1
0
mirror of https://github.com/openbsd/src.git synced 2024-12-22 07:27:59 -08:00
openbsd-src/usr.sbin/relayd/relayd.c
tb 92388deed9 relayd: add support for client certificates
This feature has been requested many times over the years. Various patches
were provided by Asherah Connor, Rivo Nurges, Markus Läll and maybe others.
These patches always stalled for various reasons.

From Sören Tempel, mostly based on Asherah's latest patch.

ok florian tb
2024-10-28 19:56:18 +00:00

1894 lines
40 KiB
C

/* $OpenBSD: relayd.c,v 1.192 2024/10/28 19:56:18 tb Exp $ */
/*
* Copyright (c) 2007 - 2016 Reyk Floeter <reyk@openbsd.org>
* Copyright (c) 2006 Pierre-Yves Ritschard <pyr@openbsd.org>
*
* Permission to use, copy, modify, and distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
* ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
* ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
*/
#include <sys/types.h>
#include <sys/queue.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <sys/wait.h>
#include <sys/resource.h>
#include <netinet/in.h>
#include <arpa/inet.h>
#include <signal.h>
#include <string.h>
#include <stdio.h>
#include <stdlib.h>
#include <fcntl.h>
#include <getopt.h>
#include <fnmatch.h>
#include <syslog.h>
#include <err.h>
#include <errno.h>
#include <event.h>
#include <unistd.h>
#include <ctype.h>
#include <pwd.h>
#include <sha1.h>
#include <md5.h>
#include <tls.h>
#include "relayd.h"
#define MAXIMUM(a, b) (((a) > (b)) ? (a) : (b))
__dead void usage(void);
int parent_configure(struct relayd *);
void parent_configure_done(struct relayd *);
void parent_reload(struct relayd *, u_int, const char *);
void parent_sig_handler(int, short, void *);
void parent_shutdown(struct relayd *);
int parent_dispatch_pfe(int, struct privsep_proc *, struct imsg *);
int parent_dispatch_hce(int, struct privsep_proc *, struct imsg *);
int parent_dispatch_relay(int, struct privsep_proc *,
struct imsg *);
int parent_dispatch_ca(int, struct privsep_proc *,
struct imsg *);
int bindany(struct ctl_bindany *);
void parent_tls_ticket_rekey(int, short, void *);
struct relayd *relayd_env;
static struct privsep_proc procs[] = {
{ "pfe", PROC_PFE, parent_dispatch_pfe, pfe },
{ "hce", PROC_HCE, parent_dispatch_hce, hce },
{ "relay", PROC_RELAY, parent_dispatch_relay, relay },
{ "ca", PROC_CA, parent_dispatch_ca, ca }
};
enum privsep_procid privsep_process;
void
parent_sig_handler(int sig, short event, void *arg)
{
struct privsep *ps = arg;
switch (sig) {
case SIGTERM:
case SIGINT:
parent_shutdown(ps->ps_env);
break;
case SIGHUP:
log_info("%s: reload requested with SIGHUP", __func__);
/*
* This is safe because libevent uses async signal handlers
* that run in the event loop and not in signal context.
*/
parent_reload(ps->ps_env, CONFIG_RELOAD, NULL);
break;
case SIGPIPE:
case SIGUSR1:
/* ignore */
break;
default:
fatalx("unexpected signal");
}
}
__dead void
usage(void)
{
extern char *__progname;
fprintf(stderr, "usage: %s [-dnv] [-D macro=value] [-f file]\n",
__progname);
exit(1);
}
int
main(int argc, char *argv[])
{
int c;
int debug = 0, verbose = 0;
u_int32_t opts = 0;
struct relayd *env;
struct privsep *ps;
const char *conffile = CONF_FILE;
enum privsep_procid proc_id = PROC_PARENT;
int proc_instance = 0;
const char *errp, *title = NULL;
int argc0 = argc;
while ((c = getopt(argc, argv, "dD:nI:P:f:v")) != -1) {
switch (c) {
case 'd':
debug = 2;
break;
case 'D':
if (cmdline_symset(optarg) < 0)
log_warnx("could not parse macro definition %s",
optarg);
break;
case 'n':
debug = 2;
opts |= RELAYD_OPT_NOACTION;
break;
case 'f':
conffile = optarg;
break;
case 'v':
verbose++;
opts |= RELAYD_OPT_VERBOSE;
break;
case 'P':
title = optarg;
proc_id = proc_getid(procs, nitems(procs), title);
if (proc_id == PROC_MAX)
fatalx("invalid process name");
break;
case 'I':
proc_instance = strtonum(optarg, 0,
PROC_MAX_INSTANCES, &errp);
if (errp)
fatalx("invalid process instance");
break;
default:
usage();
}
}
/* log to stderr until daemonized */
log_init(debug ? debug : 1, LOG_DAEMON);
argc -= optind;
if (argc > 0)
usage();
if ((env = calloc(1, sizeof(*env))) == NULL ||
(ps = calloc(1, sizeof(*ps))) == NULL)
exit(1);
relayd_env = env;
env->sc_ps = ps;
ps->ps_env = env;
TAILQ_INIT(&ps->ps_rcsocks);
env->sc_conffile = conffile;
env->sc_conf.opts = opts;
TAILQ_INIT(&env->sc_hosts);
TAILQ_INIT(&env->sc_sessions);
env->sc_rtable = getrtable();
/* initialize the TLS session id to a random key for all relay procs */
arc4random_buf(env->sc_conf.tls_sid, sizeof(env->sc_conf.tls_sid));
if (parse_config(env->sc_conffile, env) == -1)
exit(1);
if (debug)
env->sc_conf.opts |= RELAYD_OPT_LOGUPDATE;
if (geteuid())
errx(1, "need root privileges");
if ((ps->ps_pw = getpwnam(RELAYD_USER)) == NULL)
errx(1, "unknown user %s", RELAYD_USER);
log_init(debug, LOG_DAEMON);
log_setverbose(verbose);
if (env->sc_conf.opts & RELAYD_OPT_NOACTION)
