727 lines
17 KiB
C
727 lines
17 KiB
C
/* $NetBSD: npf_build.c,v 1.24 2013/05/19 20:45:34 rmind Exp $ */
|
|
|
|
/*-
|
|
* Copyright (c) 2011-2013 The NetBSD Foundation, Inc.
|
|
* All rights reserved.
|
|
*
|
|
* This material is based upon work partially supported by The
|
|
* NetBSD Foundation under a contract with Mindaugas Rasiukevicius.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
* documentation and/or other materials provided with the distribution.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
|
|
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
|
|
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
|
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
|
|
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
|
|
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
|
|
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
|
|
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
|
|
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
|
|
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
* POSSIBILITY OF SUCH DAMAGE.
|
|
*/
|
|
|
|
/*
|
|
* npfctl(8) building of the configuration.
|
|
*/
|
|
|
|
#include <sys/cdefs.h>
|
|
__RCSID("$NetBSD: npf_build.c,v 1.24 2013/05/19 20:45:34 rmind Exp $");
|
|
|
|
#include <sys/types.h>
|
|
#include <sys/ioctl.h>
|
|
|
|
#include <stdlib.h>
|
|
#include <inttypes.h>
|
|
#include <string.h>
|
|
#include <errno.h>
|
|
#include <err.h>
|
|
|
|
#include "npfctl.h"
|
|
|
|
#define MAX_RULE_NESTING 16
|
|
|
|
static nl_config_t * npf_conf = NULL;
|
|
static bool npf_debug = false;
|
|
static nl_rule_t * the_rule = NULL;
|
|
|
|
static nl_rule_t * current_group[MAX_RULE_NESTING];
|
|
static unsigned rule_nesting_level = 0;
|
|
static nl_rule_t * defgroup = NULL;
|
|
|
|
void
|
|
npfctl_config_init(bool debug)
|
|
{
|
|
npf_conf = npf_config_create();
|
|
if (npf_conf == NULL) {
|
|
errx(EXIT_FAILURE, "npf_config_create failed");
|
|
}
|
|
npf_debug = debug;
|
|
memset(current_group, 0, sizeof(current_group));
|
|
}
|
|
|
|
int
|
|
npfctl_config_send(int fd, const char *out)
|
|
{
|
|
int error;
|
|
|
|
if (out) {
|
|
_npf_config_setsubmit(npf_conf, out);
|
|
printf("\nSaving to %s\n", out);
|
|
}
|
|
if (!defgroup) {
|
|
errx(EXIT_FAILURE, "default group was not defined");
|
|
}
|
|
npf_rule_insert(npf_conf, NULL, defgroup);
|
|
error = npf_config_submit(npf_conf, fd);
|
|
if (error) {
|
|
nl_error_t ne;
|
|
_npf_config_error(npf_conf, &ne);
|
|
npfctl_print_error(&ne);
|
|
}
|
|
npf_config_destroy(npf_conf);
|
|
return error;
|
|
}
|
|
|
|
nl_config_t *
|
|
npfctl_config_ref(void)
|
|
{
|
|
return npf_conf;
|
|
}
|
|
|
|
nl_rule_t *
|
|
npfctl_rule_ref(void)
|
|
{
|
|
return the_rule;
|
|
}
|
|
|
|
unsigned long
|
|
npfctl_debug_addif(const char *ifname)
|
|
{
|
|
char tname[] = "npftest";
|
|
const size_t tnamelen = sizeof(tname) - 1;
|
|
|
|
if (!npf_debug || strncmp(ifname, tname, tnamelen) != 0) {
|
|
return 0;
|
|
}
|
|
struct ifaddrs ifa = {
|
|
.ifa_name = __UNCONST(ifname),
|
|
.ifa_flags = 0
|
|
};
|
|
unsigned long if_idx = atol(ifname + tnamelen) + 1;
|
|
_npf_debug_addif(npf_conf, &ifa, if_idx);
|
|
return if_idx;
|
|
}
|
|
|
|
bool
|
|
npfctl_table_exists_p(const char *id)
|
|
{
|
|
return npf_table_exists_p(npf_conf, atoi(id));
|
|
}
|
|
|
|
static in_port_t
|
|
npfctl_get_singleport(const npfvar_t *vp)
|
|
{
|
|
port_range_t *pr;
|
|
in_port_t *port;
|
|
|
|
if (npfvar_get_count(vp) > 1) {
|
|
yyerror("multiple ports are not valid");
|
|
}
|
|
pr = npfvar_get_data(vp, NPFVAR_PORT_RANGE, 0);
|
|
if (pr->pr_start != pr->pr_end) {
|
|
yyerror("port range is not valid");
|
|
}
|
|
port = &pr->pr_start;
|
|
return *port;
|
|
}
|
|
|
|
static fam_addr_mask_t *
|
|
npfctl_get_singlefam(const npfvar_t *vp)
|
|
{
|
|
if (npfvar_get_count(vp) > 1) {
|
|
yyerror("multiple addresses are not valid");
|
|
}
|
|
return npfvar_get_data(vp, NPFVAR_FAM, 0);
|
|
}
|
|
|
|
static bool
|
|
npfctl_build_fam(nc_ctx_t *nc, sa_family_t family,
|
|
fam_addr_mask_t *fam, int opts)
|
|
{
|
|
/*
|
|
* If family is specified, address does not match it and the
|
|
* address is extracted from the interface, then simply ignore.
