mirror of
https://git.FreeBSD.org/src.git
synced 2024-12-22 11:17:19 +00:00
06a99fe36f
databases. - Make nsswitch support caching. Submitted by: Michael Bushkov <bushman__at__rsu.ru> Sponsored by: Google Summer of Code 2005
885 lines
24 KiB
C
885 lines
24 KiB
C
/*-
|
|
* Copyright (c) 2005 Michael Bushkov <bushman@rsu.ru>
|
|
* All rights reserved.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
* notice, this list of conditions and the following disclaimer in thereg
|
|
* documentation and/or other materials provided with the distribution.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 AUTHOR 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 <sys/cdefs.h>
|
|
__FBSDID("$FreeBSD$");
|
|
|
|
#include <sys/types.h>
|
|
#include <sys/event.h>
|
|
#include <sys/socket.h>
|
|
#include <sys/time.h>
|
|
#include <sys/param.h>
|
|
#include <sys/un.h>
|
|
#include <assert.h>
|
|
#include <err.h>
|
|
#include <errno.h>
|
|
#include <fcntl.h>
|
|
#include <libutil.h>
|
|
#include <pthread.h>
|
|
#include <signal.h>
|
|
#include <stdio.h>
|
|
#include <stdlib.h>
|
|
#include <string.h>
|
|
#include <unistd.h>
|
|
|
|
#include "agents/passwd.h"
|
|
#include "agents/group.h"
|
|
#include "agents/services.h"
|
|
#include "cachedcli.h"
|
|
#include "cachelib.h"
|
|
#include "config.h"
|
|
#include "debug.h"
|
|
#include "log.h"
|
|
#include "parser.h"
|
|
#include "query.h"
|
|
#include "singletons.h"
|
|
|
|
#ifndef CONFIG_PATH
|
|
#define CONFIG_PATH "/etc/cached.conf"
|
|
#endif
|
|
#define DEFAULT_CONFIG_PATH "cached.conf"
|
|
|
|
#define MAX_SOCKET_IO_SIZE 4096
|
|
|
|
struct processing_thread_args {
|
|
cache the_cache;
|
|
struct configuration *the_configuration;
|
|
struct runtime_env *the_runtime_env;
|
|
};
|
|
|
|
static void accept_connection(struct kevent *, struct runtime_env *,
|
|
struct configuration *);
|
|
static void destroy_cache_(cache);
|
|
static void destroy_runtime_env(struct runtime_env *);
|
|
static cache init_cache_(struct configuration *);
|
|
static struct runtime_env *init_runtime_env(struct configuration *);
|
|
static void print_version_info(void);
|
|
static void processing_loop(cache, struct runtime_env *,
|
|
struct configuration *);
|
|
static void process_socket_event(struct kevent *, struct runtime_env *,
|
|
struct configuration *);
|
|
static void process_timer_event(struct kevent *, struct runtime_env *,
|
|
struct configuration *);
|
|
static void *processing_thread(void *);
|
|
static void usage(void);
|
|
|
|
void get_time_func(struct timeval *);
|
|
|
|
static void
|
|
print_version_info(void)
|
|
{
|
|
TRACE_IN(print_version_info);
|
|
printf("cached v0.2 (20 Oct 2005)\nwas developed during SoC 2005\n");
|
|
TRACE_OUT(print_version_info);
|
|
}
|
|
|
|
static void
|
|
usage(void)
|
|
{
|
|
fprintf(stderr,"usage: cached [-nstiId]\n");
|
|
exit(1);
|
|
}
|
|
|
|
static cache
|
|
init_cache_(struct configuration *config)
|
|
{
|
|
struct cache_params params;
|
|
cache retval;
|
|
|
|
struct configuration_entry *config_entry;
|
|
size_t size, i;
|
|
int res;
|
|
|
|
TRACE_IN(init_cache_);
|
|
|
|
memset(¶ms, 0, sizeof(struct cache_params));
|
|
params.get_time_func = get_time_func;
|
|
retval = init_cache(¶ms);
|
|
|
|
size = configuration_get_entries_size(config);
|
|
for (i = 0; i < size; ++i) {
|
|
config_entry = configuration_get_entry(config, i);
|
|
/*
|
|
* We should register common entries now - multipart entries
|
|
* would be registered automatically during the queries.