ps->ps_noaction = 1;
ps->ps_instances[PROC_RELAY] = env->sc_conf.prefork_relay;
ps->ps_instances[PROC_CA] = env->sc_conf.prefork_relay;
ps->ps_instance = proc_instance;
if (title != NULL)
ps->ps_title[proc_id] = title;
/* only the parent returns */
proc_init(ps, procs, nitems(procs), debug, argc0, argv, proc_id);
log_procinit("parent");
if (ps->ps_noaction == 0)
log_info("startup");
if (unveil("/", "rx") == -1)
err(1, "unveil /");
if (unveil(NULL, NULL) == -1)
err(1, "unveil");
event_init();
signal_set(&ps->ps_evsigint, SIGINT, parent_sig_handler, ps);
signal_set(&ps->ps_evsigterm, SIGTERM, parent_sig_handler, ps);
signal_set(&ps->ps_evsighup, SIGHUP, parent_sig_handler, ps);
signal_set(&ps->ps_evsigpipe, SIGPIPE, parent_sig_handler, ps);
signal_set(&ps->ps_evsigusr1, SIGUSR1, parent_sig_handler, ps);
signal_add(&ps->ps_evsigint, NULL);
signal_add(&ps->ps_evsigterm, NULL);
signal_add(&ps->ps_evsighup, NULL);
signal_add(&ps->ps_evsigpipe, NULL);
signal_add(&ps->ps_evsigusr1, NULL);
proc_connect(ps);
relay_http(NULL);
if (load_config(env->sc_conffile, env) == -1) {
proc_kill(env->sc_ps);
exit(1);
}
if (env->sc_conf.opts & RELAYD_OPT_NOACTION) {
fprintf(stderr, "configuration OK\n");
proc_kill(env->sc_ps);
exit(0);
}
/* rekey the TLS tickets before pushing the config */
parent_tls_ticket_rekey(0, 0, env);
if (parent_configure(env) == -1)
fatalx("configuration failed");
init_routes(env);
event_dispatch();
parent_shutdown(env);
/* NOTREACHED */
return (0);
}
int
parent_configure(struct relayd *env)
{
struct table *tb;
struct rdr *rdr;
struct router *rt;
struct protocol *proto;
struct relay *rlay;
int id;
int ret = -1;
TAILQ_FOREACH(tb, env->sc_tables, entry)
config_settable(env, tb);
TAILQ_FOREACH(rdr, env->sc_rdrs, entry)
config_setrdr(env, rdr);
TAILQ_FOREACH(rt, env->sc_rts, rt_entry)
config_setrt(env, rt);
TAILQ_FOREACH(proto, env->sc_protos, entry)
config_setproto(env, proto);
TAILQ_FOREACH(proto, env->sc_protos, entry)
config_setrule(env, proto);
TAILQ_FOREACH(rlay, env->sc_relays, rl_entry) {
/* Check for TLS Inspection */
if ((rlay->rl_conf.flags & (F_TLS|F_TLSCLIENT)) ==
(F_TLS|F_TLSCLIENT) && rlay->rl_tls_cacert_fd != -1)
rlay->rl_conf.flags |= F_TLSINSPECT;
config_setrelay(env, rlay);
}
/* HCE, PFE, CA and the relays need to reload their config. */
env->sc_reload = 2 + (2 * env->sc_conf.prefork_relay);
for (id = 0; id < PROC_MAX; id++) {
if (id == privsep_process)
continue;
proc_compose_imsg(env->sc_ps, id, -1, IMSG_CFG_DONE, -1,
-1, &env->sc_conf, sizeof(env->sc_conf));
}
ret = 0;
config_purge(env, CONFIG_ALL & ~CONFIG_RELAYS);
return (ret);
}
void
parent_reload(struct relayd *env, u_int reset, const char *filename)
{
if (env->sc_reload) {
log_debug("%s: already in progress: %d pending",
__func__, env->sc_reload);
return;
}
/* Switch back to the default config file */
if (filename == NULL || *filename == '\0')
filename = env->sc_conffile;
log_debug("%s: level %d config file %s", __func__, reset, filename);
config_purge(env, CONFIG_ALL);
if (reset == CONFIG_RELOAD) {
if (load_config(filename, env) == -1) {
log_debug("%s: failed to load config file %s",
__func__, filename);
}
config_setreset(env, CONFIG_ALL);
if (parent_configure(env) == -1) {
log_debug("%s: failed to commit config from %s",
__func__, filename);
}
} else
config_setreset(env, reset);
}
void
parent_configure_done(struct relayd *env)
{
int id;
if (env->sc_reload == 0) {
log_warnx("%s: configuration already finished", __func__);
return;
}
env->sc_reload--;
if (env->sc_reload == 0) {
for (id = 0; id < PROC_MAX; id++) {
if (id == privsep_process)
continue;
proc_compose(env->sc_ps, id, IMSG_CTL_START, NULL, 0);
}
}
}
void
parent_shutdown(struct relayd *env)
{
config_purge(env, CONFIG_ALL);
proc_kill(env->sc_ps);
control_cleanup(&env->sc_ps->ps_csock);
carp_demote_shutdown();
free(env->sc_ps);
free(env);
log_info("parent terminating, pid %d", getpid());
exit(0);
}
int
parent_dispatch_pfe(int fd, struct privsep_proc *p, struct imsg *imsg)
{
struct privsep *ps = p->p_ps;
struct relayd *env = ps->ps_env;
struct ctl_demote demote;
struct ctl_netroute crt;
u_int v;
char *str = NULL;
switch (imsg->hdr.type) {
case IMSG_DEMOTE:
IMSG_SIZE_CHECK(imsg, &demote);
memcpy(&demote, imsg->data, sizeof(demote));
carp_demote_set(demote.group, demote.level);
break;
case IMSG_RTMSG:
IMSG_SIZE_CHECK(imsg, &crt);
memcpy(&crt, imsg->data, sizeof(crt));
pfe_route(env, &crt);
break;
case IMSG_CTL_RESET:
IMSG_SIZE_CHECK(imsg, &v);
memcpy(&v, imsg->data, sizeof(v));
parent_reload(env, v, NULL);
break;
case IMSG_CTL_RELOAD:
if (IMSG_DATA_SIZE(imsg) > 0)
str = get_string(imsg->data, IMSG_DATA_SIZE(imsg));
parent_reload(env, CONFIG_RELOAD, str);
free(str);
break;
case IMSG_CTL_SHUTDOWN:
parent_shutdown(env);
break;
case IMSG_CFG_DONE:
parent_configure_done(env);
break;
case IMSG_AGENTXSOCK:
agentx_setsock(env, p->p_id);
break;
