_through)) {
config_changed = true;
}
crm_trace("Activating %s->%s%s",
crm_element_value(the_cib, PCMK_XA_NUM_UPDATES),
crm_element_value(result_cib, PCMK_XA_NUM_UPDATES),
(config_changed? " changed" : ""));
rc = activateCibXml(result_cib, config_changed, op);
if (rc != pcmk_ok) {
crm_err("Failed to activate new CIB: %s", pcmk_strerror(rc));
}
}
if ((rc == pcmk_ok) && contains_config_change(*cib_diff)) {
cib_read_config(config_hash, result_cib);
}
/* @COMPAT Nodes older than feature set 3.19.0 don't support
* transactions. In a mixed-version cluster with nodes <3.19.0, we must
* sync the updated CIB, so that the older nodes receive the changes.
* Any node that has already applied the transaction will ignore the
* synced CIB.
*
* To ensure the updated CIB is synced from only one node, we sync it
* from the originator.
*/
if ((operation->type == cib__op_commit_transact)
&& pcmk__str_eq(originator, OUR_NODENAME, pcmk__str_casei)
&& compare_version(crm_element_value(the_cib,
PCMK_XA_CRM_FEATURE_SET),
"3.19.0") < 0) {
sync_our_cib(request, TRUE);
}
mainloop_timer_stop(digest_timer);
mainloop_timer_start(digest_timer);
} else if (rc == -pcmk_err_schema_validation) {
CRM_ASSERT(result_cib != the_cib);
if (output != NULL) {
crm_log_xml_info(output, "cib:output");
free_xml(output);
}
output = result_cib;
} else {
crm_trace("Not activating %d %d %s", rc,
pcmk_is_set(call_options, cib_dryrun),
crm_element_value(result_cib, PCMK_XA_NUM_UPDATES));
if (result_cib != the_cib) {
free_xml(result_cib);
}
}
if (!pcmk_any_flags_set(call_options,
cib_dryrun|cib_inhibit_notify|cib_transaction)) {
crm_trace("Sending notifications %d",
pcmk_is_set(call_options, cib_dryrun));
cib_diff_notify(op, rc, call_id, client_id, client_name, originator,
input, *cib_diff);
}
pcmk__log_xml_patchset(LOG_TRACE, *cib_diff);
done:
if (!pcmk_is_set(call_options, cib_discard_reply) || cib_legacy_mode()) {
*reply = create_cib_reply(op, call_id, client_id, call_options, rc,
output);
}
if (output != the_cib) {
free_xml(output);
}
crm_trace("done");
return rc;
}
void
cib_peer_callback(xmlNode * msg, void *private_data)
{
const char *reason = NULL;
const char *originator = crm_element_value(msg, PCMK__XA_SRC);
if (cib_legacy_mode()
&& pcmk__str_eq(originator, OUR_NODENAME,
pcmk__str_casei|pcmk__str_null_matches)) {
/* message is from ourselves */
int bcast_id = 0;
if (crm_element_value_int(msg, PCMK__XA_CIB_LOCAL_NOTIFY_ID,
&bcast_id) == 0) {
check_local_notify(bcast_id);
}
return;
} else if (crm_peer_cache == NULL) {
reason = "membership not established";
goto bail;
}
if (crm_element_value(msg, PCMK__XA_CIB_CLIENTNAME) == NULL) {
crm_xml_add(msg, PCMK__XA_CIB_CLIENTNAME, originator);
}
/* crm_log_xml_trace(msg, "Peer[inbound]"); */
cib_process_request(msg, TRUE, NULL);
return;
bail:
if (reason) {
const char *op = crm_element_value(msg, PCMK__XA_CIB_OP);
crm_warn("Discarding %s message from %s: %s", op, originator, reason);
}
}
static gboolean
cib_force_exit(gpointer data)
{
crm_notice("Forcing exit!");
terminate_cib(__func__, CRM_EX_ERROR);
return FALSE;
}
static void
disconnect_remote_client(gpointer key, gpointer value, gpointer user_data)
{
pcmk__client_t *a_client = value;
crm_err("Can't disconnect client %s: Not implemented",
pcmk__client_name(a_client));
}
static void
initiate_exit(void)
{
int active = 0;
xmlNode *leaving = NULL;
active = crm_active_peers();
if (active < 2) { // This is the last active node
terminate_cib(__func__, 0);
return;
}
crm_info("Sending shutdown request to %d peers", active);
leaving = pcmk__xe_create(NULL, PCMK__XE_EXIT_NOTIFICATION);
crm_xml_add(leaving, PCMK__XA_T, PCMK__VALUE_CIB);
crm_xml_add(leaving, PCMK__XA_CIB_OP, PCMK__CIB_REQUEST_SHUTDOWN);
send_cluster_message(NULL, crm_msg_cib, leaving, TRUE);
free_xml(leaving);
g_timeout_add(EXIT_ESCALATION_MS, cib_force_exit, NULL);
}
void
cib_shutdown(int nsig)
{
struct qb_ipcs_stats srv_stats;
if (cib_shutdown_flag == FALSE) {
int disconnects = 0;
qb_ipcs_connection_t *c = NULL;
cib_shutdown_flag = TRUE;
c = qb_ipcs_connection_first_get(ipcs_rw);
while (c != NULL) {
qb_ipcs_connection_t *last = c;
c = qb_ipcs_connection_next_get(ipcs_rw, last);
crm_debug("Disconnecting r/w client %p...", last);
qb_ipcs_disconnect(last);
qb_ipcs_connection_unref(last);
disconnects++;
}
c = qb_ipcs_connection_first_get(ipcs_ro);
while (c != NULL) {
qb_ipcs_connection_t *last = c;
c = qb_ipcs_connection_next_get(ipcs_ro, last);
crm_debug("Disconnecting r/o client %p...", last);
qb_ipcs_disconnect(last);
qb_ipcs_connection_unref(last);
disconnects++;
}
c = qb_ipcs_connection_first_get(ipcs_shm);
while (c != NULL) {
qb_ipcs_connection_t *last = c;
c = qb_ipcs_connection_next_get(ipcs_shm, last);
crm_debug("Disconnecting non-blocking r/w client %p...", last);
qb_ipcs_disconnect(last);
qb_ipcs_connection_unref(last);
disconnects++;
}
disconnects += pcmk__ipc_client_count();
crm_debug("Disconnecting %d remote clients", pcmk__ipc_client_count());
pcmk__foreach_ipc_client(disconnect_remote_client, NULL);
crm_info("Disconnected %d clients", disconnects);
}
qb_ipcs_stats_get(ipcs_rw, &srv_stats, QB_FALSE);
if (pcmk__ipc_client_count() == 0) {
crm_info("All clients disconnected (%d)", srv_stats.active_connections);
initiate_exit();
} else {
crm_info("Waiting on %d clients to disconnect (%d)",
pcmk__ipc_client_count(), srv_stats.active_connections);
}
}
extern int remote_fd;
extern int remote_tls_fd;
/*!
* \internal
* \brief Close remote sockets, free the global CIB and quit
*
* \param[in] caller Name of calling function (for log message)
* \param[in] fast If -1, skip disconnect; if positive, exit that
*/
void
terminate_cib(const char *caller, int fast)
{
crm_info("%s: Exiting%s...", caller,
(fast > 0)? " fast" : mainloop ? " from mainloop" : "");
if (remote_fd > 0) {
close(remote_fd);
remote_fd = 0;
}
if (remote_tls_fd > 0) {
close(remote_tls_fd);
remote_tls_fd = 0;
}
uninitializeCib();
if (fast > 0) {
/* Quit fast on error */
pcmk__stop_based_ipc(ipcs_ro, ipcs_rw, ipcs_shm);
crm_exit(fast);
} else if ((mainloop != NULL) && g_main_loop_is_running(mainloop)) {
/* Quit via returning from the main loop. If fast == -1, we skip the
* disconnect here, and it will be done when the main loop returns
* (this allows the peer status callback to avoid messing with the
* peer caches).
