rename accel-pptp to accel-ppp

4 files changed, 1275 insertions, 0 deletions

diff --git a/accel-pppd/logs/CMakeLists.txt b/accel-pppd/logs/CMakeLists.txt
new file mode 100644
index 0000000..ce909b6
--- /dev/null
+++ b/accel-pppd/logs/CMakeLists.txt
@@ -0,0 +1,17 @@
+ADD_LIBRARY(log_file SHARED log_file.c)
+TARGET_LINK_LIBRARIES(log_file rt)
+
+ADD_LIBRARY(log_tcp SHARED log_tcp.c)
+
+INSTALL(TARGETS log_file log_tcp
+	LIBRARY DESTINATION lib/accel-ppp
+)
+
+IF(LOG_PGSQL)
+	ADD_LIBRARY(log_pgsql SHARED log_pgsql.c)
+	TARGET_LINK_LIBRARIES(log_pgsql pq)
+	INSTALL(TARGETS log_pgsql
+		LIBRARY DESTINATION lib/accel-ppp
+	)
+ENDIF(LOG_PGSQL)
+
diff --git a/accel-pppd/logs/log_file.c b/accel-pppd/logs/log_file.c
new file mode 100644
index 0000000..e435757
--- /dev/null
+++ b/accel-pppd/logs/log_file.c
@@ -0,0 +1,614 @@
+#include <stdlib.h>
+#include <stdio.h>
+#include <errno.h>
+#include <string.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <limits.h>
+#include <aio.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+
+#include "log.h"
+#include "events.h"
+#include "ppp.h"
+#include "spinlock.h"
+#include "mempool.h"
+
+#include "memdebug.h"
+
+#define LOG_BUF_SIZE 16*1024
+
+#define RED_COLOR     "\033[1;31m"
+#define GREEN_COLOR   "\033[1;32m"
+#define YELLOW_COLOR  "\033[1;33m"
+#define BLUE_COLOR  	"\033[1;34m"
+#define NORMAL_COLOR  "\033[0;39m"
+
+struct log_file_t
+{
+	struct list_head entry;
+	struct list_head msgs;
+	spinlock_t lock;
+	int need_free:1;
+	int queued:1;
+	struct log_file_pd_t *lpd;
+
+	int fd;
+	int new_fd;
+	off_t offset;
+	unsigned long magic;
+};
+
+struct log_file_pd_t
+{
+	struct ppp_pd_t pd;
+	struct log_file_t lf;
+	unsigned long tmp;
+};
+
+static int conf_color;
+static int conf_per_session;
+static char *conf_per_user_dir;
+static char *conf_per_session_dir;
+static int conf_copy;
+
+static const char* level_name[]={"  msg", "error", " warn", " info", " info", "debug"};
+static const char* level_color[]={NORMAL_COLOR, RED_COLOR, YELLOW_COLOR, GREEN_COLOR, GREEN_COLOR, BLUE_COLOR};
+
+static void *pd_key1;
+static void *pd_key2;
+static struct log_file_t *log_file;
+
+static mempool_t lpd_pool;
+static char *log_buf;
+
+static struct aiocb aiocb = {
+	.aio_lio_opcode = LIO_WRITE,
+	.aio_sigevent.sigev_notify = SIGEV_SIGNAL,
+	.aio_sigevent.sigev_signo = SIGIO,
+};
+
+static LIST_HEAD(lf_queue);
+static spinlock_t lf_queue_lock = SPINLOCK_INITIALIZER;
+static int lf_queue_sleeping = 1;
+
+static unsigned long temp_seq;
+
+static void send_next_chunk();
+
+
+static void log_file_init(struct log_file_t *lf)
+{
+	spinlock_init(&lf->lock);
+	INIT_LIST_HEAD(&lf->msgs);
+	lf->fd = -1;
+	lf->new_fd = -1;
+}
+
+static int log_file_open(struct log_file_t *lf, const char *fname)
+{
+	lf->fd = open(fname, O_WRONLY | O_CREAT, S_IRUSR | S_IWUSR);
+	if (lf->fd < 0) {
+		log_emerg("log_file: open '%s': %s\n", fname, strerror(errno));
+		return -1;
+	} 
+	
+	lf->offset = lseek(lf->fd, 0, SEEK_END);
+	
+	return 0;
+}
+
+static void sigio(int num, siginfo_t *si, void *uc)
+{
+	struct log_file_t *lf;
+	int n;
+
+	if (si->si_signo != SIGIO)
+		return;
+
+	if (si->si_code != SI_ASYNCIO) {
+		if (aio_write(&aiocb))
+			log_emerg("log_file: aio_write: %s\n", strerror(errno));
+		return;
+	}
+
+	lf = (struct log_file_t *)si->si_ptr;
+
+	n = aio_return(&aiocb);
+	if (n < 0)
+		log_emerg("log_file: %s\n", strerror(aio_error(&aiocb)));
+	else if (n != aiocb.aio_nbytes)
+		log_emerg("log_file: short write %p %i %lu\n", lf, n, aiocb.aio_nbytes);
+