default:
return (-1);
}
return (0);
}
int
parent_dispatch_hce(int fd, struct privsep_proc *p, struct imsg *imsg)
{
struct privsep *ps = p->p_ps;
struct relayd *env = ps->ps_env;
struct ctl_script scr;
switch (imsg->hdr.type) {
case IMSG_SCRIPT:
IMSG_SIZE_CHECK(imsg, &scr);
bcopy(imsg->data, &scr, sizeof(scr));
scr.retval = script_exec(env, &scr);
proc_compose(ps, PROC_HCE, IMSG_SCRIPT, &scr, sizeof(scr));
break;
case IMSG_CFG_DONE:
parent_configure_done(env);
break;
default:
return (-1);
}
return (0);
}
int
parent_dispatch_relay(int fd, struct privsep_proc *p, struct imsg *imsg)
{
struct privsep *ps = p->p_ps;
struct relayd *env = ps->ps_env;
struct ctl_bindany bnd;
int s;
switch (imsg->hdr.type) {
case IMSG_BINDANY:
IMSG_SIZE_CHECK(imsg, &bnd);
bcopy(imsg->data, &bnd, sizeof(bnd));
if (bnd.bnd_proc > env->sc_conf.prefork_relay)
fatalx("%s: invalid relay proc", __func__);
switch (bnd.bnd_proto) {
case IPPROTO_TCP:
case IPPROTO_UDP:
break;
default:
fatalx("%s: requested socket "
"for invalid protocol", __func__);
/* NOTREACHED */
}
s = bindany(&bnd);
proc_compose_imsg(ps, PROC_RELAY, bnd.bnd_proc,
IMSG_BINDANY, -1, s, &bnd.bnd_id, sizeof(bnd.bnd_id));
break;
case IMSG_CFG_DONE:
parent_configure_done(env);
break;
default:
return (-1);
}
return (0);
}
int
parent_dispatch_ca(int fd, struct privsep_proc *p, struct imsg *imsg)
{
struct privsep *ps = p->p_ps;
struct relayd *env = ps->ps_env;
switch (imsg->hdr.type) {
case IMSG_CFG_DONE:
parent_configure_done(env);
break;
default:
return (-1);
}
return (0);
}
void
purge_table(struct relayd *env, struct tablelist *head, struct table *table)
{
struct host *host;
while ((host = TAILQ_FIRST(&table->hosts)) != NULL) {
TAILQ_REMOVE(&table->hosts, host, entry);
TAILQ_REMOVE(&env->sc_hosts, host, globalentry);
if (event_initialized(&host->cte.ev)) {
event_del(&host->cte.ev);
close(host->cte.s);
}
ibuf_free(host->cte.buf);
tls_free(host->cte.tls);
free(host);
}
free(table->sendbuf);
ibuf_free(table->sendbinbuf);
tls_config_free(table->tls_cfg);
if (head != NULL)
TAILQ_REMOVE(head, table, entry);
free(table);
}
void
purge_key(char **key, off_t len)
{
freezero(*key, len);
*key = NULL;
}
void
purge_relay(struct relayd *env, struct relay *rlay)
{
struct rsession *con;
struct relay_table *rlt;
struct relay_cert *cert, *tmpcert;
/* shutdown and remove relay */
if (event_initialized(&rlay->rl_ev))
event_del(&rlay->rl_ev);
close(rlay->rl_s);
TAILQ_REMOVE(env->sc_relays, rlay, rl_entry);
/* cleanup sessions */
while ((con =
SPLAY_ROOT(&rlay->rl_sessions)) != NULL)
relay_close(con, NULL, 0);
/* cleanup relay */
if (rlay->rl_bev != NULL)
bufferevent_free(rlay->rl_bev);
if (rlay->rl_dstbev != NULL)
bufferevent_free(rlay->rl_dstbev);
purge_key(&rlay->rl_tls_cakey, rlay->rl_conf.tls_cakey_len);
if (rlay->rl_tls_pkey != NULL) {
EVP_PKEY_free(rlay->rl_tls_pkey);
rlay->rl_tls_pkey = NULL;
}
if (rlay->rl_tls_cacertx509 != NULL) {
X509_free(rlay->rl_tls_cacertx509);
rlay->rl_tls_cacertx509 = NULL;
}
if (rlay->rl_tls_capkey != NULL) {
EVP_PKEY_free(rlay->rl_tls_capkey);
rlay->rl_tls_capkey = NULL;
}
tls_free(rlay->rl_tls_ctx);
tls_config_free(rlay->rl_tls_cfg);
tls_config_free(rlay->rl_tls_client_cfg);
while ((rlt = TAILQ_FIRST(&rlay->rl_tables))) {
TAILQ_REMOVE(&rlay->rl_tables, rlt, rlt_entry);
free(rlt);
}
TAILQ_FOREACH_SAFE(cert, env->sc_certs, cert_entry, tmpcert) {
if (rlay->rl_conf.id != cert->cert_relayid)
continue;
if (cert->cert_fd != -1)
close(cert->cert_fd);
if (cert->cert_key_fd != -1)
close(cert->cert_key_fd);
if (cert->cert_ocsp_fd != -1)
close(cert->cert_ocsp_fd);
if (cert->cert_pkey != NULL)
EVP_PKEY_free(cert->cert_pkey);
TAILQ_REMOVE(env->sc_certs, cert, cert_entry);
free(cert);
}
free(rlay);
}
struct kv *
kv_add(struct kvtree *keys, char *key, char *value, int unique)
{
struct kv *kv, *oldkv;
if (key == NULL)
return (NULL);
if ((kv = calloc(1, sizeof(*kv))) == NULL)
return (NULL);
if ((kv->kv_key = strdup(key)) == NULL)
goto fail;
if (value != NULL &&
(kv->kv_value = strdup(value)) == NULL)
goto fail;
TAILQ_INIT(&kv->kv_children);
if ((oldkv = RB_INSERT(kvtree, keys, kv)) != NULL) {
/*
* return error if the key should occur only once,
* or add it to a list attached to the key's node.
*/
if (unique)
goto fail;
TAILQ_INSERT_TAIL(&oldkv->kv_children, kv, kv_entry);
kv->kv_parent = oldkv;
}
return (kv);
fail:
free(kv->kv_key);
free(kv->kv_value);
free(kv);
return (NULL);
}
int
kv_set(struct kv *kv, char *fmt, ...)
{
va_list ap;
char *value = NULL;
struct kv *ckv;
int ret;
va_start(ap, fmt);
ret = vasprintf(&value, fmt, ap);
va_end(ap);
if (ret == -1)
return (-1);
/* Remove all children */
while ((ckv = TAILQ_FIRST(&kv->kv_children)) != NULL) {
TAILQ_REMOVE(&kv->kv_children, ckv, kv_entry);
kv_free(ckv);
free(ckv);
}
/* Set the new value */
free(kv->kv_value);
kv->kv_value = value;
return (0);
}
int
kv_setkey(struct kv *kv, char *fmt, ...)