*/
if (fast == 0) {
- crm_cluster_disconnect(crm_cluster);
+ pcmk_cluster_disconnect(crm_cluster);
}
g_main_loop_quit(mainloop);
} else {
/* Quit via clean exit. Even the peer status callback can disconnect
* here, because we're not returning control to the caller. */
- crm_cluster_disconnect(crm_cluster);
+ pcmk_cluster_disconnect(crm_cluster);
pcmk__stop_based_ipc(ipcs_ro, ipcs_rw, ipcs_shm);
crm_exit(CRM_EX_OK);
}
}
diff --git a/daemons/based/pacemaker-based.c b/daemons/based/pacemaker-based.c
index 72fa289ec5..a9e7e2af6c 100644
--- a/daemons/based/pacemaker-based.c
+++ b/daemons/based/pacemaker-based.c
@@ -1,450 +1,450 @@
/*
* Copyright 2004-2024 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include <crm_internal.h>
#include <stdio.h>
#include <stdlib.h>
#include <pwd.h>
#include <grp.h>
#include <bzlib.h>
#include <sys/types.h>
#include <glib.h>
#include <libxml/tree.h>
#include <crm/crm.h>
#include <crm/cib/internal.h>
#include <crm/cluster/internal.h>
#include <crm/common/cmdline_internal.h>
#include <crm/common/mainloop.h>
#include <crm/common/output_internal.h>
#include <crm/common/xml.h>
#include <pacemaker-based.h>
#define SUMMARY "daemon for managing the configuration of a Pacemaker cluster"
extern int init_remote_listener(int port, gboolean encrypted);
gboolean cib_shutdown_flag = FALSE;
int cib_status = pcmk_ok;
crm_cluster_t *crm_cluster = NULL;
GMainLoop *mainloop = NULL;
gchar *cib_root = NULL;
static gboolean preserve_status = FALSE;
gboolean cib_writes_enabled = TRUE;
gboolean stand_alone = FALSE;
int remote_fd = 0;
int remote_tls_fd = 0;
GHashTable *config_hash = NULL;
GHashTable *local_notify_queue = NULL;
static void cib_init(void);
void cib_shutdown(int nsig);
static bool startCib(const char *filename);
extern int write_cib_contents(gpointer p);
static crm_exit_t exit_code = CRM_EX_OK;
static void
cib_enable_writes(int nsig)
{
crm_info("(Re)enabling disk writes");
cib_writes_enabled = TRUE;
}
/*!
* \internal
* \brief Set up options, users, and groups for stand-alone mode
*
* \param[out] error GLib error object
*
* \return Standard Pacemaker return code
*/
static int
setup_stand_alone(GError **error)
{
int rc = 0;
struct passwd *pwentry = NULL;
preserve_status = TRUE;
cib_writes_enabled = FALSE;
errno = 0;
pwentry = getpwnam(CRM_DAEMON_USER);
if (pwentry == NULL) {
exit_code = CRM_EX_FATAL;
if (errno != 0) {
g_set_error(error, PCMK__EXITC_ERROR, exit_code,
"Error getting password DB entry for %s: %s",
CRM_DAEMON_USER, strerror(errno));
return errno;
}
g_set_error(error, PCMK__EXITC_ERROR, exit_code,
"Password DB entry for '%s' not found", CRM_DAEMON_USER);
return ENXIO;
}
rc = setgid(pwentry->pw_gid);
if (rc < 0) {
exit_code = CRM_EX_FATAL;
g_set_error(error, PCMK__EXITC_ERROR, exit_code,
"Could not set group to %d: %s",
pwentry->pw_gid, strerror(errno));
return errno;
}
rc = initgroups(CRM_DAEMON_USER, pwentry->pw_gid);
if (rc < 0) {
exit_code = CRM_EX_FATAL;
g_set_error(error, PCMK__EXITC_ERROR, exit_code,
"Could not setup groups for user %d: %s",
pwentry->pw_uid, strerror(errno));
return errno;
}
rc = setuid(pwentry->pw_uid);
if (rc < 0) {
exit_code = CRM_EX_FATAL;
g_set_error(error, PCMK__EXITC_ERROR, exit_code,
"Could not set user to %d: %s",
pwentry->pw_uid, strerror(errno));
return errno;
}
return pcmk_rc_ok;
}
/* @COMPAT Deprecated since 2.1.8. Use pcmk_list_cluster_options() or
* crm_attribute --list-options=cluster instead of querying daemon metadata.
*/
static int
based_metadata(pcmk__output_t *out)
{
return pcmk__daemon_metadata(out, "pacemaker-based",
"Cluster Information Base manager options",
"Cluster options used by Pacemaker's Cluster "
"Information Base manager",
pcmk__opt_based);
}
static GOptionEntry entries[] = {
{ "stand-alone", 's', G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &stand_alone,
"(Advanced use only) Run in stand-alone mode", NULL },
{ "disk-writes", 'w', G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE,
&cib_writes_enabled,
"(Advanced use only) Enable disk writes (enabled by default unless in "
"stand-alone mode)", NULL },
{ "cib-root", 'r', G_OPTION_FLAG_NONE, G_OPTION_ARG_FILENAME, &cib_root,
"(Advanced use only) Directory where the CIB XML file should be located "
"(default: " CRM_CONFIG_DIR ")", NULL },
{ NULL }
};
static pcmk__supported_format_t formats[] = {
PCMK__SUPPORTED_FORMAT_NONE,
PCMK__SUPPORTED_FORMAT_TEXT,
PCMK__SUPPORTED_FORMAT_XML,
{ NULL, NULL, NULL }
};
static GOptionContext *
build_arg_context(pcmk__common_args_t *args, GOptionGroup **group)
{
GOptionContext *context = NULL;
context = pcmk__build_arg_context(args, "text (default), xml", group, NULL);
pcmk__add_main_args(context, entries);
return context;
}
int
main(int argc, char **argv)
{
int rc = pcmk_rc_ok;
crm_ipc_t *old_instance = NULL;
pcmk__output_t *out = NULL;
GError *error = NULL;
GOptionGroup *output_group = NULL;
pcmk__common_args_t *args = pcmk__new_common_args(SUMMARY);
gchar **processed_args = pcmk__cmdline_preproc(argv, "r");
GOptionContext *context = build_arg_context(args, &output_group);
crm_log_preinit(NULL, argc, argv);
pcmk__register_formats(output_group, formats);
if (!g_option_context_parse_strv(context, &processed_args, &error)) {
exit_code = CRM_EX_USAGE;
goto done;
}
rc = pcmk__output_new(&out, args->output_ty, args->output_dest, argv);
if (rc != pcmk_rc_ok) {
exit_code = CRM_EX_ERROR;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Error creating output format %s: %s",
args->output_ty, pcmk_rc_str(rc));
goto done;
}
if (args->version) {
out->version(out, false);
goto done;
}
mainloop_add_signal(SIGTERM, cib_shutdown);
mainloop_add_signal(SIGPIPE, cib_enable_writes);
cib_writer = mainloop_add_trigger(G_PRIORITY_LOW, write_cib_contents, NULL);
if ((g_strv_length(processed_args) >= 2)
&& pcmk__str_eq(processed_args[1], "metadata", pcmk__str_none)) {
rc = based_metadata(out);
if (rc != pcmk_rc_ok) {
exit_code = CRM_EX_FATAL;
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Unable to display metadata: %s", pcmk_rc_str(rc));
}
goto done;
}
pcmk__cli_init_logging("pacemaker-based", args->verbosity);
crm_log_init(NULL, LOG_INFO, TRUE, FALSE, argc, argv, FALSE);
crm_notice("Starting Pacemaker CIB manager");
old_instance = crm_ipc_new(PCMK__SERVER_BASED_RO, 0);
if (old_instance == NULL) {
/* crm_ipc_new() will have already logged an error message with
* crm_err()
*/
exit_code = CRM_EX_FATAL;
goto done;
}
if (pcmk__connect_generic_ipc(old_instance) == pcmk_rc_ok) {
/* IPC end-point already up */
crm_ipc_close(old_instance);
crm_ipc_destroy(old_instance);
crm_err("pacemaker-based is already active, aborting startup");
goto done;
} else {
/* not up or not authentic, we'll proceed either way */
crm_ipc_destroy(old_instance);
old_instance = NULL;
}
if (stand_alone) {
rc = setup_stand_alone(&error);
if (rc != pcmk_rc_ok) {
goto done;
}
}
if (cib_root == NULL) {
cib_root = g_strdup(CRM_CONFIG_DIR);
} else {
crm_notice("Using custom config location: %s", cib_root);
}
if (!pcmk__daemon_can_write(cib_root, NULL)) {
exit_code = CRM_EX_FATAL;
crm_err("Terminating due to bad permissions on %s", cib_root);
g_set_error(&error, PCMK__EXITC_ERROR, exit_code,
"Bad permissions on %s (see logs for details)", cib_root);
goto done;
}
crm_peer_init();
// Read initial CIB, connect to cluster, and start IPC servers
cib_init();
// Run the main loop
mainloop = g_main_loop_new(NULL, FALSE);
crm_notice("Pacemaker CIB manager successfully started and accepting connections");
g_main_loop_run(mainloop);
/* If main loop returned, clean up and exit. We disconnect in case
* terminate_cib() was called with fast=-1.
