No OneTemporary
Actions

Size

70 KB

Referenced Files

None

Subscribers

None

View Options

	diff --git a/src/inline-fn.h b/src/inline-fn.h
	index db1c938..539cd18 100644
	--- a/src/inline-fn.h
	+++ b/src/inline-fn.h
	@@ -1,294 +1,314 @@
	/*
	* Copyright (C) 2013-2014 Philipp Marek <philipp.marek@linbit.com>
	*
	* This program is free software; you can redistribute it and/or
	* modify it under the terms of the GNU General Public
	* License as published by the Free Software Foundation; either
	* version 2 of the License, or (at your option) any later version.
	*
	* This software is distributed in the hope that it will be useful,
	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	* General Public License for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with this program; if not, write to the Free Software Foundation, Inc.,
	* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
	*/

	#ifndef _INLINE_FN_H
	#define _INLINE_FN_H

	#include <time.h>
	#include <sys/time.h>
	#include <assert.h>
	#include <string.h>
	#include "timer.h"
	#include "config.h"
	#include "transport.h"



	inline static int get_local_id(void)
	{
	return local ? local->site_id : -1;
	}


	inline static uint32_t get_node_id(struct booth_site *node)
	{
	return node ? node->site_id : 0;
	}


	/** Returns number of seconds left, if any. */
	inline static int term_time_left(struct ticket_config *tk)
	{
	int left = 0;

	if (is_time_set(&tk->term_expires)) {
	left = time_left(&tk->term_expires);
	}
	return (left < 0) ? 0 : left;
	}


	inline static int leader_and_valid(struct ticket_config *tk)
	{
	if (tk->leader != local)
	return 0;

	return term_time_left(tk);
	}


	/** Is this some leader? */
	inline static int is_owned(const struct ticket_config *tk)
	{
	return (tk->leader && tk->leader != no_leader);
	}

	inline static int is_resend(struct ticket_config *tk)
	{
	timetype now;

	get_time(&now);
	return time_sub_int(&now, &tk->req_sent_at) >= tk->timeout;
	}


	static inline void init_header_bare(struct boothc_header *h) {
	timetype now;

	assert(local && local->site_id);
	h->magic = htonl(BOOTHC_MAGIC);
	h->version = htonl(BOOTHC_VERSION);
	h->from = htonl(local->site_id);
	if (is_auth_req()) {
	get_time(&now);
	h->opts = htonl(BOOTH_OPT_AUTH);
	h->secs = htonl(secs_since_epoch(&now));
	h->usecs = htonl(get_usecs(&now));
	} else {
	h->opts = htonl(0);
	h->secs = htonl(0);
	h->usecs = htonl(0);
	}
	}

	/* get the _real_ message length out of the header
	*/
	#define sendmsglen(msg) ntohl((msg)->header.length)

	static inline void init_header(struct boothc_header *h,
	int cmd, int request, int options,
	int result, int reason, int data_len)
	{
	init_header_bare(h);
	h->length = htonl(data_len -
	(is_auth_req() ? 0 : sizeof(struct hmac)));
	h->cmd = htonl(cmd);
	h->request = htonl(request);
	h->options = htonl(options);
	h->result = htonl(result);
	h->reason = htonl(reason);
	}

	#define my_last_term(tk) \
	(((tk)->state == ST_CANDIDATE && (tk)->last_valid_tk) ? \
	(tk)->last_valid_tk->current_term : (tk)->current_term)

	extern int TIME_RES, TIME_MULT;

	#define msg_term_time(msg) \
	ntohl((msg)->ticket.term_valid_for)*TIME_RES/TIME_MULT
	#define set_msg_term_time(msg, tk) \
	(msg)->ticket.term_valid_for = htonl(term_time_left(tk)*TIME_MULT/TIME_RES)

	static inline void init_ticket_msg(struct boothc_ticket_msg *msg,
	int cmd, int request, int rv, int reason,
	struct ticket_config *tk)
	{
	assert(sizeof(msg->ticket.id) == sizeof(tk->name));

	init_header(&msg->header, cmd, request, 0, rv, reason, sizeof(*msg));

	if (!tk) {
	memset(&msg->ticket, 0, sizeof(msg->ticket));
	} else {
	memcpy(msg->ticket.id, tk->name, sizeof(msg->ticket.id));

	msg->ticket.leader = htonl(get_node_id(
	(tk->leader && tk->leader != no_leader) ? tk->leader :
	(tk->voted_for ? tk->voted_for : no_leader)));
	msg->ticket.term = htonl(tk->current_term);
	set_msg_term_time(msg, tk);
	}
	}


	static inline struct booth_transport const *transport(void)
	{
	return booth_transport + booth_conf->proto;
	}


	-static inline const char site_string(struct booth_site site)
	+static inline const char site_string(const struct booth_site site)
	{
	return site ? site->addr_string : "NONE";
	}

	+/**
	+ * @internal
	+ * Parse booth configuration file and store as structured data
	+ *
	+ * @param[in] site subject of TCP/UDP port extraction
	+ *
	+ * @return 0 for "undefined", actual port number otherwise
	+ */
	+static inline uint16_t site_port(const struct booth_site *site)
	+{
	+ assert(site != NULL);
	+
	+ return site
	+ ? site->family == AF_INET
	+ ? ntohs(site->sa4.sin_port)
	+ : site->family == AF_INET6
	+ ? ntohs(site->sa6.sin6_port)
	+ : 0
	+ : 0;
	+}

	static inline const char ticket_leader_string(struct ticket_config tk)
	{
	return site_string(tk->leader);
	}


	/* We allow half of the uint32_t to be used;
	* half of that below, half of that above the current known "good" value.
	* 0 UINT32_MAX
	* \|--------------------------+----------------+------------\|
	* \| \| \|
	* \|--------+-------\| allowed range
	* \|
	* current commit index
	*
	* So, on overflow it looks like that:
	* UINT32_MAX 0
	* \|--------------------------+-----------\|\|---+------------\|
	* \| \| \|
	* \|--------+-------\| allowed range
	* \|
	* current commit index
	*
	* This should be possible by using the same datatype and relying
	* on the under/overflow semantics.
	*
	*
	* Having 30 bits available, and assuming an expire time of
	* one minute and a (high) commit index step of 64 == 2^6 (because
	* of weights), we get 2^24 minutes of range - which is ~750
	* years. "Should be enough for everybody."
	*/
	static inline int index_is_higher_than(uint32_t c_high, uint32_t c_low)
	{
	uint32_t diff;

	if (c_high == c_low)
	return 0;

	diff = c_high - c_low;
	if (diff < UINT32_MAX/4)
	return 1;

	diff = c_low - c_high;
	if (diff < UINT32_MAX/4)
	return 0;

	assert(!"commit index out of range - invalid");
	}


	static inline uint32_t index_max2(uint32_t a, uint32_t b)
	{
	return index_is_higher_than(a, b) ? a : b;
	}

	static inline uint32_t index_max3(uint32_t a, uint32_t b, uint32_t c)
	{
	return index_max2( index_max2(a, b), c);
	}


	/* only invoked when ticket leader */
	static inline void get_next_election_time(struct ticket_config tk, timetype next)
	{
	assert(tk->leader == local);

	/* if last_renewal is not set, which is unusual, it may mean
	* that the ticket never got updated, i.e. nobody acked
	* ticket updates (say, due to a temporary connection
	* problem)
	* we may try a bit later again */
	if (!is_time_set(&tk->last_renewal)) {
	time_reset(next);
	} else {
	interval_add(&tk->last_renewal, tk->renewal_freq, next);
	}

	/* if delay_commit is earlier than next, then set next to
	* delay_commit */
	if (is_time_set(&tk->delay_commit) &&
	time_cmp(next, &tk->delay_commit, >)) {
	copy_time(&tk->delay_commit, next);
	}
	}


	static inline void expect_replies(struct ticket_config *tk,
	int reply_type)
	{
	tk->retry_number = 0;
	tk->acks_expected = reply_type;
	tk->acks_received = local->bitmask;
	get_time(&tk->req_sent_at);
	}

	static inline void no_resends(struct ticket_config *tk)
	{
	tk->retry_number = 0;
	tk->acks_expected = 0;
	}

	static inline struct booth_site my_vote(struct ticket_config tk)
	{
	return tk->votes_for[ local->index ];
	}


	static inline int count_bits(uint64_t val) {
	return __builtin_popcount(val);
	}

	static inline int majority_of_bits(struct ticket_config *tk, uint64_t val)
	{
	/* Use ">" to get majority decision, even for an even number
	* of participants. */
	return count_bits(val) * 2 >
	booth_conf->site_count;
	}


	static inline int all_replied(struct ticket_config *tk)
	{
	return !(tk->acks_received ^ booth_conf->all_bits);
	}

	static inline int all_sites_replied(struct ticket_config *tk)
	{
	return !((tk->acks_received & booth_conf->sites_bits) ^ booth_conf->sites_bits);
	}


	#endif
	diff --git a/src/main.c b/src/main.c
	index 8d921c5..9b47ec4 100644
	--- a/src/main.c
	+++ b/src/main.c
	@@ -1,1639 +1,1636 @@
	/*
	* Copyright (C) 2011 Jiaju Zhang <jjzhang@suse.de>
	* Copyright (C) 2013-2014 Philipp Marek <philipp.marek@linbit.com>
	*
	* This program is free software; you can redistribute it and/or
	* modify it under the terms of the GNU General Public
	* License as published by the Free Software Foundation; either
	* version 2 of the License, or (at your option) any later version.
	*
	* This software is distributed in the hope that it will be useful,
	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	* General Public License for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with this program; if not, write to the Free Software Foundation, Inc.,
	* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
	*/

	#include <stdio.h>
	#include <unistd.h>
	#include <stdlib.h>
	#include <sched.h>
	#include <errno.h>
	#include <limits.h>
	#include <sys/file.h>
	#include <sys/types.h>
	#include <sys/stat.h>
	#include <sys/mman.h>
	#include <sys/time.h>
	#include <sys/resource.h>
	#include <sys/socket.h>
	#include <sys/un.h>
	#include <sys/poll.h>
	#include <sys/wait.h>
	#include <fcntl.h>
	#include <string.h>
	#include <ctype.h>
	#include <assert.h>
	#include <signal.h>
	#include <netdb.h>
	#include <arpa/inet.h>
	#include <sys/types.h>

	#include <pacemaker/crm/services.h>

	#include "b_config.h"

	#if HAVE_LIBGCRYPT
	#include <gcrypt.h>
	#endif
	#ifndef NAMETAG_LIBSYSTEMD
	#include <clplumbing/setproctitle.h>
	#else
	#include "alt/nametag_libsystemd.h"
	#endif
	#ifdef COREDUMP_NURSING
	#include <sys/prctl.h>
	#include <clplumbing/coredumps.h>
	#endif
	#include "log.h"
	#include "booth.h"
	#include "config.h"
	#include "transport.h"
	#include "inline-fn.h"
	#include "pacemaker.h"
	#include "ticket.h"
	#include "request.h"
	#include "attr.h"
	#include "handler.h"