+	spin_lock(&lf->lock);
+	lf->offset += n;
+	if (list_empty(&lf->msgs)) {
+		if (lf->need_free) {
+			spin_unlock(&lf->lock);
+			close(lf->fd);
+			mempool_free(lf->lpd);
+		} else {
+			lf->queued = 0;
+			spin_unlock(&lf->lock);
+		}
+	} else {
+		spin_unlock(&lf->lock);
+
+		spin_lock(&lf_queue_lock);
+		list_add_tail(&lf->entry, &lf_queue);
+		spin_unlock(&lf_queue_lock);
+	}
+	
+	send_next_chunk();
+}
+
+static int dequeue_log(struct log_file_t *lf)
+{
+	int n, pos = 0;
+	struct log_msg_t *msg;
+	struct log_chunk_t *chunk;
+
+	while (1) {
+		spin_lock(&lf->lock);
+		if (list_empty(&lf->msgs)) {
+			spin_unlock(&lf->lock);
+			return pos;
+		}
+		msg = list_entry(lf->msgs.next, typeof(*msg), entry);
+		list_del(&msg->entry);
+		spin_unlock(&lf->lock);
+
+		if (pos + msg->hdr->len > LOG_BUF_SIZE)
+			goto overrun;
+		memcpy(log_buf + pos, msg->hdr->msg, msg->hdr->len);
+		n = msg->hdr->len;
+
+		list_for_each_entry(chunk, msg->chunks, entry) {
+			if (pos + n + chunk->len > LOG_BUF_SIZE)
+				goto overrun;
+			memcpy(log_buf + pos + n, chunk->msg, chunk->len);
+			n += chunk->len;
+		}
+
+		log_free_msg(msg);
+		pos += n;
+	}
+
+overrun:
+	spin_lock(&lf->lock);
+	list_add(&msg->entry, &lf->msgs);
+	spin_unlock(&lf->lock);
+
+	return pos;
+}
+
+static void send_next_chunk(void)
+{
+	struct log_file_t *lf;
+	int n;
+
+	spin_lock(&lf_queue_lock);
+	if (list_empty(&lf_queue)) {
+		lf_queue_sleeping = 1;
+		spin_unlock(&lf_queue_lock);
+		return;
+	}
+	lf = list_entry(lf_queue.next, typeof(*lf), entry);
+	
+	n = log_file->entry.next == NULL;
+	list_del(&lf->entry);
+
+	spin_unlock(&lf_queue_lock);
+
+	if (lf->new_fd != -1) {
+		close(lf->fd);
+		lf->fd = lf->new_fd;
+		lf->new_fd = -1;
+		lf->offset = 0;
+	}
+
+	aiocb.aio_fildes = lf->fd;
+	aiocb.aio_offset = lf->offset;
+	aiocb.aio_sigevent.sigev_value.sival_ptr = lf;
+	aiocb.aio_nbytes = dequeue_log(lf);
+
+	if (aio_write(&aiocb))
+		log_emerg("log_file: aio_write: %s\n", strerror(errno));
+}
+
+static void queue_lf(struct log_file_t *lf)
+{
+	int r;
+
+	spin_lock(&lf_queue_lock);
+	list_add_tail(&lf->entry, &lf_queue);
+	r = lf_queue_sleeping;
+	lf_queue_sleeping = 0;
+	spin_unlock(&lf_queue_lock);
+
+	if (r)
+		send_next_chunk();
+}
+
+static void queue_log(struct log_file_t *lf, struct log_msg_t *msg)
+{
+	int r;
+
+	spin_lock(&lf->lock);
+	list_add_tail(&msg->entry, &lf->msgs);
+	if (lf->fd != -1) {
+		r = lf->queued;
+		lf->queued = 1;
+	} else
+		r = 1;
+	spin_unlock(&lf->lock);
+
+	if (!r)
+		queue_lf(lf);
+}
+
+static void set_hdr(struct log_msg_t *msg, struct ppp_t *ppp)
+{
+	struct tm tm;
+	char timestamp[32];
+
+	localtime_r(&msg->timestamp.tv_sec, &tm);
+
+	strftime(timestamp, sizeof(timestamp), "%Y-%m-%d %H:%M:%S", &tm);
+	sprintf(msg->hdr->msg, "%s[%s]: %s: %s%s%s", conf_color ? level_color[msg->level] : "", 
+		timestamp, level_name[msg->level],
+		ppp ? ppp->ifname : "",
+		ppp ? ": " : "",
+		conf_color ? NORMAL_COLOR : "");
+	msg->hdr->len = strlen(msg->hdr->msg);
+}
+
+static void general_log(struct log_target_t *t, struct log_msg_t *msg, struct ppp_t *ppp)
+{
+	if (ppp && !conf_copy) {
+		log_free_msg(msg);
+		return;
+	}
+
+	set_hdr(msg, ppp);
+	queue_log(log_file, msg);
+}
+
+static struct log_file_pd_t *find_pd(struct ppp_t *ppp, void *pd_key)
+{
+	struct ppp_pd_t *pd;
+	struct log_file_pd_t *lpd;
+
+	list_for_each_entry(pd, &ppp->pd_list, entry) {
+		if (pd->key == pd_key) {
+			lpd = container_of(pd, typeof(*lpd), pd);
+			return lpd;
+		}
+	}
+	return NULL;
+}
+
