diff --git a/exec/main.c b/exec/main.c
index 5fb4d47c..977aaf5e 100644
--- a/exec/main.c
+++ b/exec/main.c
@@ -1,1665 +1,1666 @@
/*
* Copyright (c) 2002-2006 MontaVista Software, Inc.
* Copyright (c) 2006-2021 Red Hat, Inc.
*
* All rights reserved.
*
* Author: Steven Dake (sdake@redhat.com)
*
* This software licensed under BSD license, the text of which follows:
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* - Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
* - Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
* - Neither the name of the MontaVista Software, Inc. nor the names of its
* contributors may be used to endorse or promote products derived from this
* software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
* THE POSSIBILITY OF SUCH DAMAGE.
*/
/**
* \mainpage Corosync
*
* This is the doxygen generated developer documentation for the Corosync
* project. For more information about Corosync, please see the project
* web site, corosync.org.
*
* \section license License
*
* This software licensed under BSD license, the text of which follows:
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* - Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
* - Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
* - Neither the name of the MontaVista Software, Inc. nor the names of its
* contributors may be used to endorse or promote products derived from this
* software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
* THE POSSIBILITY OF SUCH DAMAGE.
*/
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#ifdef HAVE_LIBSYSTEMD
#include
#endif
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include "quorum.h"
#include "totemsrp.h"
#include "logconfig.h"
#include "totemconfig.h"
#include "main.h"
#include "sync.h"
#include "timer.h"
#include "util.h"
#include "apidef.h"
#include "service.h"
#include "schedwrk.h"
#include "ipcs_stats.h"
#include "stats.h"
#ifdef HAVE_SMALL_MEMORY_FOOTPRINT
#define IPC_LOGSYS_SIZE 1024*64
#else
#define IPC_LOGSYS_SIZE 8192*128
#endif
/*
* LibQB adds default "*" syslog filter so we have to set syslog_priority as low
* as possible so filters applied later in _logsys_config_apply_per_file takes
* effect.
*/
LOGSYS_DECLARE_SYSTEM ("corosync",
LOGSYS_MODE_OUTPUT_STDERR | LOGSYS_MODE_OUTPUT_SYSLOG,
LOG_DAEMON,
LOG_EMERG);
LOGSYS_DECLARE_SUBSYS ("MAIN");
#define SERVER_BACKLOG 5
static int sched_priority = 0;
static unsigned int service_count = 32;
static struct totem_logging_configuration totem_logging_configuration;
static struct corosync_api_v1 *api = NULL;
static int sync_in_process = 1;
static qb_loop_t *corosync_poll_handle;
struct sched_param global_sched_param;
static corosync_timer_handle_t corosync_stats_timer_handle;
static const char *corosync_lock_file = LOCALSTATEDIR"/run/corosync.pid";
static char corosync_config_file[PATH_MAX + 1] = COROSYSCONFDIR "/corosync.conf";
static int lockfile_fd = -1;
enum move_to_root_cgroup_mode {
MOVE_TO_ROOT_CGROUP_MODE_OFF = 0,
MOVE_TO_ROOT_CGROUP_MODE_ON = 1,
MOVE_TO_ROOT_CGROUP_MODE_AUTO = 2,
};
qb_loop_t *cs_poll_handle_get (void)
{
return (corosync_poll_handle);
}
int cs_poll_dispatch_add (qb_loop_t * handle,
int fd,
int events,
void *data,
int (*dispatch_fn) (int fd,
int revents,
void *data))
{
return qb_loop_poll_add(handle, QB_LOOP_MED, fd, events, data,
dispatch_fn);
}
int cs_poll_dispatch_delete(qb_loop_t * handle, int fd)
{
return qb_loop_poll_del(handle, fd);
}
void corosync_state_dump (void)
{
int i;
for (i = 0; i < SERVICES_COUNT_MAX; i++) {
if (corosync_service[i] && corosync_service[i]->exec_dump_fn) {
corosync_service[i]->exec_dump_fn ();
}
}
}
const char *corosync_get_config_file(void)
{
return (corosync_config_file);
}
static void corosync_blackbox_write_to_file (void)
{
char fname[PATH_MAX];
char fdata_fname[PATH_MAX];
char time_str[PATH_MAX];
struct tm cur_time_tm;
time_t cur_time_t;
ssize_t res;
cur_time_t = time(NULL);
localtime_r(&cur_time_t, &cur_time_tm);
strftime(time_str, PATH_MAX, "%Y-%m-%dT%H:%M:%S", &cur_time_tm);
if (snprintf(fname, PATH_MAX, "%s/fdata-%s-%lld",
get_state_dir(),
time_str,
(long long int)getpid()) >= PATH_MAX) {
log_printf(LOGSYS_LEVEL_ERROR, "Can't snprintf blackbox file name");
return ;
}
if ((res = qb_log_blackbox_write_to_file(fname)) < 0) {
LOGSYS_PERROR(-res, LOGSYS_LEVEL_ERROR, "Can't store blackbox file");
return ;
}
snprintf(fdata_fname, sizeof(fdata_fname), "%s/fdata", get_state_dir());
unlink(fdata_fname);
if (symlink(fname, fdata_fname) == -1) {
log_printf(LOGSYS_LEVEL_ERROR, "Can't create symlink to '%s' for corosync blackbox file '%s'",
fname, fdata_fname);
}
}
static void unlink_all_completed (void)
{
api->timer_delete (corosync_stats_timer_handle);
qb_loop_stop (corosync_poll_handle);
icmap_fini();
}
void corosync_shutdown_request (void)
{
corosync_service_unlink_all (api, unlink_all_completed);
}
static int32_t sig_diag_handler (int num, void *data)
{
corosync_state_dump ();
return 0;
}
static int32_t sig_exit_handler (int num, void *data)
{
log_printf(LOGSYS_LEVEL_NOTICE, "Node was shut down by a signal");
corosync_service_unlink_all (api, unlink_all_completed);
return 0;
}
static void sigsegv_handler (int num)
{
(void)signal (num, SIG_DFL);
corosync_blackbox_write_to_file ();
qb_log_fini();
raise (num);
}
#define LOCALHOST_IP inet_addr("127.0.0.1")
static void *corosync_group_handle;
static struct totempg_group corosync_group = {
.group = "a",
.group_len = 1
};
static void serialize_lock (void)
{
}
static void serialize_unlock (void)
{
}
static void corosync_sync_completed (void)
{
log_printf (LOGSYS_LEVEL_NOTICE,
"Completed service synchronization, ready to provide service.");
sync_in_process = 0;
cs_ipcs_sync_state_changed(sync_in_process);
cs_ipc_allow_connections(1);
/*
* Inform totem to start using new message queue again
*/
totempg_trans_ack();
#ifdef HAVE_LIBSYSTEMD
sd_notify (0, "READY=1");
#endif
}
static int corosync_sync_callbacks_retrieve (
int service_id,
struct sync_callbacks *callbacks)
{
if (corosync_service[service_id] == NULL) {
return (-1);
}
if (callbacks == NULL) {
return (0);
}
callbacks->name = corosync_service[service_id]->name;
callbacks->sync_init = corosync_service[service_id]->sync_init;
callbacks->sync_process = corosync_service[service_id]->sync_process;
callbacks->sync_activate = corosync_service[service_id]->sync_activate;
callbacks->sync_abort = corosync_service[service_id]->sync_abort;
return (0);
}
static struct memb_ring_id corosync_ring_id;
static void member_object_joined (unsigned int nodeid)
{
char member_ip[ICMAP_KEYNAME_MAXLEN];
char member_join_count[ICMAP_KEYNAME_MAXLEN];
char member_status[ICMAP_KEYNAME_MAXLEN];
snprintf(member_ip, ICMAP_KEYNAME_MAXLEN,
"runtime.members.%u.ip", nodeid);
snprintf(member_join_count, ICMAP_KEYNAME_MAXLEN,
"runtime.members.%u.join_count", nodeid);
snprintf(member_status, ICMAP_KEYNAME_MAXLEN,
"runtime.members.%u.status", nodeid);
if (icmap_get(member_ip, NULL, NULL, NULL) == CS_OK) {
icmap_inc(member_join_count);
icmap_set_string(member_status, "joined");
} else {
icmap_set_string(member_ip, (char*)api->totem_ifaces_print (nodeid));
icmap_set_uint32(member_join_count, 1);
icmap_set_string(member_status, "joined");
}
log_printf (LOGSYS_LEVEL_DEBUG,
"Member joined: %s", api->totem_ifaces_print (nodeid));
}
static void member_object_left (unsigned int nodeid)
{
char member_status[ICMAP_KEYNAME_MAXLEN];
snprintf(member_status, ICMAP_KEYNAME_MAXLEN,
"runtime.members.%u.status", nodeid);
icmap_set_string(member_status, "left");
log_printf (LOGSYS_LEVEL_DEBUG,
"Member left: %s", api->totem_ifaces_print (nodeid));
}
static void confchg_fn (
enum totem_configuration_type configuration_type,
const unsigned int *member_list, size_t member_list_entries,
const unsigned int *left_list, size_t left_list_entries,
const unsigned int *joined_list, size_t joined_list_entries,
const struct memb_ring_id *ring_id)
{
int i;
int abort_activate = 0;
if (sync_in_process == 1) {
abort_activate = 1;
}
sync_in_process = 1;
cs_ipcs_sync_state_changed(sync_in_process);
memcpy (&corosync_ring_id, ring_id, sizeof (struct memb_ring_id));
for (i = 0; i < left_list_entries; i++) {
member_object_left (left_list[i]);
}
for (i = 0; i < joined_list_entries; i++) {
member_object_joined (joined_list[i]);
}
/*
* Call configuration change for all services
*/
for (i = 0; i < service_count; i++) {
if (corosync_service[i] && corosync_service[i]->confchg_fn) {
corosync_service[i]->confchg_fn (configuration_type,
member_list, member_list_entries,
left_list, left_list_entries,
joined_list, joined_list_entries, ring_id);
}
}
if (abort_activate) {
sync_abort ();
}
if (configuration_type == TOTEM_CONFIGURATION_TRANSITIONAL) {
sync_save_transitional (member_list, member_list_entries, ring_id);
}
if (configuration_type == TOTEM_CONFIGURATION_REGULAR) {
sync_start (member_list, member_list_entries, ring_id);
}
}
static void priv_drop (void)
{
return; /* TODO: we are still not dropping privs */
}
static void corosync_tty_detach (void)
{
int devnull;
/*
* Disconnect from TTY if this is not a debug run
*/
switch (fork ()) {
case -1:
corosync_exit_error (COROSYNC_DONE_FORK);
break;
case 0:
/*
* child which is disconnected, run this process
*/
break;
default:
exit (0);
break;
}
/* Create new session */
(void)setsid();
/*
* Map stdin/out/err to /dev/null.
*/
devnull = open("/dev/null", O_RDWR);
if (devnull == -1) {
corosync_exit_error (COROSYNC_DONE_STD_TO_NULL_REDIR);
}
if (dup2(devnull, 0) < 0 || dup2(devnull, 1) < 0
|| dup2(devnull, 2) < 0) {
close(devnull);
corosync_exit_error (COROSYNC_DONE_STD_TO_NULL_REDIR);
}
close(devnull);
}
static void corosync_mlockall (void)
{
int res;
struct rlimit rlimit;
rlimit.rlim_cur = RLIM_INFINITY;
rlimit.rlim_max = RLIM_INFINITY;
#ifndef RLIMIT_MEMLOCK
#define RLIMIT_MEMLOCK RLIMIT_VMEM
#endif
res = setrlimit (RLIMIT_MEMLOCK, &rlimit);
if (res == -1) {
LOGSYS_PERROR (errno, LOGSYS_LEVEL_WARNING,
"Could not increase RLIMIT_MEMLOCK, not locking memory");
return;
}
res = mlockall (MCL_CURRENT | MCL_FUTURE);
if (res == -1) {
LOGSYS_PERROR (errno, LOGSYS_LEVEL_WARNING,
"Could not lock memory of service to avoid page faults");
};
}
static void corosync_totem_stats_updater (void *data)
{
totempg_stats_t * stats;
uint32_t total_mtt_rx_token;
uint32_t total_backlog_calc;
uint32_t total_token_holdtime;
int t, prev;
int32_t token_count;
const char *cstr;
stats = api->totem_get_stats();
stats->srp->firewall_enabled_or_nic_failure = stats->srp->continuous_gather > MAX_NO_CONT_GATHER ? 1 : 0;
if (stats->srp->continuous_gather > MAX_NO_CONT_GATHER ||
stats->srp->continuous_sendmsg_failures > MAX_NO_CONT_SENDMSG_FAILURES) {
cstr = "";
if (stats->srp->continuous_sendmsg_failures > MAX_NO_CONT_SENDMSG_FAILURES) {
cstr = "number of multicast sendmsg failures is above threshold";
}
if (stats->srp->continuous_gather > MAX_NO_CONT_GATHER) {
cstr = "totem is continuously in gather state";
}
log_printf (LOGSYS_LEVEL_WARNING,
"Totem is unable to form a cluster because of an "
"operating system or network fault (reason: %s). The most common "
"cause of this message is that the local firewall is "
"configured improperly.", cstr);
stats->srp->firewall_enabled_or_nic_failure = 1;
} else {
stats->srp->firewall_enabled_or_nic_failure = 0;
}
total_mtt_rx_token = 0;
total_token_holdtime = 0;
total_backlog_calc = 0;
token_count = 0;
t = stats->srp->latest_token;
while (1) {
if (t == 0)
prev = TOTEM_TOKEN_STATS_MAX - 1;
else
prev = t - 1;
if (prev == stats->srp->earliest_token)
break;
/* if tx == 0, then dropped token (not ours) */
if (stats->srp->token[t].tx != 0 ||
(stats->srp->token[t].rx - stats->srp->token[prev].rx) > 0 ) {
total_mtt_rx_token += (stats->srp->token[t].rx - stats->srp->token[prev].rx);
total_token_holdtime += (stats->srp->token[t].tx - stats->srp->token[t].rx);
total_backlog_calc += stats->srp->token[t].backlog_calc;
token_count++;
}
t = prev;
}
if (token_count) {
stats->srp->mtt_rx_token = (total_mtt_rx_token / token_count);
stats->srp->avg_token_workload = (total_token_holdtime / token_count);
stats->srp->avg_backlog_calc = (total_backlog_calc / token_count);
}
stats->srp->time_since_token_last_received = qb_util_nano_current_get () / QB_TIME_NS_IN_MSEC -
stats->srp->token[stats->srp->latest_token].rx;
stats_trigger_trackers();
api->timer_add_duration (1500 * MILLI_2_NANO_SECONDS, NULL,
corosync_totem_stats_updater,
&corosync_stats_timer_handle);
}
static void corosync_totem_stats_init (void)
{
/* start stats timer */
api->timer_add_duration (1500 * MILLI_2_NANO_SECONDS, NULL,
corosync_totem_stats_updater,
&corosync_stats_timer_handle);
}
static void deliver_fn (
unsigned int nodeid,
const void *msg,
unsigned int msg_len,
int endian_conversion_required)
{
const struct qb_ipc_request_header *header;
int32_t service;
int32_t fn_id;
uint32_t id;
header = msg;
if (endian_conversion_required) {
id = swab32 (header->id);
} else {
id = header->id;
}
/*
* Call the proper executive handler
*/
service = id >> 16;
fn_id = id & 0xffff;
if (!corosync_service[service]) {
return;
}
if (fn_id >= corosync_service[service]->exec_engine_count) {
log_printf(LOGSYS_LEVEL_WARNING, "discarded unknown message %d for service %d (max id %d)",
fn_id, service, corosync_service[service]->exec_engine_count);
return;
}
icmap_fast_inc(service_stats_rx[service][fn_id]);
if (endian_conversion_required) {
assert(corosync_service[service]->exec_engine[fn_id].exec_endian_convert_fn != NULL);
corosync_service[service]->exec_engine[fn_id].exec_endian_convert_fn
((void *)msg);
}
corosync_service[service]->exec_engine[fn_id].exec_handler_fn
(msg, nodeid);
}
int main_mcast (
const struct iovec *iovec,
unsigned int iov_len,
unsigned int guarantee)
{
const struct qb_ipc_request_header *req = iovec->iov_base;
int32_t service;
int32_t fn_id;
service = req->id >> 16;
fn_id = req->id & 0xffff;
if (corosync_service[service]) {
icmap_fast_inc(service_stats_tx[service][fn_id]);
}
return (totempg_groups_mcast_joined (corosync_group_handle, iovec, iov_len, guarantee));
}
static void corosync_ring_id_create_or_load (
struct memb_ring_id *memb_ring_id,
unsigned int nodeid)
{
int fd;
int res = 0;
char filename[PATH_MAX];
snprintf (filename, sizeof(filename), "%s/ringid_%u",
get_state_dir(), nodeid);
fd = open (filename, O_RDONLY);
/*
* If file can be opened and read, read the ring id
*/
if (fd != -1) {