	#define RELEASE_VERSION "1.0"
	#define RELEASE_STR RELEASE_VERSION " (build " BOOTH_BUILD_VERSION ")"

	#define CLIENT_NALLOC 32

	static int daemonize = 1;
	int enable_stderr = 0;
	timetype start_time;


	/** Structure for "clients".
	* Filehandles with incoming data get registered here (and in pollfds),
	* along with their callbacks.
	* Because these can be reallocated with every new fd, addressing
	* happens _only_ by their numeric index. */
	struct client *clients = NULL;
	struct pollfd *pollfds = NULL;
	static int client_maxi;
	static int client_size = 0;


	static const struct booth_site _no_leader = {
	.addr_string = "none",
	.site_id = NO_ONE,
	.index = -1,
	};
	struct booth_site const no_leader = (struct booth_site) &_no_leader;

	typedef enum
	{
	BOOTHD_STARTED=0,
	BOOTHD_STARTING
	} BOOTH_DAEMON_STATE;

	int poll_timeout;



	struct booth_config *booth_conf;
	struct command_line cl;

	static void client_alloc(void)
	{
	int i;

	if (!(clients = realloc(
	clients, (client_size + CLIENT_NALLOC) * sizeof(*clients))
	) \|\| !(pollfds = realloc(
	pollfds, (client_size + CLIENT_NALLOC) * sizeof(*pollfds))
	)) {
	log_error("can't alloc for client array");
	exit(1);
	}

	for (i = client_size; i < client_size + CLIENT_NALLOC; i++) {
	clients[i].workfn = NULL;
	clients[i].deadfn = NULL;
	clients[i].fd = -1;
	pollfds[i].fd = -1;
	pollfds[i].revents = 0;
	}
	client_size += CLIENT_NALLOC;
	}

	static void client_dead(int ci)
	{
	struct client *c = clients + ci;

	if (c->fd != -1) {
	log_debug("removing client %d", c->fd);
	close(c->fd);
	}

	c->fd = -1;
	c->workfn = NULL;

	if (c->msg) {
	free(c->msg);
	c->msg = NULL;
	c->offset = 0;
	}

	pollfds[ci].fd = -1;
	}

	int client_add(int fd, const struct booth_transport *tpt,
	void (*workfn)(int ci),
	void (*deadfn)(int ci))
	{
	int i;
	struct client *c;


	if (client_size - 1 <= client_maxi ) {
	client_alloc();
	}

	for (i = 0; i < client_size; i++) {
	c = clients + i;
	if (c->fd != -1)
	continue;

	c->workfn = workfn;
	if (deadfn)
	c->deadfn = deadfn;
	else
	c->deadfn = client_dead;

	c->transport = tpt;
	c->fd = fd;
	c->msg = NULL;
	c->offset = 0;

	pollfds[i].fd = fd;
	pollfds[i].events = POLLIN;
	if (i > client_maxi)
	client_maxi = i;

	return i;
	}

	assert(!"no client");
	}

	int find_client_by_fd(int fd)
	{
	int i;

	if (fd < 0)
	return -1;

	for (i = 0; i <= client_maxi; i++) {
	if (clients[i].fd == fd)
	return i;
	}
	return -1;
	}

	static int format_peers(char *pdata, unsigned int len)
	{
	struct booth_site *s;
	char data, cp;
	char time_str[64];
	int i, alloc;

	*pdata = NULL;
	*len = 0;

	alloc = booth_conf->site_count * (BOOTH_NAME_LEN + 256);
	data = malloc(alloc);
	if (!data)
	return -ENOMEM;

	cp = data;
	FOREACH_NODE(i, s) {
	if (s == local)
	continue;
	strftime(time_str, sizeof(time_str), "%F %T",
	localtime(&s->last_recv));
	cp += snprintf(cp,
	alloc - (cp - data),
	"%-12s %s, last recv: %s\n",
	type_to_string(s->type),
	s->addr_string,
	time_str);
	cp += snprintf(cp,
	alloc - (cp - data),
	"\tSent pkts:%u error:%u resends:%u\n",
	s->sent_cnt,
	s->sent_err_cnt,
	s->resend_cnt);
	cp += snprintf(cp,
	alloc - (cp - data),
	"\tRecv pkts:%u error:%u authfail:%u invalid:%u\n\n",
	s->recv_cnt,
	s->recv_err_cnt,
	s->sec_cnt,
	s->invalid_cnt);
	if (alloc - (cp - data) <= 0) {
	free(data);
	return -ENOMEM;
	}
	}

	*pdata = data;
	*len = cp - data;

	return 0;
	}


	void list_peers(int fd)
	{
	char *data;
	unsigned int olen;
	struct boothc_hdr_msg hdr;

	if (format_peers(&data, &olen) < 0)
	goto out;

	init_header(&hdr.header, CL_LIST, 0, 0, RLT_SUCCESS, 0, sizeof(hdr) + olen);
	(void)send_header_plus(fd, &hdr, data, olen);

	out:
	if (data)
	free(data);
	}

	/* trim trailing spaces if the key is ascii
	*/
	static void trim_key()
	{
	char *p;
	int i;

	for (i=0, p=booth_conf->authkey; i < booth_conf->authkey_len; i++, p++)
	if (!isascii(*p))
	return;

	p = booth_conf->authkey;
	while (booth_conf->authkey_len > 0 && isspace(*p)) {
	p++;
	booth_conf->authkey_len--;
	}
	memmove(booth_conf->authkey, p, booth_conf->authkey_len);

	p = booth_conf->authkey + booth_conf->authkey_len - 1;
	while (booth_conf->authkey_len > 0 && isspace(*p)) {
	booth_conf->authkey_len--;
	p--;
	}
	}

	static int read_authkey()
	{
	int fd;

	booth_conf->authkey[0] = '\0';
	fd = open(booth_conf->authfile, O_RDONLY);
	if (fd < 0) {
	log_error("cannot open %s: %s",
	booth_conf->authfile, strerror(errno));
	return -1;
	}
	if (fstat(fd, &booth_conf->authstat) < 0) {
	log_error("cannot stat authentication file %s (%d): %s",
	booth_conf->authfile, fd, strerror(errno));
	close(fd);
	return -1;
	}
	if (booth_conf->authstat.st_mode & (S_IRGRP \| S_IROTH)) {
	log_error("%s: file shall not be readable for anyone but the owner",
	booth_conf->authfile);
	close(fd);
	return -1;
	}
	booth_conf->authkey_len = read(fd, booth_conf->authkey, BOOTH_MAX_KEY_LEN);
	close(fd);
	trim_key();
	log_debug("read key of size %d in authfile %s",
	booth_conf->authkey_len, booth_conf->authfile);
	/* make sure that the key is of minimum length */
	return (booth_conf->authkey_len >= BOOTH_MIN_KEY_LEN) ? 0 : -1;
	}

	int update_authkey()
	{
	struct stat statbuf;

	if (stat(booth_conf->authfile, &statbuf) < 0) {
	log_error("cannot stat authentication file %s: %s",
	booth_conf->authfile, strerror(errno));
	return -1;
	}
	if (statbuf.st_mtime > booth_conf->authstat.st_mtime) {
	return read_authkey();
	}
	return 0;
	}

	static int setup_config(struct booth_config **conf_pptr, int type)
	{
	int rv;

	assert(conf_pptr != NULL);

	rv = read_config(conf_pptr, cl.configfile, type);
	if (rv < 0)
	goto out;

	if (is_auth_req()) {
	rv = read_authkey();
	if (rv < 0)
	goto out;
	#if HAVE_LIBGCRYPT
	if (!gcry_check_version(NULL)) {
	log_error("gcry_check_version");
	rv = -ENOENT;
	goto out;
	}
	gcry_control(GCRYCTL_DISABLE_SECMEM, 0);
	gcry_control(GCRYCTL_INITIALIZATION_FINISHED, 0);
	#endif
	}

	/* Set "local" pointer, ignoring errors. */
	if (cl.type == DAEMON && cl.site[0]) {
	if (!find_site_by_name(booth_conf, cl.site, &local, 1)) {
	log_error("Cannot find \"%s\" in the configuration.",
	cl.site);
	return -EINVAL;
	}
	local->local = 1;
	} else
	find_myself(booth_conf, NULL,
	type == CLIENT \|\| type == GEOSTORE);


	rv = check_config(booth_conf, type);
	if (rv < 0)
	goto out;


	/* Per default the PID file name is derived from the
	* configuration name. */
	if (!cl.lockfile[0]) {
	snprintf(cl.lockfile, sizeof(cl.lockfile) - 1,
	"%s/%s.pid", BOOTH_RUN_DIR, (*conf_pptr)->name);
	}

	out:
	return rv;
	}

	static int setup_transport(void)
	{
	int rv;

	rv = transport()->init(message_recv);
	if (rv < 0) {
	log_error("failed to init booth_transport %s", transport()->name);
	goto out;
	}

	rv = booth_transport[TCP].init(NULL);
	if (rv < 0) {
	log_error("failed to init booth_transport[TCP]");
	goto out;
	}

	out:
	return rv;
	}


	static int write_daemon_state(int fd, int state)
	{
	char buffer[1024];
	int rv, size;

	size = sizeof(buffer) - 1;
	rv = snprintf(buffer, size,
	- "booth_pid=%d "
	- "booth_state=%s "
	- "booth_type=%s "
	- "booth_cfg_name='%s' "
	- "booth_id=%d "
	- "booth_addr_string='%s' "
	- "booth_port=%d\n",
	- getpid(),
	- ( state == BOOTHD_STARTED ? "started" :
	- state == BOOTHD_STARTING ? "starting" :
	- "invalid"),
	- type_to_string(local->type),
	- booth_conf->name,
	- local->site_id,
	- local->addr_string,
	- booth_conf->port);
	+ "booth_pid=%d "
	+ "booth_state=%s "
	+ "booth_type=%s "
	+ "booth_cfg_name='%s' "
	+ "booth_id=%d "
	+ "booth_addr_string='%s' "
	+ "booth_port=%d\n",
	+ getpid(),
	+ ( state == BOOTHD_STARTED ? "started" :
	+ state == BOOTHD_STARTING ? "starting" : "invalid"),
	+ type_to_string(local->type), booth_conf->name,
	+ get_local_id(), site_string(local),
	+ site_port(local));

	if (rv < 0 \|\| rv == size) {
	log_error("Buffer filled up in write_daemon_state().");
	return -1;
	}
	size = rv;