+static void per_user_log(struct log_target_t *t, struct log_msg_t *msg, struct ppp_t *ppp)
+{
+	struct log_file_pd_t *lpd;
+
+	if (!ppp) {
+		log_free_msg(msg);
+		return;
+	}
+
+	lpd = find_pd(ppp, &pd_key1);
+
+	if (!lpd) {
+		log_free_msg(msg);
+		return;
+	}
+
+	set_hdr(msg, ppp);
+	queue_log(&lpd->lf, msg);
+}
+
+static void per_session_log(struct log_target_t *t, struct log_msg_t *msg, struct ppp_t *ppp)
+{
+	struct log_file_pd_t *lpd;
+	
+	if (!ppp) {
+		log_free_msg(msg);
+		return;
+	}
+
+	lpd = find_pd(ppp, &pd_key2);
+
+	if (!lpd) {
+		log_free_msg(msg);
+		return;
+	}
+
+	set_hdr(msg, ppp);
+	queue_log(&lpd->lf, msg);
+}
+
+static void general_reopen(void)
+{
+	char *fname = conf_get_opt("log", "log-file");
+ 	int fd = open(fname, O_WRONLY | O_CREAT, S_IRUSR | S_IWUSR);
+	if (fd < 0) {
+		log_emerg("log_file: open '%s': %s\n", fname, strerror(errno));
+		return;
+	}
+	log_file->new_fd = fd;
+}
+
+static void free_lpd(struct log_file_pd_t *lpd)
+{
+	struct log_msg_t *msg;
+
+	spin_lock(&lpd->lf.lock);
+	list_del(&lpd->pd.entry);
+	lpd->lf.need_free = 1;
+	if (lpd->lf.queued)
+		spin_unlock(&lpd->lf.lock);
+	else {
+		while (!list_empty(&lpd->lf.msgs)) {
+			msg = list_entry(lpd->lf.msgs.next, typeof(*msg), entry);
+			list_del(&msg->entry);
+			log_free_msg(msg);
+		}
+		if (lpd->lf.fd != -1)
+			close(lpd->lf.fd);
+		spin_unlock(&lpd->lf.lock);
+		mempool_free(lpd);
+	}
+}
+
+static void ev_ctrl_started(struct ppp_t *ppp)
+{
+	struct log_file_pd_t *lpd;
+	char *fname;
+
+	if (conf_per_user_dir) {
+		lpd = mempool_alloc(lpd_pool);
+		if (!lpd) {
+			log_emerg("log_file: out of memory\n");
+			return;
+		}
+		memset(lpd, 0, sizeof(*lpd));
+		lpd->pd.key = &pd_key1;
+		log_file_init(&lpd->lf);
+		lpd->lf.lpd = lpd;
+		list_add_tail(&lpd->pd.entry, &ppp->pd_list);
+	}
+
+	if (conf_per_session_dir) {
+		lpd = mempool_alloc(lpd_pool);
+		if (!lpd) {
+			log_emerg("log_file: out of memory\n");
+			return;
+		}
+		memset(lpd, 0, sizeof(*lpd));
+		lpd->pd.key = &pd_key2;
+		log_file_init(&lpd->lf);
+		lpd->lf.lpd = lpd;
+
+		fname = _malloc(PATH_MAX);
+		if (!fname) {
+			mempool_free(lpd);
+			log_emerg("log_file: out of memory\n");
+			return;
+		}
+
+		lpd->tmp = temp_seq++;
+		strcpy(fname, conf_per_session_dir);
+		strcat(fname, "/tmp");
+		sprintf(fname + strlen(fname), "%lu", lpd->tmp);
+
+		if (log_file_open(&lpd->lf, fname)) {
+			mempool_free(lpd);
+			_free(fname);
+			return;
+		}
+
+		_free(fname);
+
+		list_add_tail(&lpd->pd.entry, &ppp->pd_list);
+	}
+}
+
+static void ev_ctrl_finished(struct ppp_t *ppp)
+{
+	struct log_file_pd_t *lpd;
+	char *fname;
+
+	lpd = find_pd(ppp, &pd_key1);
+	if (lpd)
+		free_lpd(lpd);
+
+
+	lpd = find_pd(ppp, &pd_key2);
+	if (lpd) {
+		if (lpd->tmp) {
+			fname = _malloc(PATH_MAX);
+			if (fname) {
+				strcpy(fname, conf_per_session_dir);
+				strcat(fname, "/tmp");
+				sprintf(fname + strlen(fname), "%lu", lpd->tmp);
+				if (unlink(fname))
+					log_emerg("log_file: unlink '%s': %s\n", fname, strerror(errno));
+				_free(fname);
+			} else
+				log_emerg("log_file: out of memory\n");
+		}
+		free_lpd(lpd);
+	}
+}
+
+static void ev_ppp_starting(struct ppp_t *ppp)
+{
+	struct log_file_pd_t *lpd;
+	char *fname1, *fname2;
+
+	lpd = find_pd(ppp, &pd_key2);
+	if (!lpd)
+		return;
+	
+	fname1 = _malloc(PATH_MAX);
+	if (!fname1) {
+		log_emerg("log_file: out of memory\n");
+		return;
+	}
+
+	fname2 = _malloc(PATH_MAX);
+	if (!fname2) {
+		log_emerg("log_file: out of memory\n");
+		_free(fname1);
+		return;
+	}
+
+	strcpy(fname1, conf_per_session_dir);
+	strcat(fname1, "/tmp");