|
|
* Otherwise, address of invalid family was passed manually.
|
|
*/
|
|
if (family != AF_UNSPEC && family != fam->fam_family) {
|
|
if (!fam->fam_ifindex) {
|
|
yyerror("specified address is not of the required "
|
|
"family %d", family);
|
|
}
|
|
return false;
|
|
}
|
|
|
|
/*
|
|
* Optimise 0.0.0.0/0 case to be NOP. Otherwise, address with
|
|
* zero mask would never match and therefore is not valid.
|
|
*/
|
|
if (fam->fam_mask == 0) {
|
|
npf_addr_t zero;
|
|
|
|
memset(&zero, 0, sizeof(npf_addr_t));
|
|
if (memcmp(&fam->fam_addr, &zero, sizeof(npf_addr_t))) {
|
|
yyerror("filter criterion would never match");
|
|
}
|
|
return false;
|
|
}
|
|
|
|
switch (fam->fam_family) {
|
|
case AF_INET:
|
|
npfctl_gennc_v4cidr(nc, opts,
|
|
&fam->fam_addr, fam->fam_mask);
|
|
break;
|
|
case AF_INET6:
|
|
npfctl_gennc_v6cidr(nc, opts,
|
|
&fam->fam_addr, fam->fam_mask);
|
|
break;
|
|
default:
|
|
yyerror("family %d is not supported", fam->fam_family);
|
|
}
|
|
return true;
|
|
}
|
|
|
|
static void
|
|
npfctl_build_vars(nc_ctx_t *nc, sa_family_t family, npfvar_t *vars, int opts)
|
|
{
|
|
const int type = npfvar_get_type(vars, 0);
|
|
size_t i;
|
|
|
|
npfctl_ncgen_group(nc);
|
|
for (i = 0; i < npfvar_get_count(vars); i++) {
|
|
void *data = npfvar_get_data(vars, type, i);
|
|
assert(data != NULL);
|
|
|
|
switch (type) {
|
|
case NPFVAR_FAM: {
|
|
fam_addr_mask_t *fam = data;
|
|
npfctl_build_fam(nc, family, fam, opts);
|
|
break;
|
|
}
|
|
case NPFVAR_PORT_RANGE: {
|
|
port_range_t *pr = data;
|
|
if (opts & NC_MATCH_TCP) {
|
|
npfctl_gennc_ports(nc, opts & ~NC_MATCH_UDP,
|
|
pr->pr_start, pr->pr_end);
|
|
}
|
|
if (opts & NC_MATCH_UDP) {
|
|
npfctl_gennc_ports(nc, opts & ~NC_MATCH_TCP,
|
|
pr->pr_start, pr->pr_end);
|
|
}
|
|
break;
|
|
}
|
|
case NPFVAR_TABLE: {
|
|
u_int tid = atoi(data);
|
|
npfctl_gennc_tbl(nc, opts, tid);
|
|
break;
|
|
}
|
|
default:
|
|
assert(false);
|
|
}
|
|
}
|
|
npfctl_ncgen_endgroup(nc);
|
|
}
|
|
|
|
static int
|
|
npfctl_build_proto(nc_ctx_t *nc, sa_family_t family,
|
|
const opt_proto_t *op, bool noaddrs, bool noports)
|
|
{
|
|
const npfvar_t *popts = op->op_opts;
|
|
const int proto = op->op_proto;
|
|
int pflag = 0;
|
|
|
|
switch (proto) {
|
|
case IPPROTO_TCP:
|
|
pflag = NC_MATCH_TCP;
|
|
if (!popts) {
|
|
break;
|
|
}
|
|
assert(npfvar_get_count(popts) == 2);
|
|
|