res = read (fd, &memb_ring_id->seq, sizeof (uint64_t));
close (fd);
}
/*
* If file could not be opened or read, create a new ring id
*/
if ((fd == -1) || (res != sizeof (uint64_t))) {
memb_ring_id->seq = 0;
fd = creat (filename, 0600);
if (fd != -1) {
res = write (fd, &memb_ring_id->seq, sizeof (uint64_t));
close (fd);
if (res == -1) {
LOGSYS_PERROR (errno, LOGSYS_LEVEL_ERROR,
"Couldn't write ringid file '%s'", filename);
corosync_exit_error (COROSYNC_DONE_STORE_RINGID);
}
} else {
LOGSYS_PERROR (errno, LOGSYS_LEVEL_ERROR,
"Couldn't create ringid file '%s'", filename);
corosync_exit_error (COROSYNC_DONE_STORE_RINGID);
}
}
memb_ring_id->rep = nodeid;
}
static void corosync_ring_id_store (
const struct memb_ring_id *memb_ring_id,
unsigned int nodeid)
{
char filename[PATH_MAX];
int fd;
int res;
snprintf (filename, sizeof(filename), "%s/ringid_%u",
get_state_dir(), nodeid);
fd = creat (filename, 0600);
if (fd == -1) {
LOGSYS_PERROR(errno, LOGSYS_LEVEL_ERROR,
"Couldn't store new ring id " CS_PRI_RING_ID_SEQ " to stable storage",
memb_ring_id->seq);
corosync_exit_error (COROSYNC_DONE_STORE_RINGID);
}
log_printf (LOGSYS_LEVEL_DEBUG,
"Storing new sequence id for ring " CS_PRI_RING_ID_SEQ, memb_ring_id->seq);
res = write (fd, &memb_ring_id->seq, sizeof(memb_ring_id->seq));
close (fd);
if (res != sizeof(memb_ring_id->seq)) {
LOGSYS_PERROR(errno, LOGSYS_LEVEL_ERROR,
"Couldn't store new ring id " CS_PRI_RING_ID_SEQ " to stable storage",
memb_ring_id->seq);
corosync_exit_error (COROSYNC_DONE_STORE_RINGID);
}
}
static qb_loop_timer_handle recheck_the_q_level_timer;
void corosync_recheck_the_q_level(void *data)
{
totempg_check_q_level(corosync_group_handle);
if (cs_ipcs_q_level_get() == TOTEM_Q_LEVEL_CRITICAL) {
qb_loop_timer_add(cs_poll_handle_get(), QB_LOOP_MED, 1*QB_TIME_NS_IN_MSEC,
NULL, corosync_recheck_the_q_level, &recheck_the_q_level_timer);
}
}
struct sending_allowed_private_data_struct {
int reserved_msgs;
};
int corosync_sending_allowed (
unsigned int service,
unsigned int id,
const void *msg,
void *sending_allowed_private_data)
{
struct sending_allowed_private_data_struct *pd =
(struct sending_allowed_private_data_struct *)sending_allowed_private_data;
struct iovec reserve_iovec;
struct qb_ipc_request_header *header = (struct qb_ipc_request_header *)msg;
int sending_allowed;
reserve_iovec.iov_base = (char *)header;
reserve_iovec.iov_len = header->size;
pd->reserved_msgs = totempg_groups_joined_reserve (
corosync_group_handle,
&reserve_iovec, 1);
if (pd->reserved_msgs == -1) {
return -EINVAL;
}
/* Message ID out of range */
if (id >= corosync_service[service]->lib_engine_count) {
return -EINVAL;
}
sending_allowed = QB_FALSE;
if (corosync_quorum_is_quorate() == 1 ||
corosync_service[service]->allow_inquorate == CS_LIB_ALLOW_INQUORATE) {
// we are quorate
// now check flow control
if (corosync_service[service]->lib_engine[id].flow_control == CS_LIB_FLOW_CONTROL_NOT_REQUIRED) {
sending_allowed = QB_TRUE;
} else if (pd->reserved_msgs && sync_in_process == 0) {
sending_allowed = QB_TRUE;
} else if (pd->reserved_msgs == 0) {
return -ENOBUFS;
} else /* (sync_in_process) */ {
return -EINPROGRESS;
}
} else {
return -EHOSTUNREACH;
}
return (sending_allowed);
}
void corosync_sending_allowed_release (void *sending_allowed_private_data)
{
struct sending_allowed_private_data_struct *pd =
(struct sending_allowed_private_data_struct *)sending_allowed_private_data;
if (pd->reserved_msgs == -1) {
return;
}
totempg_groups_joined_release (pd->reserved_msgs);
}
int message_source_is_local (const mar_message_source_t *source)
{
int ret = 0;
assert (source != NULL);
if (source->nodeid == totempg_my_nodeid_get ()) {
ret = 1;
}
return ret;
}
void message_source_set (
mar_message_source_t *source,
void *conn)
{
assert ((source != NULL) && (conn != NULL));
memset (source, 0, sizeof (mar_message_source_t));
source->nodeid = totempg_my_nodeid_get ();
source->conn = conn;
}
struct scheduler_pause_timeout_data {
struct totem_config *totem_config;
qb_loop_timer_handle handle;
unsigned long long tv_prev;
unsigned long long max_tv_diff;
};
static void timer_function_scheduler_timeout (void *data)
{
struct scheduler_pause_timeout_data *timeout_data = (struct scheduler_pause_timeout_data *)data;
unsigned long long tv_current;
unsigned long long tv_diff;
uint64_t schedmiss_event_tstamp;
tv_current = qb_util_nano_current_get ();
if (timeout_data->tv_prev == 0) {
/*
* Initial call -> just pretent everything is ok
*/
timeout_data->tv_prev = tv_current;
timeout_data->max_tv_diff = 0;
}
tv_diff = tv_current - timeout_data->tv_prev;
timeout_data->tv_prev = tv_current;
if (tv_diff > timeout_data->max_tv_diff) {
schedmiss_event_tstamp = qb_util_nano_from_epoch_get() / QB_TIME_NS_IN_MSEC;
log_printf (LOGSYS_LEVEL_WARNING, "Corosync main process was not scheduled (@%" PRIu64 ") for %0.4f ms "
"(threshold is %0.4f ms). Consider token timeout increase.",
schedmiss_event_tstamp,
(float)tv_diff / QB_TIME_NS_IN_MSEC, (float)timeout_data->max_tv_diff / QB_TIME_NS_IN_MSEC);
stats_add_schedmiss_event(schedmiss_event_tstamp, (float)tv_diff / QB_TIME_NS_IN_MSEC);
}
/*
* Set next threshold, because token_timeout can change
*/
timeout_data->max_tv_diff = timeout_data->totem_config->token_timeout * QB_TIME_NS_IN_MSEC * 0.8;
qb_loop_timer_add (corosync_poll_handle,
QB_LOOP_MED,
timeout_data->totem_config->token_timeout * QB_TIME_NS_IN_MSEC / 3,
timeout_data,
timer_function_scheduler_timeout,
&timeout_data->handle);
}
/*
* Set main pid RR scheduler.
* silent: don't log sched_get_priority_max and sched_setscheduler errors
* Returns: 0 - success, -1 failure, -2 platform doesn't support SCHED_RR
*/
static int corosync_set_rr_scheduler (int silent)
{
int ret_val = 0;
#if defined(HAVE_PTHREAD_SETSCHEDPARAM) && defined(HAVE_SCHED_GET_PRIORITY_MAX) && defined(HAVE_SCHED_SETSCHEDULER)
int res;
sched_priority = sched_get_priority_max (SCHED_RR);
if (sched_priority != -1) {
global_sched_param.sched_priority = sched_priority;
res = sched_setscheduler (0, SCHED_RR, &global_sched_param);
if (res == -1) {
if (!silent) {
LOGSYS_PERROR(errno, LOGSYS_LEVEL_WARNING,
"Could not set SCHED_RR at priority %d",
global_sched_param.sched_priority);
}
global_sched_param.sched_priority = 0;
#ifdef HAVE_QB_LOG_THREAD_PRIORITY_SET
qb_log_thread_priority_set (SCHED_OTHER, 0);
#endif
ret_val = -1;
} else {
/*
* Turn on SCHED_RR in logsys system
*/
#ifdef HAVE_QB_LOG_THREAD_PRIORITY_SET
res = qb_log_thread_priority_set (SCHED_RR, sched_priority);
#else
res = -1;
#endif
if (res == -1) {
log_printf (LOGSYS_LEVEL_ERROR,
"Could not set logsys thread priority."
" Can't continue because of priority inversions.");
corosync_exit_error (COROSYNC_DONE_LOGSETUP);
}
}
} else {
if (!silent) {
LOGSYS_PERROR (errno, LOGSYS_LEVEL_WARNING,
"Could not get maximum scheduler priority");
}
sched_priority = 0;
ret_val = -1;
}
#else
log_printf(LOGSYS_LEVEL_WARNING,
"The Platform is missing process priority setting features. Leaving at default.");
ret_val = -2;
#endif
return (ret_val);
}
/* The basename man page contains scary warnings about
thread-safety and portability, hence this */
static const char *corosync_basename(const char *file_name)
{
char *base;
base = strrchr (file_name, '/');
if (base) {
return base + 1;
}
return file_name;
}
static void
_logsys_log_printf(int level, int subsys,
const char *function_name,
const char *file_name,
int file_line,
const char *format,
...) __attribute__((format(printf, 6, 7)));
static void
_logsys_log_printf(int level, int subsys,
const char *function_name,
const char *file_name,
int file_line,
const char *format, ...)
{
va_list ap;
va_start(ap, format);
qb_log_from_external_source_va(function_name, corosync_basename(file_name),
format, level, file_line,
subsys, ap);
va_end(ap);
}
static void fplay_key_change_notify_fn (
int32_t event,
const char *key_name,
struct icmap_notify_value new_val,
struct icmap_notify_value old_val,
void *user_data)
{
if (strcmp(key_name, "runtime.blackbox.dump_flight_data") == 0) {
fprintf(stderr,"Writetofile\n");
corosync_blackbox_write_to_file ();
}
if (strcmp(key_name, "runtime.blackbox.dump_state") == 0) {
fprintf(stderr,"statefump\n");
corosync_state_dump ();
}
}
static void corosync_fplay_control_init (void)
{
icmap_track_t track = NULL;
icmap_set_string("runtime.blackbox.dump_flight_data", "no");
icmap_set_string("runtime.blackbox.dump_state", "no");
icmap_track_add("runtime.blackbox.dump_flight_data",
ICMAP_TRACK_ADD | ICMAP_TRACK_DELETE | ICMAP_TRACK_MODIFY,
fplay_key_change_notify_fn,
NULL, &track);
icmap_track_add("runtime.blackbox.dump_state",
ICMAP_TRACK_ADD | ICMAP_TRACK_DELETE | ICMAP_TRACK_MODIFY,
fplay_key_change_notify_fn,
NULL, &track);
}
static void force_gather_notify_fn(
int32_t event,
const char *key_name,
struct icmap_notify_value new_val,
struct icmap_notify_value old_val,
void *user_data)
{
char *key_val;
if (icmap_get_string(key_name, &key_val) == CS_OK && strcmp(key_val, "no") == 0)
goto out;
icmap_set_string("runtime.force_gather", "no");
if (strcmp(key_name, "runtime.force_gather") == 0) {
log_printf(LOGSYS_LEVEL_ERROR, "Forcing into GATHER state\n");
totempg_force_gather();
}
out:
free(key_val);
}
static void corosync_force_gather_init (void)
{
icmap_track_t track = NULL;
icmap_set_string("runtime.force_gather", "no");
icmap_track_add("runtime.force_gather",
ICMAP_TRACK_ADD | ICMAP_TRACK_DELETE | ICMAP_TRACK_MODIFY,
force_gather_notify_fn,
NULL, &track);
}
/*
* Set RO flag for keys, which ether doesn't make sense to change by user (statistic)
* or which when changed are not reflected by runtime (totem.crypto_cipher, ...).
*
* Also some RO keys cannot be determined in this stage, so they are set later in
* other functions (like nodelist.local_node_pos, ...)
*/
static void set_icmap_ro_keys_flag (void)
{
/*
* Set RO flag for all keys of internal configuration and runtime statistics
*/
icmap_set_ro_access("internal_configuration.", CS_TRUE, CS_TRUE);
icmap_set_ro_access("runtime.services.", CS_TRUE, CS_TRUE);
icmap_set_ro_access("runtime.config.", CS_TRUE, CS_TRUE);
icmap_set_ro_access("runtime.totem.", CS_TRUE, CS_TRUE);
icmap_set_ro_access("uidgid.config.", CS_TRUE, CS_TRUE);
icmap_set_ro_access("system.", CS_TRUE, CS_TRUE);
icmap_set_ro_access("nodelist.", CS_TRUE, CS_TRUE);
/*
* Set RO flag for constrete keys of configuration which can't be changed
* during runtime
*/
icmap_set_ro_access("totem.crypto_cipher", CS_FALSE, CS_TRUE);
icmap_set_ro_access("totem.crypto_hash", CS_FALSE, CS_TRUE);
icmap_set_ro_access("totem.crypto_model", CS_FALSE, CS_TRUE);
icmap_set_ro_access("totem.keyfile", CS_FALSE, CS_TRUE);
icmap_set_ro_access("totem.key", CS_FALSE, CS_TRUE);
icmap_set_ro_access("totem.secauth", CS_FALSE, CS_TRUE);
icmap_set_ro_access("totem.ip_version", CS_FALSE, CS_TRUE);
icmap_set_ro_access("totem.rrp_mode", CS_FALSE, CS_TRUE);
icmap_set_ro_access("totem.transport", CS_FALSE, CS_TRUE);
icmap_set_ro_access("totem.cluster_name", CS_FALSE, CS_TRUE);
icmap_set_ro_access("totem.netmtu", CS_FALSE, CS_TRUE);
icmap_set_ro_access("totem.threads", CS_FALSE, CS_TRUE);
icmap_set_ro_access("totem.version", CS_FALSE, CS_TRUE);
icmap_set_ro_access("totem.nodeid", CS_FALSE, CS_TRUE);
icmap_set_ro_access("totem.clear_node_high_bit", CS_FALSE, CS_TRUE);
icmap_set_ro_access("config.reload_in_progress", CS_FALSE, CS_TRUE);
icmap_set_ro_access("config.totemconfig_reload_in_progress", CS_FALSE, CS_TRUE);
}
static void main_service_ready (void)
{
int res;
/*
* This must occur after totempg is initialized because "this_ip" must be set
*/
res = corosync_service_defaults_link_and_init (api);
if (res == -1) {
log_printf (LOGSYS_LEVEL_ERROR, "Could not initialize default services");
corosync_exit_error (COROSYNC_DONE_INIT_SERVICES);
}
cs_ipcs_init();
corosync_totem_stats_init ();
corosync_fplay_control_init ();
corosync_force_gather_init ();
sync_init (
corosync_sync_callbacks_retrieve,
corosync_sync_completed);
}
static enum e_corosync_done corosync_flock (const char *lockfile, pid_t pid)
{
struct flock lock;
enum e_corosync_done err;
char pid_s[17];
int fd_flag;
err = COROSYNC_DONE_EXIT;
lockfile_fd = open (lockfile, O_WRONLY | O_CREAT, 0640);
if (lockfile_fd == -1) {
log_printf (LOGSYS_LEVEL_ERROR, "Corosync Executive couldn't create lock file.");
return (COROSYNC_DONE_ACQUIRE_LOCK);
}
retry_fcntl:
lock.l_type = F_WRLCK;
lock.l_start = 0;
lock.l_whence = SEEK_SET;
lock.l_len = 0;
if (fcntl (lockfile_fd, F_SETLK, &lock) == -1) {
switch (errno) {
case EINTR:
goto retry_fcntl;
break;
case EAGAIN:
case EACCES:
log_printf (LOGSYS_LEVEL_ERROR, "Another Corosync instance is already running.");
err = COROSYNC_DONE_ALREADY_RUNNING;
goto error_close;
break;
default:
log_printf (LOGSYS_LEVEL_ERROR, "Corosync Executive couldn't acquire lock. Error was %s",
strerror(errno));
err = COROSYNC_DONE_ACQUIRE_LOCK;
goto error_close;
break;
}
}
if (ftruncate (lockfile_fd, 0) == -1) {
log_printf (LOGSYS_LEVEL_ERROR, "Corosync Executive couldn't truncate lock file. Error was %s",
strerror (errno));
err = COROSYNC_DONE_ACQUIRE_LOCK;
goto error_close_unlink;
}
memset (pid_s, 0, sizeof (pid_s));
snprintf (pid_s, sizeof (pid_s) - 1, "%u\n", pid);
retry_write:
if (write (lockfile_fd, pid_s, strlen (pid_s)) != strlen (pid_s)) {
if (errno == EINTR) {
goto retry_write;
} else {
log_printf (LOGSYS_LEVEL_ERROR, "Corosync Executive couldn't write pid to lock file. "
"Error was %s", strerror (errno));
err = COROSYNC_DONE_ACQUIRE_LOCK;
goto error_close_unlink;
}
}
if ((fd_flag = fcntl (lockfile_fd, F_GETFD, 0)) == -1) {
log_printf (LOGSYS_LEVEL_ERROR, "Corosync Executive couldn't get close-on-exec flag from lock file. "
"Error was %s", strerror (errno));
err = COROSYNC_DONE_ACQUIRE_LOCK;
goto error_close_unlink;
}
fd_flag |= FD_CLOEXEC;
if (fcntl (lockfile_fd, F_SETFD, fd_flag) == -1) {
log_printf (LOGSYS_LEVEL_ERROR, "Corosync Executive couldn't set close-on-exec flag to lock file. "
"Error was %s", strerror (errno));
err = COROSYNC_DONE_ACQUIRE_LOCK;
goto error_close_unlink;
}
return (err);
error_close_unlink:
unlink (lockfile);
error_close:
close (lockfile_fd);
return (err);
}
static int corosync_move_to_root_cgroup(void) {
FILE *f;
int res = -1;
const char *cgroup_task_fname = NULL;
/*
* /sys/fs/cgroup is hardcoded, because most of Linux distributions are now
* using systemd and systemd uses hardcoded path of cgroup mount point.