|
|
*/
|
|
res = register_cache_entry(retval, (struct cache_entry_params *)
|
|
&config_entry->positive_cache_params);
|
|
config_entry->positive_cache_entry = find_cache_entry(retval,
|
|
config_entry->positive_cache_params.entry_name);
|
|
assert(config_entry->positive_cache_entry !=
|
|
INVALID_CACHE_ENTRY);
|
|
|
|
res = register_cache_entry(retval, (struct cache_entry_params *)
|
|
&config_entry->negative_cache_params);
|
|
config_entry->negative_cache_entry = find_cache_entry(retval,
|
|
config_entry->negative_cache_params.entry_name);
|
|
assert(config_entry->negative_cache_entry !=
|
|
INVALID_CACHE_ENTRY);
|
|
}
|
|
|
|
LOG_MSG_2("cache", "cache was successfully initialized");
|
|
TRACE_OUT(init_cache_);
|
|
return (retval);
|
|
}
|
|
|
|
static void
|
|
destroy_cache_(cache the_cache)
|
|
{
|
|
TRACE_IN(destroy_cache_);
|
|
destroy_cache(the_cache);
|
|
TRACE_OUT(destroy_cache_);
|
|
}
|
|
|
|
/*
|
|
* Socket and kqueues are prepared here. We have one global queue for both
|
|
* socket and timers events.
|
|
*/
|
|
static struct runtime_env *
|
|
init_runtime_env(struct configuration *config)
|
|
{
|
|
int serv_addr_len;
|
|
struct sockaddr_un serv_addr;
|
|
|
|
struct kevent eventlist;
|
|
struct timespec timeout;
|
|
|
|
struct runtime_env *retval;
|
|
|
|
TRACE_IN(init_runtime_env);
|
|
retval = (struct runtime_env *)malloc(sizeof(struct runtime_env));
|
|
assert(retval != NULL);
|
|
memset(retval, 0, sizeof(struct runtime_env));
|
|
|
|
retval->sockfd = socket(PF_LOCAL, SOCK_STREAM, 0);
|
|
|
|
if (config->force_unlink == 1)
|
|
unlink(config->socket_path);
|
|
|
|
memset(&serv_addr, 0, sizeof(struct sockaddr_un));
|
|
serv_addr.sun_family = PF_LOCAL;
|
|
strncpy(serv_addr.sun_path, config->socket_path,
|
|
sizeof(serv_addr.sun_path));
|
|
serv_addr_len = sizeof(serv_addr.sun_family) +
|
|
strlen(serv_addr.sun_path) + 1;
|
|
|
|
if (bind(retval->sockfd, (struct sockaddr *)&serv_addr,
|
|
serv_addr_len) == -1) {
|
|
close(retval->sockfd);
|
|
free(retval);
|
|
|
|
LOG_ERR_2("runtime environment", "can't bind socket to path: "
|
|
"%s", config->socket_path);
|
|
TRACE_OUT(init_runtime_env);
|
|
return (NULL);
|
|
}
|
|
LOG_MSG_2("runtime environment", "using socket %s",
|
|
config->socket_path);
|
|
|
|
/*
|
|
* Here we're marking socket as non-blocking and setting its backlog
|
|
* to the maximum value
|
|
*/
|
|
chmod(config->socket_path, config->socket_mode);
|
|
listen(retval->sockfd, -1);
|
|
fcntl(retval->sockfd, F_SETFL, O_NONBLOCK);
|
|
|
|
retval->queue = kqueue();
|
|
assert(retval->queue != -1);
|
|
|
|
EV_SET(&eventlist, retval->sockfd, EVFILT_READ, EV_ADD | EV_ONESHOT,
|
|
0, 0, 0);
|
|
memset(&timeout, 0, sizeof(struct timespec));
|
|
kevent(retval->queue, &eventlist, 1, NULL, 0, &timeout);
|
|
|
|
LOG_MSG_2("runtime environment", "successfully initialized");
|
|
TRACE_OUT(init_runtime_env);
|
|
return (retval);
|
|
}
|
|
|
|
static void
|
|
destroy_runtime_env(struct runtime_env *env)
|
|
{
|