{
va_list ap;
char *key = NULL;
int ret;
va_start(ap, fmt);
ret = vasprintf(&key, fmt, ap);
va_end(ap);
if (ret == -1)
return (-1);
free(kv->kv_key);
kv->kv_key = key;
return (0);
}
void
kv_delete(struct kvtree *keys, struct kv *kv)
{
struct kv *ckv;
RB_REMOVE(kvtree, keys, kv);
/* Remove all children */
while ((ckv = TAILQ_FIRST(&kv->kv_children)) != NULL) {
TAILQ_REMOVE(&kv->kv_children, ckv, kv_entry);
kv_free(ckv);
free(ckv);
}
kv_free(kv);
free(kv);
}
struct kv *
kv_extend(struct kvtree *keys, struct kv *kv, char *value)
{
char *newvalue;
if (kv == NULL) {
return (NULL);
} else if (kv->kv_value != NULL) {
if (asprintf(&newvalue, "%s%s", kv->kv_value, value) == -1)
return (NULL);
free(kv->kv_value);
kv->kv_value = newvalue;
} else if ((kv->kv_value = strdup(value)) == NULL)
return (NULL);
return (kv);
}
void
kv_purge(struct kvtree *keys)
{
struct kv *kv;
while ((kv = RB_MIN(kvtree, keys)) != NULL)
kv_delete(keys, kv);
}
void
kv_free(struct kv *kv)
{
/*
* This function does not clear memory referenced by
* kv_children or stuff on the tailqs. Use kv_delete() instead.
*/
free(kv->kv_key);
free(kv->kv_value);
memset(kv, 0, sizeof(*kv));
}
struct kv *
kv_inherit(struct kv *dst, struct kv *src)
{
memset(dst, 0, sizeof(*dst));
memcpy(dst, src, sizeof(*dst));
TAILQ_INIT(&dst->kv_children);
if (src->kv_key != NULL) {
if ((dst->kv_key = strdup(src->kv_key)) == NULL) {
kv_free(dst);
return (NULL);
}
}
if (src->kv_value != NULL) {
if ((dst->kv_value = strdup(src->kv_value)) == NULL) {
kv_free(dst);
return (NULL);
}
}
if (src->kv_match != NULL)
dst->kv_match = src->kv_match;
if (src->kv_matchtree != NULL)
dst->kv_matchtree = src->kv_matchtree;
return (dst);
}
int
kv_log(struct rsession *con, struct kv *kv, u_int16_t labelid,
enum direction dir)
{
char *msg;
if (con->se_log == NULL)
return (0);
if (asprintf(&msg, " %s%s%s%s%s%s%s",
dir == RELAY_DIR_REQUEST ? "[" : "{",
labelid == 0 ? "" : label_id2name(labelid),
labelid == 0 ? "" : ", ",
kv->kv_key == NULL ? "(unknown)" : kv->kv_key,
kv->kv_value == NULL ? "" : ": ",
kv->kv_value == NULL ? "" : kv->kv_value,
dir == RELAY_DIR_REQUEST ? "]" : "}") == -1)
return (-1);
if (evbuffer_add(con->se_log, msg, strlen(msg)) == -1) {
free(msg);
return (-1);
}
free(msg);
con->se_haslog = 1;
return (0);
}
struct kv *
kv_find(struct kvtree *keys, struct kv *kv)
{
struct kv *match;
const char *key;
if (kv->kv_flags & KV_FLAG_GLOBBING) {
/* Test header key using shell globbing rules */
key = kv->kv_key == NULL ? "" : kv->kv_key;
RB_FOREACH(match, kvtree, keys) {
if (fnmatch(key, match->kv_key, FNM_CASEFOLD) == 0)
break;
}
} else {
/* Fast tree-based lookup only works without globbing */
match = RB_FIND(kvtree, keys, kv);
}
return (match);
}
struct kv *
kv_find_value(struct kvtree *keys, char *key, const char *value,
const char *delim)
{
struct kv *match, kv;
char *val = NULL, *next, *ptr;
size_t len;
kv.kv_key = key;
if ((match = RB_FIND(kvtree, keys, &kv)) == NULL)
return (NULL);
if (match->kv_value == NULL)
return (NULL);
if (delim == NULL) {
if (strcasecmp(match->kv_value, value) == 0)
goto done;
} else {
if ((val = strdup(match->kv_value)) == NULL)
return (NULL);
for (next = ptr = val; ptr != NULL;
ptr = strsep(&next, delim)) {
/* strip whitespace */
ptr += strspn(ptr, " \t");
len = strcspn(ptr, " \t");
if (strncasecmp(ptr, value, len) == 0)
goto done;
}
}
/* not matched */
match = NULL;
done:
#ifdef DEBUG
if (match != NULL)
DPRINTF("%s: matched %s: %s", __func__, key, value);
#endif
free(val);
return (match);
}
int
kv_cmp(struct kv *a, struct kv *b)
{
return (strcasecmp(a->kv_key, b->kv_key));
}
RB_GENERATE(kvtree, kv, kv_node, kv_cmp);
int
rule_add(struct protocol *proto, struct relay_rule *rule, const char *rulefile)
{
struct relay_rule *r = NULL;
struct kv *kv = NULL;
FILE *fp = NULL;
char buf[BUFSIZ];
int ret = -1;
u_int i;
for (i = 0; i < KEY_TYPE_MAX; i++) {
kv = &rule->rule_kv[i];
if (kv->kv_type != i)
continue;
switch (kv->kv_option) {
case KEY_OPTION_LOG:
/* log action needs a key or a file to be specified */
if (kv->kv_key == NULL && rulefile == NULL &&
(kv->kv_key = strdup("*")) == NULL)
goto fail;
break;
default:
break;
}
switch (kv->kv_type) {
case KEY_TYPE_QUERY:
case KEY_TYPE_PATH:
case KEY_TYPE_URL:
if (rule->rule_dir != RELAY_DIR_REQUEST)
goto fail;
break;
default:
break;
}
if (kv->kv_value != NULL && strchr(kv->kv_value, '$') != NULL)
kv->kv_flags |= KV_FLAG_MACRO;
if (kv->kv_key != NULL && strpbrk(kv->kv_key, "*?[") != NULL)
kv->kv_flags |= KV_FLAG_GLOBBING;
}
if (rulefile == NULL) {