	rv = ftruncate(fd, 0);
	if (rv < 0) {
	log_error("lockfile %s truncate error %d: %s",
	cl.lockfile, errno, strerror(errno));
	return rv;
	}


	rv = lseek(fd, 0, SEEK_SET);
	if (rv < 0) {
	log_error("lseek set fd(%d) offset to 0 error, return(%d), message(%s)",
	fd, rv, strerror(errno));
	rv = -1;
	return rv;
	}


	rv = write(fd, buffer, size);

	if (rv != size) {
	log_error("write to fd(%d, %d) returned %d, errno %d, message(%s)",
	fd, size,
	rv, errno, strerror(errno));
	return -1;
	}

	return 0;
	}

	static int loop(int fd)
	{
	void (*workfn) (int ci);
	void (*deadfn) (int ci);
	int rv, i;

	rv = setup_transport();
	if (rv < 0)
	goto fail;

	rv = setup_ticket(booth_conf);
	if (rv < 0)
	goto fail;

	rv = write_daemon_state(fd, BOOTHD_STARTED);
	if (rv != 0) {
	log_error("write daemon state %d to lockfile error %s: %s",
	BOOTHD_STARTED, cl.lockfile, strerror(errno));
	goto fail;
	}

	log_info("BOOTH %s daemon started, node id is 0x%08X (%d).",
	type_to_string(local->type),
	local->site_id, local->site_id);

	while (1) {
	rv = poll(pollfds, client_maxi + 1, poll_timeout);
	if (rv == -1 && errno == EINTR)
	continue;
	if (rv < 0) {
	log_error("poll failed: %s (%d)", strerror(errno), errno);
	goto fail;
	}

	for (i = 0; i <= client_maxi; i++) {
	if (clients[i].fd < 0)
	continue;

	if (pollfds[i].revents & POLLIN) {
	workfn = clients[i].workfn;
	if (workfn)
	workfn(i);
	}
	if (pollfds[i].revents &
	(POLLERR \| POLLHUP \| POLLNVAL)) {
	deadfn = clients[i].deadfn;
	if (deadfn)
	deadfn(i);
	}
	}

	process_tickets();
	}

	return 0;

	fail:
	return -1;
	}


	static int test_reply(cmd_result_t reply_code, cmd_request_t cmd)
	{
	int rv = 0;
	const char *op_str = "";

	if (cmd == CMD_GRANT)
	op_str = "grant";
	else if (cmd == CMD_REVOKE)
	op_str = "revoke";
	else if (cmd == CMD_LIST)
	op_str = "list";
	else if (cmd == CMD_PEERS)
	op_str = "peers";
	else {
	log_error("internal error reading reply result!");
	return -1;
	}

	switch (reply_code) {
	case RLT_OVERGRANT:
	log_info("You're granting a granted ticket. "
	"If you wanted to migrate a ticket, "
	"use revoke first, then use grant.");
	rv = -1;
	break;

	case RLT_TICKET_IDLE:
	log_info("ticket is not owned");
	rv = 0;
	break;

	case RLT_ASYNC:
	log_info("%s command sent, result will be returned "
	"asynchronously. Please use \"booth list\" to "
	"see the outcome.", op_str);
	rv = 0;
	break;

	case RLT_CIB_PENDING:
	log_info("%s succeeded (CIB commit pending)", op_str);
	/* wait for the CIB commit? */
	rv = (cl.options & OPT_WAIT_COMMIT) ? 3 : 0;
	break;

	case RLT_MORE:
	rv = 2;
	break;

	case RLT_SYNC_SUCC:
	case RLT_SUCCESS:
	if (cmd != CMD_LIST && cmd != CMD_PEERS)
	log_info("%s succeeded!", op_str);
	rv = 0;
	break;

	case RLT_SYNC_FAIL:
	log_info("%s failed!", op_str);
	rv = -1;
	break;

	case RLT_INVALID_ARG:
	log_error("ticket \"%s\" does not exist",
	cl.msg.ticket.id);
	rv = -1;
	break;

	case RLT_AUTH:
	log_error("authentication error");
	rv = -1;
	break;

	case RLT_EXT_FAILED:
	log_error("before-acquire-handler for ticket \"%s\" failed, grant denied",
	cl.msg.ticket.id);
	rv = -1;
	break;

	case RLT_ATTR_PREREQ:
	log_error("attr-prereq for ticket \"%s\" failed, grant denied",
	cl.msg.ticket.id);
	rv = -1;
	break;

	case RLT_REDIRECT:
	/* talk to another site */
	rv = 1;
	break;

	default:
	log_error("got an error code: %x", rv);
	rv = -1;
	}
	return rv;
	}

	static int query_get_string_answer(cmd_request_t cmd)
	{
	struct booth_site *site;
	struct boothc_hdr_msg reply;
	struct boothc_header *header;
	char *data;
	int data_len;
	int rv;
	struct booth_transport const *tpt;
	int (*test_reply_f) (cmd_result_t reply_code, cmd_request_t cmd);
	size_t msg_size;
	void *request;

	if (cl.type == GEOSTORE) {
	test_reply_f = test_attr_reply;
	msg_size = sizeof(cl.attr_msg);
	request = &cl.attr_msg;
	} else {
	test_reply_f = test_reply;
	msg_size = sizeof(cl.msg);
	request = &cl.msg;
	}
	header = (struct boothc_header *)request;
	data = NULL;

	init_header(header, cmd, 0, cl.options, 0, 0, msg_size);

	if (!*cl.site)
	site = local;
	else if (!find_site_by_name(booth_conf, cl.site, &site, 1)) {
	log_error("cannot find site \"%s\"", cl.site);
	rv = ENOENT;
	goto out;
	}

	tpt = booth_transport + TCP;
	rv = tpt->open(site);
	if (rv < 0)
	goto out_close;

	rv = tpt->send(site, request, msg_size);
	if (rv < 0)
	goto out_close;

	rv = tpt->recv_auth(site, &reply, sizeof(reply));
	if (rv < 0)
	goto out_close;

	data_len = ntohl(reply.header.length) - rv;

	/* no attribute, or no ticket found */
	if (!data_len) {
	goto out_test_reply;
	}

	data = malloc(data_len+1);
	if (!data) {
	rv = -ENOMEM;
	goto out_close;
	}
	rv = tpt->recv(site, data, data_len);
	if (rv < 0)
	goto out_close;
	*(data+data_len) = '\0';

	*(data + data_len) = '\0';
	(void)fputs(data, stdout);
	fflush(stdout);
	rv = 0;

	out_test_reply:
	rv = test_reply_f(ntohl(reply.header.result), cmd);
	out_close:
	tpt->close(site);
	out:
	if (data)
	free(data);
	return rv;
	}


	static int do_command(cmd_request_t cmd)
	{
	struct booth_site *site;
	struct boothc_ticket_msg reply;
	struct booth_transport const *tpt;
	uint32_t leader_id;
	int rv;
	int reply_cnt = 0, msg_logged = 0;
	const char *op_str = "";

	if (cmd == CMD_GRANT)
	op_str = "grant";
	else if (cmd == CMD_REVOKE)
	op_str = "revoke";

	rv = 0;
	site = NULL;

	/* Always use TCP for client - at least for now. */
	tpt = booth_transport + TCP;

	if (!*cl.site)
	site = local;
	else {
	if (!find_site_by_name(booth_conf, cl.site, &site, 1)) {
	log_error("Site \"%s\" not configured.", cl.site);
	goto out_close;
	}
	}

	if (site->type == ARBITRATOR) {
	if (site == local) {
	log_error("We're just an arbitrator, cannot grant/revoke tickets here.");
	} else {
	log_error("%s is just an arbitrator, cannot grant/revoke tickets there.", cl.site);
	}
	goto out_close;
	}

	assert(site->type == SITE);

	/* We don't check for existence of ticket, so that asking can be
	* done without local configuration, too.
	* Although, that means that the UDP port has to be specified, too. */
	if (!cl.msg.ticket.id[0]) {
	/* If the loaded configuration has only a single ticket defined, use that. */
	if (booth_conf->ticket_count == 1) {
	strncpy(cl.msg.ticket.id, booth_conf->ticket[0].name,
	sizeof(cl.msg.ticket.id));
	} else {
	log_error("No ticket given.");
	goto out_close;
	}
	}

	redirect:
	init_header(&cl.msg.header, cmd, 0, cl.options, 0, 0, sizeof(cl.msg));

	rv = tpt->open(site);
	if (rv < 0)
	goto out_close;

	rv = tpt->send(site, &cl.msg, sendmsglen(&cl.msg));
	if (rv < 0)
	goto out_close;

	read_more:
	rv = tpt->recv_auth(site, &reply, sizeof(reply));
	if (rv < 0) {
	/* print any errors depending on the code sent by the
	* server */
	(void)test_reply(ntohl(reply.header.result), cmd);
	goto out_close;
	}

	rv = test_reply(ntohl(reply.header.result), cmd);
	if (rv == 1) {
	tpt->close(site);
	leader_id = ntohl(reply.ticket.leader);
	if (!find_site_by_id(booth_conf, leader_id, &site)) {
	log_error("Message with unknown redirect site %x received", leader_id);
	rv = -1;
	goto out_close;
	}
	goto redirect;
	} else if (rv == 2 \|\| rv == 3) {
	/* the server has more to say */
	/* don't wait too long */
	if (reply_cnt > 1 && !(cl.options & OPT_WAIT)) {
	rv = 0;
	log_info("Giving up on waiting for the definite result. "
	"Please use \"booth list\" later to "
	"see the outcome.");
	goto out_close;
	}
	if (reply_cnt == 0) {
	log_info("%s request sent, "
	"waiting for the result ...", op_str);
	msg_logged++;
	} else if (rv == 3 && msg_logged < 2) {
	log_info("waiting for the CIB commit ...");
	msg_logged++;
	}
	reply_cnt++;
	goto read_more;
	}

	out_close:
	if (site)
	tpt->close(site);
	return rv;
	}



	static int _lockfile(int mode, int fdp, pid_t locked_by)
	{
	struct flock lock;
	int fd, rv;


	/* After reboot the directory may not yet exist.
	* Try to create it, but ignore errors. */
	if (strncmp(cl.lockfile, BOOTH_RUN_DIR,
	strlen(BOOTH_RUN_DIR)) == 0)
	mkdir(BOOTH_RUN_DIR, 0775);


	if (locked_by)
	*locked_by = 0;

	*fdp = -1;
	fd = open(cl.lockfile, mode, 0664);
	if (fd < 0)
	return errno;