*
* This feature is expected to be removed as soon as systemd gets support
* for managing RT configuration.
*/
f = fopen("/sys/fs/cgroup/cpu/cpu.rt_runtime_us", "rt");
if (f == NULL) {
/*
* Try cgroup v2
*/
f = fopen("/sys/fs/cgroup/cgroup.procs", "rt");
if (f == NULL) {
log_printf(LOG_DEBUG, "cpu.rt_runtime_us or cgroup.procs doesn't exist -> "
"system without cgroup or with disabled CONFIG_RT_GROUP_SCHED");
res = 0;
goto exit_res;
} else {
log_printf(LOGSYS_LEVEL_DEBUG, "Moving main pid to cgroup v2 root cgroup");
cgroup_task_fname = "/sys/fs/cgroup/cgroup.procs";
}
} else {
log_printf(LOGSYS_LEVEL_DEBUG, "Moving main pid to cgroup v1 root cgroup");
cgroup_task_fname = "/sys/fs/cgroup/cpu/tasks";
}
(void)fclose(f);
f = fopen(cgroup_task_fname, "w");
if (f == NULL) {
log_printf(LOGSYS_LEVEL_WARNING, "Can't open cgroups tasks file for writing");
goto exit_res;
}
if (fprintf(f, "%jd\n", (intmax_t)getpid()) <= 0) {
log_printf(LOGSYS_LEVEL_WARNING, "Can't write corosync pid into cgroups tasks file");
goto close_and_exit_res;
}
close_and_exit_res:
if (fclose(f) != 0) {
log_printf(LOGSYS_LEVEL_WARNING, "Can't close cgroups tasks file");
goto exit_res;
}
exit_res:
return (res);
}
static void show_version_info_crypto(void)
{
const char *error_string;
const char *list_str;
if (util_is_valid_knet_crypto_model(NULL, &list_str, 1, "", &error_string) != -1) {
printf("Available crypto models: %s\n", list_str);
} else {
perror(error_string);
}
}
static void show_version_info_compress(void)
{
const char *error_string;
const char *list_str;
if (util_is_valid_knet_compress_model(NULL, &list_str, 1, "", &error_string) != -1) {
printf("Available compression models: %s\n", list_str);
} else {
perror(error_string);
}
}
static void show_version_info(void)
{
printf ("Corosync Cluster Engine, version '%s'\n", VERSION);
printf ("Copyright (c) 2006-2021 Red Hat, Inc.\n");
printf ("\nBuilt-in features:" PACKAGE_FEATURES "\n");
show_version_info_crypto();
show_version_info_compress();
}
int main (int argc, char **argv, char **envp)
{
const char *error_string;
struct totem_config totem_config;
int res, ch;
int background, sched_rr, prio, testonly;
enum move_to_root_cgroup_mode move_to_root_cgroup;
enum e_corosync_done flock_err;
uint64_t totem_config_warnings;
struct scheduler_pause_timeout_data scheduler_pause_timeout_data;
long int tmpli;
char *ep;
char *tmp_str;
int log_subsys_id_totem;
int silent;
/* default configuration
*/
background = 1;
testonly = 0;
while ((ch = getopt (argc, argv, "c:ftv")) != EOF) {
switch (ch) {
case 'c':
res = snprintf(corosync_config_file, sizeof(corosync_config_file), "%s", optarg);
if (res >= sizeof(corosync_config_file)) {
fprintf (stderr, "Config file path too long.\n");
syslog (LOGSYS_LEVEL_ERROR, "Config file path too long.");
logsys_system_fini();
return EXIT_FAILURE;
}
break;
case 'f':
background = 0;
break;
case 't':
testonly = 1;
break;
case 'v':
show_version_info();
logsys_system_fini();
return EXIT_SUCCESS;
break;
default:
fprintf(stderr, \
"usage:\n"\
" -c : Corosync config file path.\n"\
" -f : Start application in foreground.\n"\
" -t : Test configuration and exit.\n"\
" -v : Display version, git revision and some useful information about Corosync and exit.\n");
logsys_system_fini();
return EXIT_FAILURE;
}
}
/*
* Other signals are registered later via qb_loop_signal_add
*/
(void)signal (SIGSEGV, sigsegv_handler);
(void)signal (SIGABRT, sigsegv_handler);
#if MSG_NOSIGNAL != 0
(void)signal (SIGPIPE, SIG_IGN);
#endif
if (icmap_init() != CS_OK) {
fprintf (stderr, "Corosync Executive couldn't initialize configuration component.\n");
syslog (LOGSYS_LEVEL_ERROR, "Corosync Executive couldn't initialize configuration component.");
corosync_exit_error (COROSYNC_DONE_ICMAP);
}
set_icmap_ro_keys_flag();
/*
* Initialize the corosync_api_v1 definition
*/
api = apidef_get ();
res = coroparse_configparse(icmap_get_global_map(), &error_string);
if (res == -1) {
/*
* Logsys can't log properly at this early stage, and we need to get this message out
*
*/
fprintf (stderr, "%s\n", error_string);
syslog (LOGSYS_LEVEL_ERROR, "%s", error_string);
corosync_exit_error (COROSYNC_DONE_MAINCONFIGREAD);
}
if (stats_map_init(api) != CS_OK) {
fprintf (stderr, "Corosync Executive couldn't initialize statistics component.\n");
syslog (LOGSYS_LEVEL_ERROR, "Corosync Executive couldn't initialize statistics component.");
corosync_exit_error (COROSYNC_DONE_STATS);
}
res = corosync_log_config_read (&error_string);
if (res == -1) {
/*
* if we are here, we _must_ flush the logsys queue
* and try to inform that we couldn't read the config.
* this is a desperate attempt before certain death
* and there is no guarantee that we can print to stderr
* nor that logsys is sending the messages where we expect.
*/
log_printf (LOGSYS_LEVEL_ERROR, "%s", error_string);
fprintf(stderr, "%s", error_string);
syslog (LOGSYS_LEVEL_ERROR, "%s", error_string);
corosync_exit_error (COROSYNC_DONE_LOGCONFIGREAD);
}
if (!testonly) {
log_printf (LOGSYS_LEVEL_NOTICE, "Corosync Cluster Engine %s starting up", VERSION);
log_printf (LOGSYS_LEVEL_INFO, "Corosync built-in features:" PACKAGE_FEATURES "");
}
/*
* Create totem logsys subsys before totem_config_read so log functions can be used
*/
log_subsys_id_totem = _logsys_subsys_create("TOTEM", "totem,"
"totemip.c,totemconfig.c,totemcrypto.c,totemsrp.c,"
"totempg.c,totemudp.c,totemudpu.c,totemnet.c,totemknet.c");
res = chdir(get_state_dir());
if (res == -1) {
log_printf (LOGSYS_LEVEL_ERROR, "Cannot chdir to state directory %s. %s", get_state_dir(), strerror(errno));
corosync_exit_error (COROSYNC_DONE_DIR_NOT_PRESENT);
}
res = totem_config_read (&totem_config, &error_string, &totem_config_warnings);
if (res == -1) {
log_printf (LOGSYS_LEVEL_ERROR, "%s", error_string);
corosync_exit_error (COROSYNC_DONE_MAINCONFIGREAD);
}
if (totem_config_warnings & TOTEM_CONFIG_WARNING_MEMBERS_IGNORED) {
log_printf (LOGSYS_LEVEL_WARNING, "member section is used together with nodelist. Members ignored.");
}
if (totem_config_warnings & TOTEM_CONFIG_WARNING_MEMBERS_DEPRECATED) {
log_printf (LOGSYS_LEVEL_WARNING, "member section is deprecated.");
}
- if (totem_config_warnings & TOTEM_CONFIG_WARNING_TOTEM_NODEID_IGNORED) {
- log_printf (LOGSYS_LEVEL_WARNING, "nodeid appears both in totem section and nodelist. Nodelist one is used.");
+ if (totem_config_warnings & TOTEM_CONFIG_WARNING_TOTEM_NODEID_SET) {
+ log_printf (LOGSYS_LEVEL_WARNING, "nodeid in totem section is deprecated and ignored. "
+ "Nodelist (or autogenerated) nodeid is going to be used.");
}
if (totem_config_warnings & TOTEM_CONFIG_BINDNETADDR_NODELIST_SET) {
log_printf (LOGSYS_LEVEL_WARNING, "interface section bindnetaddr is used together with nodelist. "
"Nodelist one is going to be used.");
}
if (totem_config_warnings != 0) {
log_printf (LOGSYS_LEVEL_WARNING, "Please migrate config file to nodelist.");
}
res = totem_config_validate (&totem_config, &error_string);
if (res == -1) {
log_printf (LOGSYS_LEVEL_ERROR, "%s", error_string);
corosync_exit_error (COROSYNC_DONE_MAINCONFIGREAD);
}
if (testonly) {
corosync_exit_error (COROSYNC_DONE_EXIT);
}
move_to_root_cgroup = MOVE_TO_ROOT_CGROUP_MODE_AUTO;
if (icmap_get_string("system.move_to_root_cgroup", &tmp_str) == CS_OK) {
/*
* Validity of move_to_root_cgroup values checked in coroparse.c
*/
if (strcmp(tmp_str, "yes") == 0) {
move_to_root_cgroup = MOVE_TO_ROOT_CGROUP_MODE_ON;
} else if (strcmp(tmp_str, "no") == 0) {
move_to_root_cgroup = MOVE_TO_ROOT_CGROUP_MODE_OFF;
}
free(tmp_str);
}
sched_rr = 1;
if (icmap_get_string("system.sched_rr", &tmp_str) == CS_OK) {
if (strcmp(tmp_str, "yes") != 0) {
sched_rr = 0;
}
free(tmp_str);
}
prio = 0;
if (icmap_get_string("system.priority", &tmp_str) == CS_OK) {
if (strcmp(tmp_str, "max") == 0) {
prio = INT_MIN;
} else if (strcmp(tmp_str, "min") == 0) {
prio = INT_MAX;
} else {
errno = 0;
tmpli = strtol(tmp_str, &ep, 10);
if (errno != 0 || *ep != '\0' || tmpli > INT_MAX || tmpli < INT_MIN) {
log_printf (LOGSYS_LEVEL_ERROR, "Priority value %s is invalid", tmp_str);
corosync_exit_error (COROSYNC_DONE_MAINCONFIGREAD);
}
prio = tmpli;
}
free(tmp_str);
}
if (move_to_root_cgroup == MOVE_TO_ROOT_CGROUP_MODE_ON) {
/*
* Try to move corosync into root cpu cgroup. Failure is not fatal and
* error is deliberately ignored.
*/
(void)corosync_move_to_root_cgroup();
}
/*
* Set round robin realtime scheduling with priority 99
*/
if (sched_rr) {
silent = (move_to_root_cgroup == MOVE_TO_ROOT_CGROUP_MODE_AUTO);
res = corosync_set_rr_scheduler (silent);
if (res == -1 && move_to_root_cgroup == MOVE_TO_ROOT_CGROUP_MODE_AUTO) {
/*
* Try to move process to root cgroup and try set priority again
*/
(void)corosync_move_to_root_cgroup();
res = corosync_set_rr_scheduler (0);
}
if (res != 0) {
prio = INT_MIN;
} else {
prio = 0;
}
}
if (prio != 0) {
if (setpriority(PRIO_PGRP, 0, prio) != 0) {
LOGSYS_PERROR(errno, LOGSYS_LEVEL_WARNING,
"Could not set priority %d", prio);
}
}
totem_config.totem_memb_ring_id_create_or_load = corosync_ring_id_create_or_load;
totem_config.totem_memb_ring_id_store = corosync_ring_id_store;
totem_config.totem_logging_configuration = totem_logging_configuration;
totem_config.totem_logging_configuration.log_subsys_id = log_subsys_id_totem;
totem_config.totem_logging_configuration.log_level_security = LOGSYS_LEVEL_WARNING;
totem_config.totem_logging_configuration.log_level_error = LOGSYS_LEVEL_ERROR;
totem_config.totem_logging_configuration.log_level_warning = LOGSYS_LEVEL_WARNING;
totem_config.totem_logging_configuration.log_level_notice = LOGSYS_LEVEL_NOTICE;
totem_config.totem_logging_configuration.log_level_debug = LOGSYS_LEVEL_DEBUG;
totem_config.totem_logging_configuration.log_level_trace = LOGSYS_LEVEL_TRACE;
totem_config.totem_logging_configuration.log_printf = _logsys_log_printf;
logsys_config_apply();
/*
* Now we are fully initialized.