|
TRACE_IN(destroy_runtime_env);
|
|
close(env->queue);
|
|
close(env->sockfd);
|
|
free(env);
|
|
TRACE_OUT(destroy_runtime_env);
|
|
}
|
|
|
|
static void
|
|
accept_connection(struct kevent *event_data, struct runtime_env *env,
|
|
struct configuration *config)
|
|
{
|
|
struct kevent eventlist[2];
|
|
struct timespec timeout;
|
|
struct query_state *qstate;
|
|
|
|
int fd;
|
|
int res;
|
|
|
|
uid_t euid;
|
|
gid_t egid;
|
|
|
|
TRACE_IN(accept_connection);
|
|
fd = accept(event_data->ident, NULL, NULL);
|
|
if (fd == -1) {
|
|
LOG_ERR_2("accept_connection", "error %d during accept()",
|
|
errno);
|
|
TRACE_OUT(accept_connection);
|
|
return;
|
|
}
|
|
|
|
if (getpeereid(fd, &euid, &egid) != 0) {
|
|
LOG_ERR_2("accept_connection", "error %d during getpeereid()",
|
|
errno);
|
|
TRACE_OUT(accept_connection);
|
|
return;
|
|
}
|
|
|
|
qstate = init_query_state(fd, sizeof(int), euid, egid);
|
|
if (qstate == NULL) {
|
|
LOG_ERR_2("accept_connection", "can't init query_state");
|
|
TRACE_OUT(accept_connection);
|
|
return;
|
|
}
|
|
|
|
memset(&timeout, 0, sizeof(struct timespec));
|
|
EV_SET(&eventlist[0], fd, EVFILT_TIMER, EV_ADD | EV_ONESHOT,
|
|
0, qstate->timeout.tv_sec * 1000, qstate);
|
|
EV_SET(&eventlist[1], fd, EVFILT_READ, EV_ADD | EV_ONESHOT,
|
|
NOTE_LOWAT, qstate->kevent_watermark, qstate);
|
|
res = kevent(env->queue, eventlist, 2, NULL, 0, &timeout);
|
|
if (res < 0)
|
|
LOG_ERR_2("accept_connection", "kevent error");
|
|
|
|
TRACE_OUT(accept_connection);
|
|
}
|
|
|
|
static void
|
|
process_socket_event(struct kevent *event_data, struct runtime_env *env,
|
|
struct configuration *config)
|
|
{
|
|
struct kevent eventlist[2];
|
|
struct timeval query_timeout;
|
|
struct timespec kevent_timeout;
|
|
int nevents;
|
|
int eof_res, res;
|
|
ssize_t io_res;
|
|
struct query_state *qstate;
|
|
|
|
TRACE_IN(process_socket_event);
|
|
eof_res = event_data->flags & EV_EOF ? 1 : 0;
|
|
res = 0;
|
|
|
|
memset(&kevent_timeout, 0, sizeof(struct timespec));
|
|
EV_SET(&eventlist[0], event_data->ident, EVFILT_TIMER, EV_DELETE,
|
|
0, 0, NULL);
|
|
nevents = kevent(env->queue, eventlist, 1, NULL, 0, &kevent_timeout);
|
|
if (nevents == -1) {
|
|
if (errno == ENOENT) {
|
|
/* the timer is already handling this event */
|
|
TRACE_OUT(process_socket_event);
|
|
return;
|
|
} else {
|
|
/* some other error happened */
|
|
LOG_ERR_2("process_socket_event", "kevent error, errno"
|
|
" is %d", errno);
|
|
TRACE_OUT(process_socket_event);
|
|
return;
|
|
}
|
|
}
|
|
qstate = (struct query_state *)event_data->udata;
|
|
|
|
/*
|
|
* If the buffer that is to be send/received is too large,
|
|
* we send it implicitly, by using query_io_buffer_read and
|
|
* query_io_buffer_write functions in the query_state. These functions
|
|
* use the temporary buffer, which is later send/received in parts.
|
|
* The code below implements buffer splitting/mergind for send/receive
|
|
* operations. It also does the actual socket IO operations.