	*fdp = fd;

	lock.l_type = F_WRLCK;
	lock.l_start = 0;
	lock.l_whence = SEEK_SET;
	lock.l_len = 0;
	lock.l_pid = 0;


	if (fcntl(fd, F_SETLK, &lock) == 0)
	return 0;

	rv = errno;

	if (locked_by)
	if (fcntl(fd, F_GETLK, &lock) == 0)
	*locked_by = lock.l_pid;

	return rv;
	}


	static inline int is_root(void)
	{
	return geteuid() == 0;
	}


	static int create_lockfile(void)
	{
	int rv, fd;

	fd = -1;
	rv = _lockfile(O_CREAT \| O_WRONLY, &fd, NULL);

	if (fd == -1) {
	log_error("lockfile %s open error %d: %s",
	cl.lockfile, rv, strerror(rv));
	return -1;
	}

	if (rv < 0) {
	log_error("lockfile %s setlk error %d: %s",
	cl.lockfile, rv, strerror(rv));
	goto fail;
	}

	rv = write_daemon_state(fd, BOOTHD_STARTING);
	if (rv != 0) {
	log_error("write daemon state %d to lockfile error %s: %s",
	BOOTHD_STARTING, cl.lockfile, strerror(errno));
	goto fail;
	}

	if (is_root()) {
	if (fchown(fd, booth_conf->uid, booth_conf->gid) < 0)
	log_error("fchown() on lockfile said %d: %s",
	errno, strerror(errno));
	}

	return fd;

	fail:
	close(fd);
	return -1;
	}

	static void unlink_lockfile(int fd)
	{
	unlink(cl.lockfile);
	close(fd);
	}

	static void print_usage(void)
	{
	printf(
	"Usage:\n"
	" booth list [options]\n"
	" booth {grant\|revoke} [options] <ticket>\n"
	" booth status [options]\n"
	"\n"
	" list: List all tickets\n"
	" grant: Grant ticket to site\n"
	" revoke: Revoke ticket\n"
	"\n"
	"Options:\n"
	" -c FILE Specify config file [default " BOOTH_DEFAULT_CONF "]\n"
	" Can be a path or just a name without \".conf\" suffix\n"
	" -s <site> Connect/grant to a different site\n"
	" -F Try to grant the ticket immediately\n"
	" even if not all sites are reachable\n"
	" For manual tickets:\n"
	" grant a manual ticket even if it has been already granted\n"
	" -w Wait forever for the outcome of the request\n"
	" -C Wait until the ticket is committed to the CIB (grant only)\n"
	" -h Print this help\n"
	"\n"
	"Examples:\n"
	"\n"
	" # booth list (list tickets)\n"
	" # booth grant ticket-A (grant ticket here)\n"
	" # booth grant -s 10.121.8.183 ticket-A (grant ticket to site 10.121.8.183)\n"
	" # booth revoke ticket-A (revoke ticket)\n"
	"\n"
	"See the booth(8) man page for more details.\n"
	);
	}

	#define OPTION_STRING "c:Dl:t:s:FhSwC"
	#define ATTR_OPTION_STRING "c:Dt:s:h"

	void safe_copy(char dest, char value, size_t buflen, const char *description) {
	int content_len = buflen - 1;

	if (strlen(value) >= content_len) {
	fprintf(stderr, "'%s' exceeds maximum %s length of %d\n",
	value, description, content_len);
	exit(EXIT_FAILURE);
	}
	strncpy(dest, value, content_len);
	dest[content_len] = 0;
	}

	static int host_convert(char hostname, char ip_str, size_t ip_size)
	{
	struct addrinfo *result = NULL, hints = {0};
	int re = -1;

	memset(&hints, 0, sizeof(hints));
	hints.ai_family = AF_INET;
	hints.ai_socktype = SOCK_DGRAM;

	re = getaddrinfo(hostname, NULL, &hints, &result);

	if (re == 0) {
	struct in_addr addr = ((struct sockaddr_in *)result->ai_addr)->sin_addr;
	const char *re_ntop = inet_ntop(AF_INET, &addr, ip_str, ip_size);
	if (re_ntop == NULL) {
	re = -1;
	}
	}

	freeaddrinfo(result);
	return re;
	}

	#define cparg(dest, descr) do { \
	if (optind >= argc) \
	goto missingarg; \
	safe_copy(dest, argv[optind], sizeof(dest), descr); \
	optind++; \
	} while(0)

	static int read_arguments(int argc, char **argv)
	{
	int optchar;
	char *arg1 = argv[1];
	char *op = NULL;
	char *cp;
	const char *opt_string = OPTION_STRING;
	char site_arg[INET_ADDRSTRLEN] = {0};
	int left;

	cl.type = 0;
	if ((cp = strstr(argv[0], ATTR_PROG)) &&
	!strcmp(cp, ATTR_PROG)) {
	cl.type = GEOSTORE;
	op = argv[1];
	optind = 2;
	opt_string = ATTR_OPTION_STRING;
	} else if (argc > 1 && (strcmp(arg1, "arbitrator") == 0 \|\|
	strcmp(arg1, "site") == 0 \|\|
	strcmp(arg1, "start") == 0 \|\|
	strcmp(arg1, "daemon") == 0)) {
	cl.type = DAEMON;
	optind = 2;
	} else if (argc > 1 && (strcmp(arg1, "status") == 0)) {
	cl.type = STATUS;
	optind = 2;
	} else if (argc > 1 && (strcmp(arg1, "client") == 0)) {
	cl.type = CLIENT;
	if (argc < 3) {
	print_usage();
	exit(EXIT_FAILURE);
	}
	op = argv[2];
	optind = 3;
	}
	if (!cl.type) {
	cl.type = CLIENT;
	op = argv[1];
	optind = 2;
	}

	if (argc < 2 \|\| !strcmp(arg1, "help") \|\| !strcmp(arg1, "--help") \|\|
	!strcmp(arg1, "-h")) {
	if (cl.type == GEOSTORE)
	print_geostore_usage();
	else
	print_usage();
	exit(EXIT_SUCCESS);
	}

	if (!strcmp(arg1, "version") \|\| !strcmp(arg1, "--version") \|\|
	!strcmp(arg1, "-V")) {
	printf("%s %s\n", argv[0], RELEASE_STR);
	exit(EXIT_SUCCESS);
	}

	if (cl.type == CLIENT) {
	if (!strcmp(op, "list"))
	cl.op = CMD_LIST;
	else if (!strcmp(op, "grant"))
	cl.op = CMD_GRANT;
	else if (!strcmp(op, "revoke"))
	cl.op = CMD_REVOKE;
	else if (!strcmp(op, "peers"))
	cl.op = CMD_PEERS;
	else {
	fprintf(stderr, "client operation \"%s\" is unknown\n",
	op);
	exit(EXIT_FAILURE);
	}
	} else if (cl.type == GEOSTORE) {
	if (!strcmp(op, "list"))
	cl.op = ATTR_LIST;
	else if (!strcmp(op, "set"))
	cl.op = ATTR_SET;
	else if (!strcmp(op, "get"))
	cl.op = ATTR_GET;
	else if (!strcmp(op, "delete"))
	cl.op = ATTR_DEL;
	else {
	fprintf(stderr, "attribute operation \"%s\" is unknown\n",
	op);
	exit(EXIT_FAILURE);
	}
	}

	while (optind < argc) {
	optchar = getopt(argc, argv, opt_string);

	switch (optchar) {
	case 'c':
	if (strchr(optarg, '/')) {
	safe_copy(cl.configfile, optarg,
	sizeof(cl.configfile), "config file");
	} else {
	/* If no "/" in there, use with default directory. */
	strcpy(cl.configfile, BOOTH_DEFAULT_CONF_DIR);
	cp = cl.configfile + strlen(BOOTH_DEFAULT_CONF_DIR);
	assert(cp > cl.configfile);
	assert(*(cp-1) == '/');

	/* Write at the \0, ie. after the "/" */
	safe_copy(cp, optarg,
	(sizeof(cl.configfile) -
	(cp - cl.configfile) -
	strlen(BOOTH_DEFAULT_CONF_EXT)),
	"config name");

	/* If no extension, append ".conf".
	* Space is available, see -strlen() above. */
	if (!strchr(cp, '.'))
	strcat(cp, BOOTH_DEFAULT_CONF_EXT);
	}
	break;

	case 'D':
	debug_level++;
	break;

	case 'S':
	daemonize = 0;
	enable_stderr = 1;
	break;

	case 'l':
	safe_copy(cl.lockfile, optarg, sizeof(cl.lockfile), "lock file");
	break;
	case 't':
	if (cl.op == CMD_GRANT \|\| cl.op == CMD_REVOKE) {
	safe_copy(cl.msg.ticket.id, optarg,
	sizeof(cl.msg.ticket.id), "ticket name");
	} else if (cl.type == GEOSTORE) {
	safe_copy(cl.attr_msg.attr.tkt_id, optarg,
	sizeof(cl.attr_msg.attr.tkt_id), "ticket name");
	} else {
	print_usage();
	exit(EXIT_FAILURE);
	}
	break;

	case 's':
	/* For testing and debugging: allow "-s site" also for
	* daemon start, so that the address that should be used
	* can be set manually.
	* This makes it easier to start multiple processes
	* on one machine. */
	if (cl.type == CLIENT \|\| cl.type == GEOSTORE \|\|
	(cl.type == DAEMON && debug_level)) {
	if (strcmp(optarg, OTHER_SITE) &&
	host_convert(optarg, site_arg, INET_ADDRSTRLEN) == 0) {
	safe_copy(cl.site, site_arg, sizeof(cl.site), "site name");
	} else {
	safe_copy(cl.site, optarg, sizeof(cl.site), "site name");
	}
	} else {
	log_error("\"-s\" not allowed in daemon mode.");
	exit(EXIT_FAILURE);
	}
	break;

	case 'F':
	if (cl.type != CLIENT \|\| cl.op != CMD_GRANT) {
	log_error("use \"-F\" only for client grant");
	exit(EXIT_FAILURE);
	}
	cl.options \|= OPT_IMMEDIATE;
	break;

	case 'w':
	if (cl.type != CLIENT \|\|
	(cl.op != CMD_GRANT && cl.op != CMD_REVOKE)) {
	log_error("use \"-w\" only for grant and revoke");
	exit(EXIT_FAILURE);
	}
	cl.options \|= OPT_WAIT;
	break;

	case 'C':
	if (cl.type != CLIENT \|\| cl.op != CMD_GRANT) {
	log_error("use \"-C\" only for grant");
	exit(EXIT_FAILURE);
	}
	cl.options \|= OPT_WAIT \| OPT_WAIT_COMMIT;
	break;

	case 'h':
	if (cl.type == GEOSTORE)
	print_geostore_usage();
	else
	print_usage();
	exit(EXIT_SUCCESS);
	break;

	case ':':
	case '?':
	fprintf(stderr, "Please use '-h' for usage.\n");
	exit(EXIT_FAILURE);
	break;

	case -1:
	/* No more parameters on cmdline, only arguments. */
	goto extra_args;