*/
if (background) {
logsys_blackbox_prefork();
corosync_tty_detach ();
logsys_blackbox_postfork();
log_printf (LOGSYS_LEVEL_DEBUG, "Corosync TTY detached");
}
/*
* Lock all memory to avoid page faults which may interrupt
* application healthchecking
*/
corosync_mlockall ();
corosync_poll_handle = qb_loop_create ();
memset(&scheduler_pause_timeout_data, 0, sizeof(scheduler_pause_timeout_data));
scheduler_pause_timeout_data.totem_config = &totem_config;
timer_function_scheduler_timeout (&scheduler_pause_timeout_data);
qb_loop_signal_add(corosync_poll_handle, QB_LOOP_LOW,
SIGUSR2, NULL, sig_diag_handler, NULL);
qb_loop_signal_add(corosync_poll_handle, QB_LOOP_HIGH,
SIGINT, NULL, sig_exit_handler, NULL);
qb_loop_signal_add(corosync_poll_handle, QB_LOOP_HIGH,
SIGQUIT, NULL, sig_exit_handler, NULL);
qb_loop_signal_add(corosync_poll_handle, QB_LOOP_HIGH,
SIGTERM, NULL, sig_exit_handler, NULL);
if (logsys_thread_start() != 0) {
log_printf (LOGSYS_LEVEL_ERROR, "Can't initialize log thread");
corosync_exit_error (COROSYNC_DONE_LOGCONFIGREAD);
}
if ((flock_err = corosync_flock (corosync_lock_file, getpid ())) != COROSYNC_DONE_EXIT) {
corosync_exit_error (flock_err);
}
/*
* if totempg_initialize doesn't have root priveleges, it cannot
* bind to a specific interface. This only matters if
* there is more then one interface in a system, so
* in this case, only a warning is printed
*/
/*
* Join multicast group and setup delivery
* and configuration change functions
*/
if (totempg_initialize (
corosync_poll_handle,
&totem_config) != 0) {
log_printf (LOGSYS_LEVEL_ERROR, "Can't initialize TOTEM layer");
corosync_exit_error (COROSYNC_DONE_FATAL_ERR);
}
totempg_service_ready_register (
main_service_ready);
totempg_groups_initialize (
&corosync_group_handle,
deliver_fn,
confchg_fn);
totempg_groups_join (
corosync_group_handle,
&corosync_group,
1);
/*
* Drop root privleges to user 'corosync'
* TODO: Don't really need full root capabilities;
* needed capabilities are:
* CAP_NET_RAW (bindtodevice)
* CAP_SYS_NICE (setscheduler)
* CAP_IPC_LOCK (mlockall)
*/
priv_drop ();
schedwrk_init (
serialize_lock,
serialize_unlock);
/*
* Start main processing loop
*/
qb_loop_run (corosync_poll_handle);
/*
* Exit was requested
*/
totempg_finalize ();
/*
* free the loop resources
*/
qb_loop_destroy (corosync_poll_handle);
/*
* free up the icmap
*/
/*
* Remove pid lock file
*/
close (lockfile_fd);
unlink (corosync_lock_file);
corosync_exit_error (COROSYNC_DONE_EXIT);
return EXIT_SUCCESS;
}
diff --git a/exec/totemconfig.c b/exec/totemconfig.c
index 77c7ee03..21bfbae4 100644
--- a/exec/totemconfig.c
+++ b/exec/totemconfig.c
@@ -1,2409 +1,2411 @@
/*
* Copyright (c) 2002-2005 MontaVista Software, Inc.
* Copyright (c) 2006-2019 Red Hat, Inc.
*
* All rights reserved.
*
* Author: Steven Dake (sdake@redhat.com)
* Jan Friesse (jfriesse@redhat.com)
* Chrissie Caulfield (ccaulfie@redhat.com)
*
* This software licensed under BSD license, the text of which follows:
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* - Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
* - Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
* - Neither the name of the MontaVista Software, Inc. nor the names of its
* contributors may be used to endorse or promote products derived from this
* software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
* THE POSSIBILITY OF SUCH DAMAGE.
*/
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include "util.h"
#include "totemconfig.h"
#define TOKEN_RETRANSMITS_BEFORE_LOSS_CONST 4
#define TOKEN_TIMEOUT 3000
#define TOKEN_WARNING 75
#define TOKEN_COEFFICIENT 650
#define JOIN_TIMEOUT 50
#define MERGE_TIMEOUT 200
#define DOWNCHECK_TIMEOUT 1000
#define FAIL_TO_RECV_CONST 2500
#define SEQNO_UNCHANGED_CONST 30
#define MINIMUM_TIMEOUT (int)(1000/HZ)*3
#define MINIMUM_TIMEOUT_HOLD (int)(MINIMUM_TIMEOUT * 0.8 - (1000/HZ))
#define MAX_NETWORK_DELAY 50
#define WINDOW_SIZE 50
#define MAX_MESSAGES 17
#define MISS_COUNT_CONST 5
#define BLOCK_UNLISTED_IPS 1
/* This constant is not used for knet */
#define UDP_NETMTU 1500
/* Currently all but PONG_COUNT match the defaults in libknet.h */
#define KNET_PING_INTERVAL 1000
#define KNET_PING_TIMEOUT 2000
#define KNET_PING_PRECISION 2048
#define KNET_PONG_COUNT 2
#define KNET_PMTUD_INTERVAL 30
#define KNET_DEFAULT_TRANSPORT KNET_TRANSPORT_UDP
#define DEFAULT_PORT 5405
static char error_string_response[768];
static void add_totem_config_notification(struct totem_config *totem_config);
static void *totem_get_param_by_name(struct totem_config *totem_config, const char *param_name)
{
if (strcmp(param_name, "totem.token") == 0)
return &totem_config->token_timeout;
if (strcmp(param_name, "totem.token_warning") == 0)
return &totem_config->token_warning;
if (strcmp(param_name, "totem.token_retransmit") == 0)
return &totem_config->token_retransmit_timeout;
if (strcmp(param_name, "totem.hold") == 0)
return &totem_config->token_hold_timeout;
if (strcmp(param_name, "totem.token_retransmits_before_loss_const") == 0)
return &totem_config->token_retransmits_before_loss_const;
if (strcmp(param_name, "totem.join") == 0)
return &totem_config->join_timeout;
if (strcmp(param_name, "totem.send_join") == 0)
return &totem_config->send_join_timeout;
if (strcmp(param_name, "totem.consensus") == 0)
return &totem_config->consensus_timeout;
if (strcmp(param_name, "totem.merge") == 0)
return &totem_config->merge_timeout;
if (strcmp(param_name, "totem.downcheck") == 0)
return &totem_config->downcheck_timeout;
if (strcmp(param_name, "totem.fail_recv_const") == 0)
return &totem_config->fail_to_recv_const;
if (strcmp(param_name, "totem.seqno_unchanged_const") == 0)
return &totem_config->seqno_unchanged_const;
if (strcmp(param_name, "totem.heartbeat_failures_allowed") == 0)
return &totem_config->heartbeat_failures_allowed;
if (strcmp(param_name, "totem.max_network_delay") == 0)
return &totem_config->max_network_delay;
if (strcmp(param_name, "totem.window_size") == 0)
return &totem_config->window_size;
if (strcmp(param_name, "totem.max_messages") == 0)
return &totem_config->max_messages;
if (strcmp(param_name, "totem.miss_count_const") == 0)
return &totem_config->miss_count_const;
if (strcmp(param_name, "totem.knet_pmtud_interval") == 0)
return &totem_config->knet_pmtud_interval;
if (strcmp(param_name, "totem.knet_compression_threshold") == 0)
return &totem_config->knet_compression_threshold;
if (strcmp(param_name, "totem.knet_compression_level") == 0)
return &totem_config->knet_compression_level;
if (strcmp(param_name, "totem.knet_compression_model") == 0)
return totem_config->knet_compression_model;
if (strcmp(param_name, "totem.block_unlisted_ips") == 0)
return &totem_config->block_unlisted_ips;
return NULL;
}
/*
* Read key_name from icmap. If key is not found or key_name == delete_key or if allow_zero is false
* and readed value is zero, default value is used and stored into totem_config.
*/
static void totem_volatile_config_set_uint32_value (struct totem_config *totem_config, icmap_map_t map,
const char *key_name, const char *deleted_key, unsigned int default_value,
int allow_zero_value)
{
char runtime_key_name[ICMAP_KEYNAME_MAXLEN];
if (icmap_get_uint32_r(map, key_name, totem_get_param_by_name(totem_config, key_name)) != CS_OK ||
(deleted_key != NULL && strcmp(deleted_key, key_name) == 0) ||
(!allow_zero_value && *(uint32_t *)totem_get_param_by_name(totem_config, key_name) == 0)) {
*(uint32_t *)totem_get_param_by_name(totem_config, key_name) = default_value;
}
/*
* Store totem_config value to cmap runtime section
*/
if (strlen("runtime.config.") + strlen(key_name) >= ICMAP_KEYNAME_MAXLEN) {
/*
* This shouldn't happen
*/
return ;
}
strcpy(runtime_key_name, "runtime.config.");
strcat(runtime_key_name, key_name);
icmap_set_uint32_r(map, runtime_key_name, *(uint32_t *)totem_get_param_by_name(totem_config, key_name));
}
static void totem_volatile_config_set_int32_value (struct totem_config *totem_config, icmap_map_t map,
const char *key_name, const char *deleted_key, int default_value,
int allow_zero_value)
{
char runtime_key_name[ICMAP_KEYNAME_MAXLEN];
if (icmap_get_int32_r(map, key_name, totem_get_param_by_name(totem_config, key_name)) != CS_OK ||
(deleted_key != NULL && strcmp(deleted_key, key_name) == 0) ||
(!allow_zero_value && *(int32_t *)totem_get_param_by_name(totem_config, key_name) == 0)) {
*(int32_t *)totem_get_param_by_name(totem_config, key_name) = default_value;
}
/*
* Store totem_config value to cmap runtime section
*/
if (strlen("runtime.config.") + strlen(key_name) >= ICMAP_KEYNAME_MAXLEN) {
/*
* This shouldn't happen
*/
return ;
}
strcpy(runtime_key_name, "runtime.config.");
strcat(runtime_key_name, key_name);
icmap_set_int32_r(map, runtime_key_name, *(int32_t *)totem_get_param_by_name(totem_config, key_name));
}
static void totem_volatile_config_set_string_value (struct totem_config *totem_config, icmap_map_t map,
const char *key_name, const char *deleted_key, const char *default_value)
{
char runtime_key_name[ICMAP_KEYNAME_MAXLEN];
int res;
char *new_config_value;
const void *config_value;
config_value = totem_get_param_by_name(totem_config, key_name);
res = icmap_get_string_r(map, key_name, (char **)&new_config_value);
if (res != CS_OK ||
(deleted_key != NULL && strcmp(deleted_key, key_name) == 0)) {
/* Slightly pointless use of strncpy but it keeps coverity happy */
strncpy((char *)config_value, default_value, CONFIG_STRING_LEN_MAX);
} else {
strncpy((char *)config_value, new_config_value, CONFIG_STRING_LEN_MAX);
}
if (res == CS_OK) {
free(new_config_value);
}
/*
* Store totem_config value to cmap runtime section
*/
if (strlen("runtime.config.") + strlen(key_name) >= ICMAP_KEYNAME_MAXLEN) {
/*
* This shouldn't happen
*/
return ;
}
strcpy(runtime_key_name, "runtime.config.");
strcat(runtime_key_name, key_name);
(void)icmap_set_string_r(map, runtime_key_name, (char *)config_value);
}
/*
* Read string value stored in key_name from icmap, use it as a boolean (yes/no) type, convert it
* to integer value (1/0) and store into totem_config.
*
* If key is not found or key_name == delete_key default value is used
* and stored into totem_config.
*/
static void totem_volatile_config_set_boolean_value (struct totem_config *totem_config, icmap_map_t map,
const char *key_name, const char *deleted_key, unsigned int default_value)
{
char runtime_key_name[ICMAP_KEYNAME_MAXLEN];
char *str;
int val;
str = NULL;
val = default_value;
if ((deleted_key != NULL && strcmp(deleted_key, key_name) == 0) ||
(icmap_get_string_r(map, key_name, &str) != CS_OK)) {
/*
* Do nothing. str is NULL (icmap_get_string ether not called or
* not changed str).
*/
} else {
if (strcmp(str, "yes") == 0) {
val = 1;
} else if (strcmp(str, "no") == 0) {
val = 0;
}
free(str);
}
/*
* Store totem_config value to cmap runtime section
*/
if (strlen("runtime.config.") + strlen(key_name) >= ICMAP_KEYNAME_MAXLEN) {
/*
* This shouldn't happen
*/
return ;
}
strcpy(runtime_key_name, "runtime.config.");
strcat(runtime_key_name, key_name);
*(uint32_t *)totem_get_param_by_name(totem_config, key_name) = val;
icmap_set_uint32_r(map, runtime_key_name, val);
}
/*
* Read and validate config values from cmap and store them into totem_config. If key doesn't exists,
* default value is stored. deleted_key is name of key beeing processed by delete operation
* from cmap. It is considered as non existing even if it can be read. Can be NULL.