TAILQ_INSERT_TAIL(&proto->rules, rule, rule_entry);
return (0);
}
if ((fp = fopen(rulefile, "r")) == NULL)
goto fail;
while (fgets(buf, sizeof(buf), fp) != NULL) {
/* strip whitespace and newline characters */
buf[strcspn(buf, "\r\n\t ")] = '\0';
if (!strlen(buf) || buf[0] == '#')
continue;
if ((r = rule_inherit(rule)) == NULL)
goto fail;
for (i = 0; i < KEY_TYPE_MAX; i++) {
kv = &r->rule_kv[i];
if (kv->kv_type != i)
continue;
free(kv->kv_key);
if ((kv->kv_key = strdup(buf)) == NULL) {
rule_free(r);
free(r);
goto fail;
}
}
TAILQ_INSERT_TAIL(&proto->rules, r, rule_entry);
}
ret = 0;
rule_free(rule);
free(rule);
fail:
if (fp != NULL)
fclose(fp);
return (ret);
}
struct relay_rule *
rule_inherit(struct relay_rule *rule)
{
struct relay_rule *r;
u_int i;
struct kv *kv;
if ((r = calloc(1, sizeof(*r))) == NULL)
return (NULL);
memcpy(r, rule, sizeof(*r));
for (i = 0; i < KEY_TYPE_MAX; i++) {
kv = &rule->rule_kv[i];
if (kv->kv_type != i)
continue;
if (kv_inherit(&r->rule_kv[i], kv) == NULL) {
free(r);
return (NULL);
}
}
if (r->rule_label > 0)
label_ref(r->rule_label);
if (r->rule_tag > 0)
tag_ref(r->rule_tag);
if (r->rule_tagged > 0)
tag_ref(r->rule_tagged);
return (r);
}
void
rule_free(struct relay_rule *rule)
{
u_int i;
for (i = 0; i < KEY_TYPE_MAX; i++)
kv_free(&rule->rule_kv[i]);
if (rule->rule_label > 0)
label_unref(rule->rule_label);
if (rule->rule_tag > 0)
tag_unref(rule->rule_tag);
if (rule->rule_tagged > 0)
tag_unref(rule->rule_tagged);
}
void
rule_delete(struct relay_rules *rules, struct relay_rule *rule)
{
TAILQ_REMOVE(rules, rule, rule_entry);
rule_free(rule);
free(rule);
}
void
rule_settable(struct relay_rules *rules, struct relay_table *rlt)
{
struct relay_rule *r;
char pname[TABLE_NAME_SIZE];
if (rlt->rlt_table == NULL || strlcpy(pname, rlt->rlt_table->conf.name,
sizeof(pname)) >= sizeof(pname))
return;
pname[strcspn(pname, ":")] = '\0';
TAILQ_FOREACH(r, rules, rule_entry) {
if (r->rule_tablename[0] &&
strcmp(pname, r->rule_tablename) == 0)
r->rule_table = rlt;
}
}
/*
* Utility functions
*/
struct host *
host_find(struct relayd *env, objid_t id)
{
struct table *table;
struct host *host;
TAILQ_FOREACH(table, env->sc_tables, entry)
TAILQ_FOREACH(host, &table->hosts, entry)
if (host->conf.id == id)
return (host);
return (NULL);
}
struct table *
table_find(struct relayd *env, objid_t id)
{
struct table *table;
TAILQ_FOREACH(table, env->sc_tables, entry)
if (table->conf.id == id)
return (table);
return (NULL);
}
struct rdr *
rdr_find(struct relayd *env, objid_t id)
{
struct rdr *rdr;
TAILQ_FOREACH(rdr, env->sc_rdrs, entry)
if (rdr->conf.id == id)
return (rdr);
return (NULL);
}
struct relay *
relay_find(struct relayd *env, objid_t id)
{
struct relay *rlay;
TAILQ_FOREACH(rlay, env->sc_relays, rl_entry)
if (rlay->rl_conf.id == id)
return (rlay);
return (NULL);
}
struct protocol *
proto_find(struct relayd *env, objid_t id)
{
struct protocol *p;
TAILQ_FOREACH(p, env->sc_protos, entry)
if (p->id == id)
return (p);
return (NULL);
}
struct rsession *
session_find(struct relayd *env, objid_t id)
{
struct relay *rlay;
struct rsession *con;
TAILQ_FOREACH(rlay, env->sc_relays, rl_entry)
SPLAY_FOREACH(con, session_tree, &rlay->rl_sessions)
if (con->se_id == id)
return (con);
return (NULL);
}
struct netroute *
route_find(struct relayd *env, objid_t id)
{
struct netroute *nr;
TAILQ_FOREACH(nr, env->sc_routes, nr_route)
if (nr->nr_conf.id == id)
return (nr);
return (NULL);
}
struct router *
router_find(struct relayd *env, objid_t id)
{
struct router *rt;
TAILQ_FOREACH(rt, env->sc_rts, rt_entry)
if (rt->rt_conf.id == id)
return (rt);
return (NULL);
}
struct host *
host_findbyname(struct relayd *env, const char *name)
{
struct table *table;
struct host *host;
TAILQ_FOREACH(table, env->sc_tables, entry)
TAILQ_FOREACH(host, &table->hosts, entry)
if (strcmp(host->conf.name, name) == 0)
return (host);
return (NULL);
}
struct table *
table_findbyname(struct relayd *env, const char *name)
{
struct table *table;
TAILQ_FOREACH(table, env->sc_tables, entry)
if (strcmp(table->conf.name, name) == 0)
return (table);
return (NULL);
}
struct table *
table_findbyconf(struct relayd *env, struct table *tb)
{
struct table *table;
struct table_config a, b;
bcopy(&tb->conf, &a, sizeof(a));
a.id = a.rdrid = 0;
a.flags &= ~(F_USED|F_BACKUP);
TAILQ_FOREACH(table, env->sc_tables, entry) {
bcopy(&table->conf, &b, sizeof(b));
b.id = b.rdrid = 0;
b.flags &= ~(F_USED|F_BACKUP);
/*
* Compare two tables and return the existing table if
* the configuration seems to be the same.