	default:
	goto unknown;
	};
	}

	return 0;

	extra_args:
	if (cl.type == CLIENT && !cl.msg.ticket.id[0]) {
	cparg(cl.msg.ticket.id, "ticket name");
	} else if (cl.type == GEOSTORE) {
	if (cl.op != ATTR_LIST) {
	cparg(cl.attr_msg.attr.name, "attribute name");
	}
	if (cl.op == ATTR_SET) {
	cparg(cl.attr_msg.attr.val, "attribute value");
	}
	}

	if (optind == argc)
	return 0;


	left = argc - optind;
	fprintf(stderr, "Superfluous argument%s: %s%s\n",
	left == 1 ? "" : "s",
	argv[optind],
	left == 1 ? "" : "...");
	exit(EXIT_FAILURE);

	unknown:
	fprintf(stderr, "unknown option: %s\n", argv[optind]);
	exit(EXIT_FAILURE);

	missingarg:
	fprintf(stderr, "not enough arguments\n");
	exit(EXIT_FAILURE);
	}


	static void set_scheduler(void)
	{
	struct sched_param sched_param;
	struct rlimit rlimit;
	int rv;

	rlimit.rlim_cur = RLIM_INFINITY;
	rlimit.rlim_max = RLIM_INFINITY;
	setrlimit(RLIMIT_MEMLOCK, &rlimit);
	rv = mlockall(MCL_CURRENT \| MCL_FUTURE);
	if (rv < 0) {
	log_error("mlockall failed");
	}

	rv = sched_get_priority_max(SCHED_RR);
	if (rv != -1) {
	sched_param.sched_priority = rv;
	rv = sched_setscheduler(0, SCHED_RR, &sched_param);
	if (rv == -1)
	log_error("could not set SCHED_RR priority %d: %s (%d)",
	sched_param.sched_priority,
	strerror(errno), errno);
	} else {
	log_error("could not get maximum scheduler priority err %d",
	errno);
	}
	}

	static int set_procfs_val(const char path, const char val)
	{
	int rc = -1;
	FILE *fp = fopen(path, "w");

	if (fp) {
	if (fprintf(fp, "%s", val) > 0)
	rc = 0;
	fclose(fp);
	}
	return rc;
	}

	static int do_status(struct booth_config **conf_pptr, int type)
	{
	pid_t pid;
	int rv, status_lock_fd, ret;
	const char *reason = NULL;
	char lockfile_data[1024], *cp;

	assert(conf_pptr != NULL);

	ret = PCMK_OCF_NOT_RUNNING;

	rv = setup_config(conf_pptr, type);
	if (rv) {
	reason = "Error reading configuration.";
	ret = PCMK_OCF_UNKNOWN_ERROR;
	goto quit;
	}


	if (!local) {
	reason = "No Service IP active here.";
	goto quit;
	}


	rv = _lockfile(O_RDWR, &status_lock_fd, &pid);
	if (status_lock_fd == -1) {
	reason = "No PID file.";
	goto quit;
	}
	if (rv == 0) {
	close(status_lock_fd);
	reason = "PID file not locked.";
	goto quit;
	}
	if (pid) {
	fprintf(stdout, "booth_lockpid=%d ", pid);
	fflush(stdout);
	}

	rv = read(status_lock_fd, lockfile_data, sizeof(lockfile_data) - 1);
	if (rv < 4) {
	close(status_lock_fd);
	reason = "Cannot read lockfile data.";
	ret = PCMK_LSB_UNKNOWN_ERROR;
	goto quit;
	}
	lockfile_data[rv] = 0;

	close(status_lock_fd);


	/* Make sure it's only a single line */
	cp = strchr(lockfile_data, '\r');
	if (cp)
	*cp = 0;
	cp = strchr(lockfile_data, '\n');
	if (cp)
	*cp = 0;



	rv = setup_tcp_listener(1);
	if (rv == 0) {
	reason = "TCP port not in use.";
	goto quit;
	}


	fprintf(stdout, "booth_lockfile='%s' %s\n",
	cl.lockfile, lockfile_data);
	if (!daemonize)
	fprintf(stderr, "Booth at %s port %d seems to be running.\n",
	- local->addr_string, (*conf_pptr)->port);
	+ site_string(local), site_port(local));
	return 0;


	quit:
	log_debug("not running: %s", reason);
	/* Ie. "DEBUG" */
	if (!daemonize)
	fprintf(stderr, "not running: %s\n", reason);
	return ret;
	}


	static int limit_this_process(void)
	{
	int rv;
	if (!is_root())
	return 0;

	if (setregid(booth_conf->gid, booth_conf->gid) < 0) {
	rv = errno;
	log_error("setregid() didn't work: %s", strerror(rv));
	return rv;
	}

	if (setreuid(booth_conf->uid, booth_conf->uid) < 0) {
	rv = errno;
	log_error("setreuid() didn't work: %s", strerror(rv));
	return rv;
	}

	return 0;
	}

	static int lock_fd = -1;

	static void server_exit(void)
	{
	int rv;

	if (lock_fd >= 0) {
	/* We might not be able to delete it, but at least
	* make it empty. */
	rv = ftruncate(lock_fd, 0);
	(void)rv;
	unlink_lockfile(lock_fd);
	}
	log_info("exiting");
	}

	static void sig_exit_handler(int sig)
	{
	log_info("caught signal %d", sig);
	exit(0);
	}

	static int do_server(struct booth_config **conf_pptr, int type)
	{
	int rv = -1;
	static char log_ent[128] = DAEMON_NAME "-";

	assert(conf_pptr != NULL);

	rv = setup_config(conf_pptr, type);
	if (rv < 0)
	return rv;

	if (!local) {
	log_error("Cannot find myself in the configuration.");
	exit(EXIT_FAILURE);
	}

	if (daemonize) {
	if (daemon(0, 0) < 0) {
	perror("daemon error");
	exit(EXIT_FAILURE);
	}
	}

	/* The lockfile must be written to _after_ the call to daemon(), so
	* that the lockfile contains the pid of the daemon, not the parent. */
	lock_fd = create_lockfile();
	if (lock_fd < 0)
	return lock_fd;

	atexit(server_exit);

	strcat(log_ent, type_to_string(local->type));
	cl_log_set_entity(log_ent);
	cl_log_enable_stderr(enable_stderr ? TRUE : FALSE);
	cl_log_set_facility(HA_LOG_FACILITY);
	cl_inherit_logging_environment(0);

	log_info("BOOTH %s %s daemon is starting",
	type_to_string(local->type), RELEASE_STR);

	signal(SIGUSR1, (__sighandler_t)tickets_log_info);
	signal(SIGTERM, (__sighandler_t)sig_exit_handler);
	signal(SIGINT, (__sighandler_t)sig_exit_handler);
	/* we'll handle errors there and then */
	signal(SIGPIPE, SIG_IGN);

	set_scheduler();
	/* we don't want to be killed by the OOM-killer */
	if (set_procfs_val("/proc/self/oom_score_adj", "-999"))
	(void)set_procfs_val("/proc/self/oom_adj", "-16");
	set_proc_title("%s %s %s for [%s]:%d",
	DAEMON_NAME, cl.configfile, type_to_string(local->type),
	- local->addr_string, (*conf_pptr) ->port);
	+ site_string(local), site_port(local));

	rv = limit_this_process();
	if (rv)
	return rv;

	#ifdef COREDUMP_NURSING
	if (cl_enable_coredumps(TRUE) < 0){
	log_error("enabling core dump failed");
	}
	cl_cdtocoredir();
	prctl(PR_SET_DUMPABLE, (unsigned long)TRUE, 0UL, 0UL, 0UL);
	#else
	if (chdir(BOOTH_CORE_DIR) < 0) {
	log_error("cannot change working directory to %s", BOOTH_CORE_DIR);
	}
	#endif

	signal(SIGCHLD, (__sighandler_t)wait_child);
	rv = loop(lock_fd);

	return rv;
	}

	static int do_client(struct booth_config **conf_pptr)
	{
	int rv;

	rv = setup_config(conf_pptr, CLIENT);
	if (rv < 0) {
	log_error("cannot read config");
	goto out;
	}

	switch (cl.op) {
	case CMD_LIST:
	case CMD_PEERS:
	rv = query_get_string_answer(cl.op);
	break;

	case CMD_GRANT:
	case CMD_REVOKE:
	rv = do_command(cl.op);
	break;
	}

	out:
	return rv;
	}

	static int do_attr(struct booth_config **conf_pptr)
	{
	int rv = -1;

	assert(conf_pptr != NULL);

	rv = setup_config(conf_pptr, GEOSTORE);
	if (rv < 0) {
	log_error("cannot read config");
	goto out;
	}

	/* We don't check for existence of ticket, so that asking can be
	* done without local configuration, too.
	* Although, that means that the UDP port has to be specified, too. */
	if (!cl.attr_msg.attr.tkt_id[0]) {
	/* If the loaded configuration has only a single ticket defined, use that. */
	if ((*conf_pptr)->ticket_count == 1) {
	strncpy(cl.attr_msg.attr.tkt_id,
	(*conf_pptr)->ticket[0].name,
	sizeof(cl.attr_msg.attr.tkt_id));
	} else {
	rv = 1;
	log_error("No ticket given.");
	goto out;
	}
	}

	switch (cl.op) {
	case ATTR_LIST:
	case ATTR_GET:
	rv = query_get_string_answer(cl.op);
	break;

	case ATTR_SET:
	case ATTR_DEL:
	rv = do_attr_command(booth_conf, cl.op);
	break;
	}

	out:
	return rv;
	}

	int main(int argc, char argv[], char envp[])
	{
	int rv;
	const char *cp;
	#ifdef LOGGING_LIBQB
	enum qb_log_target_slot i;
	#endif

	init_set_proc_title(argc, argv, envp);
	get_time(&start_time);

	memset(&cl, 0, sizeof(cl));
	strncpy(cl.configfile,
	BOOTH_DEFAULT_CONF, BOOTH_PATH_LEN - 1);
	cl.lockfile[0] = 0;
	debug_level = 0;


	cp = ((cp = strstr(argv[0], ATTR_PROG)) && !strcmp(cp, ATTR_PROG)
	? ATTR_PROG
	: "booth");
	#ifndef LOGGING_LIBQB
	cl_log_set_entity(cp);
	#else
	qb_log_init(cp, LOG_USER, LOG_DEBUG); /* prio driven by debug_level */
	for (i = QB_LOG_TARGET_START; i < QB_LOG_TARGET_MAX; i++) {
	if (i == QB_LOG_SYSLOG \|\| i == QB_LOG_BLACKBOX)
	continue;
	qb_log_format_set(i, "%t %H %N: [%P]: %p: %b");
	}
	(void) qb_log_filter_ctl(QB_LOG_STDERR, QB_LOG_FILTER_ADD,
	QB_LOG_FILTER_FILE, "*", LOG_DEBUG);
	#endif
	cl_log_enable_stderr(TRUE);
	cl_log_set_facility(0);