|
/* Build TCP flags block (optional). */
|
|
uint8_t *tf, *tf_mask;
|
|
|
|
tf = npfvar_get_data(popts, NPFVAR_TCPFLAG, 0);
|
|
tf_mask = npfvar_get_data(popts, NPFVAR_TCPFLAG, 1);
|
|
npfctl_gennc_tcpfl(nc, *tf, *tf_mask);
|
|
noports = false;
|
|
break;
|
|
case IPPROTO_UDP:
|
|
pflag = NC_MATCH_UDP;
|
|
break;
|
|
case IPPROTO_ICMP:
|
|
/*
|
|
* Build ICMP block.
|
|
*/
|
|
if (!noports) {
|
|
goto invop;
|
|
}
|
|
assert(npfvar_get_count(popts) == 2);
|
|
|
|
int *icmp_type, *icmp_code;
|
|
icmp_type = npfvar_get_data(popts, NPFVAR_ICMP, 0);
|
|
icmp_code = npfvar_get_data(popts, NPFVAR_ICMP, 1);
|
|
npfctl_gennc_icmp(nc, *icmp_type, *icmp_code);
|
|
noports = false;
|
|
break;
|
|
case IPPROTO_ICMPV6:
|
|
/*
|
|
* Build ICMP block.
|
|
*/
|
|
if (!noports) {
|
|
goto invop;
|
|
}
|
|
assert(npfvar_get_count(popts) == 2);
|
|
|
|
int *icmp6_type, *icmp6_code;
|
|
icmp6_type = npfvar_get_data(popts, NPFVAR_ICMP6, 0);
|
|
icmp6_code = npfvar_get_data(popts, NPFVAR_ICMP6, 1);
|
|
npfctl_gennc_icmp6(nc, *icmp6_type, *icmp6_code);
|
|
noports = false;
|
|
break;
|
|
case -1:
|
|
pflag = NC_MATCH_TCP | NC_MATCH_UDP;
|
|
noports = false;
|
|
break;
|
|
default:
|
|
/*
|
|
* No filter options are supported for other protocols,
|
|
* only the IP addresses are allowed.
|
|
*/
|
|
if (noports) {
|
|
break;
|
|
}
|
|
invop:
|
|
yyerror("invalid filter options for protocol %d", proto);
|
|
}
|
|
|
|
/*
|
|
* Build the protocol block, unless other blocks will implicitly
|
|
* perform the family/protocol checks for us.
|
|
*/
|
|
if ((family != AF_UNSPEC && noaddrs) || (proto != -1 && noports)) {
|
|
uint8_t addrlen;
|
|
|
|
switch (family) {
|
|
case AF_INET:
|
|
addrlen = sizeof(struct in_addr);
|
|
break;
|
|
case AF_INET6:
|
|
addrlen = sizeof(struct in6_addr);
|
|
break;
|
|
default:
|
|
addrlen = 0;
|
|
}
|
|
npfctl_gennc_proto(nc,
|
|
noaddrs ? addrlen : 0,
|
|
noports ? proto : 0xff);
|
|
}
|
|
return pflag;
|
|
}
|
|
|
|
static bool
|
|
npfctl_build_ncode(nl_rule_t *rl, sa_family_t family, const opt_proto_t *op,
|
|
const filt_opts_t *fopts, bool invert)
|
|
{
|
|
const addr_port_t *apfrom = &fopts->fo_from;
|
|
const addr_port_t *apto = &fopts->fo_to;
|
|
const int proto = op->op_proto;
|
|
bool noaddrs, noports;
|
|
nc_ctx_t *nc;
|
|
void *code;
|
|
size_t len;
|
|
|
|
/*
|
|
* If none specified, no n-code.