*/
if (bcmp(&a, &b, sizeof(b)) == 0 &&
((tb->sendbuf == NULL && table->sendbuf == NULL) ||
(tb->sendbuf != NULL && table->sendbuf != NULL &&
strcmp(tb->sendbuf, table->sendbuf) == 0)))
return (table);
}
return (NULL);
}
struct rdr *
rdr_findbyname(struct relayd *env, const char *name)
{
struct rdr *rdr;
TAILQ_FOREACH(rdr, env->sc_rdrs, entry)
if (strcmp(rdr->conf.name, name) == 0)
return (rdr);
return (NULL);
}
struct relay *
relay_findbyname(struct relayd *env, const char *name)
{
struct relay *rlay;
TAILQ_FOREACH(rlay, env->sc_relays, rl_entry)
if (strcmp(rlay->rl_conf.name, name) == 0)
return (rlay);
return (NULL);
}
struct relay *
relay_findbyaddr(struct relayd *env, struct relay_config *rc)
{
struct relay *rlay;
TAILQ_FOREACH(rlay, env->sc_relays, rl_entry)
if (bcmp(&rlay->rl_conf.ss, &rc->ss, sizeof(rc->ss)) == 0 &&
rlay->rl_conf.port == rc->port)
return (rlay);
return (NULL);
}
EVP_PKEY *
pkey_find(struct relayd *env, char * hash)
{
struct ca_pkey *pkey;
TAILQ_FOREACH(pkey, env->sc_pkeys, pkey_entry)
if (strcmp(hash, pkey->pkey_hash) == 0)
return (pkey->pkey);
return (NULL);
}
struct ca_pkey *
pkey_add(struct relayd *env, EVP_PKEY *pkey, char *hash)
{
struct ca_pkey *ca_pkey;
if (env->sc_pkeys == NULL)
fatalx("pkeys");
if ((ca_pkey = calloc(1, sizeof(*ca_pkey))) == NULL)
return (NULL);
ca_pkey->pkey = pkey;
if (strlcpy(ca_pkey->pkey_hash, hash, sizeof(ca_pkey->pkey_hash)) >=
sizeof(ca_pkey->pkey_hash)) {
free(ca_pkey);
return (NULL);
}
TAILQ_INSERT_TAIL(env->sc_pkeys, ca_pkey, pkey_entry);
return (ca_pkey);
}
struct relay_cert *
cert_add(struct relayd *env, objid_t id)
{
static objid_t last_cert_id = 0;
struct relay_cert *cert;
if ((cert = calloc(1, sizeof(*cert))) == NULL)
return (NULL);
if (id == 0)
id = ++last_cert_id;
if (id == INT_MAX) {
log_warnx("too many tls keypairs defined");
free(cert);
return (NULL);
}
cert->cert_id = id;
cert->cert_fd = -1;
cert->cert_key_fd = -1;
cert->cert_ocsp_fd = -1;
TAILQ_INSERT_TAIL(env->sc_certs, cert, cert_entry);
return (cert);
}
struct relay_cert *
cert_find(struct relayd *env, objid_t id)
{
struct relay_cert *cert;
TAILQ_FOREACH(cert, env->sc_certs, cert_entry)
if (cert->cert_id == id)
return (cert);
return (NULL);
}
char *
relay_load_fd(int fd, off_t *len)
{
char *buf = NULL;
struct stat st;
off_t size;
ssize_t rv;
int err;
if (fstat(fd, &st) != 0)
goto fail;
size = st.st_size;
if ((buf = calloc(1, size + 1)) == NULL)
goto fail;
if ((rv = pread(fd, buf, size, 0)) != size)
goto fail;
close(fd);
*len = size;
return (buf);
fail:
err = errno;
free(buf);
close(fd);
errno = err;
return (NULL);
}
int
relay_load_certfiles(struct relayd *env, struct relay *rlay, const char *name)
{
char certfile[PATH_MAX];
char hbuf[PATH_MAX];
struct protocol *proto = rlay->rl_proto;
struct relay_cert *cert;
int useport = htons(rlay->rl_conf.port);
int cert_fd = -1, key_fd = -1, ocsp_fd = -1;
if (rlay->rl_conf.flags & F_TLSCLIENT) {
if (strlen(proto->tlsca) && rlay->rl_tls_ca_fd == -1) {
if ((rlay->rl_tls_ca_fd =
open(proto->tlsca, O_RDONLY)) == -1)
return (-1);
log_debug("%s: using ca %s", __func__, proto->tlsca);
}
if (strlen(proto->tlscacert) && rlay->rl_tls_cacert_fd == -1) {
if ((rlay->rl_tls_cacert_fd =
open(proto->tlscacert, O_RDONLY)) == -1)
return (-1);
log_debug("%s: using ca certificate %s", __func__,
proto->tlscacert);
}
if (strlen(proto->tlscakey) && !rlay->rl_conf.tls_cakey_len &&
proto->tlscapass != NULL) {
if ((rlay->rl_tls_cakey =
ssl_load_key(env, proto->tlscakey,
&rlay->rl_conf.tls_cakey_len,
proto->tlscapass)) == NULL)
return (-1);
log_debug("%s: using ca key %s", __func__,
proto->tlscakey);
}
}
if ((rlay->rl_conf.flags & F_TLS) == 0)
return (0);
if (strlen(proto->tlsclientca) && rlay->rl_tls_client_ca_fd == -1) {
if ((rlay->rl_tls_client_ca_fd =
open(proto->tlsclientca, O_RDONLY)) == -1)
return (-1);
log_debug("%s: using client ca %s", __func__,
proto->tlsclientca);
}
if (name == NULL &&
print_host(&rlay->rl_conf.ss, hbuf, sizeof(hbuf)) == NULL)
goto fail;
else if (name != NULL &&
strlcpy(hbuf, name, sizeof(hbuf)) >= sizeof(hbuf))
goto fail;
if (snprintf(certfile, sizeof(certfile),
"/etc/ssl/%s:%u.crt", hbuf, useport) == -1)
goto fail;
if ((cert_fd = open(certfile, O_RDONLY)) == -1) {
if (snprintf(certfile, sizeof(certfile),
"/etc/ssl/%s.crt", hbuf) == -1)
goto fail;
if ((cert_fd = open(certfile, O_RDONLY)) == -1)
goto fail;
useport = 0;
}
log_debug("%s: using certificate %s", __func__, certfile);
if (useport) {
if (snprintf(certfile, sizeof(certfile),
"/etc/ssl/private/%s:%u.key", hbuf, useport) == -1)
goto fail;
} else {
if (snprintf(certfile, sizeof(certfile),
"/etc/ssl/private/%s.key", hbuf) == -1)
goto fail;
}
if ((key_fd = open(certfile, O_RDONLY)) == -1)
goto fail;
log_debug("%s: using private key %s", __func__, certfile);
if (useport) {
if (snprintf(certfile, sizeof(certfile),
"/etc/ssl/%s:%u.ocsp", hbuf, useport) == -1)
goto fail;
} else {
if (snprintf(certfile, sizeof(certfile),
"/etc/ssl/%s.ocsp", hbuf) == -1)
goto fail;
}
if ((ocsp_fd = open(certfile, O_RDONLY)) != -1)
log_debug("%s: using OCSP staple file %s", __func__, certfile);
if ((cert = cert_add(env, 0)) == NULL)
goto fail;
cert->cert_relayid = rlay->rl_conf.id;
cert->cert_fd = cert_fd;
cert->cert_key_fd = key_fd;
cert->cert_ocsp_fd = ocsp_fd;
return (0);
fail:
if (cert_fd != -1)
close(cert_fd);
if (key_fd != -1)
close(key_fd);
if (ocsp_fd != -1)
close(ocsp_fd);
return (-1);
}
void
event_again(struct event *ev, int fd, short event,