*/
- crm_cluster_disconnect(crm_cluster);
+ pcmk_cluster_disconnect(crm_cluster);
pcmk__stop_based_ipc(ipcs_ro, ipcs_rw, ipcs_shm);
done:
g_strfreev(processed_args);
pcmk__free_arg_context(context);
crm_peer_destroy();
if (local_notify_queue != NULL) {
g_hash_table_destroy(local_notify_queue);
}
if (config_hash != NULL) {
g_hash_table_destroy(config_hash);
}
pcmk__client_cleanup();
pcmk_cluster_free(crm_cluster);
g_free(cib_root);
pcmk__output_and_clear_error(&error, out);
if (out != NULL) {
out->finish(out, exit_code, true, NULL);
pcmk__output_free(out);
}
pcmk__unregister_formats();
crm_exit(exit_code);
}
#if SUPPORT_COROSYNC
static void
cib_cs_dispatch(cpg_handle_t handle,
const struct cpg_name *groupName,
uint32_t nodeid, uint32_t pid, void *msg, size_t msg_len)
{
uint32_t kind = 0;
xmlNode *xml = NULL;
const char *from = NULL;
char *data = pcmk_message_common_cs(handle, nodeid, pid, msg, &kind, &from);
if(data == NULL) {
return;
}
if (kind == crm_class_cluster) {
xml = pcmk__xml_parse(data);
if (xml == NULL) {
crm_err("Invalid XML: '%.120s'", data);
free(data);
return;
}
crm_xml_add(xml, PCMK__XA_SRC, from);
cib_peer_callback(xml, NULL);
}
free_xml(xml);
free(data);
}
static void
cib_cs_destroy(gpointer user_data)
{
if (cib_shutdown_flag) {
crm_info("Corosync disconnection complete");
} else {
crm_crit("Lost connection to cluster layer, shutting down");
terminate_cib(__func__, CRM_EX_DISCONNECT);
}
}
#endif
static void
cib_peer_update_callback(enum crm_status_type type, crm_node_t * node, const void *data)
{
switch (type) {
case crm_status_processes:
if (cib_legacy_mode()
&& !pcmk_is_set(node->processes, crm_get_cluster_proc())) {
uint32_t old = data? *(const uint32_t *)data : 0;
if ((node->processes ^ old) & crm_proc_cpg) {
crm_info("Attempting to disable legacy mode after %s left the cluster",
node->uname);
legacy_mode = FALSE;
}
}
break;
case crm_status_uname:
case crm_status_nstate:
if (cib_shutdown_flag && (crm_active_peers() < 2)
&& (pcmk__ipc_client_count() == 0)) {
crm_info("No more peers");
terminate_cib(__func__, -1);
}
break;
}
}
static void
cib_init(void)
{
crm_cluster = pcmk_cluster_new();
#if SUPPORT_COROSYNC
if (is_corosync_cluster()) {
crm_cluster->destroy = cib_cs_destroy;
crm_cluster->cpg.cpg_deliver_fn = cib_cs_dispatch;
crm_cluster->cpg.cpg_confchg_fn = pcmk_cpg_membership;
}
#endif // SUPPORT_COROSYNC
config_hash = pcmk__strkey_table(free, free);
if (startCib("cib.xml") == FALSE) {
crm_crit("Cannot start CIB... terminating");
crm_exit(CRM_EX_NOINPUT);
}
if (!stand_alone) {
crm_set_status_callback(&cib_peer_update_callback);
if (pcmk_cluster_connect(crm_cluster) != pcmk_rc_ok) {
crm_crit("Cannot sign in to the cluster... terminating");
crm_exit(CRM_EX_FATAL);
}
}
pcmk__serve_based_ipc(&ipcs_ro, &ipcs_rw, &ipcs_shm, &ipc_ro_callbacks,
&ipc_rw_callbacks);
if (stand_alone) {
based_is_primary = true;
}
}
static bool
startCib(const char *filename)
{
gboolean active = FALSE;
xmlNode *cib = readCibXmlFile(cib_root, filename, !preserve_status);
if (activateCibXml(cib, TRUE, "start") == 0) {
int port = 0;
active = TRUE;
cib_read_config(config_hash, cib);
pcmk__scan_port(crm_element_value(cib, PCMK_XA_REMOTE_TLS_PORT), &port);
if (port >= 0) {
remote_tls_fd = init_remote_listener(port, TRUE);
}
pcmk__scan_port(crm_element_value(cib, PCMK_XA_REMOTE_CLEAR_PORT),
&port);
if (port >= 0) {
remote_fd = init_remote_listener(port, FALSE);
}
}
return active;
}
diff --git a/daemons/controld/controld_control.c b/daemons/controld/controld_control.c
index 953e73b456..4e33e38da0 100644
--- a/daemons/controld/controld_control.c
+++ b/daemons/controld/controld_control.c
@@ -1,689 +1,689 @@
/*
* Copyright 2004-2024 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include <crm_internal.h>
#include <sys/param.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <crm/crm.h>
#include <crm/common/xml.h>
#include <crm/pengine/rules.h>
#include <crm/cluster/internal.h>
#include <crm/cluster/election_internal.h>
#include <crm/common/ipc_internal.h>
#include <pacemaker-controld.h>
static qb_ipcs_service_t *ipcs = NULL;
static crm_trigger_t *config_read_trigger = NULL;
#if SUPPORT_COROSYNC
extern gboolean crm_connect_corosync(crm_cluster_t * cluster);
#endif
static void crm_shutdown(int nsig);
static gboolean crm_read_options(gpointer user_data);
/* A_HA_CONNECT */
void
do_ha_control(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state,
enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
gboolean registered = FALSE;
static crm_cluster_t *cluster = NULL;
if (cluster == NULL) {
cluster = pcmk_cluster_new();
}
if (action & A_HA_DISCONNECT) {
- crm_cluster_disconnect(cluster);
+ pcmk_cluster_disconnect(cluster);
crm_info("Disconnected from the cluster");
controld_set_fsa_input_flags(R_HA_DISCONNECTED);
}
if (action & A_HA_CONNECT) {
crm_set_status_callback(&peer_update_callback);
crm_set_autoreap(FALSE);
#if SUPPORT_COROSYNC
if (is_corosync_cluster()) {
registered = crm_connect_corosync(cluster);
}
#endif // SUPPORT_COROSYNC
if (registered) {
controld_election_init(cluster->uname);
controld_globals.our_nodename = cluster->uname;
controld_globals.our_uuid = cluster->uuid;
if(cluster->uuid == NULL) {
crm_err("Could not obtain local uuid");
registered = FALSE;
}
}
if (!registered) {
controld_set_fsa_input_flags(R_HA_DISCONNECTED);
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
return;
}
populate_cib_nodes(node_update_none, __func__);
controld_clear_fsa_input_flags(R_HA_DISCONNECTED);
crm_info("Connected to the cluster");
}
if (action & ~(A_HA_CONNECT | A_HA_DISCONNECT)) {
crm_err("Unexpected action %s in %s", fsa_action2string(action),
__func__);
}
}
/* A_SHUTDOWN */
void
do_shutdown(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
/* just in case */
controld_set_fsa_input_flags(R_SHUTDOWN);
controld_disconnect_fencer(FALSE);
}
/* A_SHUTDOWN_REQ */
void
do_shutdown_req(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state,
enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
xmlNode *msg = NULL;
controld_set_fsa_input_flags(R_SHUTDOWN);
//controld_set_fsa_input_flags(R_STAYDOWN);
crm_info("Sending shutdown request to all peers (DC is %s)",
pcmk__s(controld_globals.dc_name, "not set"));
msg = create_request(CRM_OP_SHUTDOWN_REQ, NULL, NULL, CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL);
if (send_cluster_message(NULL, crm_msg_crmd, msg, TRUE) == FALSE) {
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
}
free_xml(msg);
}
void
crmd_fast_exit(crm_exit_t exit_code)
{
if (pcmk_is_set(controld_globals.fsa_input_register, R_STAYDOWN)) {
crm_warn("Inhibiting respawn "CRM_XS" remapping exit code %d to %d",
exit_code, CRM_EX_FATAL);
exit_code = CRM_EX_FATAL;
} else if ((exit_code == CRM_EX_OK)
&& pcmk_is_set(controld_globals.fsa_input_register,
R_IN_RECOVERY)) {
crm_err("Could not recover from internal error");
exit_code = CRM_EX_ERROR;
}
if (controld_globals.logger_out != NULL) {
controld_globals.logger_out->finish(controld_globals.logger_out,
exit_code, true, NULL);
pcmk__output_free(controld_globals.logger_out);
controld_globals.logger_out = NULL;
}
crm_exit(exit_code);
}
crm_exit_t
crmd_exit(crm_exit_t exit_code)
{
GMainLoop *mloop = controld_globals.mainloop;
static bool in_progress = FALSE;
if (in_progress && (exit_code == CRM_EX_OK)) {
crm_debug("Exit is already in progress");
return exit_code;
} else if(in_progress) {
crm_notice("Error during shutdown process, exiting now with status %d (%s)",
exit_code, crm_exit_str(exit_code));
crm_write_blackbox(SIGTRAP, NULL);
crmd_fast_exit(exit_code);
}
in_progress = TRUE;
crm_trace("Preparing to exit with status %d (%s)",
exit_code, crm_exit_str(exit_code));
/* Suppress secondary errors resulting from us disconnecting everything */
controld_set_fsa_input_flags(R_HA_DISCONNECTED);
/* Close all IPC servers and clients to ensure any and all shared memory files are cleaned up */
if(ipcs) {
crm_trace("Closing IPC server");
mainloop_del_ipc_server(ipcs);
ipcs = NULL;
}
controld_close_attrd_ipc();
controld_shutdown_schedulerd_ipc();
controld_disconnect_fencer(TRUE);
if ((exit_code == CRM_EX_OK) && (controld_globals.mainloop == NULL)) {
crm_debug("No mainloop detected");
exit_code = CRM_EX_ERROR;
}
/* On an error, just get out.
*
* Otherwise, make the effort to have mainloop exit gracefully so
* that it (mostly) cleans up after itself and valgrind has less
* to report on - allowing real errors stand out
*/
if (exit_code != CRM_EX_OK) {
crm_notice("Forcing immediate exit with status %d (%s)",
exit_code, crm_exit_str(exit_code));
crm_write_blackbox(SIGTRAP, NULL);
crmd_fast_exit(exit_code);
}
/* Clean up as much memory as possible for valgrind */
for (GList *iter = controld_globals.fsa_message_queue; iter != NULL;
iter = iter->next) {
fsa_data_t *fsa_data = (fsa_data_t *) iter->data;
crm_info("Dropping %s: [ state=%s cause=%s origin=%s ]",
fsa_input2string(fsa_data->fsa_input),
fsa_state2string(controld_globals.fsa_state),
fsa_cause2string(fsa_data->fsa_cause), fsa_data->origin);
delete_fsa_input(fsa_data);
}
controld_clear_fsa_input_flags(R_MEMBERSHIP);
g_list_free(controld_globals.fsa_message_queue);
controld_globals.fsa_message_queue = NULL;
controld_free_node_pending_timers();
controld_election_fini();
/* Tear down the CIB manager connection, but don't free it yet -- it could
* be used when we drain the mainloop later.
*/
controld_disconnect_cib_manager();
verify_stopped(controld_globals.fsa_state, LOG_WARNING);
controld_clear_fsa_input_flags(R_LRM_CONNECTED);
lrm_state_destroy_all();
mainloop_destroy_trigger(config_read_trigger);
config_read_trigger = NULL;
controld_destroy_fsa_trigger();
controld_destroy_transition_trigger();
pcmk__client_cleanup();
crm_peer_destroy();
controld_free_fsa_timers();
te_cleanup_stonith_history_sync(NULL, TRUE);
controld_free_sched_timer();
free(controld_globals.our_nodename);
controld_globals.our_nodename = NULL;
free(controld_globals.our_uuid);
controld_globals.our_uuid = NULL;
free(controld_globals.dc_name);
controld_globals.dc_name = NULL;
free(controld_globals.dc_version);
controld_globals.dc_version = NULL;
free(controld_globals.cluster_name);
controld_globals.cluster_name = NULL;
free(controld_globals.te_uuid);
controld_globals.te_uuid = NULL;
free_max_generation();
controld_destroy_failed_sync_table();
controld_destroy_outside_events_table();
mainloop_destroy_signal(SIGPIPE);
mainloop_destroy_signal(SIGUSR1);
mainloop_destroy_signal(SIGTERM);
mainloop_destroy_signal(SIGTRAP);
/* leave SIGCHLD engaged as we might still want to drain some service-actions */
if (mloop) {
GMainContext *ctx = g_main_loop_get_context(controld_globals.mainloop);
/* Don't re-enter this block */
controld_globals.mainloop = NULL;
/* no signals on final draining anymore */
mainloop_destroy_signal(SIGCHLD);
crm_trace("Draining mainloop %d %d", g_main_loop_is_running(mloop), g_main_context_pending(ctx));
{
int lpc = 0;
while((g_main_context_pending(ctx) && lpc < 10)) {
lpc++;
crm_trace("Iteration %d", lpc);
g_main_context_dispatch(ctx);
}
}
crm_trace("Closing mainloop %d %d", g_main_loop_is_running(mloop), g_main_context_pending(ctx));
g_main_loop_quit(mloop);
/* Won't do anything yet, since we're inside it now */
g_main_loop_unref(mloop);
} else {
mainloop_destroy_signal(SIGCHLD);
}
cib_delete(controld_globals.cib_conn);
controld_globals.cib_conn = NULL;
throttle_fini();
/* Graceful */
crm_trace("Done preparing for exit with status %d (%s)",
exit_code, crm_exit_str(exit_code));
return exit_code;
}
/* A_EXIT_0, A_EXIT_1 */
void
do_exit(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
crm_exit_t exit_code = CRM_EX_OK;
if (pcmk_is_set(action, A_EXIT_1)) {
exit_code = CRM_EX_ERROR;
crm_err("Exiting now due to errors");
}
verify_stopped(cur_state, LOG_ERR);
crmd_exit(exit_code);
}
static void sigpipe_ignore(int nsig) { return; }
/* A_STARTUP */
void
do_startup(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