|
|
*/
|
|
if (((qstate->use_alternate_io == 0) &&
|
|
(qstate->kevent_watermark <= event_data->data)) ||
|
|
((qstate->use_alternate_io != 0) &&
|
|
(qstate->io_buffer_watermark <= event_data->data))) {
|
|
if (qstate->use_alternate_io != 0) {
|
|
switch (qstate->io_buffer_filter) {
|
|
case EVFILT_READ:
|
|
io_res = query_socket_read(qstate,
|
|
qstate->io_buffer_p,
|
|
qstate->io_buffer_watermark);
|
|
if (io_res < 0) {
|
|
qstate->use_alternate_io = 0;
|
|
qstate->process_func = NULL;
|
|
} else {
|
|
qstate->io_buffer_p += io_res;
|
|
if (qstate->io_buffer_p ==
|
|
qstate->io_buffer +
|
|
qstate->io_buffer_size) {
|
|
qstate->io_buffer_p =
|
|
qstate->io_buffer;
|
|
qstate->use_alternate_io = 0;
|
|
}
|
|
}
|
|
break;
|
|
default:
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (qstate->use_alternate_io == 0) {
|
|
do {
|
|
res = qstate->process_func(qstate);
|
|
} while ((qstate->kevent_watermark == 0) &&
|
|
(qstate->process_func != NULL) &&
|
|
(res == 0));
|
|
|
|
if (res != 0)
|
|
qstate->process_func = NULL;
|
|
}
|
|
|
|
if ((qstate->use_alternate_io != 0) &&
|
|
(qstate->io_buffer_filter == EVFILT_WRITE)) {
|
|
io_res = query_socket_write(qstate, qstate->io_buffer_p,
|
|
qstate->io_buffer_watermark);
|
|
if (io_res < 0) {
|
|
qstate->use_alternate_io = 0;
|
|
qstate->process_func = NULL;
|
|
} else
|
|
qstate->io_buffer_p += io_res;
|
|
}
|
|
} else {
|
|
/* assuming that socket was closed */
|
|
qstate->process_func = NULL;
|
|
qstate->use_alternate_io = 0;
|
|
}
|
|
|
|
if (((qstate->process_func == NULL) &&
|
|
(qstate->use_alternate_io == 0)) ||
|
|
(eof_res != 0) || (res != 0)) {
|
|
destroy_query_state(qstate);
|
|
close(event_data->ident);
|
|
TRACE_OUT(process_socket_event);
|
|
return;
|
|
}
|
|
|
|
/* updating the query_state lifetime variable */
|
|
get_time_func(&query_timeout);
|
|
query_timeout.tv_usec = 0;
|
|
query_timeout.tv_sec -= qstate->creation_time.tv_sec;
|
|
if (query_timeout.tv_sec > qstate->timeout.tv_sec)
|
|
query_timeout.tv_sec = 0;
|
|
else
|
|
query_timeout.tv_sec = qstate->timeout.tv_sec -
|
|
query_timeout.tv_sec;
|
|
|
|
if ((qstate->use_alternate_io != 0) && (qstate->io_buffer_p ==
|
|
qstate->io_buffer + qstate->io_buffer_size))
|
|
qstate->use_alternate_io = 0;
|
|
|
|
if (qstate->use_alternate_io == 0) {
|
|
/*
|
|
* If we must send/receive the large block of data,
|
|
* we should prepare the query_state's io_XXX fields.
|
|
* We should also substitute its write_func and read_func
|
|
* with the query_io_buffer_write and query_io_buffer_read,
|
|
* which will allow us to implicitly send/receive this large
|
|
* buffer later (in the subsequent calls to the
|
|
* process_socket_event).