void (*fn)(int, short, void *),
struct timeval *start, struct timeval *end, void *arg)
{
struct timeval tv_next, tv_now, tv;
getmonotime(&tv_now);
bcopy(end, &tv_next, sizeof(tv_next));
timersub(&tv_now, start, &tv_now);
timersub(&tv_next, &tv_now, &tv_next);
bzero(&tv, sizeof(tv));
if (timercmp(&tv_next, &tv, >))
bcopy(&tv_next, &tv, sizeof(tv));
event_del(ev);
event_set(ev, fd, event, fn, arg);
event_add(ev, &tv);
}
int
expand_string(char *label, size_t len, const char *srch, const char *repl)
{
char *tmp;
char *p, *q;
if ((tmp = calloc(1, len)) == NULL) {
log_debug("%s: calloc", __func__);
return (-1);
}
p = q = label;
while ((q = strstr(p, srch)) != NULL) {
*q = '\0';
if ((strlcat(tmp, p, len) >= len) ||
(strlcat(tmp, repl, len) >= len)) {
log_debug("%s: string too long", __func__);
free(tmp);
return (-1);
}
q += strlen(srch);
p = q;
}
if (strlcat(tmp, p, len) >= len) {
log_debug("%s: string too long", __func__);
free(tmp);
return (-1);
}
(void)strlcpy(label, tmp, len); /* always fits */
free(tmp);
return (0);
}
void
translate_string(char *str)
{
char *reader;
char *writer;
reader = writer = str;
while (*reader) {
if (*reader == '\\') {
reader++;
switch (*reader) {
case 'n':
*writer++ = '\n';
break;
case 'r':
*writer++ = '\r';
break;
default:
*writer++ = *reader;
}
} else
*writer++ = *reader;
reader++;
}
*writer = '\0';
}
char *
digeststr(enum digest_type type, const u_int8_t *data, size_t len, char *buf)
{
switch (type) {
case DIGEST_SHA1:
return (SHA1Data(data, len, buf));
break;
case DIGEST_MD5:
return (MD5Data(data, len, buf));
break;
default:
break;
}
return (NULL);
}
const char *
canonicalize_host(const char *host, char *name, size_t len)
{
struct sockaddr_in sin4;
struct sockaddr_in6 sin6;
size_t i, j;
size_t plen;
char c;
if (len < 2)
goto fail;
/*
* Canonicalize an IPv4/6 address
*/
if (inet_pton(AF_INET, host, &sin4) == 1)
return (inet_ntop(AF_INET, &sin4, name, len));
if (inet_pton(AF_INET6, host, &sin6) == 1)
return (inet_ntop(AF_INET6, &sin6, name, len));
/*
* Canonicalize a hostname
*/
/* 1. remove repeated dots and convert upper case to lower case */
plen = strlen(host);
bzero(name, len);
for (i = j = 0; i < plen; i++) {
if (j >= (len - 1))
goto fail;
c = tolower((unsigned char)host[i]);
if ((c == '.') && (j == 0 || name[j - 1] == '.'))
continue;
name[j++] = c;
}
/* 2. remove trailing dots */
for (i = j; i > 0; i--) {
if (name[i - 1] != '.')
break;
name[i - 1] = '\0';
j--;
}
if (j <= 0)
goto fail;
return (name);
fail:
errno = EINVAL;
return (NULL);
}
int
parse_url(const char *url, char **protoptr, char **hostptr, char **pathptr)
{
char *p, *proto = NULL, *host = NULL, *path = NULL;
/* return error if it is not a URL */
if ((p = strstr(url, ":/")) == NULL ||
(strcspn(url, ":/") != (size_t)(p - url)))
return (-1);
/* get protocol */
if ((proto = strdup(url)) == NULL)
goto fail;
p = proto + (p - url);
/* get host */
p += strspn(p, ":/");
if (*p == '\0' || (host = strdup(p)) == NULL)
goto fail;
*p = '\0';
/* find and copy path or default to "/" */
if ((p = strchr(host, '/')) == NULL)
p = "/";
if ((path = strdup(p)) == NULL)
goto fail;
/* strip path after host */
host[strcspn(host, "/")] = '\0';
DPRINTF("%s: %s proto %s, host %s, path %s", __func__,
url, proto, host, path);
*protoptr = proto;
*hostptr = host;
*pathptr = path;
return (0);
fail:
free(proto);
free(host);
free(path);
return (-1);
}
int
bindany(struct ctl_bindany *bnd)
{
int s, v;
s = -1;
v = 1;
if (relay_socket_af(&bnd->bnd_ss, bnd->bnd_port) == -1)
goto fail;
if ((s = socket(bnd->bnd_ss.ss_family,
bnd->bnd_proto == IPPROTO_TCP ? SOCK_STREAM : SOCK_DGRAM,
bnd->bnd_proto)) == -1)
goto fail;
if (setsockopt(s, SOL_SOCKET, SO_BINDANY,
&v, sizeof(v)) == -1)
goto fail;
if (bind(s, (struct sockaddr *)&bnd->bnd_ss,
bnd->bnd_ss.ss_len) == -1)
goto fail;
return (s);
fail:
if (s != -1)
close(s);
return (-1);
}
int
map6to4(struct sockaddr_storage *in6)
{
struct sockaddr_storage out4;
struct sockaddr_in *sin4 = (struct sockaddr_in *)&out4;
struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)in6;
bzero(sin4, sizeof(*sin4));
sin4->sin_len = sizeof(*sin4);
sin4->sin_family = AF_INET;
sin4->sin_port = sin6->sin6_port;
bcopy(&sin6->sin6_addr.s6_addr[12], &sin4->sin_addr.s_addr,
sizeof(sin4->sin_addr));
if (sin4->sin_addr.s_addr == INADDR_ANY ||
sin4->sin_addr.s_addr == INADDR_BROADCAST ||
IN_MULTICAST(ntohl(sin4->sin_addr.s_addr)))
return (-1);
bcopy(&out4, in6, sizeof(*in6));
return (0);
}
int
map4to6(struct sockaddr_storage *in4, struct sockaddr_storage *map)
{
struct sockaddr_storage out6;
struct sockaddr_in *sin4 = (struct sockaddr_in *)in4;
struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&out6;
struct sockaddr_in6 *map6 = (struct sockaddr_in6 *)map;
if (sin4->sin_addr.s_addr == INADDR_ANY ||
sin4->sin_addr.s_addr == INADDR_BROADCAST ||
IN_MULTICAST(ntohl(sin4->sin_addr.s_addr)))
return (-1);
bcopy(map6, sin6, sizeof(*sin6));
sin6->sin6_len = sizeof(*sin6);
sin6->sin6_family = AF_INET6;
sin6->sin6_port = sin4->sin_port;
bcopy(&sin4->sin_addr.s_addr, &sin6->sin6_addr.s6_addr[12],
sizeof(sin4->sin_addr));
bcopy(&out6, in4, sizeof(*in4));
return (0);
}
void
socket_rlimit(int maxfd)
{
struct rlimit rl;
if (getrlimit(RLIMIT_NOFILE, &rl) == -1)
fatal("%s: failed to get resource limit", __func__);
log_debug("%s: max open files %llu", __func__, rl.rlim_max);
/*
* Allow the maximum number of open file descriptors for this
* login class (which should be the class "daemon" by default).