crm_debug("Registering Signal Handlers");
mainloop_add_signal(SIGTERM, crm_shutdown);
mainloop_add_signal(SIGPIPE, sigpipe_ignore);
config_read_trigger = mainloop_add_trigger(G_PRIORITY_HIGH,
crm_read_options, NULL);
controld_init_fsa_trigger();
controld_init_transition_trigger();
crm_debug("Creating CIB manager and executor objects");
controld_globals.cib_conn = cib_new();
lrm_state_init_local();
if (controld_init_fsa_timers() == FALSE) {
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
}
}
// \return libqb error code (0 on success, -errno on error)
static int32_t
accept_controller_client(qb_ipcs_connection_t *c, uid_t uid, gid_t gid)
{
crm_trace("Accepting new IPC client connection");
if (pcmk__new_client(c, uid, gid) == NULL) {
return -ENOMEM;
}
return 0;
}
// \return libqb error code (0 on success, -errno on error)
static int32_t
dispatch_controller_ipc(qb_ipcs_connection_t * c, void *data, size_t size)
{
uint32_t id = 0;
uint32_t flags = 0;
pcmk__client_t *client = pcmk__find_client(c);
xmlNode *msg = pcmk__client_data2xml(client, data, &id, &flags);
if (msg == NULL) {
pcmk__ipc_send_ack(client, id, flags, PCMK__XE_ACK, NULL,
CRM_EX_PROTOCOL);
return 0;
}
pcmk__ipc_send_ack(client, id, flags, PCMK__XE_ACK, NULL,
CRM_EX_INDETERMINATE);
CRM_ASSERT(client->user != NULL);
pcmk__update_acl_user(msg, PCMK__XA_CRM_USER, client->user);
crm_xml_add(msg, PCMK__XA_CRM_SYS_FROM, client->id);
if (controld_authorize_ipc_message(msg, client, NULL)) {
crm_trace("Processing IPC message from client %s",
pcmk__client_name(client));
route_message(C_IPC_MESSAGE, msg);
}
controld_trigger_fsa();
free_xml(msg);
return 0;
}
static int32_t
ipc_client_disconnected(qb_ipcs_connection_t *c)
{
pcmk__client_t *client = pcmk__find_client(c);
if (client) {
crm_trace("Disconnecting %sregistered client %s (%p/%p)",
(client->userdata? "" : "un"), pcmk__client_name(client),
c, client);
free(client->userdata);
pcmk__free_client(client);
controld_trigger_fsa();
}
return 0;
}
static void
ipc_connection_destroyed(qb_ipcs_connection_t *c)
{
crm_trace("Connection %p", c);
ipc_client_disconnected(c);
}
/* A_STOP */
void
do_stop(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
crm_trace("Closing IPC server");
mainloop_del_ipc_server(ipcs); ipcs = NULL;
register_fsa_input(C_FSA_INTERNAL, I_TERMINATE, NULL);
}
/* A_STARTED */
void
do_started(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
static struct qb_ipcs_service_handlers crmd_callbacks = {
.connection_accept = accept_controller_client,
.connection_created = NULL,
.msg_process = dispatch_controller_ipc,
.connection_closed = ipc_client_disconnected,
.connection_destroyed = ipc_connection_destroyed
};
if (cur_state != S_STARTING) {
crm_err("Start cancelled... %s", fsa_state2string(cur_state));
return;
} else if (!pcmk_is_set(controld_globals.fsa_input_register,
R_MEMBERSHIP)) {
crm_info("Delaying start, no membership data (%.16llx)", R_MEMBERSHIP);
crmd_fsa_stall(TRUE);
return;
} else if (!pcmk_is_set(controld_globals.fsa_input_register,
R_LRM_CONNECTED)) {
crm_info("Delaying start, not connected to executor (%.16llx)", R_LRM_CONNECTED);
crmd_fsa_stall(TRUE);
return;
} else if (!pcmk_is_set(controld_globals.fsa_input_register,
R_CIB_CONNECTED)) {
crm_info("Delaying start, CIB not connected (%.16llx)", R_CIB_CONNECTED);
crmd_fsa_stall(TRUE);
return;
} else if (!pcmk_is_set(controld_globals.fsa_input_register,
R_READ_CONFIG)) {
crm_info("Delaying start, Config not read (%.16llx)", R_READ_CONFIG);
crmd_fsa_stall(TRUE);
return;
} else if (!pcmk_is_set(controld_globals.fsa_input_register, R_PEER_DATA)) {
crm_info("Delaying start, No peer data (%.16llx)", R_PEER_DATA);
crmd_fsa_stall(TRUE);
return;
}
crm_debug("Init server comms");
ipcs = pcmk__serve_controld_ipc(&crmd_callbacks);
if (ipcs == NULL) {
crm_err("Failed to create IPC server: shutting down and inhibiting respawn");
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
} else {
crm_notice("Pacemaker controller successfully started and accepting connections");
}
controld_set_fsa_input_flags(R_ST_REQUIRED);
controld_timer_fencer_connect(GINT_TO_POINTER(TRUE));
controld_clear_fsa_input_flags(R_STARTING);
register_fsa_input(msg_data->fsa_cause, I_PENDING, NULL);
}
/* A_RECOVER */
void
do_recover(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
controld_set_fsa_input_flags(R_IN_RECOVERY);
crm_warn("Fast-tracking shutdown in response to errors");
register_fsa_input(C_FSA_INTERNAL, I_TERMINATE, NULL);
}
static void
config_query_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data)
{
const char *value = NULL;
GHashTable *config_hash = NULL;
crm_time_t *now = crm_time_new(NULL);
xmlNode *crmconfig = NULL;
xmlNode *alerts = NULL;
if (rc != pcmk_ok) {
fsa_data_t *msg_data = NULL;
crm_err("Local CIB query resulted in an error: %s", pcmk_strerror(rc));
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
if (rc == -EACCES || rc == -pcmk_err_schema_validation) {
crm_err("The cluster is mis-configured - shutting down and staying down");
controld_set_fsa_input_flags(R_STAYDOWN);
}
goto bail;
}
crmconfig = output;
if ((crmconfig != NULL) && !pcmk__xe_is(crmconfig, PCMK_XE_CRM_CONFIG)) {
crmconfig = pcmk__xe_first_child(crmconfig, PCMK_XE_CRM_CONFIG, NULL,
NULL);
}
if (!crmconfig) {
fsa_data_t *msg_data = NULL;
crm_err("Local CIB query for " PCMK_XE_CRM_CONFIG " section failed");
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
goto bail;
}
crm_debug("Call %d : Parsing CIB options", call_id);
config_hash = pcmk__strkey_table(free, free);
pe_unpack_nvpairs(crmconfig, crmconfig, PCMK_XE_CLUSTER_PROPERTY_SET, NULL,
config_hash, PCMK_VALUE_CIB_BOOTSTRAP_OPTIONS, FALSE, now,
NULL);
// Validate all options, and use defaults if not already present in hash
pcmk__validate_cluster_options(config_hash);
/* Validate the watchdog timeout in the context of the local node
* environment. If invalid, the controller will exit with a fatal error.