|
|
*/
|
|
noaddrs = !apfrom->ap_netaddr && !apto->ap_netaddr;
|
|
noports = !apfrom->ap_portrange && !apto->ap_portrange;
|
|
if (family == AF_UNSPEC && proto == -1 && !op->op_opts &&
|
|
noaddrs && noports)
|
|
return false;
|
|
|
|
int srcflag = NC_MATCH_SRC;
|
|
int dstflag = NC_MATCH_DST;
|
|
|
|
if (invert) {
|
|
srcflag = NC_MATCH_DST;
|
|
dstflag = NC_MATCH_SRC;
|
|
}
|
|
|
|
nc = npfctl_ncgen_create();
|
|
|
|
/* Build layer 4 protocol blocks. */
|
|
int pflag = npfctl_build_proto(nc, family, op, noaddrs, noports);
|
|
|
|
/* Build IP address blocks. */
|
|
npfctl_build_vars(nc, family, apfrom->ap_netaddr, srcflag);
|
|
npfctl_build_vars(nc, family, apto->ap_netaddr, dstflag);
|
|
|
|
/* Build port-range blocks. */
|
|
npfctl_build_vars(nc, family, apfrom->ap_portrange, srcflag | pflag);
|
|
npfctl_build_vars(nc, family, apto->ap_portrange, dstflag | pflag);
|
|
|
|
/*
|
|
* Complete n-code (destroys the context) and pass to the rule.
|
|
*/
|
|
code = npfctl_ncgen_complete(nc, &len);
|
|
if (npf_debug) {
|
|
extern char *yytext;
|
|
extern int yylineno;
|
|
|
|
printf("RULE AT LINE %d\n", yylineno - (int)(*yytext == '\n'));
|
|
npfctl_ncgen_print(code, len);
|
|
}
|
|
assert(code && len > 0);
|
|
|
|
if (npf_rule_setcode(rl, NPF_CODE_NC, code, len) == -1) {
|
|
errx(EXIT_FAILURE, "npf_rule_setcode failed");
|
|
}
|
|
free(code);
|
|
return true;
|
|
}
|
|
|
|
static void
|
|
npfctl_build_rpcall(nl_rproc_t *rp, const char *name, npfvar_t *args)
|
|
{
|
|
npf_extmod_t *extmod;
|
|
nl_ext_t *extcall;
|
|
int error;
|
|
|
|
extmod = npf_extmod_get(name, &extcall);
|
|
if (extmod == NULL) {
|
|
yyerror("unknown rule procedure '%s'", name);
|
|
}
|
|
|
|
for (size_t i = 0; i < npfvar_get_count(args); i++) {
|
|
const char *param, *value;
|
|
proc_param_t *p;
|
|
|
|
p = npfvar_get_data(args, NPFVAR_PROC_PARAM, i);
|
|
param = p->pp_param;
|
|
value = p->pp_value;
|
|
|
|
error = npf_extmod_param(extmod, extcall, param, value);
|
|
switch (error) {
|
|
case EINVAL:
|
|
yyerror("invalid parameter '%s'", param);
|
|
default:
|
|
break;
|
|
}
|
|
}
|
|
error = npf_rproc_extcall(rp, extcall);
|
|
if (error) {
|
|
yyerror(error == EEXIST ?
|
|
"duplicate procedure call" : "unexpected error");
|
|
}
|
|
}
|
|
|
|
/*
|
|
* npfctl_build_rproc: create and insert a rule procedure.