+	sprintf(fname1 + strlen(fname1), "%lu", lpd->tmp);
+
+	strcpy(fname2, conf_per_session_dir);
+	strcat(fname2, "/");
+	strcat(fname2, ppp->sessionid);
+	strcat(fname2, ".log");
+
+	if (rename(fname1, fname2))
+		log_emerg("log_file: rename '%s' to '%s': %s\n", fname1, fname2, strerror(errno));
+
+	lpd->tmp = 0;
+
+	_free(fname1);
+	_free(fname2);
+}
+
+static void ev_ppp_authorized(struct ppp_t *ppp)
+{
+	struct log_file_pd_t *lpd;
+	char *fname;
+
+	lpd = find_pd(ppp, &pd_key1);
+	if (!lpd)
+		return;
+
+	fname = _malloc(PATH_MAX);
+	if (!fname) {
+		log_emerg("log_file: out of memory\n");
+		return;
+	}
+
+	strcpy(fname, conf_per_user_dir);
+	strcat(fname, "/");
+	strcat(fname, ppp->username);
+	if (conf_per_session) {
+		if (mkdir(fname, S_IRWXU) && errno != EEXIST) {
+			log_emerg("log_file: mkdir '%s': %s'\n", fname, strerror(errno));
+			goto out_err;
+		}
+		strcat(fname, "/");
+		strcat(fname, ppp->sessionid);
+	}
+	strcat(fname, ".log");
+
+	if (log_file_open(&lpd->lf, fname))
+		goto out_err;
+
+	_free(fname);
+
+	if (!list_empty(&lpd->lf.msgs)) {
+		lpd->lf.queued = 1;
+		queue_lf(&lpd->lf);
+	}
+
+	return;
+
+out_err:
+	_free(fname);
+	list_del(&lpd->pd.entry);
+	free_lpd(lpd);
+}
+
+static struct log_target_t general_target = 
+{
+	.log = general_log,
+	.reopen = general_reopen,
+};
+
+static struct log_target_t per_user_target = 
+{
+	.log = per_user_log,
+};
+
+static struct log_target_t per_session_target = 
+{
+	.log = per_session_log,
+};
+
+static void __init init(void)
+{
+	char *opt;
+	
+	sigset_t set;
+	sigemptyset(&set);
+	sigaddset(&set, SIGIO);
+
+	struct sigaction sa = {
+		.sa_sigaction = sigio,
+		.sa_flags = SA_SIGINFO,
+		.sa_mask = set,
+	};
+
+	lpd_pool = mempool_create(sizeof(struct log_file_pd_t));
+	log_buf = malloc(LOG_BUF_SIZE);
+	aiocb.aio_buf = log_buf;
+
+	if (sigaction(SIGIO, &sa, NULL)) {
+		log_emerg("log_file: sigaction: %s\n", strerror(errno));
+		return;
+	}
+
+	opt = conf_get_opt("log", "log-file");
+	if (opt) {
+		log_file = malloc(sizeof(*log_file));
+		memset(log_file, 0, sizeof(*log_file));
+		log_file_init(log_file);
+		if (log_file_open(log_file, opt)) {
+			free(log_file);
+			_exit(EXIT_FAILURE);
+		}
+	}
+	
+	opt = conf_get_opt("log","color");
+	if (opt && atoi(opt) > 0)
+		conf_color = 1;
+	
+	opt = conf_get_opt("log", "per-user-dir");
+	if (opt)
+		conf_per_user_dir = _strdup(opt);
+
+	opt = conf_get_opt("log", "per-session-dir");
+	if (opt)
+		conf_per_session_dir = _strdup(opt);
+
+	opt = conf_get_opt("log", "per-session");
+	if (opt && atoi(opt) > 0)
+		conf_per_session = 1;
+	
+	opt = conf_get_opt("log", "copy");
+	if (opt && atoi(opt) > 0)
+		conf_copy = 1;
+
+	log_register_target(&general_target);
+	
+	if (conf_per_user_dir)
+		log_register_target(&per_user_target);
+	
+	if (conf_per_session_dir)
+		log_register_target(&per_session_target);
+
+	triton_event_register_handler(EV_CTRL_STARTED, (triton_event_func)ev_ctrl_started);
+	triton_event_register_handler(EV_CTRL_FINISHED, (triton_event_func)ev_ctrl_finished);
+	triton_event_register_handler(EV_PPP_STARTING, (triton_event_func)ev_ppp_starting);
+	triton_event_register_handler(EV_PPP_AUTHORIZED, (triton_event_func)ev_ppp_authorized);
+}
diff --git a/accel-pppd/logs/log_pgsql.c b/accel-pppd/logs/log_pgsql.c
new file mode 100644
index 0000000..af67e0b
--- /dev/null
+++ b/accel-pppd/logs/log_pgsql.c
@@ -0,0 +1,321 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include <postgresql/libpq-fe.h>
+
+#include "triton.h"
+#include "spinlock.h"