*
* We do this via a wrapper in the controller, so that we call
* pcmk__valid_stonith_watchdog_timeout() only if watchdog fencing is
* enabled for the local node. Otherwise, we may exit unnecessarily.
*
* A validator function in libcrmcommon can't act as such a wrapper, because
* it doesn't have a stonith API connection or the local node name.
*/
value = g_hash_table_lookup(config_hash, PCMK_OPT_STONITH_WATCHDOG_TIMEOUT);
controld_verify_stonith_watchdog_timeout(value);
value = g_hash_table_lookup(config_hash, PCMK_OPT_NO_QUORUM_POLICY);
if (pcmk__str_eq(value, PCMK_VALUE_FENCE_LEGACY, pcmk__str_casei)
&& (pcmk__locate_sbd() != 0)) {
controld_set_global_flags(controld_no_quorum_suicide);
}
value = g_hash_table_lookup(config_hash, PCMK_OPT_SHUTDOWN_LOCK);
if (crm_is_true(value)) {
controld_set_global_flags(controld_shutdown_lock_enabled);
} else {
controld_clear_global_flags(controld_shutdown_lock_enabled);
}
value = g_hash_table_lookup(config_hash, PCMK_OPT_SHUTDOWN_LOCK_LIMIT);
pcmk_parse_interval_spec(value, &controld_globals.shutdown_lock_limit);
controld_globals.shutdown_lock_limit /= 1000;
value = g_hash_table_lookup(config_hash, PCMK_OPT_NODE_PENDING_TIMEOUT);
pcmk_parse_interval_spec(value, &controld_globals.node_pending_timeout);
controld_globals.node_pending_timeout /= 1000;
value = g_hash_table_lookup(config_hash, PCMK_OPT_CLUSTER_NAME);
pcmk__str_update(&(controld_globals.cluster_name), value);
// Let subcomponents initialize their own static variables
controld_configure_election(config_hash);
controld_configure_fencing(config_hash);
controld_configure_fsa_timers(config_hash);
controld_configure_throttle(config_hash);
alerts = pcmk__xe_first_child(output, PCMK_XE_ALERTS, NULL, NULL);
crmd_unpack_alerts(alerts);
controld_set_fsa_input_flags(R_READ_CONFIG);
controld_trigger_fsa();
g_hash_table_destroy(config_hash);
bail:
crm_time_free(now);
}
/*!
* \internal
* \brief Trigger read and processing of the configuration
*
* \param[in] fn Calling function name
* \param[in] line Line number where call occurred
*/
void
controld_trigger_config_as(const char *fn, int line)
{
if (config_read_trigger != NULL) {
crm_trace("%s:%d - Triggered config processing", fn, line);
mainloop_set_trigger(config_read_trigger);
}
}
gboolean
crm_read_options(gpointer user_data)
{
cib_t *cib_conn = controld_globals.cib_conn;
int call_id = cib_conn->cmds->query(cib_conn,
"//" PCMK_XE_CRM_CONFIG
" | //" PCMK_XE_ALERTS,
NULL, cib_xpath|cib_scope_local);
fsa_register_cib_callback(call_id, NULL, config_query_callback);
crm_trace("Querying the CIB... call %d", call_id);
return TRUE;
}
/* A_READCONFIG */
void
do_read_config(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state,
enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
throttle_init();
controld_trigger_config();
}
static void
crm_shutdown(int nsig)
{
const char *value = NULL;
guint default_period_ms = 0;
if ((controld_globals.mainloop == NULL)
|| !g_main_loop_is_running(controld_globals.mainloop)) {
crmd_exit(CRM_EX_OK);
return;
}
if (pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)) {
crm_err("Escalating shutdown");
register_fsa_input_before(C_SHUTDOWN, I_ERROR, NULL);
return;
}
controld_set_fsa_input_flags(R_SHUTDOWN);
register_fsa_input(C_SHUTDOWN, I_SHUTDOWN, NULL);
/* If shutdown timer doesn't have a period set, use the default
*
* @TODO: Evaluate whether this is still necessary. As long as
* config_query_callback() has been run at least once, it doesn't look like
* anything could have changed the timer period since then.
*/
value = pcmk__cluster_option(NULL, PCMK_OPT_SHUTDOWN_ESCALATION);
pcmk_parse_interval_spec(value, &default_period_ms);
controld_shutdown_start_countdown(default_period_ms);
}
diff --git a/include/crm/cluster.h b/include/crm/cluster.h
index 1962e6a455..117070225d 100644
--- a/include/crm/cluster.h
+++ b/include/crm/cluster.h
@@ -1,241 +1,242 @@
/*
* Copyright 2004-2024 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU Lesser General Public License
* version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY.
*/
#ifndef PCMK__CRM_CLUSTER__H
# define PCMK__CRM_CLUSTER__H
# include <stdint.h> // uint32_t, uint64_t
# include <glib.h> // gboolean, GHashTable
# include <libxml/tree.h> // xmlNode
# include <crm/common/xml.h>
# include <crm/common/util.h>
#ifdef __cplusplus
extern "C" {
#endif
# if SUPPORT_COROSYNC
# include <corosync/cpg.h>
# endif
extern gboolean crm_have_quorum;
extern GHashTable *crm_peer_cache;
extern GHashTable *crm_remote_peer_cache;
extern unsigned long long crm_peer_seq;
#define CRM_NODE_LOST "lost"
#define CRM_NODE_MEMBER "member"
enum crm_join_phase {
/* @COMPAT: crm_join_nack_quiet can be replaced by crm_node_t:user_data
* at a compatibility break.
*/
//! Not allowed to join, but don't send a nack message
crm_join_nack_quiet = -2,
crm_join_nack = -1,
crm_join_none = 0,
crm_join_welcomed = 1,
crm_join_integrated = 2,
crm_join_finalized = 3,
crm_join_confirmed = 4,
};
enum crm_node_flags {
/* node is not a cluster node and should not be considered for cluster membership */
crm_remote_node = 0x0001,
/* node's cache entry is dirty */
crm_node_dirty = 0x0010,
};
typedef struct crm_peer_node_s {
char *uname; // Node name as known to cluster
/* @COMPAT This is less than ideal since the value is not a valid XML ID
* (for Corosync, it's the string equivalent of the node's numeric node ID,
* but XML IDs can't start with a number) and the three elements should have
* different IDs.
*
* Ideally, we would use something like node-NODEID, node_state-NODEID, and
* transient_attributes-NODEID as the element IDs. Unfortunately changing it
* would be impractical due to backward compatibility; older nodes in a
* rolling upgrade will always write and expect the value in the old format.
*
* This is also named poorly, since the value is not a UUID, but at least
* that can be changed at an API compatibility break.
*/
/*! Value of the PCMK_XA_ID XML attribute to use with the node's
* PCMK_XE_NODE, PCMK_XE_NODE_STATE, and PCMK_XE_TRANSIENT_ATTRIBUTES
* XML elements in the CIB
*/
char *uuid;
char *state; // @TODO change to enum
uint64_t flags; // Bitmask of crm_node_flags
uint64_t last_seen; // Only needed by cluster nodes
uint32_t processes; // @TODO most not needed, merge into flags
/* @TODO When we can break public API compatibility, we can make the rest of
* these members separate structs and use void *cluster_data and
* void *user_data here instead, to abstract the cluster layer further.
*/
// Currently only needed by corosync stack
uint32_t id; // Node ID
time_t when_lost; // When CPG membership was last lost
// Only used by controller
enum crm_join_phase join;
char *expected;
time_t peer_lost;
char *conn_host;
time_t when_member; // Since when node has been a cluster member
time_t when_online; // Since when peer has been online in CPG
} crm_node_t;
void crm_peer_init(void);
void crm_peer_destroy(void);
typedef struct crm_cluster_s {
char *uuid;
char *uname;
uint32_t nodeid;
void (*destroy) (gpointer);
# if SUPPORT_COROSYNC
/* @TODO When we can break public API compatibility, make these members a
* separate struct and use void *cluster_data here instead, to abstract the
* cluster layer further.