*/
void totem_volatile_config_read (struct totem_config *totem_config, icmap_map_t temp_map, const char *deleted_key)
{
uint32_t u32;
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.token_retransmits_before_loss_const", deleted_key,
TOKEN_RETRANSMITS_BEFORE_LOSS_CONST, 0);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.token", deleted_key, TOKEN_TIMEOUT, 0);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.token_warning", deleted_key, TOKEN_WARNING, 1);
if (totem_config->interfaces[0].member_count > 2) {
u32 = TOKEN_COEFFICIENT;
icmap_get_uint32_r(temp_map, "totem.token_coefficient", &u32);
totem_config->token_timeout += (totem_config->interfaces[0].member_count - 2) * u32;
/*
* Store totem_config value to cmap runtime section
*/
icmap_set_uint32_r(temp_map, "runtime.config.totem.token", totem_config->token_timeout);
}
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.max_network_delay", deleted_key, MAX_NETWORK_DELAY, 0);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.window_size", deleted_key, WINDOW_SIZE, 0);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.max_messages", deleted_key, MAX_MESSAGES, 0);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.miss_count_const", deleted_key, MISS_COUNT_CONST, 0);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.knet_pmtud_interval", deleted_key, KNET_PMTUD_INTERVAL, 0);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.token_retransmit", deleted_key,
(int)(totem_config->token_timeout / (totem_config->token_retransmits_before_loss_const + 0.2)), 0);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.hold", deleted_key,
(int)(totem_config->token_retransmit_timeout * 0.8 - (1000/HZ)), 0);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.join", deleted_key, JOIN_TIMEOUT, 0);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.consensus", deleted_key,
(int)(float)(1.2 * totem_config->token_timeout), 0);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.merge", deleted_key, MERGE_TIMEOUT, 0);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.downcheck", deleted_key, DOWNCHECK_TIMEOUT, 0);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.fail_recv_const", deleted_key, FAIL_TO_RECV_CONST, 0);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.seqno_unchanged_const", deleted_key,
SEQNO_UNCHANGED_CONST, 0);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.send_join", deleted_key, 0, 1);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.heartbeat_failures_allowed", deleted_key, 0, 1);
totem_volatile_config_set_uint32_value(totem_config, temp_map, "totem.knet_compression_threshold", deleted_key, 0, 1);
totem_volatile_config_set_int32_value(totem_config, temp_map, "totem.knet_compression_level", deleted_key, 0, 1);
totem_volatile_config_set_string_value(totem_config, temp_map, "totem.knet_compression_model", deleted_key, "none");
totem_volatile_config_set_boolean_value(totem_config, temp_map, "totem.block_unlisted_ips", deleted_key,
BLOCK_UNLISTED_IPS);
}
int totem_volatile_config_validate (
struct totem_config *totem_config,
icmap_map_t temp_map,
const char **error_string)
{
/* Static just to keep them off the stack */
static char local_error_reason[512];
static char addr_str_buf[INET6_ADDRSTRLEN];
const char *error_reason = local_error_reason;
char name_key[ICMAP_KEYNAME_MAXLEN];
char *name_str;
int i, j, num_configured, members;
uint32_t tmp_config_value;
if (totem_config->max_network_delay < MINIMUM_TIMEOUT) {
snprintf (local_error_reason, sizeof(local_error_reason),
"The max_network_delay parameter (%d ms) may not be less than (%d ms).",
totem_config->max_network_delay, MINIMUM_TIMEOUT);
goto parse_error;
}
if (totem_config->token_timeout < MINIMUM_TIMEOUT) {
snprintf (local_error_reason, sizeof(local_error_reason),
"The token timeout parameter (%d ms) may not be less than (%d ms).",
totem_config->token_timeout, MINIMUM_TIMEOUT);
goto parse_error;
}
if (totem_config->token_warning > 100 || totem_config->token_warning < 0) {
snprintf (local_error_reason, sizeof(local_error_reason),
"The token warning parameter (%d%%) must be between 0 (disabled) and 100.",
totem_config->token_warning);
goto parse_error;
}
if (totem_config->token_retransmit_timeout < MINIMUM_TIMEOUT) {
if (icmap_get_uint32_r(temp_map, "totem.token_retransmit", &tmp_config_value) == CS_OK) {
snprintf (local_error_reason, sizeof(local_error_reason),
"The token retransmit timeout parameter (%d ms) may not be less than (%d ms).",
totem_config->token_retransmit_timeout, MINIMUM_TIMEOUT);
goto parse_error;
} else {
snprintf (local_error_reason, sizeof(local_error_reason),
"Not appropriate token or token_retransmits_before_loss_const value set");
goto parse_error;
}
}
if (totem_config->token_hold_timeout < MINIMUM_TIMEOUT_HOLD) {
snprintf (local_error_reason, sizeof(local_error_reason),
"The token hold timeout parameter (%d ms) may not be less than (%d ms).",
totem_config->token_hold_timeout, MINIMUM_TIMEOUT_HOLD);
goto parse_error;
}
if (totem_config->join_timeout < MINIMUM_TIMEOUT) {
snprintf (local_error_reason, sizeof(local_error_reason),
"The join timeout parameter (%d ms) may not be less than (%d ms).",
totem_config->join_timeout, MINIMUM_TIMEOUT);
goto parse_error;
}
if (totem_config->consensus_timeout < MINIMUM_TIMEOUT) {
snprintf (local_error_reason, sizeof(local_error_reason),
"The consensus timeout parameter (%d ms) may not be less than (%d ms).",
totem_config->consensus_timeout, MINIMUM_TIMEOUT);
goto parse_error;
}
if (totem_config->consensus_timeout < totem_config->join_timeout) {
snprintf (local_error_reason, sizeof(local_error_reason),
"The consensus timeout parameter (%d ms) may not be less than join timeout (%d ms).",
totem_config->consensus_timeout, totem_config->join_timeout);
goto parse_error;
}
if (totem_config->merge_timeout < MINIMUM_TIMEOUT) {
snprintf (local_error_reason, sizeof(local_error_reason),
"The merge timeout parameter (%d ms) may not be less than (%d ms).",
totem_config->merge_timeout, MINIMUM_TIMEOUT);
goto parse_error;
}
if (totem_config->downcheck_timeout < MINIMUM_TIMEOUT) {
snprintf (local_error_reason, sizeof(local_error_reason),
"The downcheck timeout parameter (%d ms) may not be less than (%d ms).",
totem_config->downcheck_timeout, MINIMUM_TIMEOUT);
goto parse_error;
}
/* Check that we have nodelist 'name' if there is more than one link */
num_configured = 0;
members = -1;
for (i = 0; i < INTERFACE_MAX; i++) {
if (totem_config->interfaces[i].configured) {
if (num_configured == 0) {
members = totem_config->interfaces[i].member_count;
}
num_configured++;
}
}
if (num_configured > 1) {
/*
* This assert is here just to make compiler happy
*/
assert(members != -1);
for (i=0; i < members; i++) {
snprintf(name_key, sizeof(name_key), "nodelist.node.%d.name", i);
if (icmap_get_string_r(temp_map, name_key, &name_str) != CS_OK) {
snprintf (local_error_reason, sizeof(local_error_reason),
"for a multi-link configuration, all nodes must have a 'name' attribute");
goto parse_error;
}
free(name_str);
}
for (i=0; i < INTERFACE_MAX; i++) {
if (!totem_config->interfaces[i].configured) {
continue;
}
if (totem_config->interfaces[i].member_count != members) {
snprintf (local_error_reason, sizeof(local_error_reason),
"Not all nodes have the same number of links");
goto parse_error;
}
}
}
/* Verify that all nodes on the same link have the same IP family */
for (i=0; i < INTERFACE_MAX; i++) {
for (j=1; jinterfaces[i].member_count; j++) {
if (totem_config->interfaces[i].configured) {
if (totem_config->interfaces[i].member_list[j].family !=
totem_config->interfaces[i].member_list[0].family) {
memcpy(addr_str_buf,
totemip_print(&(totem_config->interfaces[i].member_list[j])),
sizeof(addr_str_buf));
snprintf (local_error_reason, sizeof(local_error_reason),
"Nodes for link %d have different IP families "
"(compared %s with %s)", i,
addr_str_buf,
totemip_print(&(totem_config->interfaces[i].member_list[0])));
goto parse_error;
}
}
}
}
return 0;
parse_error:
snprintf (error_string_response, sizeof(error_string_response),
"parse error in config: %s\n", error_reason);
*error_string = error_string_response;
return (-1);
}
static int totem_get_crypto(struct totem_config *totem_config, icmap_map_t map, const char **error_string)
{
char *str;
const char *tmp_cipher;
const char *tmp_hash;
const char *tmp_model;
char *crypto_model_str;
int res = 0;
tmp_hash = "none";
tmp_cipher = "none";
tmp_model = "none";
crypto_model_str = NULL;
if (icmap_get_string_r(map, "totem.crypto_model", &crypto_model_str) == CS_OK) {
tmp_model = crypto_model_str;
} else {
tmp_model = "nss";
}
if (icmap_get_string_r(map, "totem.secauth", &str) == CS_OK) {
if (strcmp(str, "on") == 0) {
tmp_cipher = "aes256";
tmp_hash = "sha256";
}
free(str);
}
if (icmap_get_string_r(map, "totem.crypto_cipher", &str) == CS_OK) {
if (strcmp(str, "none") == 0) {
tmp_cipher = "none";
}
if (strcmp(str, "aes256") == 0) {
tmp_cipher = "aes256";
}
if (strcmp(str, "aes192") == 0) {
tmp_cipher = "aes192";
}
if (strcmp(str, "aes128") == 0) {
tmp_cipher = "aes128";
}
free(str);
}
if (icmap_get_string_r(map, "totem.crypto_hash", &str) == CS_OK) {
if (strcmp(str, "none") == 0) {
tmp_hash = "none";
}
if (strcmp(str, "md5") == 0) {
tmp_hash = "md5";
}
if (strcmp(str, "sha1") == 0) {
tmp_hash = "sha1";
}
if (strcmp(str, "sha256") == 0) {
tmp_hash = "sha256";
}
if (strcmp(str, "sha384") == 0) {
tmp_hash = "sha384";
}
if (strcmp(str, "sha512") == 0) {
tmp_hash = "sha512";
}
free(str);
}
if ((strcmp(tmp_cipher, "none") != 0) &&
(strcmp(tmp_hash, "none") == 0)) {
*error_string = "crypto_cipher requires crypto_hash with value other than none";
res = -1;
goto out_free_crypto_model_str;
}
if (strcmp(tmp_model, "none") == 0) {
/*
* Shouldn't happen because it is handled by coroparse
*/
*error_string = "invalid crypto_model";
res = -1;
goto out_free_crypto_model_str;
}
if (strcmp(tmp_cipher, totem_config->crypto_cipher_type) ||
strcmp(tmp_hash, totem_config->crypto_hash_type) ||
strcmp(tmp_model, totem_config->crypto_model)) {
totem_config->crypto_changed = 1;
}
strncpy(totem_config->crypto_cipher_type, tmp_cipher, CONFIG_STRING_LEN_MAX - 1);
totem_config->crypto_cipher_type[CONFIG_STRING_LEN_MAX - 1] = '\0';
strncpy(totem_config->crypto_hash_type, tmp_hash, CONFIG_STRING_LEN_MAX - 1);
totem_config->crypto_hash_type[CONFIG_STRING_LEN_MAX - 1] = '\0';
strncpy(totem_config->crypto_model, tmp_model, CONFIG_STRING_LEN_MAX - 1);
totem_config->crypto_model[CONFIG_STRING_LEN_MAX - 1] = '\0';
out_free_crypto_model_str:
free(crypto_model_str);
return (res);
}
static int nodelist_byname(icmap_map_t map, const char *find_name, int strip_domain)
{
icmap_iter_t iter;
const char *iter_key;
char name_str[ICMAP_KEYNAME_MAXLEN];
int res = 0;
unsigned int node_pos;
char *name;
unsigned int namelen;
iter = icmap_iter_init_r(map, "nodelist.node.");
while ((iter_key = icmap_iter_next(iter, NULL, NULL)) != NULL) {
res = sscanf(iter_key, "nodelist.node.%u.%s", &node_pos, name_str);
if (res != 2) {
continue;
}
/* ring0_addr is allowed as a fallback */
if (strcmp(name_str, "name") && strcmp(name_str, "ring0_addr")) {
continue;
}
if (icmap_get_string_r(map, iter_key, &name) != CS_OK) {
continue;
}
namelen = strlen(name);
if (strip_domain) {
char *dot;
dot = strchr(name, '.');
if (dot) {
namelen = dot - name;
}
}
if (strncmp(find_name, name, namelen) == 0 &&
strlen(find_name) == namelen) {
icmap_iter_finalize(iter);
return node_pos;
}
}
icmap_iter_finalize(iter);
return -1;
}
/* Compare two addresses - only address part (sin_addr/sin6_addr) is checked */
static int ipaddr_equal(const struct sockaddr *addr1, const struct sockaddr *addr2)
{
int addrlen = 0;
const void *addr1p, *addr2p;
if (addr1->sa_family != addr2->sa_family)
return 0;
switch (addr1->sa_family) {
case AF_INET:
addrlen = sizeof(struct in_addr);
addr1p = &((struct sockaddr_in *)addr1)->sin_addr;
addr2p = &((struct sockaddr_in *)addr2)->sin_addr;
break;
case AF_INET6:
addrlen = sizeof(struct in6_addr);
addr1p = &((struct sockaddr_in6 *)addr1)->sin6_addr;
addr2p = &((struct sockaddr_in6 *)addr2)->sin6_addr;
break;
default:
assert(0);
}
return (memcmp(addr1p, addr2p, addrlen) == 0);
}
/* Finds the local node and returns its position in the nodelist.
* Uses nodelist.local_node_pos as a cache to save effort
*/
static int find_local_node(icmap_map_t map, int use_cache)
{
char nodename2[PATH_MAX];
char name_str[ICMAP_KEYNAME_MAXLEN];
icmap_iter_t iter;
const char *iter_key;
unsigned int cached_pos;
char *dot = NULL;
const char *node;
struct ifaddrs *ifa, *ifa_list;
struct sockaddr *sa;
int found = 0;
int node_pos = -1;
int res;
struct utsname utsname;
/* Check for cached value first */
if (use_cache) {
if (icmap_get_uint32("nodelist.local_node_pos", &cached_pos) == CS_OK) {
return cached_pos;
}
}
res = uname(&utsname);
if (res) {
return -1;
}
node = utsname.nodename;
/* 1. Exact match */
node_pos = nodelist_byname(map, node, 0);
if (node_pos > -1) {
found = 1;
goto ret_found;
}
/* 2. Try to match with increasingly more
* specific versions of it
*/
strcpy(nodename2, node);
dot = strrchr(nodename2, '.');
while (dot) {
*dot = '\0';
node_pos = nodelist_byname(map, nodename2, 0);
if (node_pos > -1) {
found = 1;
goto ret_found;
}
dot = strrchr(nodename2, '.');
}
node_pos = nodelist_byname(map, nodename2, 1);
if (node_pos > -1) {
found = 1;
goto ret_found;
}
/*
* The corosync.conf name may not be related to uname at all,
* they may match a hostname on some network interface.
*/
if (getifaddrs(&ifa_list))
return -1;
for (ifa = ifa_list; ifa; ifa = ifa->ifa_next) {
socklen_t salen = 0;
/* Restore this */
strcpy(nodename2, node);
sa = ifa->ifa_addr;
if (!sa) {
continue;
}
if (sa->sa_family != AF_INET && sa->sa_family != AF_INET6) {
continue;
}
if (sa->sa_family == AF_INET) {
salen = sizeof(struct sockaddr_in);
}
if (sa->sa_family == AF_INET6) {
salen = sizeof(struct sockaddr_in6);
}
if (getnameinfo(sa, salen,
nodename2, sizeof(nodename2),
NULL, 0, 0) == 0) {
node_pos = nodelist_byname(map, nodename2, 0);
if (node_pos > -1) {
found = 1;
goto out;
}
/* Truncate this name and try again */
dot = strchr(nodename2, '.');
if (dot) {
*dot = '\0';
node_pos = nodelist_byname(map, nodename2, 0);
if (node_pos > -1) {
found = 1;
goto out;
}
}
}
/* See if it's the IP address that's in corosync.conf */
if (getnameinfo(sa, sizeof(*sa),
nodename2, sizeof(nodename2),
NULL, 0, NI_NUMERICHOST))
continue;
node_pos = nodelist_byname(map, nodename2, 0);
if (node_pos > -1) {
found = 1;
goto out;
}
}
out:
if (found) {
freeifaddrs(ifa_list);
goto ret_found;
}
/*
* This section covers the usecase where the nodename specified in cluster.conf
* is an alias specified in /etc/hosts. For example:
* hostname alias1 alias2
* and
* the above calls use uname and getnameinfo does not return aliases.
* here we take the name specified in cluster.conf, resolve it to an address
* and then compare against all known local ip addresses.
* if we have a match, we found our nodename. In theory this chunk of code
* could replace all the checks above, but let's avoid any possible regressions
* and use it as last.
*/
iter = icmap_iter_init_r(map, "nodelist.node.");
while ((iter_key = icmap_iter_next(iter, NULL, NULL)) != NULL) {
char *dbnodename = NULL;
struct addrinfo hints;
struct addrinfo *result = NULL, *rp = NULL;
res = sscanf(iter_key, "nodelist.node.%u.%s", &node_pos, name_str);
if (res != 2) {
continue;
}
/* 'ring0_addr' is allowed as a fallback, but 'name' will be found first
* because the names are in alpha order.
*/
if (strcmp(name_str, "name") && strcmp(name_str, "ring0_addr")) {
continue;
}
if (icmap_get_string_r(map, iter_key, &dbnodename) != CS_OK) {
continue;
}
memset(&hints, 0, sizeof(struct addrinfo));
hints.ai_family = AF_UNSPEC;
hints.ai_socktype = SOCK_DGRAM;
hints.ai_flags = 0;
hints.ai_protocol = IPPROTO_UDP;
if (getaddrinfo(dbnodename, NULL, &hints, &result)) {
continue;
}
for (rp = result; rp != NULL; rp = rp->ai_next) {
for (ifa = ifa_list; ifa; ifa = ifa->ifa_next) {
if (ifa->ifa_addr &&
ipaddr_equal(rp->ai_addr, ifa->ifa_addr)) {
freeaddrinfo(result);
found = 1;
goto out2;
}
}
}
freeaddrinfo(result);
}
out2:
icmap_iter_finalize(iter);
freeifaddrs(ifa_list);
ret_found:
if (found) {
res = icmap_set_uint32_r(map, "nodelist.local_node_pos", node_pos);
}
return node_pos;
}
static enum totem_ip_version_enum totem_config_get_ip_version(struct totem_config *totem_config)
{
enum totem_ip_version_enum res;
char *str;
res = TOTEM_IP_VERSION_6_4;
if (totem_config->transport_number == TOTEM_TRANSPORT_UDP) {
res = TOTEM_IP_VERSION_4;
}
if (icmap_get_string("totem.ip_version", &str) == CS_OK) {
if (strcmp(str, "ipv4") == 0) {
res = TOTEM_IP_VERSION_4;
}
if (strcmp(str, "ipv6") == 0) {
res = TOTEM_IP_VERSION_6;
}
if (strcmp(str, "ipv6-4") == 0) {
res = TOTEM_IP_VERSION_6_4;
}
if (strcmp(str, "ipv4-6") == 0) {
res = TOTEM_IP_VERSION_4_6;
}
free(str);
}
return (res);
}
static uint16_t generate_cluster_id (const char *cluster_name)
{
int i;
int value = 0;
for (i = 0; i < strlen(cluster_name); i++) {
value <<= 1;
value += cluster_name[i];
}
return (value & 0xFFFF);
}
static int get_cluster_mcast_addr (
const char *cluster_name,
unsigned int linknumber,
enum totem_ip_version_enum ip_version,
struct totem_ip_address *res)
{
uint16_t clusterid;
char addr[INET6_ADDRSTRLEN + 1];
int err;
if (cluster_name == NULL) {
return (-1);
}
clusterid = generate_cluster_id(cluster_name) + linknumber;
memset (res, 0, sizeof(*res));
switch (ip_version) {
case TOTEM_IP_VERSION_4:
case TOTEM_IP_VERSION_4_6:
snprintf(addr, sizeof(addr), "239.192.%d.%d", clusterid >> 8, clusterid % 0xFF);
break;
case TOTEM_IP_VERSION_6:
case TOTEM_IP_VERSION_6_4:
snprintf(addr, sizeof(addr), "ff15::%x", clusterid);
break;
default:
/*
* Unknown family
*/
return (-1);
}
err = totemip_parse (res, addr, ip_version);
return (err);
}
static unsigned int generate_nodeid(
struct totem_config *totem_config,
char *addr)
{
unsigned int nodeid;
struct totem_ip_address totemip;
/* AF_INET hard-coded here because auto-generated nodeids
are only for IPv4 */
if (totemip_parse(&totemip, addr, TOTEM_IP_VERSION_4) != 0)
return -1;
memcpy (&nodeid, &totemip.addr, sizeof (unsigned int));
#if __BYTE_ORDER == __LITTLE_ENDIAN
nodeid = swab32 (nodeid);
#endif
if (totem_config->clear_node_high_bit) {
nodeid &= 0x7FFFFFFF;
}
return nodeid;
}
static int check_for_duplicate_nodeids(
struct totem_config *totem_config,
const char **error_string)
{
icmap_iter_t iter;
icmap_iter_t subiter;
const char *iter_key;
int res = 0;
int retval = 0;
char tmp_key[ICMAP_KEYNAME_MAXLEN];
char *ring0_addr=NULL;
char *ring0_addr1=NULL;
unsigned int node_pos;
unsigned int node_pos1;
unsigned int last_node_pos = -1;
unsigned int nodeid;
unsigned int nodeid1;
int autogenerated;
iter = icmap_iter_init("nodelist.node.");
while ((iter_key = icmap_iter_next(iter, NULL, NULL)) != NULL) {
res = sscanf(iter_key, "nodelist.node.%u.%s", &node_pos, tmp_key);
if (res != 2) {
continue;
}
/*
* This relies on the fact the icmap keys are always returned in order
* so all of the keys for a node will be grouped together. We're basically
* just running the code below once for each node.