|
|
*/
|
|
if (qstate->kevent_watermark > MAX_SOCKET_IO_SIZE) {
|
|
if (qstate->io_buffer != NULL)
|
|
free(qstate->io_buffer);
|
|
|
|
qstate->io_buffer = (char *)malloc(
|
|
qstate->kevent_watermark);
|
|
assert(qstate->io_buffer != NULL);
|
|
memset(qstate->io_buffer, 0, qstate->kevent_watermark);
|
|
|
|
qstate->io_buffer_p = qstate->io_buffer;
|
|
qstate->io_buffer_size = qstate->kevent_watermark;
|
|
qstate->io_buffer_filter = qstate->kevent_filter;
|
|
|
|
qstate->write_func = query_io_buffer_write;
|
|
qstate->read_func = query_io_buffer_read;
|
|
|
|
if (qstate->kevent_filter == EVFILT_READ)
|
|
qstate->use_alternate_io = 1;
|
|
|
|
qstate->io_buffer_watermark = MAX_SOCKET_IO_SIZE;
|
|
EV_SET(&eventlist[1], event_data->ident,
|
|
qstate->kevent_filter, EV_ADD | EV_ONESHOT,
|
|
NOTE_LOWAT, MAX_SOCKET_IO_SIZE, qstate);
|
|
} else {
|
|
EV_SET(&eventlist[1], event_data->ident,
|
|
qstate->kevent_filter, EV_ADD | EV_ONESHOT,
|
|
NOTE_LOWAT, qstate->kevent_watermark, qstate);
|
|
}
|
|
} else {
|
|
if (qstate->io_buffer + qstate->io_buffer_size -
|
|
qstate->io_buffer_p <
|
|
MAX_SOCKET_IO_SIZE) {
|
|
qstate->io_buffer_watermark = qstate->io_buffer +
|
|
qstate->io_buffer_size - qstate->io_buffer_p;
|
|
EV_SET(&eventlist[1], event_data->ident,
|
|
qstate->io_buffer_filter,
|
|
EV_ADD | EV_ONESHOT, NOTE_LOWAT,
|
|
qstate->io_buffer_watermark,
|
|
qstate);
|
|
} else {
|
|
qstate->io_buffer_watermark = MAX_SOCKET_IO_SIZE;
|
|
EV_SET(&eventlist[1], event_data->ident,
|
|
qstate->io_buffer_filter, EV_ADD | EV_ONESHOT,
|
|
NOTE_LOWAT, MAX_SOCKET_IO_SIZE, qstate);
|
|
}
|
|
}
|
|
EV_SET(&eventlist[0], event_data->ident, EVFILT_TIMER,
|
|
EV_ADD | EV_ONESHOT, 0, query_timeout.tv_sec * 1000, qstate);
|
|
kevent(env->queue, eventlist, 2, NULL, 0, &kevent_timeout);
|
|
|
|
TRACE_OUT(process_socket_event);
|
|
}
|
|
|
|
/*
|
|
* This routine is called if timer event has been signaled in the kqueue. It
|
|
* just closes the socket and destroys the query_state.
|
|
*/
|
|
static void
|
|
process_timer_event(struct kevent *event_data, struct runtime_env *env,
|
|
struct configuration *config)
|
|
{
|
|
struct query_state *qstate;
|
|
|
|
TRACE_IN(process_timer_event);
|
|
qstate = (struct query_state *)event_data->udata;
|
|
destroy_query_state(qstate);
|
|
close(event_data->ident);
|
|
TRACE_OUT(process_timer_event);
|
|
}
|
|
|
|
/*
|
|
* Processing loop is the basic processing routine, that forms a body of each
|
|
* procssing thread
|
|
*/
|
|
static void
|
|
processing_loop(cache the_cache, struct runtime_env *env,
|
|
struct configuration *config)
|
|
{
|
|
struct timespec timeout;
|
|
const int eventlist_size = 1;
|
|
struct kevent eventlist[eventlist_size];
|
|
int nevents, i;
|
|
|
|
TRACE_MSG("=> processing_loop");
|
|
memset(&timeout, 0, sizeof(struct timespec));
|
|
memset(&eventlist, 0, sizeof(struct kevent) * eventlist_size);
|
|
|
|
for (;;) {
|
|
nevents = kevent(env->queue, NULL, 0, eventlist,
|
|
eventlist_size, NULL);
|
|
/*
|
|
* we can only receive 1 event on success
|
|
*/
|
|
if (nevents == 1) {
|
|
struct kevent *event_data;
|
|
event_data = &eventlist[0];
|
|
|
|
if (event_data->ident == env->sockfd) {
|
|
for (i = 0; i < event_data->data; ++i)
|
|
accept_connection(event_data, env, config);
|
|
|
|
EV_SET(eventlist, s_runtime_env->sockfd,
|
|
EVFILT_READ, EV_ADD | EV_ONESHOT,
|
|
0, 0, 0);
|
|
memset(&timeout, 0,
|
|
sizeof(struct timespec));
|
|
kevent(s_runtime_env->queue, eventlist,
|
|
1, NULL, 0, &timeout);
|
|
|
|
} else {
|
|
switch (event_data->filter) {
|
|
case EVFILT_READ:
|
|
case EVFILT_WRITE:
|
|
process_socket_event(event_data,
|
|
env, config);
|
|
break;
|
|
case EVFILT_TIMER:
|
|
process_timer_event(event_data,
|
|
env, config);
|
|
break;
|
|
default:
|
|
break;
|
|
}
|
|
}
|
|
} else {
|
|
/* this branch shouldn't be currently executed */
|
|
}
|
|
}
|
|
|
|
TRACE_MSG("<= processing_loop");
|
|
}
|
|
|
|
/*
|
|
* Wrapper above the processing loop function. It sets the thread signal mask
|
|
* to avoid SIGPIPE signals (which can happen if the client works incorrectly).