*/
struct cpg_name group;
cpg_callbacks_t cpg;
cpg_handle_t cpg_handle;
# endif
} crm_cluster_t;
int pcmk_cluster_connect(crm_cluster_t *cluster);
+int pcmk_cluster_disconnect(crm_cluster_t *cluster);
void crm_cluster_disconnect(crm_cluster_t *cluster);
crm_cluster_t *pcmk_cluster_new(void);
void pcmk_cluster_free(crm_cluster_t *cluster);
enum crm_ais_msg_class {
crm_class_cluster = 0,
};
enum crm_ais_msg_types {
crm_msg_none = 0,
crm_msg_ais = 1,
crm_msg_lrmd = 2,
crm_msg_cib = 3,
crm_msg_crmd = 4,
crm_msg_attrd = 5,
crm_msg_stonithd = 6,
crm_msg_te = 7,
crm_msg_pe = 8,
crm_msg_stonith_ng = 9,
};
gboolean send_cluster_message(const crm_node_t *node,
enum crm_ais_msg_types service,
const xmlNode *data, gboolean ordered);
int crm_remote_peer_cache_size(void);
/* Initialize and refresh the remote peer cache from a cib config */
void crm_remote_peer_cache_refresh(xmlNode *cib);
crm_node_t *crm_remote_peer_get(const char *node_name);
void crm_remote_peer_cache_remove(const char *node_name);
guint crm_active_peers(void);
gboolean crm_is_peer_active(const crm_node_t * node);
guint reap_crm_member(uint32_t id, const char *name);
# if SUPPORT_COROSYNC
uint32_t get_local_nodeid(cpg_handle_t handle);
void pcmk_cpg_membership(cpg_handle_t handle,
const struct cpg_name *groupName,
const struct cpg_address *member_list, size_t member_list_entries,
const struct cpg_address *left_list, size_t left_list_entries,
const struct cpg_address *joined_list, size_t joined_list_entries);
gboolean crm_is_corosync_peer_active(const crm_node_t * node);
gboolean send_cluster_text(enum crm_ais_msg_class msg_class, const char *data,
gboolean local, const crm_node_t *node,
enum crm_ais_msg_types dest);
char *pcmk_message_common_cs(cpg_handle_t handle, uint32_t nodeid, uint32_t pid, void *msg,
uint32_t *kind, const char **from);
# endif
const char *crm_peer_uuid(crm_node_t *node);
const char *crm_peer_uname(const char *uuid);
enum crm_status_type {
crm_status_uname,
crm_status_nstate,
crm_status_processes,
};
enum crm_ais_msg_types text2msg_type(const char *text);
void crm_set_status_callback(void (*dispatch) (enum crm_status_type, crm_node_t *, const void *));
void crm_set_autoreap(gboolean autoreap);
enum cluster_type_e {
pcmk_cluster_unknown = 0x0001,
pcmk_cluster_invalid = 0x0002,
// 0x0004 was heartbeat
// 0x0010 was corosync 1 with plugin
pcmk_cluster_corosync = 0x0020,
// 0x0040 was corosync 1 with CMAN
};
enum cluster_type_e get_cluster_type(void);
const char *name_for_cluster_type(enum cluster_type_e type);
gboolean is_corosync_cluster(void);
const char *get_local_node_name(void);
char *get_node_name(uint32_t nodeid);
/*!
* \brief Get log-friendly string equivalent of a join phase
*
* \param[in] phase Join phase
*
* \return Log-friendly string equivalent of \p phase
*/
static inline const char *
crm_join_phase_str(enum crm_join_phase phase)
{
switch (phase) {
case crm_join_nack_quiet: return "nack_quiet";
case crm_join_nack: return "nack";
case crm_join_none: return "none";
case crm_join_welcomed: return "welcomed";
case crm_join_integrated: return "integrated";
case crm_join_finalized: return "finalized";
case crm_join_confirmed: return "confirmed";
default: return "invalid";
}
}
#if !defined(PCMK_ALLOW_DEPRECATED) || (PCMK_ALLOW_DEPRECATED == 1)
#include <crm/cluster/compat.h>
#endif
#ifdef __cplusplus
}
#endif
#endif
diff --git a/lib/cluster/cluster.c b/lib/cluster/cluster.c
index a355c38244..5e1407bab5 100644
--- a/lib/cluster/cluster.c
+++ b/lib/cluster/cluster.c
@@ -1,415 +1,435 @@
/*
* Copyright 2004-2024 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU Lesser General Public License
* version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY.
*/
#include <crm_internal.h>
#include <dlfcn.h>
#include <stdio.h>
#include <unistd.h>
#include <string.h>
#include <stdlib.h>
#include <time.h>
#include <sys/param.h>
#include <sys/types.h>
#include <crm/crm.h>
#include <crm/common/ipc.h>
#include <crm/common/xml.h>
#include <crm/cluster/internal.h>
#include "crmcluster_private.h"
CRM_TRACE_INIT_DATA(cluster);
/*!
* \brief Get (and set if needed) a node's UUID
*
* \param[in,out] peer Node to check
*
* \return Node UUID of \p peer, or NULL if unknown
*/
const char *
crm_peer_uuid(crm_node_t *peer)
{
char *uuid = NULL;
// Check simple cases first, to avoid any calls that might block
if (peer == NULL) {
return NULL;
}
if (peer->uuid != NULL) {
return peer->uuid;
}
switch (get_cluster_type()) {
case pcmk_cluster_corosync:
#if SUPPORT_COROSYNC
uuid = pcmk__corosync_uuid(peer);
#endif
break;
case pcmk_cluster_unknown:
case pcmk_cluster_invalid:
crm_err("Unsupported cluster type");
break;
}
peer->uuid = uuid;
return peer->uuid;
}
/*!
* \internal
* \brief Connect to the cluster layer
*
* \param[in,out] cluster Initialized cluster object to connect
*
* \return Standard Pacemaker return code
*/
int
pcmk_cluster_connect(crm_cluster_t *cluster)
{
enum cluster_type_e type = get_cluster_type();
crm_notice("Connecting to %s cluster infrastructure",
name_for_cluster_type(type));
switch (type) {
case pcmk_cluster_corosync:
#if SUPPORT_COROSYNC
crm_peer_init();
return pcmk__corosync_connect(cluster);
#else
break;
#endif // SUPPORT_COROSYNC
default:
break;
}
crm_err("Failed to connect to unsupported cluster type %s",
name_for_cluster_type(type));
return EPROTONOSUPPORT;
}
/*!
* \brief Disconnect from the cluster layer
*
* \param[in,out] cluster Cluster object to disconnect
+ *
+ * \return Standard Pacemaker return code
*/
-void
-crm_cluster_disconnect(crm_cluster_t *cluster)
+int
+pcmk_cluster_disconnect(crm_cluster_t *cluster)
{
enum cluster_type_e type = get_cluster_type();
crm_info("Disconnecting from %s cluster infrastructure",
name_for_cluster_type(type));
+
switch (type) {
case pcmk_cluster_corosync:
#if SUPPORT_COROSYNC
crm_peer_destroy();
pcmk__corosync_disconnect(cluster);
-#endif // SUPPORT_COROSYNC
+ return pcmk_rc_ok;
+#else
break;
+#endif // SUPPORT_COROSYNC
default:
break;
}
+
+ crm_err("Failed to disconnect from unsupported cluster type %s",
+ name_for_cluster_type(type));
+ return EPROTONOSUPPORT;
+}
+
+/*!