	rv = read_arguments(argc, argv);
	if (rv < 0)
	goto out;


	switch (cl.type) {
	case STATUS:
	rv = do_status(&booth_conf, cl.type);
	break;

	case ARBITRATOR:
	case DAEMON:
	case SITE:
	rv = do_server(&booth_conf, cl.type);
	break;

	case CLIENT:
	rv = do_client(&booth_conf);
	break;

	case GEOSTORE:
	rv = do_attr(&booth_conf);
	break;
	}

	out:
	#ifdef LOGGING_LIBQB
	qb_log_fini();
	#endif
	/* Normalize values. 0x100 would be seen as "OK" by waitpid(). */
	return (rv >= 0 && rv < 0x70) ? rv : 1;
	}
	diff --git a/src/transport.c b/src/transport.c
	index f8dfc46..2efb442 100644
	--- a/src/transport.c
	+++ b/src/transport.c
	@@ -1,1132 +1,1132 @@
	/*
	* Copyright (C) 2011 Jiaju Zhang <jjzhang@suse.de>
	* Copyright (C) 2013-2014 Philipp Marek <philipp.marek@linbit.com>
	*
	* This program is free software; you can redistribute it and/or
	* modify it under the terms of the GNU General Public
	* License as published by the Free Software Foundation; either
	* version 2 of the License, or (at your option) any later version.
	*
	* This software is distributed in the hope that it will be useful,
	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	* General Public License for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with this program; if not, write to the Free Software Foundation, Inc.,
	* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
	*/

	#include <string.h>
	#include <stdlib.h>
	#include <unistd.h>
	#include <assert.h>
	#include <errno.h>
	#include <fcntl.h>
	#include <netdb.h> /* getnameinfo */
	#include <poll.h>
	#include <arpa/inet.h>
	#include <asm/types.h>
	#include <linux/rtnetlink.h>
	#include <net/if.h>
	#include <netinet/in.h>
	#include <netinet/tcp.h>
	#include <sys/socket.h> /* getnameinfo */
	#include "b_config.h"
	#include "config.h"
	#include "transport.h"
	#include "attr.h"
	#include "auth.h"
	#include "booth.h"
	#include "inline-fn.h"
	#include "log.h"
	#include "ticket.h"

	#define BOOTH_IPADDR_LEN (sizeof(struct in6_addr))

	#define NETLINK_BUFSIZE 16384
	#define SOCKET_BUFFER_SIZE 160000
	#define FRAME_SIZE_MAX 10000



	struct booth_site *local = NULL;

	/* function to be called when handling booth-group-internal messages;
	* it's expected to return 0 to indicate success, negative integer
	* to indicate silent (or possibly already complained about) error,
	* or positive integer to indicate sender's ID that will then be
	* emitted in the error log message together with the real source
	* address if this is available */
	static int (deliver_fn) (void msg, int msglen);


	static void parse_rtattr(struct rtattr *tb[],
	int max, struct rtattr *rta, int len)
	{
	while (RTA_OK(rta, len)) {
	if (rta->rta_type <= max)
	tb[rta->rta_type] = rta;
	rta = RTA_NEXT(rta,len);
	}
	}

	enum match_type {
	NO_MATCH = 0,
	FUZZY_MATCH,
	EXACT_MATCH,
	};

	static int find_address(struct booth_config *conf_ptr,
	unsigned char ipaddr[BOOTH_IPADDR_LEN],
	int family, int prefixlen, int fuzzy_allowed,
	struct booth_site *me, int address_bits_matched)
	{
	int i;
	struct booth_site *node;
	int bytes, bits_left, mask;
	unsigned char node_bits, ip_bits;
	uint8_t *n_a;
	int matched;
	enum match_type did_match = NO_MATCH;

	assert(conf_ptr != NULL);

	bytes = prefixlen / 8;
	bits_left = prefixlen % 8;
	/* One bit left to check means ignore 7 lowest bits. */
	mask = ~( (1 << (8 - bits_left)) -1);

	for (i = 0; i < conf_ptr->site_count; i++) {
	node = conf_ptr->site + i;
	if (family != node->family)
	continue;
	n_a = node_to_addr_pointer(node);

	for(matched = 0; matched < node->addrlen; matched++)
	if (ipaddr[matched] != n_a[matched])
	break;

	if (matched == node->addrlen) {
	address_bits_matched = matched 8;
	*me = node;
	did_match = EXACT_MATCH;
	break;
	}

	if (!fuzzy_allowed)
	continue;


	/* Check prefix, whole bytes */
	if (matched < bytes)
	continue;
	if (matched * 8 < *address_bits_matched)
	continue;

	node_bits = n_a[bytes];
	ip_bits = ipaddr[bytes];
	if (((node_bits ^ ip_bits) & mask) == 0) {
	/* _At_least_ prefixlen bits matched. */
	if (did_match < EXACT_MATCH) {
	*address_bits_matched = prefixlen;
	*me = node;
	did_match = FUZZY_MATCH;
	}
	}
	}

	return did_match;
	}


	static int _find_myself(struct booth_config *conf_ptr, int family,
	struct booth_site **mep, int fuzzy_allowed)
	{
	int fd;
	struct sockaddr_nl nladdr;
	struct booth_site *me;
	unsigned char ipaddr[BOOTH_IPADDR_LEN];
	static char rcvbuf[NETLINK_BUFSIZE];
	struct {
	struct nlmsghdr nlh;
	struct rtgenmsg g;
	} req;
	int address_bits_matched;


	if (local)
	goto found;


	me = NULL;
	address_bits_matched = 0;
	if (mep)
	*mep = NULL;
	fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
	if (fd < 0) {
	log_error("failed to create netlink socket");
	return 0;
	}

	setsockopt(fd, SOL_SOCKET, SO_RCVBUF, &rcvbuf, sizeof(rcvbuf));

	memset(&nladdr, 0, sizeof(nladdr));
	nladdr.nl_family = AF_NETLINK;

	memset(&req, 0, sizeof(req));
	req.nlh.nlmsg_len = sizeof(req);
	req.nlh.nlmsg_type = RTM_GETADDR;
	req.nlh.nlmsg_flags = NLM_F_ROOT\|NLM_F_MATCH\|NLM_F_REQUEST;
	req.nlh.nlmsg_pid = 0;
	req.nlh.nlmsg_seq = 1;
	req.g.rtgen_family = family;

	if (sendto(fd, (void *)&req, sizeof(req), 0,
	(struct sockaddr*)&nladdr, sizeof(nladdr)) < 0) {
	close(fd);
	log_error("failed to send data to netlink socket");
	return 0;
	}

	while (1) {
	int status;
	struct nlmsghdr *h;
	struct iovec iov = { rcvbuf, sizeof(rcvbuf) };
	struct msghdr msg = {
	(void *)&nladdr, sizeof(nladdr),
	&iov, 1,
	NULL, 0,
	0
	};

	status = recvmsg(fd, &msg, 0);
	if (!status) {
	close(fd);
	log_error("failed to recvmsg from netlink socket");
	return 0;
	}

	h = (struct nlmsghdr *)rcvbuf;
	if (h->nlmsg_type == NLMSG_DONE)
	break;

	if (h->nlmsg_type == NLMSG_ERROR) {
	close(fd);
	log_error("netlink socket recvmsg error");
	return 0;
	}

	while (NLMSG_OK(h, status)) {
	if (h->nlmsg_type == RTM_NEWADDR) {
	struct ifaddrmsg *ifa = NLMSG_DATA(h);
	struct rtattr *tb[IFA_MAX+1];
	int len = h->nlmsg_len
	- NLMSG_LENGTH(sizeof(*ifa));

	memset(tb, 0, sizeof(tb));
	parse_rtattr(tb, IFA_MAX, IFA_RTA(ifa), len);
	memset(ipaddr, 0, BOOTH_IPADDR_LEN);
	/* prefer IFA_LOCAL if it exists, for p-t-p
	* interfaces, otherwise use IFA_ADDRESS */
	if (tb[IFA_LOCAL]) {
	memcpy(ipaddr, RTA_DATA(tb[IFA_LOCAL]),
	BOOTH_IPADDR_LEN);
	} else {
	memcpy(ipaddr, RTA_DATA(tb[IFA_ADDRESS]),
	BOOTH_IPADDR_LEN);
	}

	/* Try to find the exact address or the address with subnet matching.
	* The function find_address will be called for each address received
	* from NLMSG_DATA above.
	* The exact match will be preferred. If no exact match is found,
	* the function find_address will try to return another, most similar
	* address (with the longest possible number of same bytes). */
	if (ifa->ifa_prefixlen > address_bits_matched) {
	find_address(conf_ptr, ipaddr,
	ifa->ifa_family, ifa->ifa_prefixlen,
	fuzzy_allowed, &me, &address_bits_matched);

	if (me) {
	log_debug("found myself at %s (%d bits matched)",
	site_string(me), address_bits_matched);
	}
	}
	/* If the previous NLMSG_DATA calls have already allowed us
	* to find an address with address_bits_matched matching bits,
	* then no other better non-exact address can be found.
	* But we can still try to find an exact match, so let us
	* call the function find_address with disabled searching of
	* similar addresses (fuzzy_allowed == 0) */
	else if (ifa->ifa_prefixlen == address_bits_matched) {
	find_address(conf_ptr, ipaddr,
	ifa->ifa_family, ifa->ifa_prefixlen,
	0 /* fuzzy_allowed */, &me, &address_bits_matched);

	if (me) {
	log_debug("found myself at %s (exact match)",
	site_string(me));
	}
	}
	}
	h = NLMSG_NEXT(h, status);
	}
	}

	close(fd);

	if (!me)
	return 0;

	me->local = 1;
	local = me;
	found:
	if (mep)
	*mep = local;
	return 1;
	}

	int find_myself(struct booth_config conf_ptr, struct booth_site *mep,
	int fuzzy_allowed)
	{
	return _find_myself(conf_ptr, AF_INET6, mep, fuzzy_allowed) \|\|
	_find_myself(conf_ptr, AF_INET, mep, fuzzy_allowed);
	}