+#include "log.h"
+#include "list.h"
+#include "ppp.h"
+
+#include "memdebug.h"
+
+static char *conf_conninfo;
+static int conf_queue_max = 1000;
+static char *conf_query;
+#define QUERY_TEMPLATE "insert into %s (timestamp, username, sessionid, msg) values ($1, $2, $3, $4)"
+
+static void start_connect(void);
+static void start_connect_timer(struct triton_timer_t *);
+static void pgsql_close(struct triton_context_t *ctx);
+
+static struct triton_context_t pgsql_ctx = {
+	.close = pgsql_close,
+	.before_switch = log_switch,
+};
+static struct triton_md_handler_t pgsql_hnd;
+static struct triton_timer_t connect_timer = {
+	.period = 5000,
+	.expire = start_connect_timer,
+};
+
+static PGconn *conn;
+
+static LIST_HEAD(msg_queue);
+static int queue_size;
+static int sleeping = 0;
+static spinlock_t queue_lock = SPINLOCK_INITIALIZER;
+static char *log_buf;
+static int need_close;
+
+static void unpack_msg(struct log_msg_t *msg)
+{
+	struct log_chunk_t *chunk;
+	int pos = 0;
+
+	list_for_each_entry(chunk, msg->chunks, entry) {
+		memcpy(log_buf + pos, chunk->msg, chunk->len);
+		pos += chunk->len;
+	}
+	if (pos > 1)
+		log_buf[pos - 1] = 0;
+	else
+		log_buf[0] = 0;
+}
+
+static void set_hdr(struct log_msg_t *msg, struct ppp_t *ppp)
+{
+	struct tm tm;
+
+	localtime_r(&msg->timestamp.tv_sec, &tm);
+
+	strftime(msg->hdr->msg, LOG_CHUNK_SIZE, "%Y-%m-%d %H:%M:%S", &tm);
+	msg->hdr->len = strlen(msg->hdr->msg) + 1;
+	if (ppp && ppp->username) {
+		strcpy(msg->hdr->msg + msg->hdr->len, ppp->username);
+		msg->hdr->len += strlen(ppp->username) + 1;
+		strcpy(msg->hdr->msg + msg->hdr->len, ppp->sessionid);
+		msg->hdr->len += strlen(ppp->sessionid) + 1;
+	} else 
+		memset(msg->hdr->msg + msg->hdr->len, 0, 2);
+		
+}
+
+static void write_next_msg(void)
+{
+	struct log_msg_t *msg;
+	const char *paramValues[4];
+	int paramFormats[4] = {0, 0, 0, 0};
+	char *ptr1, *ptr2;
+	int r;
+
+	spin_lock(&queue_lock);
+	if (list_empty(&msg_queue)) {
+		sleeping = 1;
+		spin_unlock(&queue_lock);
+		if (need_close) {
+			triton_md_unregister_handler(&pgsql_hnd);
+			PQfinish(conn);
+			conn = NULL;
+			triton_context_unregister(&pgsql_ctx);
+		}
+		return;
+	}
+
+	msg = list_entry(msg_queue.next, typeof(*msg), entry);
+	list_del(&msg->entry);
+	--queue_size;
+	spin_unlock(&queue_lock);
+
+	unpack_msg(msg);
+
+	ptr1 = strchr(msg->hdr->msg, 0);
+	ptr2 = strchr(ptr1 + 1, 0);
+
+	paramValues[1] = ptr1[1] ? ptr1 + 1 : NULL;
+	paramValues[2] = ptr2[1] ? ptr2 + 1 : NULL;
+	paramValues[0] = msg->hdr->msg;
+	paramValues[3] = log_buf;
+
+	if (!PQsendQueryParams(conn, conf_query, 4, NULL, paramValues, NULL, paramFormats, 0))
+		log_emerg("log_pgsql: %s\n", PQerrorMessage(conn));
+
+	log_free_msg(msg);
+
+	r = PQflush(conn);
+	if (r == -1)
+		log_emerg("log_pgsql: %s\n", PQerrorMessage(conn));
+	if (r == 0)
+		triton_md_enable_handler(&pgsql_hnd, MD_MODE_WRITE);
+}
+
+static int pgsql_check_ready(struct triton_md_handler_t *h)
+{
+	PGresult *res;
+
+	if (!PQconsumeInput(conn)) {
+		log_emerg("log_pgsql: %s\n", PQerrorMessage(conn));
+		if (PQstatus(conn) == CONNECTION_BAD) {
+			PQfinish(conn);
+			start_connect();
+		}
+	}
+
+	if (PQisBusy(conn))
+		return 0;
+
+	while (1) {
+		res = PQgetResult(conn);
+		if (!res)
+			break;
+		if (PQresultStatus(res) != PGRES_COMMAND_OK)
+			log_emerg("log_pgsql: %s\n", PQerrorMessage(conn));
+		PQclear(res);
+	}
+
+	write_next_msg();
+
+	return 0;
+}
+
+static int pgsql_flush(struct triton_md_handler_t *h)
+{
+	int r;
+
+	r = PQflush(conn);
+	if (r == -1)