|
|
*/
|
|
static void *
|
|
processing_thread(void *data)
|
|
{
|
|
struct processing_thread_args *args;
|
|
sigset_t new;
|
|
|
|
TRACE_MSG("=> processing_thread");
|
|
args = (struct processing_thread_args *)data;
|
|
|
|
sigemptyset(&new);
|
|
sigaddset(&new, SIGPIPE);
|
|
if (pthread_sigmask(SIG_BLOCK, &new, NULL) != 0)
|
|
LOG_ERR_1("processing thread",
|
|
"thread can't block the SIGPIPE signal");
|
|
|
|
processing_loop(args->the_cache, args->the_runtime_env,
|
|
args->the_configuration);
|
|
free(args);
|
|
TRACE_MSG("<= processing_thread");
|
|
|
|
return (NULL);
|
|
}
|
|
|
|
void
|
|
get_time_func(struct timeval *time)
|
|
{
|
|
struct timespec res;
|
|
memset(&res, 0, sizeof(struct timespec));
|
|
clock_gettime(CLOCK_MONOTONIC, &res);
|
|
|
|
time->tv_sec = res.tv_sec;
|
|
time->tv_usec = 0;
|
|
}
|
|
|
|
/*
|
|
* The idea of _nss_cache_cycle_prevention_function is that nsdispatch will
|
|
* search for this symbol in the executable. This symbol is the attribute of
|
|
* the caching daemon. So, if it exists, nsdispatch won't try to connect to
|
|
* the caching daemon and will just ignore the 'cache' source in the
|
|
* nsswitch.conf. This method helps to avoid cycles and organize
|
|
* self-performing requests.
|
|
*/
|
|
void
|
|
_nss_cache_cycle_prevention_function(void)
|
|
{
|
|
}
|
|
|
|
int
|
|
main(int argc, char *argv[])
|
|
{
|
|
struct processing_thread_args *thread_args;
|
|
pthread_t *threads;
|
|
|
|
struct pidfh *pidfile;
|
|
pid_t pid;
|
|
|
|
char const *config_file;
|
|
char const *error_str;
|
|
int error_line;
|
|
int i, res;
|
|
|
|
int trace_mode_enabled;
|
|
int force_single_threaded;
|
|
int do_not_daemonize;
|
|
int clear_user_cache_entries, clear_all_cache_entries;
|
|
char *user_config_entry_name, *global_config_entry_name;
|
|
int show_statistics;
|
|
int daemon_mode, interactive_mode;
|
|
|
|
|
|
/* by default all debug messages are omitted */
|
|
TRACE_OFF();
|
|
|
|
/* startup output */
|
|
print_version_info();
|
|
|
|
/* parsing command line arguments */
|
|
trace_mode_enabled = 0;
|
|
force_single_threaded = 0;
|
|
do_not_daemonize = 0;
|
|
clear_user_cache_entries = 0;
|
|
clear_all_cache_entries = 0;
|
|
show_statistics = 0;
|
|
user_config_entry_name = NULL;
|
|
global_config_entry_name = NULL;
|
|
while ((res = getopt(argc, argv, "nstdi:I:")) != -1) {
|
|
switch (res) {
|
|
case 'n':
|
|
do_not_daemonize = 1;
|
|
break;
|
|
case 's':
|
|
force_single_threaded = 1;
|
|
break;
|
|
case 't':
|
|
trace_mode_enabled = 1;
|
|
break;
|
|
case 'i':
|
|
clear_user_cache_entries = 1;
|
|
if (optarg != NULL)
|
|
if (strcmp(optarg, "all") != 0)
|
|
user_config_entry_name = strdup(optarg);
|
|
break;
|
|
case 'I':
|
|
clear_all_cache_entries = 1;
|
|
if (optarg != NULL)
|
|
if (strcmp(optarg, "all") != 0)
|
|
global_config_entry_name =
|
|
strdup(optarg);
|
|
break;
|
|
case 'd':
|
|
show_statistics = 1;
|
|
break;
|
|
case '?':
|
|
default:
|
|
usage();
|
|
/* NOT REACHED */
|
|
}
|
|
}
|
|
|
|
daemon_mode = do_not_daemonize | force_single_threaded |
|
|
trace_mode_enabled;
|
|
interactive_mode = clear_user_cache_entries | clear_all_cache_entries |