*/
if (last_node_pos == node_pos) {
continue;
}
last_node_pos = node_pos;
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "nodelist.node.%u.nodeid", node_pos);
autogenerated = 0;
/* Generated nodeids are only allowed for UDP/UDPU so ring0_addr is valid here */
if (icmap_get_uint32(tmp_key, &nodeid) != CS_OK) {
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "nodelist.node.%u.ring0_addr", node_pos);
if (icmap_get_string(tmp_key, &ring0_addr) != CS_OK) {
continue;
}
/* Generate nodeid so we can check that auto-generated nodeids don't clash either */
nodeid = generate_nodeid(totem_config, ring0_addr);
if (nodeid == -1) {
continue;
}
autogenerated = 1;
}
node_pos1 = 0;
subiter = icmap_iter_init("nodelist.node.");
while (((iter_key = icmap_iter_next(subiter, NULL, NULL)) != NULL) && (node_pos1 < node_pos)) {
res = sscanf(iter_key, "nodelist.node.%u.%s", &node_pos1, tmp_key);
if ((res != 2) || (node_pos1 >= node_pos)) {
continue;
}
if (strcmp(tmp_key, "nodeid") != 0) {
continue;
}
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "nodelist.node.%u.nodeid", node_pos1);
if (icmap_get_uint32(tmp_key, &nodeid1) != CS_OK) {
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "nodelist.node.%u.ring0_addr", node_pos1);
if (icmap_get_string(tmp_key, &ring0_addr1) != CS_OK) {
continue;
}
nodeid1 = generate_nodeid(totem_config, ring0_addr1);
if (nodeid1 == -1) {
continue;
}
}
if (nodeid == nodeid1) {
retval = -1;
snprintf (error_string_response, sizeof(error_string_response),
"Nodeid %u%s%s%s appears twice in corosync.conf", nodeid,
autogenerated?"(autogenerated from ":"",
autogenerated?ring0_addr:"",
autogenerated?")":"");
*error_string = error_string_response;
break;
}
}
icmap_iter_finalize(subiter);
}
icmap_iter_finalize(iter);
return retval;
}
/*
* This needs to be done last of all. It would be nice to do it when reading the
* interface params, but the totem params need to have them to be read first. We
* need both, so this is a way round that circular dependancy.
*/
static void calc_knet_ping_timers(struct totem_config *totem_config)
{
char runtime_key_name[ICMAP_KEYNAME_MAXLEN];
int interface;
for (interface = 0; interface < INTERFACE_MAX; interface++) {
if (totem_config->interfaces[interface].configured) {
if (!totem_config->interfaces[interface].knet_pong_count) {
totem_config->interfaces[interface].knet_pong_count = KNET_PONG_COUNT;
}
if (!totem_config->interfaces[interface].knet_ping_timeout) {
totem_config->interfaces[interface].knet_ping_timeout =
totem_config->token_timeout / totem_config->interfaces[interface].knet_pong_count;
}
snprintf(runtime_key_name, sizeof(runtime_key_name),
"runtime.config.totem.interface.%d.knet_ping_timeout", interface);
icmap_set_uint32(runtime_key_name, totem_config->interfaces[interface].knet_ping_timeout);
if (!totem_config->interfaces[interface].knet_ping_interval) {
totem_config->interfaces[interface].knet_ping_interval =
totem_config->token_timeout / (totem_config->interfaces[interface].knet_pong_count * 2);
}
snprintf(runtime_key_name, sizeof(runtime_key_name),
"runtime.config.totem.interface.%d.knet_ping_interval", interface);
icmap_set_uint32(runtime_key_name, totem_config->interfaces[interface].knet_ping_interval);
}
}
}
/*
* Compute difference between two set of totem interface arrays and commit it.
* set1 and set2
* are changed so for same ring, ip existing in both set1 and set2 are cleared
* (set to 0), and ips which are only in set1 or set2 remains untouched.
* totempg_node_add/remove is called.
*/
static void compute_and_set_totempg_interfaces(struct totem_interface *set1,
struct totem_interface *set2)
{
int ring_no, set1_pos, set2_pos;
struct totem_ip_address empty_ip_address;
memset(&empty_ip_address, 0, sizeof(empty_ip_address));
for (ring_no = 0; ring_no < INTERFACE_MAX; ring_no++) {
if (!set1[ring_no].configured && !set2[ring_no].configured) {
continue;
}
for (set1_pos = 0; set1_pos < set1[ring_no].member_count; set1_pos++) {
for (set2_pos = 0; set2_pos < set2[ring_no].member_count; set2_pos++) {
/*
* For current ring_no remove all set1 items existing
* in set2
*/
if (memcmp(&set1[ring_no].member_list[set1_pos],
&set2[ring_no].member_list[set2_pos],
sizeof(struct totem_ip_address)) == 0) {
memset(&set1[ring_no].member_list[set1_pos], 0,
sizeof(struct totem_ip_address));
memset(&set2[ring_no].member_list[set2_pos], 0,
sizeof(struct totem_ip_address));
}
}
}
}
for (ring_no = 0; ring_no < INTERFACE_MAX; ring_no++) {
for (set1_pos = 0; set1_pos < set1[ring_no].member_count; set1_pos++) {
/*
* All items which remain in set1 and don't exist in set2 any more
* have to be removed.
*/
if (memcmp(&set1[ring_no].member_list[set1_pos], &empty_ip_address, sizeof(empty_ip_address)) != 0) {
log_printf(LOGSYS_LEVEL_DEBUG,
"removing dynamic member %s for ring %u",
totemip_print(&set1[ring_no].member_list[set1_pos]),
ring_no);
totempg_member_remove(&set1[ring_no].member_list[set1_pos], ring_no);
}
}
if (!set2[ring_no].configured) {
continue;
}
for (set2_pos = 0; set2_pos < set2[ring_no].member_count; set2_pos++) {
/*
* All items which remain in set2 and don't exist in set1 are new nodes
* and have to be added.
*/
if (memcmp(&set2[ring_no].member_list[set2_pos], &empty_ip_address, sizeof(empty_ip_address)) != 0) {
log_printf(LOGSYS_LEVEL_DEBUG,
"adding dynamic member %s for ring %u",
totemip_print(&set2[ring_no].member_list[set2_pos]),
ring_no);
totempg_member_add(&set2[ring_no].member_list[set2_pos], ring_no);
}
}
}
}
/*
* Configure parameters for links
*/
static void configure_link_params(struct totem_config *totem_config, icmap_map_t map)
{
int i;
char tmp_key[ICMAP_KEYNAME_MAXLEN];
char *addr_string;
int err;
int local_node_pos = find_local_node(map, 0);
for (i = 0; iinterfaces[i].configured) {
continue;
}
log_printf(LOGSYS_LEVEL_DEBUG, "Configuring link %d params\n", i);
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "nodelist.node.%u.ring%u_addr", local_node_pos, i);
if (icmap_get_string_r(map, tmp_key, &addr_string) != CS_OK) {
continue;
}
err = totemip_parse(&totem_config->interfaces[i].local_ip, addr_string, totem_config->ip_version);
if (err != 0) {
continue;
}
totem_config->interfaces[i].local_ip.nodeid = totem_config->node_id;
/* In case this is a new link, fill in the defaults if there was no interface{} section for it */
if (!totem_config->interfaces[i].knet_link_priority)
totem_config->interfaces[i].knet_link_priority = 1;
/* knet_ping_interval & knet_ping_timeout are set later once we know all the other params */
if (!totem_config->interfaces[i].knet_ping_precision)
totem_config->interfaces[i].knet_ping_precision = KNET_PING_PRECISION;
if (!totem_config->interfaces[i].knet_pong_count)
totem_config->interfaces[i].knet_pong_count = KNET_PONG_COUNT;
if (!totem_config->interfaces[i].knet_transport)
totem_config->interfaces[i].knet_transport = KNET_TRANSPORT_UDP;
if (!totem_config->interfaces[i].ip_port)
totem_config->interfaces[i].ip_port = DEFAULT_PORT + i;
}
}
static void configure_totem_links(struct totem_config *totem_config, icmap_map_t map)
{
int i;
for (i = 0; iinterfaces[i].configured) {
continue;
}
log_printf(LOGSYS_LEVEL_INFO, "Configuring link %d\n", i);
totempg_iface_set(&totem_config->interfaces[i].local_ip, totem_config->interfaces[i].ip_port, i);
}
}
/* Check for differences in config that can't be done on-the-fly and print an error */
static int check_things_have_not_changed(struct totem_config *totem_config, const char **error_string)
{
int i,j,k;
const char *ip_str;
char addr_buf[INET6_ADDRSTRLEN];
int changed = 0;
for (i = 0; iinterfaces[i].configured &&
totem_config->orig_interfaces[i].configured) {
if (totem_config->interfaces[i].knet_transport !=
totem_config->orig_interfaces[i].knet_transport) {
log_printf(LOGSYS_LEVEL_ERROR,
"New config has different knet transport for link %d. Internal value was NOT changed.\n", i);
changed = 1;
}
/* Check each nodeid in the new configuration and make sure its IP address on this link has not changed */
for (j=0; j < totem_config->interfaces[i].member_count; j++) {
for (k=0; k < totem_config->orig_interfaces[i].member_count; k++) {
if (totem_config->interfaces[i].member_list[j].nodeid ==
totem_config->orig_interfaces[i].member_list[k].nodeid) {
/* Found our nodeid - check the IP address */
if (memcmp(&totem_config->interfaces[i].member_list[j],
&totem_config->orig_interfaces[i].member_list[k],
sizeof(struct totem_ip_address))) {
ip_str = totemip_print(&totem_config->orig_interfaces[i].member_list[k]);
/* if ip_str is NULL then the old address was invalid and is allowed to change */
if (ip_str) {
strncpy(addr_buf, ip_str, sizeof(addr_buf));
addr_buf[sizeof(addr_buf) - 1] = '\0';
log_printf(LOGSYS_LEVEL_ERROR,
"new config has different address for link %d (addr changed from %s to %s). Internal value was NOT changed.\n",
i, addr_buf, totemip_print(&totem_config->interfaces[i].member_list[j]));
changed = 1;
}
}
}
}
}
}
}
if (changed) {
snprintf (error_string_response, sizeof(error_string_response),
"To reconfigure an interface it must be deleted and recreated. A working interface needs to be available to corosync at all times");
*error_string = error_string_response;
return -1;
}
return 0;
}
static int put_nodelist_members_to_config(struct totem_config *totem_config, icmap_map_t map,
int reload, const char **error_string)
{
icmap_iter_t iter, iter2;
const char *iter_key, *iter_key2;
int res = 0;
unsigned int node_pos;
char tmp_key[ICMAP_KEYNAME_MAXLEN];
char tmp_key2[ICMAP_KEYNAME_MAXLEN];
char *node_addr_str;
int member_count;
unsigned int linknumber = 0;
int i, j;
int last_node_pos = -1;
/* Clear out nodelist so we can put the new one in if needed */
for (i = 0; i < INTERFACE_MAX; i++) {
for (j = 0; j < PROCESSOR_COUNT_MAX; j++) {
memset(&totem_config->interfaces[i].member_list[j], 0, sizeof(struct totem_ip_address));
}
totem_config->interfaces[i].member_count = 0;
}
iter = icmap_iter_init_r(map, "nodelist.node.");
while ((iter_key = icmap_iter_next(iter, NULL, NULL)) != NULL) {
res = sscanf(iter_key, "nodelist.node.%u.%s", &node_pos, tmp_key);
if (res != 2) {
continue;
}
/* If it's the same as the last node_pos then skip it */
if (node_pos == last_node_pos) {
continue;
}
last_node_pos = node_pos;
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "nodelist.node.%u.", node_pos);
iter2 = icmap_iter_init_r(map, tmp_key);
while ((iter_key2 = icmap_iter_next(iter2, NULL, NULL)) != NULL) {
unsigned int nodeid;
char *str;
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "nodelist.node.%u.nodeid", node_pos);
if (icmap_get_uint32_r(map, tmp_key, &nodeid) != CS_OK) {
nodeid = 0;
}
res = sscanf(iter_key2, "nodelist.node.%u.ring%u%s", &node_pos, &linknumber, tmp_key2);
if (res != 3 || strcmp(tmp_key2, "_addr") != 0) {
continue;
}
if (linknumber >= INTERFACE_MAX) {
snprintf (error_string_response, sizeof(error_string_response),
"parse error in config: interface ring number %u is bigger than allowed maximum %u\n",
linknumber, INTERFACE_MAX - 1);
*error_string = error_string_response;
icmap_iter_finalize(iter2);
icmap_iter_finalize(iter);
return (-1);
}
if (icmap_get_string_r(map, iter_key2, &node_addr_str) != CS_OK) {
continue;
}
/* Generate nodeids if they are not provided and transport is UDP/U */
if (!nodeid &&
(totem_config->transport_number == TOTEM_TRANSPORT_UDP ||
totem_config->transport_number == TOTEM_TRANSPORT_UDPU)) {
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "nodelist.node.%u.ring0_addr", node_pos);
if (icmap_get_string_r(map, tmp_key, &str) == CS_OK) {
nodeid = generate_nodeid(totem_config, str);
if (nodeid == -1) {
sprintf(error_string_response,
"An IPV6 network requires that a node ID be specified "
"for address '%s'.", node_addr_str);
*error_string = error_string_response;
free(str);
return (-1);
}
log_printf(LOGSYS_LEVEL_DEBUG,
"Generated nodeid = " CS_PRI_NODE_ID " for %s", nodeid, str);
free(str);
}
}
member_count = totem_config->interfaces[linknumber].member_count;
res = totemip_parse(&totem_config->interfaces[linknumber].member_list[member_count],
node_addr_str, totem_config->ip_version);
if (res == 0) {
totem_config->interfaces[linknumber].member_list[member_count].nodeid = nodeid;
totem_config->interfaces[linknumber].member_count++;
totem_config->interfaces[linknumber].configured = 1;
} else {
sprintf(error_string_response, "failed to parse node address '%s'\n", node_addr_str);
*error_string = error_string_response;
memset(&totem_config->interfaces[linknumber].member_list[member_count], 0,
sizeof(struct totem_ip_address));
free(node_addr_str);
icmap_iter_finalize(iter2);
icmap_iter_finalize(iter);
return -1;
}
free(node_addr_str);
}
icmap_iter_finalize(iter2);
}
icmap_iter_finalize(iter);
configure_link_params(totem_config, map);
if (reload) {
log_printf(LOGSYS_LEVEL_DEBUG, "About to reconfigure links from nodelist.\n");
if (check_things_have_not_changed(totem_config, error_string) == -1) {
return -1;
}
}
return 0;
}
static void config_convert_nodelist_to_interface(icmap_map_t map, struct totem_config *totem_config)
{
int res = 0;
int node_pos;
char tmp_key[ICMAP_KEYNAME_MAXLEN];
char tmp_key2[ICMAP_KEYNAME_MAXLEN];
char *node_addr_str;
unsigned int linknumber = 0;
icmap_iter_t iter;
const char *iter_key;
node_pos = find_local_node(map, 1);
if (node_pos > -1) {
/*
* We found node, so create interface section
*/
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "nodelist.node.%u.", node_pos);
iter = icmap_iter_init_r(map, tmp_key);
while ((iter_key = icmap_iter_next(iter, NULL, NULL)) != NULL) {
res = sscanf(iter_key, "nodelist.node.%u.ring%u%s", &node_pos, &linknumber, tmp_key2);
if (res != 3 || strcmp(tmp_key2, "_addr") != 0) {
continue ;
}
if (icmap_get_string_r(map, iter_key, &node_addr_str) != CS_OK) {
continue;
}
snprintf(tmp_key2, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.bindnetaddr", linknumber);
icmap_set_string_r(map, tmp_key2, node_addr_str);
free(node_addr_str);
}
icmap_iter_finalize(iter);
}
}
static int get_interface_params(struct totem_config *totem_config, icmap_map_t map,
const char **error_string, uint64_t *warnings,
int reload)
{
int res = 0;
unsigned int linknumber = 0;
int member_count = 0;
int i;
icmap_iter_t iter, member_iter;
const char *iter_key;
const char *member_iter_key;
char linknumber_key[ICMAP_KEYNAME_MAXLEN];
char tmp_key[ICMAP_KEYNAME_MAXLEN];
uint8_t u8;
uint32_t u32;
char *str;
char *cluster_name = NULL;
enum totem_ip_version_enum tmp_ip_version = TOTEM_IP_VERSION_4;
int ret = 0;
if (reload) {
for (i=0; iinterfaces[i].configured = 0;
totem_config->interfaces[i].knet_ping_timeout = 0;
totem_config->interfaces[i].knet_ping_interval = 0;
totem_config->interfaces[i].knet_ping_precision = KNET_PING_PRECISION;
totem_config->interfaces[i].knet_pong_count = KNET_PONG_COUNT;
}
}
if (icmap_get_string_r(map, "totem.cluster_name", &cluster_name) != CS_OK) {
cluster_name = NULL;
}
iter = icmap_iter_init_r(map, "totem.interface.");
while ((iter_key = icmap_iter_next(iter, NULL, NULL)) != NULL) {
res = sscanf(iter_key, "totem.interface.%[^.].%s", linknumber_key, tmp_key);
if (res != 2) {
continue;
}
if (strcmp(tmp_key, "bindnetaddr") != 0 && totem_config->transport_number == TOTEM_TRANSPORT_UDP) {
continue;
}
member_count = 0;
linknumber = atoi(linknumber_key);
if (linknumber >= INTERFACE_MAX) {
snprintf (error_string_response, sizeof(error_string_response),
"parse error in config: interface ring number %u is bigger than allowed maximum %u\n",
linknumber, INTERFACE_MAX - 1);
*error_string = error_string_response;
ret = -1;
goto out;
}
/* These things are only valid for the initial read */
if (!reload) {
/*
* Get the bind net address
*/
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.bindnetaddr", linknumber);
if (icmap_get_string_r(map, tmp_key, &str) == CS_OK) {
res = totemip_parse (&totem_config->interfaces[linknumber].bindnet, str,
totem_config->ip_version);
if (res) {
sprintf(error_string_response, "failed to parse bindnet address '%s'\n", str);
*error_string = error_string_response;
free(str);
ret = -1;
goto out;
}
free(str);
}
/*
* Get interface multicast address
*/
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.mcastaddr", linknumber);
if (icmap_get_string_r(map, tmp_key, &str) == CS_OK) {
res = totemip_parse (&totem_config->interfaces[linknumber].mcast_addr, str,
totem_config->ip_version);
if (res) {
sprintf(error_string_response, "failed to parse mcast address '%s'\n", str);
*error_string = error_string_response;
free(str);
ret = -1;
goto out;
}
free(str);
} else if (totem_config->transport_number == TOTEM_TRANSPORT_UDP) {
/*
* User not specified address -> autogenerate one from cluster_name key
* (if available). Return code is intentionally ignored, because
* udpu doesn't need mcastaddr and validity of mcastaddr for udp is
* checked later anyway.