|
|
*/
|
|
void
|
|
npfctl_build_rproc(const char *name, npfvar_t *procs)
|
|
{
|
|
nl_rproc_t *rp;
|
|
size_t i;
|
|
|
|
rp = npf_rproc_create(name);
|
|
if (rp == NULL) {
|
|
errx(EXIT_FAILURE, "%s failed", __func__);
|
|
}
|
|
npf_rproc_insert(npf_conf, rp);
|
|
|
|
for (i = 0; i < npfvar_get_count(procs); i++) {
|
|
proc_call_t *pc = npfvar_get_data(procs, NPFVAR_PROC, i);
|
|
npfctl_build_rpcall(rp, pc->pc_name, pc->pc_opts);
|
|
}
|
|
}
|
|
|
|
void
|
|
npfctl_build_maprset(const char *name, int attr, u_int if_idx)
|
|
{
|
|
const int attr_di = (NPF_RULE_IN | NPF_RULE_OUT);
|
|
nl_rule_t *rl;
|
|
|
|
/* If no direction is not specified, then both. */
|
|
if ((attr & attr_di) == 0) {
|
|
attr |= attr_di;
|
|
}
|
|
/* Allow only "in/out" attributes. */
|
|
attr = NPF_RULE_GROUP | NPF_RULE_GROUP | (attr & attr_di);
|
|
rl = npf_rule_create(name, attr, if_idx);
|
|
npf_nat_insert(npf_conf, rl, NPF_PRI_LAST);
|
|
}
|
|
|
|
/*
|
|
* npfctl_build_group: create a group, insert into the global ruleset,
|
|
* update the current group pointer and increase the nesting level.
|
|
*/
|
|
void
|
|
npfctl_build_group(const char *name, int attr, u_int if_idx, bool def)
|
|
{
|
|
const int attr_di = (NPF_RULE_IN | NPF_RULE_OUT);
|
|
nl_rule_t *rl;
|
|
|
|
if (def || (attr & attr_di) == 0) {
|
|
attr |= attr_di;
|
|
}
|
|
|
|
rl = npf_rule_create(name, attr | NPF_RULE_GROUP, if_idx);
|
|
npf_rule_setprio(rl, NPF_PRI_LAST);
|
|
if (def) {
|
|
if (defgroup) {
|
|
yyerror("multiple default groups are not valid");
|
|
}
|
|
if (rule_nesting_level) {
|
|
yyerror("default group can only be at the top level");
|
|
}
|
|
defgroup = rl;
|
|
} else {
|
|
nl_rule_t *cg = current_group[rule_nesting_level];
|
|
npf_rule_insert(npf_conf, cg, rl);
|
|
}
|
|
|
|
/* Set the current group and increase the nesting level. */
|
|
if (rule_nesting_level >= MAX_RULE_NESTING) {
|
|
yyerror("rule nesting limit reached");
|
|
}
|
|
current_group[++rule_nesting_level] = rl;
|
|
}
|
|
|
|
void
|
|
npfctl_build_group_end(void)
|
|
{
|
|
assert(rule_nesting_level > 0);
|
|
current_group[rule_nesting_level--] = NULL;
|
|
}
|
|
|
|
/*
|
|
* npfctl_build_rule: create a rule, build n-code from filter options,
|
|
* if any, and insert into the ruleset of current group, or set the rule.
|
|
*/
|
|
void
|
|
npfctl_build_rule(uint32_t attr, u_int if_idx, sa_family_t family,
|
|
const opt_proto_t *op, const filt_opts_t *fopts, const char *rproc)
|
|
{
|
|
nl_rule_t *rl;
|
|
|
|
attr |= (npf_conf ? 0 : NPF_RULE_DYNAMIC);
|
|
|
|
rl = npf_rule_create(NULL, attr, if_idx);
|
|
npfctl_build_ncode(rl, family, op, fopts, false);
|
|
if (rproc) {
|
|
npf_rule_setproc(rl, rproc);
|
|
}
|
|
|
|
if (npf_conf) {
|
|
nl_rule_t *cg = current_group[rule_nesting_level];
|
|
|
|
if (rproc && !npf_rproc_exists_p(npf_conf, rproc)) {
|
|
yyerror("rule procedure '%s' is not defined", rproc);
|
|
}
|
|
assert(cg != NULL);
|
|
npf_rule_setprio(rl, NPF_PRI_LAST);
|
|
npf_rule_insert(npf_conf, cg, rl);
|
|
} else {
|
|
/* We have parsed a single rule - set it. */
|
|
the_rule = rl;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* npfctl_build_nat: create a single NAT policy of a specified
|
|
* type with a given filter options.