|
|
show_statistics;
|
|
|
|
if ((daemon_mode != 0) && (interactive_mode != 0)) {
|
|
LOG_ERR_1("main", "daemon mode and interactive_mode arguments "
|
|
"can't be used together");
|
|
usage();
|
|
}
|
|
|
|
if (interactive_mode != 0) {
|
|
FILE *pidfin = fopen(DEFAULT_PIDFILE_PATH, "r");
|
|
char pidbuf[256];
|
|
|
|
struct cached_connection_params connection_params;
|
|
cached_connection connection;
|
|
|
|
int result;
|
|
|
|
if (pidfin == NULL)
|
|
errx(EXIT_FAILURE, "There is no daemon running.");
|
|
|
|
memset(pidbuf, 0, sizeof(pidbuf));
|
|
fread(pidbuf, sizeof(pidbuf) - 1, 1, pidfin);
|
|
fclose(pidfin);
|
|
|
|
if (ferror(pidfin) != 0)
|
|
errx(EXIT_FAILURE, "Can't read from pidfile.");
|
|
|
|
if (sscanf(pidbuf, "%d", &pid) != 1)
|
|
errx(EXIT_FAILURE, "Invalid pidfile.");
|
|
LOG_MSG_1("main", "daemon PID is %d", pid);
|
|
|
|
|
|
memset(&connection_params, 0,
|
|
sizeof(struct cached_connection_params));
|
|
connection_params.socket_path = DEFAULT_SOCKET_PATH;
|
|
connection = open_cached_connection__(&connection_params);
|
|
if (connection == INVALID_CACHED_CONNECTION)
|
|
errx(EXIT_FAILURE, "Can't connect to the daemon.");
|
|
|
|
if (clear_user_cache_entries != 0) {
|
|
result = cached_transform__(connection,
|
|
user_config_entry_name, TT_USER);
|
|
if (result != 0)
|
|
LOG_MSG_1("main",
|
|
"user cache transformation failed");
|
|
else
|
|
LOG_MSG_1("main",
|
|
"user cache_transformation "
|
|
"succeeded");
|
|
}
|
|
|
|
if (clear_all_cache_entries != 0) {
|
|
if (geteuid() != 0)
|
|
errx(EXIT_FAILURE, "Only root can initiate "
|
|
"global cache transformation.");
|
|
|
|
result = cached_transform__(connection,
|
|
global_config_entry_name, TT_ALL);
|
|
if (result != 0)
|
|
LOG_MSG_1("main",
|
|
"global cache transformation "
|
|
"failed");
|
|
else
|
|
LOG_MSG_1("main",
|
|
"global cache transformation "
|
|
"succeeded");
|
|
}
|
|
|
|
close_cached_connection__(connection);
|
|
|
|
free(user_config_entry_name);
|
|
free(global_config_entry_name);
|
|
return (EXIT_SUCCESS);
|
|
}
|
|
|
|
pidfile = pidfile_open(DEFAULT_PIDFILE_PATH, 0644, &pid);
|
|
if (pidfile == NULL) {
|
|
if (errno == EEXIST)
|
|
errx(EXIT_FAILURE, "Daemon already running, pid: %d.",
|
|
pid);
|
|
warn("Cannot open or create pidfile");
|
|
}
|
|
|
|
if (trace_mode_enabled == 1)
|
|
TRACE_ON();
|
|
|
|
/* blocking the main thread from receiving SIGPIPE signal */
|
|
sigblock(sigmask(SIGPIPE));
|
|
|
|
/* daemonization */
|
|
if (do_not_daemonize == 0) {
|
|
res = daemon(0, trace_mode_enabled == 0 ? 0 : 1);
|
|
if (res != 0) {
|
|
LOG_ERR_1("main", "can't daemonize myself: %s",
|
|
strerror(errno));
|
|
pidfile_remove(pidfile);
|
|
goto fin;
|
|
} else
|
|
LOG_MSG_1("main", "successfully daemonized");
|
|
}
|
|
|
|
pidfile_write(pidfile);
|
|
|
|
s_agent_table = init_agent_table();
|
|
register_agent(s_agent_table, init_passwd_agent());
|
|
register_agent(s_agent_table, init_passwd_mp_agent());
|
|
register_agent(s_agent_table, init_group_agent());
|
|
register_agent(s_agent_table, init_group_mp_agent());
|
|
register_agent(s_agent_table, init_services_agent());
|
|