*/
if (totem_config->interfaces[0].bindnet.family == AF_INET) {
tmp_ip_version = TOTEM_IP_VERSION_4;
} else if (totem_config->interfaces[0].bindnet.family == AF_INET6) {
tmp_ip_version = TOTEM_IP_VERSION_6;
}
(void)get_cluster_mcast_addr (cluster_name,
linknumber,
tmp_ip_version,
&totem_config->interfaces[linknumber].mcast_addr);
}
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.broadcast", linknumber);
if (icmap_get_string(tmp_key, &str) == CS_OK) {
if (strcmp (str, "yes") == 0) {
totem_config->broadcast_use = 1;
}
free(str);
}
}
/* These things are only valid for the initial read OR a newly-defined link */
if (!reload || (totem_config->interfaces[linknumber].configured == 0)) {
/*
* Get mcast port
*/
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.mcastport", linknumber);
if (icmap_get_uint16_r(map, tmp_key, &totem_config->interfaces[linknumber].ip_port) != CS_OK) {
if (totem_config->broadcast_use) {
totem_config->interfaces[linknumber].ip_port = DEFAULT_PORT + (2 * linknumber);
} else {
totem_config->interfaces[linknumber].ip_port = DEFAULT_PORT + linknumber;
}
}
/*
* Get the TTL
*/
totem_config->interfaces[linknumber].ttl = 1;
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.ttl", linknumber);
if (icmap_get_uint8_r(map, tmp_key, &u8) == CS_OK) {
totem_config->interfaces[linknumber].ttl = u8;
}
totem_config->interfaces[linknumber].knet_transport = KNET_DEFAULT_TRANSPORT;
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.knet_transport", linknumber);
if (icmap_get_string_r(map, tmp_key, &str) == CS_OK) {
if (strcmp(str, "sctp") == 0) {
totem_config->interfaces[linknumber].knet_transport = KNET_TRANSPORT_SCTP;
}
else if (strcmp(str, "udp") == 0) {
totem_config->interfaces[linknumber].knet_transport = KNET_TRANSPORT_UDP;
}
else {
*error_string = "Unrecognised knet_transport. expected 'udp' or 'sctp'";
ret = -1;
goto out;
}
}
}
totem_config->interfaces[linknumber].configured = 1;
/*
* Get the knet link params
*/
totem_config->interfaces[linknumber].knet_link_priority = 1;
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.knet_link_priority", linknumber);
if (icmap_get_uint8_r(map, tmp_key, &u8) == CS_OK) {
totem_config->interfaces[linknumber].knet_link_priority = u8;
}
totem_config->interfaces[linknumber].knet_ping_interval = 0; /* real default applied later */
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.knet_ping_interval", linknumber);
if (icmap_get_uint32_r(map, tmp_key, &u32) == CS_OK) {
totem_config->interfaces[linknumber].knet_ping_interval = u32;
}
totem_config->interfaces[linknumber].knet_ping_timeout = 0; /* real default applied later */
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.knet_ping_timeout", linknumber);
if (icmap_get_uint32_r(map, tmp_key, &u32) == CS_OK) {
totem_config->interfaces[linknumber].knet_ping_timeout = u32;
}
totem_config->interfaces[linknumber].knet_ping_precision = KNET_PING_PRECISION;
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.knet_ping_precision", linknumber);
if (icmap_get_uint32_r(map, tmp_key, &u32) == CS_OK) {
totem_config->interfaces[linknumber].knet_ping_precision = u32;
}
totem_config->interfaces[linknumber].knet_pong_count = KNET_PONG_COUNT;
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.knet_pong_count", linknumber);
if (icmap_get_uint32_r(map, tmp_key, &u32) == CS_OK) {
totem_config->interfaces[linknumber].knet_pong_count = u32;
}
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.member.", linknumber);
member_iter = icmap_iter_init_r(map, tmp_key);
while ((member_iter_key = icmap_iter_next(member_iter, NULL, NULL)) != NULL) {
if (member_count == 0) {
if (icmap_get_string("nodelist.node.0.ring0_addr", &str) == CS_OK) {
free(str);
*warnings |= TOTEM_CONFIG_WARNING_MEMBERS_IGNORED;
break;
} else {
*warnings |= TOTEM_CONFIG_WARNING_MEMBERS_DEPRECATED;
}
}
if (icmap_get_string_r(map, member_iter_key, &str) == CS_OK) {
res = totemip_parse (&totem_config->interfaces[linknumber].member_list[member_count++],
str, totem_config->ip_version);
if (res) {
sprintf(error_string_response, "failed to parse node address '%s'\n", str);
*error_string = error_string_response;
icmap_iter_finalize(member_iter);
free(str);
ret = -1;
goto out;
}
free(str);
}
}
icmap_iter_finalize(member_iter);
totem_config->interfaces[linknumber].member_count = member_count;
}
out:
icmap_iter_finalize(iter);
free(cluster_name);
return (ret);
}
extern int totem_config_read (
struct totem_config *totem_config,
const char **error_string,
uint64_t *warnings)
{
int res = 0;
char *str, *ring0_addr_str;
char tmp_key[ICMAP_KEYNAME_MAXLEN];
uint16_t u16;
int i;
int local_node_pos;
- int nodeid_set;
+ uint32_t u32;
*warnings = 0;
memset (totem_config, 0, sizeof (struct totem_config));
totem_config->interfaces = malloc (sizeof (struct totem_interface) * INTERFACE_MAX);
if (totem_config->interfaces == 0) {
*error_string = "Out of memory trying to allocate ethernet interface storage area";
return -1;
}
totem_config->transport_number = TOTEM_TRANSPORT_KNET;
if (icmap_get_string("totem.transport", &str) == CS_OK) {
if (strcmp (str, "udpu") == 0) {
totem_config->transport_number = TOTEM_TRANSPORT_UDPU;
} else if (strcmp (str, "udp") == 0) {
totem_config->transport_number = TOTEM_TRANSPORT_UDP;
} else if (strcmp (str, "knet") == 0) {
totem_config->transport_number = TOTEM_TRANSPORT_KNET;
} else {
*error_string = "Invalid transport type. Should be udpu, udp or knet";
free(str);
return -1;
}
free(str);
}
memset (totem_config->interfaces, 0,
sizeof (struct totem_interface) * INTERFACE_MAX);
strcpy (totem_config->link_mode, "passive");
icmap_get_uint32("totem.version", (uint32_t *)&totem_config->version);
/* initial crypto load */
if (totem_get_crypto(totem_config, icmap_get_global_map(), error_string) != 0) {
return -1;
}
if (totem_config_keyread(totem_config, icmap_get_global_map(), error_string) != 0) {
return -1;
}
totem_config->crypto_index = 1;
totem_config->crypto_changed = 0;
if (icmap_get_string("totem.link_mode", &str) == CS_OK) {
if (strlen(str) >= TOTEM_LINK_MODE_BYTES) {
*error_string = "totem.link_mode is too long";
free(str);
return -1;
}
strcpy (totem_config->link_mode, str);
free(str);
}
- icmap_get_uint32("totem.nodeid", &totem_config->node_id);
+ if (icmap_get_uint32("totem.nodeid", &u32) == CS_OK) {
+ *warnings |= TOTEM_CONFIG_WARNING_TOTEM_NODEID_SET;
+ }
totem_config->clear_node_high_bit = 0;
if (icmap_get_string("totem.clear_node_high_bit", &str) == CS_OK) {
if (strcmp (str, "yes") == 0) {
totem_config->clear_node_high_bit = 1;
}
free(str);
}
icmap_get_uint32("totem.threads", &totem_config->threads);
icmap_get_uint32("totem.netmtu", &totem_config->net_mtu);
totem_config->ip_version = totem_config_get_ip_version(totem_config);
if (icmap_get_string("totem.interface.0.bindnetaddr", &str) != CS_OK) {
/*
* We were not able to find ring 0 bindnet addr. Try to use nodelist informations
*/
config_convert_nodelist_to_interface(icmap_get_global_map(), totem_config);
} else {
if (icmap_get_string("nodelist.node.0.ring0_addr", &ring0_addr_str) == CS_OK) {
/*
* Both bindnetaddr and ring0_addr are set.
* Log warning information, and use nodelist instead
*/
*warnings |= TOTEM_CONFIG_BINDNETADDR_NODELIST_SET;
config_convert_nodelist_to_interface(icmap_get_global_map(), totem_config);
free(ring0_addr_str);
}
free(str);
}
/*
* Broadcast option is global but set in interface section,
* so reset before processing interfaces.