	/** Checks the header fields for validity.
	* cf. init_header().
	* For @len_incl_data < 0 the length is not checked.
	* Return <0 if error, else bytes read. */
	int check_boothc_header(struct boothc_header *h, int len_incl_data)
	{
	int l;

	if (h->magic != htonl(BOOTHC_MAGIC)) {
	log_error("magic error %x", ntohl(h->magic));
	return -EINVAL;
	}
	if (h->version != htonl(BOOTHC_VERSION)) {
	log_error("version error %x", ntohl(h->version));
	return -EINVAL;
	}


	l = ntohl(h->length);
	if (l < sizeof(*h)) {
	log_error("length %d out of range", l);
	return -EINVAL;
	}


	if (len_incl_data < 0)
	return 0;

	if (l != len_incl_data) {
	log_error("length error - got %d, wanted %d",
	len_incl_data, l);
	return -EINVAL;
	}

	return len_incl_data;
	}

	static int do_read(int fd, void *buf, size_t count)
	{
	int rv, off = 0;

	while (off < count) {
	rv = read(fd, (char *)buf + off, count - off);
	if (rv == 0)
	return -1;
	if (rv == -1 && errno == EINTR)
	continue;
	if (rv == -1 && errno == EWOULDBLOCK)
	break;
	if (rv == -1)
	return -1;
	off += rv;
	}
	return off;
	}

	static int do_write(int fd, void *buf, size_t count)
	{
	int rv, off = 0;

	retry:
	rv = send(fd, (char *)buf + off, count, MSG_NOSIGNAL);
	if (rv == -1 && errno == EINTR)
	goto retry;
	/* If we cannot write _any_ data, we'd be in an (potential) loop. */
	if (rv <= 0) {
	log_error("send failed: %s (%d)", strerror(errno), errno);
	return rv;
	}

	if (rv != count) {
	count -= rv;
	off += rv;
	goto retry;
	}
	return 0;
	}


	/* Only used for client requests (tcp) */
	int read_client(struct client *req_cl)
	{
	char *msg;
	struct boothc_header *header;
	int rv, fd;
	int len = MAX_MSG_LEN;

	if (!req_cl->msg) {
	msg = malloc(MAX_MSG_LEN);
	if (!msg) {
	log_error("out of memory for client messages");
	return -1;
	}
	req_cl->msg = (void *)msg;
	} else {
	msg = (char *)req_cl->msg;
	}
	header = (struct boothc_header *)msg;

	/* update len if we read enough */
	if (req_cl->offset >= sizeof(*header)) {
	len = min(ntohl(header->length), MAX_MSG_LEN);
	}

	fd = req_cl->fd;
	rv = do_read(fd, msg+req_cl->offset, len-req_cl->offset);
	if (rv < 0) {
	if (errno == ECONNRESET)
	log_debug("client connection reset for fd %d", fd);
	return -1;
	}
	req_cl->offset += rv;

	/* update len if we read enough */
	if (req_cl->offset >= sizeof(*header)) {
	len = min(ntohl(header->length), MAX_MSG_LEN);
	}

	if (req_cl->offset < len) {
	/* client promised to send more */
	return 1;
	}

	if (check_boothc_header(header, len) < 0) {
	return -1;
	}

	return 0;
	}


	/* Only used for client requests (tcp) */
	static void process_connection(int ci)
	{
	struct client *req_cl;
	void *msg = NULL;
	struct boothc_header *header;
	struct boothc_hdr_msg err_reply;
	cmd_result_t errc;
	void (*deadfn) (int ci);

	req_cl = clients + ci;
	switch (read_client(req_cl)) {
	case -1: /* error */
	goto kill;
	case 1: /* more to read */
	return;
	case 0:
	/* we can process the request now */
	msg = req_cl->msg;
	}

	header = (struct boothc_header *)msg;
	if (check_auth(NULL, msg, ntohl(header->length))) {
	errc = RLT_AUTH;
	goto send_err;
	}

	/* For CMD_GRANT and CMD_REVOKE:
	* Don't close connection immediately, but send
	* result a second later? */
	switch (ntohl(header->cmd)) {
	case CMD_LIST:
	ticket_answer_list(req_cl->fd);
	goto kill;
	case CMD_PEERS:
	list_peers(req_cl->fd);
	goto kill;

	case CMD_GRANT:
	case CMD_REVOKE:
	if (process_client_request(req_cl, msg) == 1)
	goto kill; /* request processed definitely, close connection */
	else
	return;

	case ATTR_LIST:
	case ATTR_GET:
	case ATTR_SET:
	case ATTR_DEL:
	if (process_attr_request(req_cl, msg) == 1)
	goto kill; /* request processed definitely, close connection */
	else
	return;

	default:
	log_error("connection %d cmd %x unknown",
	ci, ntohl(header->cmd));
	errc = RLT_INVALID_ARG;
	goto send_err;
	}

	assert(0);
	return;

	send_err:
	init_header(&err_reply.header, CL_RESULT, 0, 0, errc, 0, sizeof(err_reply));
	send_client_msg(req_cl->fd, &err_reply);

	kill:
	deadfn = req_cl->deadfn;
	if(deadfn) {
	deadfn(ci);
	}
	return;
	}


	static void process_tcp_listener(int ci)
	{
	int fd, i, flags, one = 1;
	socklen_t addrlen = sizeof(struct sockaddr);
	struct sockaddr addr;

	fd = accept(clients[ci].fd, &addr, &addrlen);
	if (fd < 0) {
	log_error("process_tcp_listener: accept error %d %d",
	fd, errno);
	return;
	}
	setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, (char *)&one, sizeof(one));

	flags = fcntl(fd, F_GETFL, 0);
	fcntl(fd, F_SETFL, flags \| O_NONBLOCK);

	i = client_add(fd, clients[ci].transport,
	process_connection, NULL);

	log_debug("client connection %d fd %d", i, fd);
	}

	int setup_tcp_listener(int test_only)
	{
	int s, rv;
	int one = 1;

	s = socket(local->family, SOCK_STREAM, 0);
	if (s == -1) {
	log_error("failed to create tcp socket %s", strerror(errno));
	return s;
	}

	rv = setsockopt(s, SOL_SOCKET, SO_REUSEADDR, (char *)&one, sizeof(one));
	if (rv == -1) {
	close(s);
	log_error("failed to set the SO_REUSEADDR option");
	return rv;
	}

	rv = bind(s, &local->sa6, local->saddrlen);
	if (test_only) {
	rv = (rv == -1) ? errno : 0;
	close(s);
	return rv;
	}

	if (rv == -1) {
	close(s);
	log_error("failed to bind socket %s", strerror(errno));
	return rv;
	}

	rv = listen(s, 5);
	if (rv == -1) {
	close(s);
	log_error("failed to listen on socket %s", strerror(errno));
	return rv;
	}

	return s;
	}

	static int booth_tcp_init(void * unused __attribute__((unused)))
	{
	int rv;

	if (get_local_id() < 0)
	return -1;

	rv = setup_tcp_listener(0);
	if (rv < 0)
	return rv;

	client_add(rv, booth_transport + TCP,
	process_tcp_listener, NULL);

	return 0;
	}

	static int connect_nonb(int sockfd, const struct sockaddr *saptr,
	socklen_t salen, int sec)
	{
	int flags, n, error;
	socklen_t len;
	fd_set rset, wset;
	struct timeval tval;

	flags = fcntl(sockfd, F_GETFL, 0);
	fcntl(sockfd, F_SETFL, flags \| O_NONBLOCK);

	error = 0;
	if ( (n = connect(sockfd, saptr, salen)) < 0)
	if (errno != EINPROGRESS)
	return -1;

	if (n == 0)
	goto done; /* connect completed immediately */

	FD_ZERO(&rset);
	FD_SET(sockfd, &rset);
	wset = rset;
	tval.tv_sec = sec;
	tval.tv_usec = 0;

	if ((n = select(sockfd + 1, &rset, &wset, NULL,
	sec ? &tval : NULL)) == 0) {
	/* leave outside function to close */
	/* timeout */
	/* close(sockfd); */
	errno = ETIMEDOUT;
	return -1;
	}

	if (FD_ISSET(sockfd, &rset) \|\| FD_ISSET(sockfd, &wset)) {
	len = sizeof(error);
	if (getsockopt(sockfd, SOL_SOCKET, SO_ERROR, &error, &len) < 0)
	return -1; /* Solaris pending error */
	} else {
	log_error("select error: sockfd not set");
	return -1;
	}

	done:
	fcntl(sockfd, F_SETFL, flags); /* restore file status flags */

	if (error) {
	/* leave outside function to close */
	/* close(sockfd); */
	errno = error;
	return -1;
	}

	return 0;
	}

	int booth_tcp_open(struct booth_site *to)
	{
	int s, rv;

	if (to->tcp_fd >= STDERR_FILENO)
	goto found;

	s = socket(to->family, SOCK_STREAM, 0);
	if (s == -1) {
	log_error("cannot create socket of family %d", to->family);
	return -1;
	}


	rv = connect_nonb(s, (struct sockaddr *)&to->sa6, to->saddrlen, 10);
	if (rv == -1) {
	if( errno == ETIMEDOUT)
	log_error("connect to %s got a timeout", site_string(to));
	else
	log_error("connect to %s got an error: %s", site_string(to),
	strerror(errno));
	goto error;
	}

	to->tcp_fd = s;

	found:
	return 1;

	error:
	if (s >= 0)
	close(s);
	return -1;
	}

	int booth_tcp_send(struct booth_site to, void buf, int len)
	{
	int rv;

	rv = add_hmac(buf, len);
	if (!rv)
	rv = do_write(to->tcp_fd, buf, len);

	return rv;
	}

	static int booth_tcp_recv(struct booth_site from, void buf, int len)
	{
	int got;
	/* Needs timeouts! */
	got = do_read(from->tcp_fd, buf, len);
	if (got < 0) {
	log_error("read failed (%d): %s", errno, strerror(errno));
	return got;
	}
	return got;
	}

	static int booth_tcp_recv_auth(struct booth_site from, void buf, int len)
	{
	int got, total;
	int payload_len;
	/* Needs timeouts! */

	payload_len = len - sizeof(struct hmac);
	got = booth_tcp_recv(from, buf, payload_len);
	if (got < 0) {
	return got;
	}
	total = got;
	if (is_auth_req()) {
	got = booth_tcp_recv(from, (unsigned char *)buf+payload_len, sizeof(struct hmac));
	if (got != sizeof(struct hmac) \|\| check_auth(from, buf, len)) {
	return -1;
	}
	total += got;
	}
	return total;
	}

	static int booth_tcp_close(struct booth_site *to)
	{
	if (to) {
	if (to->tcp_fd > STDERR_FILENO)
	close(to->tcp_fd);
	to->tcp_fd = -1;
	}
	return 0;
	}

	static int booth_tcp_exit(void)
	{
	return 0;
	}

	static int setup_udp_server(void)
	{
	int rv, fd;
	int one = 1;
	unsigned int recvbuf_size;

	fd = socket(local->family, SOCK_DGRAM, 0);
	if (fd == -1) {
	log_error("failed to create UDP socket %s", strerror(errno));
	goto ex;
	}

	rv = fcntl(fd, F_SETFL, O_NONBLOCK);
	if (rv == -1) {
	log_error("failed to set non-blocking operation "
	"on UDP socket: %s", strerror(errno));
	goto ex;
	}