+		log_emerg("log_pgsql: %s\n", PQerrorMessage(conn));
+	if (r == 1)
+		return 0;
+	
+	triton_md_disable_handler(&pgsql_hnd, MD_MODE_WRITE);
+	return 0;
+}
+
+static void wakeup_log(void)
+{
+	write_next_msg();
+}
+
+static void queue_log(struct log_msg_t *msg)
+{
+	int r = 0, f = 0;
+	spin_lock(&queue_lock);
+	if (!conn) {
+		log_free_msg(msg);
+		spin_unlock(&queue_lock);
+		return;
+	}
+	if (queue_size < conf_queue_max) {
+		list_add_tail(&msg->entry, &msg_queue);
+		++queue_size;
+		r = sleeping;
+		sleeping = 0;
+	} else
+		f = 1;
+	spin_unlock(&queue_lock);
+
+	if (r)
+		triton_context_call(&pgsql_ctx, (void (*)(void*))wakeup_log, NULL);
+	else if (f)
+		log_free_msg(msg);
+}
+
+
+static void general_log(struct log_target_t *t, struct log_msg_t *msg, struct ppp_t *ppp)
+{
+	set_hdr(msg, ppp);
+	queue_log(msg);
+}
+
+static int wait_connect(struct triton_md_handler_t *h)
+{
+	PostgresPollingStatusType status = PQconnectPoll(conn);
+	char *err_msg;
+
+	switch(status) {
+		case PGRES_POLLING_READING:
+			triton_md_enable_handler(h, MD_MODE_READ);
+			triton_md_disable_handler(h, MD_MODE_WRITE);
+			break;
+		case PGRES_POLLING_WRITING:
+			triton_md_enable_handler(h, MD_MODE_WRITE);
+			triton_md_disable_handler(h, MD_MODE_READ);
+			break;
+		case PGRES_POLLING_FAILED:
+			err_msg = PQerrorMessage(conn);
+			log_emerg("log_pgsql: %s\n", err_msg);
+			triton_md_disable_handler(h, MD_MODE_READ | MD_MODE_WRITE);
+			PQfinish(conn);
+			h->read = NULL;
+			h->write = NULL;
+			if (!connect_timer.tpd)
+				triton_timer_add(&pgsql_ctx, &connect_timer, 0);
+			break;
+		case PGRES_POLLING_OK:
+			//triton_md_disable_handler(h, MD_MODE_READ | MD_MODE_WRITE);
+			PQsetnonblocking(conn, 1);
+			h->write = pgsql_flush;
+			h->read = pgsql_check_ready;
+			triton_md_enable_handler(&pgsql_hnd, MD_MODE_READ);
+			wakeup_log();
+			break;
+		default:
+			break;
+	}
+	return 0;
+}
+
+static void start_connect(void)
+{
+	conn = PQconnectStart(conf_conninfo);
+	if (!conn) {
+		log_emerg("log_pgsql: out of memory\n");
+		return;
+	}
+
+	if (PQstatus(conn) == CONNECTION_BAD) {
+		log_emerg("log_pgsql: PQconnectStart failed\n");
+	}
+
+	pgsql_hnd.fd = PQsocket(conn);
+	pgsql_hnd.read = wait_connect;
+	pgsql_hnd.write = wait_connect;
+
+	wait_connect(&pgsql_hnd);
+}
+
+static void start_connect_timer(struct triton_timer_t *t)
+{
+	triton_timer_del(t);
+	start_connect();
+}
+
+static void pgsql_close(struct triton_context_t *ctx)
+{
+	spin_lock(&queue_lock);
+	if (sleeping) {
+		triton_md_unregister_handler(&pgsql_hnd);
+		PQfinish(conn);
+		conn = NULL;
+		triton_context_unregister(&pgsql_ctx);
+	} else
+		need_close = 1;	
+	spin_unlock(&queue_lock);
+}
+
+static struct log_target_t target = {
+	.log = general_log,
+};
+
+static void __init init(void)
+{
+	char *opt;
+
+	opt = conf_get_opt("log-pgsql", "conninfo");
+	if (!opt)
+		return;
+	conf_conninfo = _strdup(opt);
+
+	opt = conf_get_opt("log-pgsql", "connect-inteval");
+	if (opt && atoi(opt) > 0)
+		connect_timer.period = atoi(opt) * 1000;
+	
+	opt = conf_get_opt("log-pgsql", "log-query");
+	if (opt)
+		conf_query = _strdup(opt);
+	else {
+		opt = conf_get_opt("log-pgsql", "log-table");
+		if (!opt || strlen(opt) > 32)
+			opt = "log";
+		conf_query = _malloc(sizeof(QUERY_TEMPLATE) + strlen(opt));
+		sprintf(conf_query, QUERY_TEMPLATE, opt);
+	}
+
+	log_buf = _malloc(LOG_MAX_SIZE + 1);
+	if (!log_buf) {
+		log_emerg("log_pgsql: out of memory\n");
+		return;
+	}
+
+	triton_context_register(&pgsql_ctx, NULL);
+	triton_md_register_handler(&pgsql_ctx, &pgsql_hnd);