*/
if (maxfd == -1)
rl.rlim_cur = rl.rlim_max;
else
rl.rlim_cur = MAXIMUM(rl.rlim_max, (rlim_t)maxfd);
if (setrlimit(RLIMIT_NOFILE, &rl) == -1)
fatal("%s: failed to set resource limit", __func__);
}
char *
get_string(u_int8_t *ptr, size_t len)
{
size_t i;
for (i = 0; i < len; i++)
if (!(isprint((unsigned char)ptr[i]) ||
isspace((unsigned char)ptr[i])))
break;
return strndup(ptr, i);
}
void *
get_data(u_int8_t *ptr, size_t len)
{
u_int8_t *data;
if ((data = malloc(len)) == NULL)
return (NULL);
memcpy(data, ptr, len);
return (data);
}
int
sockaddr_cmp(struct sockaddr *a, struct sockaddr *b, int prefixlen)
{
struct sockaddr_in *a4, *b4;
struct sockaddr_in6 *a6, *b6;
u_int32_t av[4], bv[4], mv[4];
if (a->sa_family == AF_UNSPEC || b->sa_family == AF_UNSPEC)
return (0);
else if (a->sa_family > b->sa_family)
return (1);
else if (a->sa_family < b->sa_family)
return (-1);
if (prefixlen == -1)
memset(&mv, 0xff, sizeof(mv));
switch (a->sa_family) {
case AF_INET:
a4 = (struct sockaddr_in *)a;
b4 = (struct sockaddr_in *)b;
av[0] = a4->sin_addr.s_addr;
bv[0] = b4->sin_addr.s_addr;
if (prefixlen != -1)
mv[0] = prefixlen2mask(prefixlen);
if ((av[0] & mv[0]) > (bv[0] & mv[0]))
return (1);
if ((av[0] & mv[0]) < (bv[0] & mv[0]))
return (-1);
break;
case AF_INET6:
a6 = (struct sockaddr_in6 *)a;
b6 = (struct sockaddr_in6 *)b;
memcpy(&av, &a6->sin6_addr.s6_addr, 16);
memcpy(&bv, &b6->sin6_addr.s6_addr, 16);
if (prefixlen != -1)
prefixlen2mask6(prefixlen, mv);
if ((av[3] & mv[3]) > (bv[3] & mv[3]))
return (1);
if ((av[3] & mv[3]) < (bv[3] & mv[3]))
return (-1);
if ((av[2] & mv[2]) > (bv[2] & mv[2]))
return (1);
if ((av[2] & mv[2]) < (bv[2] & mv[2]))
return (-1);
if ((av[1] & mv[1]) > (bv[1] & mv[1]))
return (1);
if ((av[1] & mv[1]) < (bv[1] & mv[1]))
return (-1);
if ((av[0] & mv[0]) > (bv[0] & mv[0]))
return (1);
if ((av[0] & mv[0]) < (bv[0] & mv[0]))
return (-1);
break;
}
return (0);
}
u_int32_t
prefixlen2mask(u_int8_t prefixlen)
{
if (prefixlen == 0)
return (0);
if (prefixlen > 32)
prefixlen = 32;
return (htonl(0xffffffff << (32 - prefixlen)));
}
struct in6_addr *
prefixlen2mask6(u_int8_t prefixlen, u_int32_t *mask)
{
static struct in6_addr s6;
int i;
if (prefixlen > 128)
prefixlen = 128;
bzero(&s6, sizeof(s6));
for (i = 0; i < prefixlen / 8; i++)
s6.s6_addr[i] = 0xff;
i = prefixlen % 8;
if (i)
s6.s6_addr[prefixlen / 8] = 0xff00 >> i;
memcpy(mask, &s6, sizeof(s6));
return (&s6);
}
int
accept_reserve(int sockfd, struct sockaddr *addr, socklen_t *addrlen,
int reserve, volatile int *counter)
{
int ret;
if (getdtablecount() + reserve +
*counter >= getdtablesize()) {
errno = EMFILE;
return (-1);
}
if ((ret = accept4(sockfd, addr, addrlen, SOCK_NONBLOCK)) > -1) {
(*counter)++;
DPRINTF("%s: inflight incremented, now %d",__func__, *counter);
}
return (ret);
}
void
parent_tls_ticket_rekey(int fd, short events, void *arg)
{
static struct event rekeyev;
struct relayd *env = arg;
struct timeval tv;
struct relay_ticket_key key;
log_debug("%s: rekeying tickets", __func__);
key.tt_keyrev = arc4random();
arc4random_buf(key.tt_key, sizeof(key.tt_key));
proc_compose_imsg(env->sc_ps, PROC_RELAY, -1, IMSG_TLSTICKET_REKEY,
-1, -1, &key, sizeof(key));
evtimer_set(&rekeyev, parent_tls_ticket_rekey, env);
timerclear(&tv);
tv.tv_sec = TLS_SESSION_LIFETIME / 4;
evtimer_add(&rekeyev, &tv);
}