	rv = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, (char *)&one, sizeof(one));
	if (rv == -1) {
	log_error("failed to set the SO_REUSEADDR option");
	goto ex;
	}

	rv = bind(fd, (struct sockaddr *)&local->sa6, local->saddrlen);

	if (rv == -1) {
	log_error("failed to bind UDP socket to [%s]:%d: %s",
	- site_string(local), booth_conf->port,
	- strerror(errno));
	+ site_string(local), site_port(local),
	+ strerror(errno));
	goto ex;
	}

	recvbuf_size = SOCKET_BUFFER_SIZE;
	rv = setsockopt(fd, SOL_SOCKET, SO_RCVBUF,
	&recvbuf_size, sizeof(recvbuf_size));
	if (rv == -1) {
	log_error("failed to set recvbuf size");
	goto ex;
	}

	local->udp_fd = fd;
	return 0;

	ex:
	if (fd >= 0)
	close(fd);
	return -1;
	}


	/* Receive/process callback for UDP */
	static void process_recv(int ci)
	{
	struct sockaddr_storage sa;
	int rv;
	socklen_t sa_len;
	/* beware, the buffer needs to be large enough to accept
	* a packet */
	char buffer[MAX_MSG_LEN];
	/* Used for unit tests */
	struct boothc_ticket_msg *msg;


	sa_len = sizeof(sa);
	msg = (void*)buffer;
	rv = recvfrom(clients[ci].fd,
	buffer, sizeof(buffer),
	MSG_NOSIGNAL \| MSG_DONTWAIT,
	(struct sockaddr *)&sa, &sa_len);
	if (rv == -1)
	return;

	rv = deliver_fn((void*)msg, rv);
	if (rv > 0) {
	if (getnameinfo((struct sockaddr *)&sa, sa_len,
	buffer, sizeof(buffer), NULL, 0,
	NI_NUMERICHOST) == 0)
	log_error("unknown sender: %08x (real: %s)", rv, buffer);
	else
	log_error("unknown sender: %08x", rv);
	}
	}

	static int booth_udp_init(void *f)
	{
	int rv;

	rv = setup_udp_server();
	if (rv < 0)
	return rv;

	deliver_fn = f;
	client_add(local->udp_fd,
	booth_transport + UDP,
	process_recv, NULL);

	return 0;
	}

	int booth_udp_send(struct booth_site to, void buf, int len)
	{
	int rv;

	to->sent_cnt++;
	rv = sendto(local->udp_fd, buf, len, MSG_NOSIGNAL,
	(struct sockaddr *)&to->sa6, to->saddrlen);
	if (rv == len) {
	rv = 0;
	} else if (rv < 0) {
	to->sent_err_cnt++;
	log_error("Cannot send to %s: %d %s",
	site_string(to),
	errno,
	strerror(errno));
	} else {
	rv = -1;
	to->sent_err_cnt++;
	log_error("Packet sent to %s got truncated",
	site_string(to));
	}

	return rv;
	}

	int booth_udp_send_auth(struct booth_site to, void buf, int len)
	{
	int rv;

	rv = add_hmac(buf, len);
	if (rv < 0)
	return rv;
	return booth_udp_send(to, buf, len);
	}

	static int booth_udp_broadcast_auth(void *buf, int len)
	{
	int i, rv, rvs;
	struct booth_site *site;


	if (!booth_conf \|\| !booth_conf->site_count)
	return -1;

	rv = add_hmac(buf, len);
	if (rv < 0)
	return rv;

	rvs = 0;
	FOREACH_NODE(i, site) {
	if (site != local) {
	rv = booth_udp_send(site, buf, len);
	if (!rvs)
	rvs = rv;
	}
	}

	return rvs;
	}

	static int booth_udp_exit(void)
	{
	return 0;
	}

	/* SCTP transport layer has not been developed yet */
	static int booth_sctp_init(void *f __attribute__((unused)))
	{
	return 0;
	}

	static int booth_sctp_send(struct booth_site * to __attribute__((unused)),
	void *buf __attribute__((unused)),
	int len __attribute__((unused)))
	{
	return 0;
	}

	static int booth_sctp_broadcast(void *buf __attribute__((unused)),
	int len __attribute__((unused)))
	{
	return 0;
	}

	static int return_0_booth_site(struct booth_site *v __attribute((unused)))
	{
	return 0;
	}

	static int return_0(void)
	{
	return 0;
	}
	const struct booth_transport booth_transport[TRANSPORT_ENTRIES] = {
	[TCP] = {
	.name = "TCP",
	.init = booth_tcp_init,
	.open = booth_tcp_open,
	.send = booth_tcp_send,
	.recv = booth_tcp_recv,
	.recv_auth = booth_tcp_recv_auth,
	.close = booth_tcp_close,
	.exit = booth_tcp_exit
	},
	[UDP] = {
	.name = "UDP",
	.init = booth_udp_init,
	.open = return_0_booth_site,
	.send = booth_udp_send,
	.send_auth = booth_udp_send_auth,
	.close = return_0_booth_site,
	.broadcast_auth = booth_udp_broadcast_auth,
	.exit = booth_udp_exit
	},
	[SCTP] = {
	.name = "SCTP",
	.init = booth_sctp_init,
	.open = return_0_booth_site,
	.send = booth_sctp_send,
	.broadcast = booth_sctp_broadcast,
	.exit = return_0,
	}
	};

	/* data + (datalen-sizeof(struct hmac)) points to struct hmac
	* i.e. struct hmac is always tacked on the payload
	*/
	int add_hmac(void *data, int len)
	{
	int rv = 0;
	#if HAVE_LIBGCRYPT \|\| HAVE_LIBMHASH
	int payload_len;
	struct hmac *hp;

	if (!is_auth_req())
	return 0;

	payload_len = len - sizeof(struct hmac);
	hp = (struct hmac )((unsigned char )data + payload_len);
	hp->hid = htonl(BOOTH_HASH);
	memset(hp->hash, 0, BOOTH_MAC_SIZE);
	rv = calc_hmac(data, payload_len, BOOTH_HASH, hp->hash,
	booth_conf->authkey, booth_conf->authkey_len);
	if (rv < 0) {
	log_error("internal error: cannot calculate mac");
	}
	#endif
	return rv;
	}

	#if HAVE_LIBGCRYPT \|\| HAVE_LIBMHASH

	/* TODO: we need some client identification for logging */
	#define peer_string(p) (p ? site_string(p) : "client")

	/* verify the validity of timestamp from the header
	* the timestamp needs to be either greater than the one already
	* recorded for the site or, and this is checked for clients,
	* not to be older than booth_conf->maxtimeskew
	* update the timestamp for the site, if this packet is from a
	* site
	*/
	static int verify_ts(struct booth_site from, void buf, int len)
	{
	struct boothc_header *h;
	struct timeval tv, curr_tv, now;

	if (len < sizeof(*h)) {
	log_error("%s: packet too short", peer_string(from));
	return -1;
	}

	h = (struct boothc_header *)buf;
	tv.tv_sec = ntohl(h->secs);
	tv.tv_usec = ntohl(h->usecs);
	if (from) {
	curr_tv.tv_sec = from->last_secs;
	curr_tv.tv_usec = from->last_usecs;
	if (timercmp(&tv, &curr_tv, >))
	goto accept;
	log_warn("%s: packet timestamp older than previous one",
	site_string(from));
	}

	gettimeofday(&now, NULL);
	now.tv_sec -= booth_conf->maxtimeskew;
	if (timercmp(&tv, &now, >))
	goto accept;
	log_error("%s: packet timestamp older than %d seconds",
	peer_string(from), booth_conf->maxtimeskew);
	return -1;

	accept:
	if (from) {
	from->last_secs = tv.tv_sec;
	from->last_usecs = tv.tv_usec;
	}
	return 0;
	}
	#endif

	int check_auth(struct booth_site from, void buf, int len)
	{
	int rv = 0;
	#if HAVE_LIBGCRYPT \|\| HAVE_LIBMHASH
	int payload_len;
	struct hmac *hp;

	if (!is_auth_req())
	return 0;

	payload_len = len - sizeof(struct hmac);
	if (payload_len < 0) {
	log_error("%s: failed to authenticate, packet too short (size:%d)",
	peer_string(from), len);
	return -1;
	}
	hp = (struct hmac )((unsigned char )buf + payload_len);
	rv = verify_hmac(buf, payload_len, ntohl(hp->hid), hp->hash,
	booth_conf->authkey, booth_conf->authkey_len);
	if (!rv) {
	rv = verify_ts(from, buf, len);
	}
	if (rv != 0) {
	log_error("%s: failed to authenticate", peer_string(from));
	}
	#endif
	return rv;
	}

	int send_data(int fd, void *data, int datalen)
	{
	int rv = 0;

	rv = add_hmac(data, datalen);
	if (!rv)
	rv = do_write(fd, data, datalen);

	return rv;
	}

	int send_header_plus(int fd, struct boothc_hdr_msg msg, void data, int len)
	{
	int rv;

	rv = send_data(fd, msg, sendmsglen(msg)-len);

	if (rv >= 0 && len)
	rv = do_write(fd, data, len);

	return rv;
	}

	/* UDP message receiver (see also deliver_fn declaration's comment) */
	int message_recv(struct booth_config conf_ptr, void msg, int msglen)
	{
	uint32_t from;
	struct boothc_header *header;
	struct booth_site *source;

	header = (struct boothc_header *)msg;

	from = ntohl(header->from);
	if (!find_site_by_id(conf_ptr, from, &source)) {
	/* caller knows the actual source address, pass
	the (assuredly) positive number and let it report */
	from = from ? from : ~from; /* avoid 0 (success) */
	return from & (~0U >> 1); /* avoid negative (error code} */
	}

	time(&source->last_recv);
	source->recv_cnt++;

	if (check_boothc_header(header, msglen) < 0) {
	log_error("message from %s receive error", site_string(source));
	source->recv_err_cnt++;
	return -1;
	}

	if (check_auth(source, msg, msglen)) {
	log_error("%s failed to authenticate", site_string(source));
	source->sec_cnt++;
	return -1;
	}

	if (ntohl(header->opts) & BOOTH_OPT_ATTR) {
	/* not used, clients send/retrieve attributes directly
	* from sites
	*/
	return attr_recv(msg, source);
	} else {
	return ticket_recv(conf_ptr, msg, source);
	}
	}

File Metadata

Mime Type: text/x-diff
Expires: Wed, Feb 26, 12:17 PM (19 h, 47 m)
Storage Engine: blob
Storage Format: Raw Data
Storage Handle: 1465340
Default Alt Text: (70 KB)

No OneTemporaryActions

View Options

File Metadata

Event Timeline

No OneTemporary
Actions