+	triton_md_set_trig(&pgsql_hnd, MD_TRIG_LEVEL);
+	triton_context_wakeup(&pgsql_ctx);
+
+	start_connect();
+
+	log_register_target(&target);
+}
diff --git a/accel-pppd/logs/log_tcp.c b/accel-pppd/logs/log_tcp.c
new file mode 100644
index 0000000..306c450
--- /dev/null
+++ b/accel-pppd/logs/log_tcp.c
@@ -0,0 +1,323 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <errno.h>
+#include <sys/socket.h>
+#include <arpa/inet.h>
+#include <netinet/in.h>
+
+#include "log.h"
+#include "triton.h"
+#include "events.h"
+#include "ppp.h"
+#include "spinlock.h"
+#include "mempool.h"
+
+#include "memdebug.h"
+
+struct tcp_target_t
+{
+	struct log_target_t target;
+	struct list_head entry;
+	struct triton_md_handler_t hnd;
+	struct triton_timer_t conn_timer;
+  struct sockaddr_in addr;
+	char *buf;
+	int buf_size;
+	int buf_pos;
+	spinlock_t lock;
+	struct list_head queue;
+	int queue_len;
+	int connected:1;
+	int wait:1;
+};
+
+static int conf_connect_interval = 5;
+static int conf_queue_len = 1000;
+
+static struct triton_context_t tcp_ctx;
+
+static const char* level_name[]={"  msg", "error", " warn", " info", " info", "debug"};
+
+static void start_connect(struct tcp_target_t *t);
+
+static LIST_HEAD(targets);
+
+static void disconnect(struct tcp_target_t *t)
+{
+	triton_md_unregister_handler(&t->hnd);
+	close(t->hnd.fd);
+
+	start_connect(t);
+}
+
+static void unpack_msg(struct tcp_target_t *t, struct log_msg_t *msg)
+{
+	struct log_chunk_t *chunk;
+	int pos = strlen(msg->hdr->msg);
+
+	strcpy(t->buf, msg->hdr->msg);
+
+	list_for_each_entry(chunk, msg->chunks, entry) {
+		memcpy(t->buf + pos, chunk->msg, chunk->len);
+		pos += chunk->len;
+	}
+
+	t->buf_size = pos;
+	t->buf_pos = 0;
+}
+
+static int send_log(struct tcp_target_t *t)
+{
+	struct log_msg_t *msg;
+	int n;
+
+	while (1) {
+		spin_lock(&t->lock);
+		if (!t->queue_len) {
+			t->wait = 0;
+			spin_unlock(&t->lock);
+			return 0;
+		}
+		msg = list_entry(t->queue.next, typeof(*msg), entry);
+		list_del(&msg->entry);
+		t->queue_len--;
+		spin_unlock(&t->lock);
+
+		unpack_msg(t, msg);
+
+		log_free_msg(msg);
+
+		while (t->buf_pos != t->buf_size) {
+			n = write(t->hnd.fd, t->buf + t->buf_pos, t->buf_size - t->buf_pos);
+			if (n < 0) {
+				if (errno == EAGAIN)
+					return 1;
+				if (errno != EPIPE)
+					log_emerg("log-tcp: write: %s\n", strerror(errno));
+				disconnect(t);
+				return 0;
+			}
+			t->buf_pos += n;
+		}
+	}
+}
+
+static void queue_log(struct tcp_target_t *t, struct log_msg_t *msg)
+{
+	int r;
+
+	spin_lock(&t->lock);
+	if (t->queue_len == conf_queue_len) {
+		spin_unlock(&t->lock);
+		log_free_msg(msg);
+		return;
+	}
+	list_add_tail(&msg->entry, &t->queue);
+	t->queue_len++;
+	if (t->connected) {
+		r = t->wait;
+		t->wait = 1;
+	} else
+		r = 1;
+	spin_unlock(&t->lock);
+
+	if (!r) {
+		if (send_log(t))
+			triton_md_enable_handler(&t->hnd, MD_MODE_WRITE);
+	}
+}
+
+static void set_hdr(struct log_msg_t *msg, struct ppp_t *ppp)
+{
+	struct tm tm;
+	char timestamp[32];
+
+	localtime_r(&msg->timestamp.tv_sec, &tm);
+
+	strftime(timestamp, sizeof(timestamp), "%Y-%m-%d %H:%M:%S", &tm);
+	sprintf(msg->hdr->msg, "[%s]: %s: %s: ", timestamp, level_name[msg->level],	ppp ? ppp->ifname : "");
+	msg->hdr->len = strlen(msg->hdr->msg);
+}
+
+static void general_log(struct log_target_t *lt, struct log_msg_t *msg, struct ppp_t *ppp)
+{
+	struct tcp_target_t *t = container_of(lt, typeof(*t), target);
+
+	set_hdr(msg, ppp);
+	queue_log(t, msg);
+}
+
+static int log_tcp_write(struct triton_md_handler_t *h)
+{