+ * \brief Disconnect from the cluster layer
+ *
+ * \param[in,out] cluster Cluster object to disconnect
+ */
+void
+crm_cluster_disconnect(crm_cluster_t *cluster)
+{
+ pcmk_cluster_disconnect(cluster);
}
/*!
* \brief Allocate a new \p crm_cluster_t object
*
* \return A newly allocated \p crm_cluster_t object (guaranteed not \c NULL)
* \note The caller is responsible for freeing the return value using
* \p pcmk_cluster_free().
*/
crm_cluster_t *
pcmk_cluster_new(void)
{
return (crm_cluster_t *) pcmk__assert_alloc(1, sizeof(crm_cluster_t));
}
/*!
* \brief Free a \p crm_cluster_t object and its dynamically allocated members
*
* \param[in,out] cluster Cluster object to free
*/
void
pcmk_cluster_free(crm_cluster_t *cluster)
{
if (cluster == NULL) {
return;
}
free(cluster->uuid);
free(cluster->uname);
free(cluster);
}
/*!
* \brief Send an XML message via the cluster messaging layer
*
* \param[in] node Cluster node to send message to
* \param[in] service Message type to use in message host info
* \param[in] data XML message to send
* \param[in] ordered Ignored for currently supported messaging layers
*
* \return TRUE on success, otherwise FALSE
*/
gboolean
send_cluster_message(const crm_node_t *node, enum crm_ais_msg_types service,
const xmlNode *data, gboolean ordered)
{
switch (get_cluster_type()) {
case pcmk_cluster_corosync:
#if SUPPORT_COROSYNC
return pcmk__cpg_send_xml(data, node, service);
#endif
break;
default:
break;
}
return FALSE;
}
/*!
* \brief Get the local node's name
*
* \return Local node's name
* \note This will fatally exit if local node name cannot be known.
*/
const char *
get_local_node_name(void)
{
static char *name = NULL;
if (name == NULL) {
name = get_node_name(0);
}
return name;
}
/*!
* \brief Get the node name corresponding to a cluster node ID
*
* \param[in] nodeid Node ID to check (or 0 for local node)
*
* \return Node name corresponding to \p nodeid
* \note This will fatally exit if \p nodeid is 0 and local node name cannot be
* known.
*/
char *
get_node_name(uint32_t nodeid)
{
char *name = NULL;
enum cluster_type_e stack = get_cluster_type();
switch (stack) {
case pcmk_cluster_corosync:
#if SUPPORT_COROSYNC
name = pcmk__corosync_name(0, nodeid);
break;
#endif // SUPPORT_COROSYNC
default:
crm_err("Unknown cluster type: %s (%d)", name_for_cluster_type(stack), stack);
}
if ((name == NULL) && (nodeid == 0)) {
name = pcmk_hostname();
if (name == NULL) {
// @TODO Maybe let the caller decide what to do
crm_err("Could not obtain the local %s node name",
name_for_cluster_type(stack));
crm_exit(CRM_EX_FATAL);
}
crm_notice("Defaulting to uname -n for the local %s node name",
name_for_cluster_type(stack));
}
if (name == NULL) {
crm_notice("Could not obtain a node name for %s node with "
PCMK_XA_ID " %u", name_for_cluster_type(stack), nodeid);
}
return name;
}
/*!
* \brief Get the node name corresponding to a node UUID
*
* \param[in] uuid UUID of desired node
*
* \return name of desired node
*
* \note This relies on the remote peer cache being populated with all
* remote nodes in the cluster, so callers should maintain that cache.
*/
const char *
crm_peer_uname(const char *uuid)
{
GHashTableIter iter;
crm_node_t *node = NULL;
CRM_CHECK(uuid != NULL, return NULL);
/* remote nodes have the same uname and uuid */
if (g_hash_table_lookup(crm_remote_peer_cache, uuid)) {
return uuid;
}
/* avoid blocking calls where possible */
g_hash_table_iter_init(&iter, crm_peer_cache);
while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) {
if (pcmk__str_eq(node->uuid, uuid, pcmk__str_casei)) {
if (node->uname != NULL) {
return node->uname;
}
break;
}
}
node = NULL;
if (is_corosync_cluster()) {
long long id;
if ((pcmk__scan_ll(uuid, &id, 0LL) != pcmk_rc_ok)
|| (id < 1LL) || (id > UINT32_MAX)) {
crm_err("Invalid Corosync node ID '%s'", uuid);
return NULL;
}
node = pcmk__search_node_caches((uint32_t) id, NULL,
pcmk__node_search_cluster);
if (node != NULL) {
crm_info("Setting uuid for node %s[%u] to %s",
node->uname, node->id, uuid);
node->uuid = strdup(uuid);
return node->uname;
}
return NULL;
}
return NULL;
}
/*!
* \brief Get a log-friendly string equivalent of a cluster type
*
* \param[in] type Cluster type
*
* \return Log-friendly string corresponding to \p type
*/
const char *
name_for_cluster_type(enum cluster_type_e type)
{
switch (type) {
case pcmk_cluster_corosync:
return "corosync";
case pcmk_cluster_unknown:
return "unknown";
case pcmk_cluster_invalid:
return "invalid";
}
crm_err("Invalid cluster type: %d", type);
return "invalid";
}
/*!
* \brief Get (and validate) the local cluster type
*
* \return Local cluster type
* \note This will fatally exit if the local cluster type is invalid.
*/
enum cluster_type_e
get_cluster_type(void)
{
bool detected = false;
const char *cluster = NULL;
static enum cluster_type_e cluster_type = pcmk_cluster_unknown;
/* Return the previous calculation, if any */
if (cluster_type != pcmk_cluster_unknown) {
return cluster_type;
}
cluster = pcmk__env_option(PCMK__ENV_CLUSTER_TYPE);
#if SUPPORT_COROSYNC
/* If nothing is defined in the environment, try corosync (if supported) */
if (cluster == NULL) {
crm_debug("Testing with Corosync");
cluster_type = pcmk__corosync_detect();
if (cluster_type != pcmk_cluster_unknown) {
detected = true;
goto done;
}
}
#endif
/* Something was defined in the environment, test it against what we support */
crm_info("Verifying cluster type: '%s'",
((cluster == NULL)? "-unspecified-" : cluster));
if (cluster == NULL) {
#if SUPPORT_COROSYNC
} else if (pcmk__str_eq(cluster, "corosync", pcmk__str_casei)) {
cluster_type = pcmk_cluster_corosync;
#endif
} else {
cluster_type = pcmk_cluster_invalid;
goto done; /* Keep the compiler happy when no stacks are supported */
}
done:
if (cluster_type == pcmk_cluster_unknown) {
crm_notice("Could not determine the current cluster type");
} else if (cluster_type == pcmk_cluster_invalid) {
crm_notice("This installation does not support the '%s' cluster infrastructure: terminating.",
cluster);
crm_exit(CRM_EX_FATAL);
} else {
crm_info("%s an active '%s' cluster",
(detected? "Detected" : "Assuming"),
name_for_cluster_type(cluster_type));
}
return cluster_type;
}
/*!
* \brief Check whether the local cluster is a Corosync cluster
*
* \return TRUE if the local cluster is a Corosync cluster, otherwise FALSE
*/
gboolean
is_corosync_cluster(void)
{
return get_cluster_type() == pcmk_cluster_corosync;
}
// Deprecated functions kept only for backward API compatibility
// LCOV_EXCL_START
#include <crm/cluster/compat.h>
void
set_uuid(xmlNode *xml, const char *attr, crm_node_t *node)
{
crm_xml_add(xml, attr, crm_peer_uuid(node));
}
gboolean
crm_cluster_connect(crm_cluster_t *cluster)
{
return pcmk_cluster_connect(cluster) == pcmk_rc_ok;
}
// LCOV_EXCL_STOP
// End deprecated API