*/
totem_config->broadcast_use = 0;
res = get_interface_params(totem_config, icmap_get_global_map(), error_string, warnings, 0);
if (res < 0) {
return res;
}
/*
* Use broadcast is global, so if set, make sure to fill mcast addr correctly
* broadcast is only supported for UDP so just do interface 0;
*/
if (totem_config->broadcast_use) {
totemip_parse (&totem_config->interfaces[0].mcast_addr,
"255.255.255.255", TOTEM_IP_VERSION_4);
}
/*
* Store automatically generated items back to icmap only for UDP
*/
if (totem_config->transport_number == TOTEM_TRANSPORT_UDP) {
for (i = 0; i < INTERFACE_MAX; i++) {
if (!totem_config->interfaces[i].configured) {
continue;
}
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.mcastaddr", i);
if (icmap_get_string(tmp_key, &str) == CS_OK) {
free(str);
} else {
str = (char *)totemip_print(&totem_config->interfaces[i].mcast_addr);
icmap_set_string(tmp_key, str);
}
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.mcastport", i);
if (icmap_get_uint16(tmp_key, &u16) != CS_OK) {
icmap_set_uint16(tmp_key, totem_config->interfaces[i].ip_port);
}
}
}
/*
* Check existence of nodelist
*/
if ((icmap_get_string("nodelist.node.0.name", &str) == CS_OK) ||
(icmap_get_string("nodelist.node.0.ring0_addr", &str) == CS_OK)) {
free(str);
/*
* find local node
*/
local_node_pos = find_local_node(icmap_get_global_map(), 1);
if (local_node_pos != -1) {
+ assert(totem_config->node_id == 0);
+
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "nodelist.node.%u.nodeid", local_node_pos);
+ (void)icmap_get_uint32(tmp_key, &totem_config->node_id);
+
- nodeid_set = (totem_config->node_id != 0);
- if (icmap_get_uint32(tmp_key, &totem_config->node_id) == CS_OK && nodeid_set) {
- *warnings |= TOTEM_CONFIG_WARNING_TOTEM_NODEID_IGNORED;
- }
if ((totem_config->transport_number == TOTEM_TRANSPORT_KNET) && (!totem_config->node_id)) {
*error_string = "Knet requires an explicit nodeid for the local node";
return -1;
}
if ((totem_config->transport_number == TOTEM_TRANSPORT_UDP ||
totem_config->transport_number == TOTEM_TRANSPORT_UDPU) && (!totem_config->node_id)) {
snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "nodelist.node.%u.ring0_addr", local_node_pos);
icmap_get_string(tmp_key, &str);
totem_config->node_id = generate_nodeid(totem_config, str);
if (totem_config->node_id == -1) {
*error_string = "An IPV6 network requires that a node ID be specified";
free(str);
return (-1);
}
totem_config->interfaces[0].member_list[local_node_pos].nodeid = totem_config->node_id;
free(str);
}
/* Users must not change this */
icmap_set_ro_access("nodelist.local_node_pos", 0, 1);
}
if (put_nodelist_members_to_config(totem_config, icmap_get_global_map(), 0, error_string)) {
return -1;
}
}
/*
* Get things that might change in the future (and can depend on totem_config->interfaces);
*/
totem_volatile_config_read(totem_config, icmap_get_global_map(), NULL);
calc_knet_ping_timers(totem_config);
/* This is now done in the totemknet interface callback */
/* configure_totem_links(totem_config, icmap_get_global_map()); */
add_totem_config_notification(totem_config);
return 0;
}
int totem_config_validate (
struct totem_config *totem_config,
const char **error_string)
{
static char local_error_reason[512];
char parse_error[512];
const char *error_reason = local_error_reason;
int i;
uint32_t u32;
int num_configured = 0;
unsigned int interface_max = INTERFACE_MAX;
for (i = 0; i < INTERFACE_MAX; i++) {
if (totem_config->interfaces[i].configured) {
num_configured++;
}
}
if (num_configured == 0) {
error_reason = "No interfaces defined";
goto parse_error;
}
/* Check we found a local node name */
if (icmap_get_uint32("nodelist.local_node_pos", &u32) != CS_OK) {
error_reason = "No valid name found for local host";
goto parse_error;
}
for (i = 0; i < INTERFACE_MAX; i++) {
/*
* Some error checking of parsed data to make sure its valid
*/
struct totem_ip_address null_addr;
if (!totem_config->interfaces[i].configured) {
continue;
}
memset (&null_addr, 0, sizeof (struct totem_ip_address));
if ((totem_config->transport_number == TOTEM_TRANSPORT_UDP) &&
memcmp (&totem_config->interfaces[i].mcast_addr, &null_addr,
sizeof (struct totem_ip_address)) == 0) {
snprintf (local_error_reason, sizeof(local_error_reason),
"No multicast address specified for interface %u", i);
goto parse_error;
}
if (totem_config->interfaces[i].ip_port == 0) {
snprintf (local_error_reason, sizeof(local_error_reason),
"No multicast port specified for interface %u", i);
goto parse_error;
}
if (totem_config->interfaces[i].ttl > 255) {
snprintf (local_error_reason, sizeof(local_error_reason),
"Invalid TTL (should be 0..255) for interface %u", i);
goto parse_error;
}
if (totem_config->transport_number != TOTEM_TRANSPORT_UDP &&
totem_config->interfaces[i].ttl != 1) {
snprintf (local_error_reason, sizeof(local_error_reason),
"Can only set ttl on multicast transport types for interface %u", i);
goto parse_error;
}
if (totem_config->interfaces[i].knet_link_priority > 255) {
snprintf (local_error_reason, sizeof(local_error_reason),
"Invalid link priority (should be 0..255) for interface %u", i);
goto parse_error;
}
if (totem_config->transport_number != TOTEM_TRANSPORT_KNET &&
totem_config->interfaces[i].knet_link_priority != 1) {
snprintf (local_error_reason, sizeof(local_error_reason),
"Can only set link priority on knet transport type for interface %u", i);
goto parse_error;
}
if (totem_config->interfaces[i].mcast_addr.family == AF_INET6 &&
totem_config->node_id == 0) {
snprintf (local_error_reason, sizeof(local_error_reason),
"An IPV6 network requires that a node ID be specified for interface %u", i);
goto parse_error;
}
if (totem_config->broadcast_use == 0 && totem_config->transport_number == TOTEM_TRANSPORT_UDP) {
if (totem_config->interfaces[i].mcast_addr.family != totem_config->interfaces[i].bindnet.family) {
snprintf (local_error_reason, sizeof(local_error_reason),
"Multicast address family does not match bind address family for interface %u", i);
goto parse_error;
}
if (totemip_is_mcast (&totem_config->interfaces[i].mcast_addr) != 0) {
snprintf (local_error_reason, sizeof(local_error_reason),
"mcastaddr is not a correct multicast address for interface %u", i);
goto parse_error;
}
}
}
if (totem_config->version != 2) {
error_reason = "This totem parser can only parse version 2 configurations.";
goto parse_error;
}
if (totem_volatile_config_validate(totem_config, icmap_get_global_map(), error_string) == -1) {
return (-1);
}
if (check_for_duplicate_nodeids(totem_config, error_string) == -1) {
return (-1);
}
/*
* KNET Link values validation
*/
if (strcmp (totem_config->link_mode, "active") &&
strcmp (totem_config->link_mode, "rr") &&
strcmp (totem_config->link_mode, "passive")) {
snprintf (local_error_reason, sizeof(local_error_reason),
"The Knet link mode \"%s\" specified is invalid. It must be active, passive or rr.\n", totem_config->link_mode);
goto parse_error;
}
/* Only Knet does multiple interfaces */
if (totem_config->transport_number != TOTEM_TRANSPORT_KNET) {
interface_max = 1;
}
if (interface_max < num_configured) {
snprintf (parse_error, sizeof(parse_error),
"%d is too many configured interfaces for non-Knet transport.",
num_configured);
error_reason = parse_error;
goto parse_error;
}
/* Only knet allows crypto */
if (totem_config->transport_number != TOTEM_TRANSPORT_KNET) {
if ((strcmp(totem_config->crypto_cipher_type, "none") != 0) ||
(strcmp(totem_config->crypto_hash_type, "none") != 0)) {
snprintf (parse_error, sizeof(parse_error),
"crypto_cipher & crypto_hash are only valid for the Knet transport.");
error_reason = parse_error;
goto parse_error;
}
}
if (totem_config->net_mtu == 0) {
if (totem_config->transport_number == TOTEM_TRANSPORT_KNET) {
totem_config->net_mtu = KNET_MAX_PACKET_SIZE;
}
else {
totem_config->net_mtu = UDP_NETMTU;
}
}
return 0;
parse_error:
snprintf (error_string_response, sizeof(error_string_response),
"parse error in config: %s\n", error_reason);
*error_string = error_string_response;
return (-1);
}
static int read_keyfile (
const char *key_location,
struct totem_config *totem_config,
const char **error_string)
{
int fd;
int res;
int saved_errno;
char error_str[100];
const char *error_ptr;
fd = open (key_location, O_RDONLY);
if (fd == -1) {
error_ptr = qb_strerror_r(errno, error_str, sizeof(error_str));
snprintf (error_string_response, sizeof(error_string_response),
"Could not open %s: %s\n",
key_location, error_ptr);
goto parse_error;
}
res = read (fd, totem_config->private_key, TOTEM_PRIVATE_KEY_LEN_MAX);
saved_errno = errno;
close (fd);
if (res == -1) {
error_ptr = qb_strerror_r (saved_errno, error_str, sizeof(error_str));
snprintf (error_string_response, sizeof(error_string_response),
"Could not read %s: %s\n",
key_location, error_ptr);
goto parse_error;
}
if (res < TOTEM_PRIVATE_KEY_LEN_MIN) {
snprintf (error_string_response, sizeof(error_string_response),
"Could only read %d bits of minimum %u bits from %s.\n",
res * 8, TOTEM_PRIVATE_KEY_LEN_MIN * 8, key_location);
goto parse_error;
}
totem_config->private_key_len = res;
return 0;
parse_error:
*error_string = error_string_response;
return (-1);
}
int totem_config_keyread (
struct totem_config *totem_config,
icmap_map_t map,
const char **error_string)
{
int got_key = 0;
char *key_location = NULL;
int res;
size_t key_len;
char old_key[TOTEM_PRIVATE_KEY_LEN_MAX];
size_t old_key_len;
/* Take a copy so we can see if it has changed */
memcpy(old_key, totem_config->private_key, sizeof(totem_config->private_key));
old_key_len = totem_config->private_key_len;
memset (totem_config->private_key, 0, sizeof(totem_config->private_key));
totem_config->private_key_len = 0;
if (strcmp(totem_config->crypto_cipher_type, "none") == 0 &&
strcmp(totem_config->crypto_hash_type, "none") == 0) {
return (0);
}
/* cmap may store the location of the key file */
if (icmap_get_string_r(map, "totem.keyfile", &key_location) == CS_OK) {
res = read_keyfile(key_location, totem_config, error_string);
free(key_location);
if (res) {
goto key_error;
}
got_key = 1;
} else { /* Or the key itself may be in the cmap */
if (icmap_get_r(map, "totem.key", NULL, &key_len, NULL) == CS_OK) {
if (key_len > sizeof(totem_config->private_key)) {
sprintf(error_string_response, "key is too long");
goto key_error;
}
if (key_len < TOTEM_PRIVATE_KEY_LEN_MIN) {
sprintf(error_string_response, "key is too short");
goto key_error;
}
if (icmap_get_r(map, "totem.key", totem_config->private_key, &key_len, NULL) == CS_OK) {
totem_config->private_key_len = key_len;
got_key = 1;
} else {
sprintf(error_string_response, "can't load private key");
goto key_error;
}
}
}
/* In desperation we read the default filename */
if (!got_key) {
res = read_keyfile(COROSYSCONFDIR "/authkey", totem_config, error_string);
if (res)
goto key_error;
}
if (old_key_len != totem_config->private_key_len ||
memcmp(old_key, totem_config->private_key, sizeof(totem_config->private_key))) {
totem_config->crypto_changed = 1;
}
return (0);
key_error:
*error_string = error_string_response;
return (-1);
}
int totem_reread_crypto_config(struct totem_config *totem_config, icmap_map_t map, const char **error_string)
{
if (totem_get_crypto(totem_config, map, error_string) != 0) {
return -1;
}
if (totem_config_keyread(totem_config, map, error_string) != 0) {
return -1;
}
return 0;
}
static void debug_dump_totem_config(const struct totem_config *totem_config)
{
log_printf(LOGSYS_LEVEL_DEBUG, "Token Timeout (%d ms) retransmit timeout (%d ms)",
totem_config->token_timeout, totem_config->token_retransmit_timeout);
if (totem_config->token_warning) {
uint32_t token_warning_ms = totem_config->token_warning * totem_config->token_timeout / 100;
log_printf(LOGSYS_LEVEL_DEBUG, "Token warning every %d ms (%d%% of Token Timeout)",
token_warning_ms, totem_config->token_warning);
if (token_warning_ms < totem_config->token_retransmit_timeout)
log_printf (LOGSYS_LEVEL_DEBUG,
"The token warning interval (%d ms) is less than the token retransmit timeout (%d ms) "
"which can lead to spurious token warnings. Consider increasing the token_warning parameter.",
token_warning_ms, totem_config->token_retransmit_timeout);
} else
log_printf(LOGSYS_LEVEL_DEBUG, "Token warnings disabled");
log_printf(LOGSYS_LEVEL_DEBUG, "token hold (%d ms) retransmits before loss (%d retrans)",
totem_config->token_hold_timeout, totem_config->token_retransmits_before_loss_const);
log_printf(LOGSYS_LEVEL_DEBUG, "join (%d ms) send_join (%d ms) consensus (%d ms) merge (%d ms)",
totem_config->join_timeout, totem_config->send_join_timeout, totem_config->consensus_timeout,
totem_config->merge_timeout);
log_printf(LOGSYS_LEVEL_DEBUG, "downcheck (%d ms) fail to recv const (%d msgs)",
totem_config->downcheck_timeout, totem_config->fail_to_recv_const);
log_printf(LOGSYS_LEVEL_DEBUG,
"seqno unchanged const (%d rotations) Maximum network MTU %d",
totem_config->seqno_unchanged_const, totem_config->net_mtu);
log_printf(LOGSYS_LEVEL_DEBUG,
"window size per rotation (%d messages) maximum messages per rotation (%d messages)",
totem_config->window_size, totem_config->max_messages);
log_printf(LOGSYS_LEVEL_DEBUG, "missed count const (%d messages)", totem_config->miss_count_const);
log_printf(LOGSYS_LEVEL_DEBUG, "heartbeat_failures_allowed (%d)",
totem_config->heartbeat_failures_allowed);
log_printf(LOGSYS_LEVEL_DEBUG, "max_network_delay (%d ms)", totem_config->max_network_delay);
}
static void totem_change_notify(
int32_t event,
const char *key_name,
struct icmap_notify_value new_val,
struct icmap_notify_value old_val,
void *user_data)
{
struct totem_config *totem_config = (struct totem_config *)user_data;
uint32_t *param;
uint8_t reloading;
const char *deleted_key = NULL;
const char *error_string;
/*
* If a full reload is in progress then don't do anything until it's done and
* can reconfigure it all atomically
*/
if (icmap_get_uint8("config.reload_in_progress", &reloading) == CS_OK && reloading)
return;
param = totem_get_param_by_name((struct totem_config *)user_data, key_name);
/*
* Process change only if changed key is found in totem_config (-> param is not NULL)
* or for special key token_coefficient. token_coefficient key is not stored in
* totem_config, but it is used for computation of token timeout.
*/
if (!param && strcmp(key_name, "totem.token_coefficient") != 0)
return;
/*
* Values other than UINT32 are not supported, or needed (yet)
*/
switch (event) {
case ICMAP_TRACK_DELETE:
deleted_key = key_name;
break;
case ICMAP_TRACK_ADD:
case ICMAP_TRACK_MODIFY:
deleted_key = NULL;
break;
default:
break;
}
totem_volatile_config_read (totem_config, icmap_get_global_map(), deleted_key);
log_printf(LOGSYS_LEVEL_DEBUG, "Totem related config key changed. Dumping actual totem config.");
debug_dump_totem_config(totem_config);
if (totem_volatile_config_validate(totem_config, icmap_get_global_map(), &error_string) == -1) {
log_printf (LOGSYS_LEVEL_ERROR, "%s", error_string);
/*
* TODO: Consider corosync exit and/or load defaults for volatile
* values. For now, log error seems to be enough
*/
}
}
int totemconfig_configure_new_params(
struct totem_config *totem_config,
icmap_map_t map,
const char **error_string)
{
uint64_t warnings = 0LL;
get_interface_params(totem_config, map, error_string, &warnings, 1);
if (put_nodelist_members_to_config (totem_config, map, 1, error_string)) {
return -1;
}
calc_knet_ping_timers(totem_config);
log_printf(LOGSYS_LEVEL_DEBUG, "Configuration reloaded. Dumping actual totem config.");
debug_dump_totem_config(totem_config);
/* Reinstate the local_node_pos */
(void)find_local_node(map, 0);
return 0;
}
void totemconfig_commit_new_params(
struct totem_config *totem_config,
icmap_map_t map)
{
struct totem_interface *new_interfaces = NULL;
new_interfaces = malloc (sizeof (struct totem_interface) * INTERFACE_MAX);
assert(new_interfaces != NULL);
memcpy(new_interfaces, totem_config->interfaces, sizeof (struct totem_interface) * INTERFACE_MAX);
/* Set link parameters including local_ip */
configure_totem_links(totem_config, map);
/* Add & remove nodes */
compute_and_set_totempg_interfaces(totem_config->orig_interfaces, new_interfaces);
/* Does basic global params (like compression) */
totempg_reconfigure();
free(new_interfaces);
}
static void add_totem_config_notification(struct totem_config *totem_config)
{
icmap_track_t icmap_track;
icmap_track_add("totem.",
ICMAP_TRACK_ADD | ICMAP_TRACK_DELETE | ICMAP_TRACK_MODIFY | ICMAP_TRACK_PREFIX,
totem_change_notify,
totem_config,
&icmap_track);
}
diff --git a/exec/totemconfig.h b/exec/totemconfig.h
index 58621765..2ea338fe 100644
--- a/exec/totemconfig.h
+++ b/exec/totemconfig.h
@@ -1,93 +1,93 @@
/*
* Copyright (c) 2005 MontaVista Software, Inc.
* Copyright (c) 2006-2012 Red Hat, Inc.
*
* All rights reserved.
*
* Author: Steven Dake (sdake@redhat.com)
*
* This software licensed under BSD license, the text of which follows:
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* - Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
* - Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
* - Neither the name of the MontaVista Software, Inc. nor the names of its
* contributors may be used to endorse or promote products derived from this
* software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
* THE POSSIBILITY OF SUCH DAMAGE.
*/
#ifndef TOTEMCONFIG_H_DEFINED
#define TOTEMCONFIG_H_DEFINED
#include
#include
#include
#include
#include "totemsrp.h"
#define TOTEM_CONFIG_WARNING_MEMBERS_IGNORED (1<<1)
#define TOTEM_CONFIG_WARNING_MEMBERS_DEPRECATED (1<<2)
-#define TOTEM_CONFIG_WARNING_TOTEM_NODEID_IGNORED (1<<3)
-#define TOTEM_CONFIG_BINDNETADDR_NODELIST_SET (1<<4)
+#define TOTEM_CONFIG_WARNING_TOTEM_NODEID_SET (1<<3)
+#define TOTEM_CONFIG_BINDNETADDR_NODELIST_SET (1<<4)
extern int totem_config_read (
struct totem_config *totem_config,
const char **error_string,
uint64_t *warnings);
extern int totem_config_validate (
struct totem_config *totem_config,
const char **error_string);
extern int totem_config_keyread (
struct totem_config *totem_config,
icmap_map_t map,
const char **error_string);
extern int totem_config_find_local_addr_in_nodelist(
struct totem_config *totem_config,
const char *ipaddr_key_prefix,
unsigned int *node_pos);
extern void totem_volatile_config_read(
struct totem_config *totem_config,
icmap_map_t temp_map,
const char *deleted_key);
extern int totem_reread_crypto_config(
struct totem_config *totem_config,
icmap_map_t map,
const char **error_string);
extern int totem_volatile_config_validate(
struct totem_config *totem_config,
icmap_map_t temp_map,
const char **error_string);
extern int totemconfig_configure_new_params(
struct totem_config *totem_config,
icmap_map_t map,
const char **error_string);
extern void totemconfig_commit_new_params(
struct totem_config *totem_config,
icmap_map_t map);
#endif /* TOTEMCONFIG_H_DEFINED */