+	struct tcp_target_t *t = container_of(h, typeof(*t), hnd);
+
+	if (!send_log(t))
+		triton_md_disable_handler(h, MD_MODE_WRITE);
+
+	return 0;
+}
+
+static int log_tcp_connect(struct triton_md_handler_t *h)
+{
+	struct tcp_target_t *t = container_of(h, typeof(*t), hnd);
+
+	if (connect(t->hnd.fd, &t->addr, sizeof(t->addr))) {
+		if (errno == EAGAIN)
+			return 0;
+		if (errno == EINPROGRESS)
+			return 0;
+		log_emerg("log-tcp: connect: %s\n", strerror(errno));
+		triton_md_unregister_handler(&t->hnd);
+		close(t->hnd.fd);
+		triton_timer_add(&tcp_ctx, &t->conn_timer, 0);
+		return 0;
+	}
+
+	t->hnd.write = log_tcp_write;
+
+	triton_md_disable_handler(&t->hnd, MD_MODE_WRITE);
+
+	spin_lock(&t->lock);
+	t->connected = 1;
+	t->wait = 1;
+	spin_unlock(&t->lock);
+
+	if (send_log(t))
+		triton_md_enable_handler(&t->hnd, MD_MODE_WRITE);
+
+	return 0;
+}
+
+static void connect_timer(struct triton_timer_t *timer)
+{
+	struct tcp_target_t *t = container_of(timer, typeof(*t), conn_timer);
+
+	triton_timer_del(timer);
+
+	start_connect(t);
+}
+
+static void start_connect(struct tcp_target_t *t)
+{
+	t->hnd.write = log_tcp_connect;
+	t->hnd.fd = socket(PF_INET, SOCK_STREAM, 0);
+
+	if (!t->hnd.fd) {
+		log_emerg("log-tcp: socket: %s\n", strerror(errno));
+		return;
+	}
+	
+	if (fcntl(t->hnd.fd, F_SETFL, O_NONBLOCK)) {
+    log_emerg("log-tcp: failed to set nonblocking mode: %s\n", strerror(errno));
+		close(t->hnd.fd);
+    return;
+	}
+
+	if (connect(t->hnd.fd, &t->addr, sizeof(t->addr))) {
+		if (errno != EINPROGRESS) {
+			log_emerg("log-tcp: connect: %s\n", strerror(errno));
+			close(t->hnd.fd);
+			return;
+		}
+	}
+
+	triton_md_register_handler(&tcp_ctx, &t->hnd);
+	triton_md_enable_handler(&t->hnd, MD_MODE_WRITE);
+}
+
+static void log_tcp_close(struct triton_context_t *ctx)
+{
+	struct tcp_target_t *t;
+
+	while (!list_empty(&targets)) {
+		t = list_entry(targets.next, typeof(*t), entry);
+		list_del(&t->entry);
+		if (t->conn_timer.tpd)
+			triton_timer_del(&t->conn_timer);
+		else {
+			t->connected = 0;
+			triton_md_unregister_handler(&t->hnd);
+			close(t->hnd.fd);
+		}
+	}
+
+	triton_context_unregister(&tcp_ctx);
+}
+
+static int start_log(const char *_opt)
+{
+	struct tcp_target_t *t;
+	char *opt = strdup(_opt);
+	int port;
+	char *d;
+
+	d = strchr(opt, ':');
+	if (!d)
+		goto err;
+
+	*d = 0;
+
+	port = atoi(d + 1);
+	if (port <= 0)
+		goto err;
+
+	t = _malloc(sizeof(*t));
+	memset(t, 0, sizeof(*t));
+
+	t->buf = _malloc(LOG_MAX_SIZE + 64);
+
+	t->conn_timer.expire_tv.tv_sec = conf_connect_interval;
+	t->conn_timer.expire = connect_timer;
+
+	t->target.log = general_log;
+
+	memset(&t->addr, 0, sizeof(t->addr));
+  t->addr.sin_family = AF_INET;
+  t->addr.sin_port = htons(port);
+	t->addr.sin_addr.s_addr = inet_addr(opt);
+
+	INIT_LIST_HEAD(&t->queue);
+
+	spinlock_init(&t->lock);
+
+	start_connect(t);
+
+	log_register_target(&t->target);
+
+	list_add_tail(&t->entry, &targets);
+
+	return 0;
+
+err:
+	free(opt);
+	return -1;
+}
+
+static struct triton_context_t tcp_ctx ={
+	.close = log_tcp_close,
+	.before_switch = log_switch,
+};
+
+static void __init init(void)
+{
+	struct conf_sect_t *s =	conf_get_section("log");
+	struct conf_option_t *opt;
+	
+	if (!s)
+		return;
+
+	triton_context_register(&tcp_ctx, NULL);
+
+	list_for_each_entry(opt, &s->items, entry) {
+		if (strcmp(opt->name, "log-tcp"))
+			continue;
+		if (!opt->val || start_log(opt->val))
+			log_emerg("log: log-tcp: invalid format: '%s'\n", opt->val);
+	}
+
+	triton_context_wakeup(&tcp_ctx);
+}
+