register_agent(s_agent_table, init_services_mp_agent());
|
|
LOG_MSG_1("main", "request agents registered successfully");
|
|
|
|
/*
|
|
* Hosts agent can't work properly until we have access to the
|
|
* appropriate dtab structures, which are used in nsdispatch
|
|
* calls
|
|
*
|
|
register_agent(s_agent_table, init_hosts_agent());
|
|
*/
|
|
|
|
/* configuration initialization */
|
|
s_configuration = init_configuration();
|
|
fill_configuration_defaults(s_configuration);
|
|
|
|
error_str = NULL;
|
|
error_line = 0;
|
|
config_file = CONFIG_PATH;
|
|
|
|
res = parse_config_file(s_configuration, config_file, &error_str,
|
|
&error_line);
|
|
if ((res != 0) && (error_str == NULL)) {
|
|
config_file = DEFAULT_CONFIG_PATH;
|
|
res = parse_config_file(s_configuration, config_file,
|
|
&error_str, &error_line);
|
|
}
|
|
|
|
if (res != 0) {
|
|
if (error_str != NULL) {
|
|
LOG_ERR_1("main", "error in configuration file(%s, %d): %s\n",
|
|
config_file, error_line, error_str);
|
|
} else {
|
|
LOG_ERR_1("main", "no configuration file found "
|
|
"- was looking for %s and %s",
|
|
CONFIG_PATH, DEFAULT_CONFIG_PATH);
|
|
}
|
|
destroy_configuration(s_configuration);
|
|
return (-1);
|
|
}
|
|
|
|
if (force_single_threaded == 1)
|
|
s_configuration->threads_num = 1;
|
|
|
|
/* cache initialization */
|
|
s_cache = init_cache_(s_configuration);
|
|
if (s_cache == NULL) {
|
|
LOG_ERR_1("main", "can't initialize the cache");
|
|
destroy_configuration(s_configuration);
|
|
return (-1);
|
|
}
|
|
|
|
/* runtime environment initialization */
|
|
s_runtime_env = init_runtime_env(s_configuration);
|
|
if (s_runtime_env == NULL) {
|
|
LOG_ERR_1("main", "can't initialize the runtime environment");
|
|
destroy_configuration(s_configuration);
|
|
destroy_cache_(s_cache);
|
|
return (-1);
|
|
}
|
|
|
|
if (s_configuration->threads_num > 1) {
|
|
threads = (pthread_t *)malloc(sizeof(pthread_t) *
|
|
s_configuration->threads_num);
|
|
memset(threads, 0, sizeof(pthread_t) *
|
|
s_configuration->threads_num);
|
|
for (i = 0; i < s_configuration->threads_num; ++i) {
|
|
thread_args = (struct processing_thread_args *)malloc(
|
|
sizeof(struct processing_thread_args));
|
|
thread_args->the_cache = s_cache;
|
|
thread_args->the_runtime_env = s_runtime_env;
|
|
thread_args->the_configuration = s_configuration;
|
|
|
|
LOG_MSG_1("main", "thread #%d was successfully created",
|
|
i);
|
|
pthread_create(&threads[i], NULL, processing_thread,
|
|
thread_args);
|
|
|
|
thread_args = NULL;
|
|
}
|
|
|
|
for (i = 0; i < s_configuration->threads_num; ++i)
|
|
pthread_join(threads[i], NULL);
|
|
} else {
|
|
LOG_MSG_1("main", "working in single-threaded mode");
|
|
processing_loop(s_cache, s_runtime_env, s_configuration);
|
|
}
|
|
|
|
fin:
|
|
/* runtime environment destruction */
|
|
destroy_runtime_env(s_runtime_env);
|
|
|
|
/* cache destruction */
|
|
destroy_cache_(s_cache);
|
|
|
|
/* configuration destruction */
|
|
destroy_configuration(s_configuration);
|
|
|
|
/* agents table destruction */
|
|
destroy_agent_table(s_agent_table);
|
|
|
|
pidfile_remove(pidfile);
|
|
return (EXIT_SUCCESS);
|
|
}
|