|
|
*/
|
|
static void
|
|
npfctl_build_nat(int type, u_int if_idx, sa_family_t family,
|
|
const addr_port_t *ap, const filt_opts_t *fopts, bool binat)
|
|
{
|
|
const opt_proto_t op = { .op_proto = -1, .op_opts = NULL };
|
|
fam_addr_mask_t *am;
|
|
in_port_t port;
|
|
nl_nat_t *nat;
|
|
|
|
if (!ap->ap_netaddr) {
|
|
yyerror("%s network segment is not specified",
|
|
type == NPF_NATIN ? "inbound" : "outbound");
|
|
}
|
|
am = npfctl_get_singlefam(ap->ap_netaddr);
|
|
if (am->fam_family != family) {
|
|
yyerror("IPv6 NAT is not supported");
|
|
}
|
|
|
|
switch (type) {
|
|
case NPF_NATOUT:
|
|
/*
|
|
* Outbound NAT (or source NAT) policy, usually used for the
|
|
* traditional NAPT. If it is a half for bi-directional NAT,
|
|
* then no port translation with mapping.
|
|
*/
|
|
nat = npf_nat_create(NPF_NATOUT, !binat ?
|
|
(NPF_NAT_PORTS | NPF_NAT_PORTMAP) : 0,
|
|
if_idx, &am->fam_addr, am->fam_family, 0);
|
|
break;
|
|
case NPF_NATIN:
|
|
/*
|
|
* Inbound NAT (or destination NAT). Unless bi-NAT, a port
|
|
* must be specified, since it has to be redirection.
|
|
*/
|
|
port = 0;
|
|
if (!binat) {
|
|
if (!ap->ap_portrange) {
|
|
yyerror("inbound port is not specified");
|
|
}
|
|
port = npfctl_get_singleport(ap->ap_portrange);
|
|
}
|
|
nat = npf_nat_create(NPF_NATIN, !binat ? NPF_NAT_PORTS : 0,
|
|
if_idx, &am->fam_addr, am->fam_family, port);
|
|
break;
|
|
default:
|
|
assert(false);
|
|
}
|
|
|
|
npfctl_build_ncode(nat, family, &op, fopts, false);
|
|
npf_nat_insert(npf_conf, nat, NPF_PRI_LAST);
|
|
}
|
|
|
|
/*
|
|
* npfctl_build_natseg: validate and create NAT policies.
|
|
*/
|
|
void
|
|
npfctl_build_natseg(int sd, int type, u_int if_idx, const addr_port_t *ap1,
|
|
const addr_port_t *ap2, const filt_opts_t *fopts)
|
|
{
|
|
sa_family_t af = AF_INET;
|
|
filt_opts_t imfopts;
|
|
bool binat;
|
|
|
|
if (sd == NPFCTL_NAT_STATIC) {
|
|
yyerror("static NAT is not yet supported");
|
|
}
|
|
assert(sd == NPFCTL_NAT_DYNAMIC);
|
|
assert(if_idx != 0);
|
|
|
|
/*
|
|
* Bi-directional NAT is a combination of inbound NAT and outbound
|
|
* NAT policies. Note that the translation address is local IP and
|
|
* the filter criteria is inverted accordingly.
|
|
*/
|
|
binat = (NPF_NATIN | NPF_NATOUT) == type;
|
|
|
|
/*
|
|
* If the filter criteria is not specified explicitly, apply implicit
|
|
* filtering according to the given network segments.
|
|
*
|
|
* Note: filled below, depending on the type.
|
|
*/
|
|
if (__predict_true(!fopts)) {
|
|
fopts = &imfopts;
|
|
}
|
|
|
|
if (type & NPF_NATIN) {
|
|
memset(&imfopts, 0, sizeof(filt_opts_t));
|
|
memcpy(&imfopts.fo_to, ap2, sizeof(addr_port_t));
|
|
npfctl_build_nat(NPF_NATIN, if_idx, af, ap1, fopts, binat);
|
|
}
|
|
if (type & NPF_NATOUT) {
|
|
memset(&imfopts, 0, sizeof(filt_opts_t));
|
|
memcpy(&imfopts.fo_from, ap1, sizeof(addr_port_t));
|
|
npfctl_build_nat(NPF_NATOUT, if_idx, af, ap2, fopts, binat);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* npfctl_fill_table: fill NPF table with entries from a specified file.
|
|
*/
|
|
static void
|
|
npfctl_fill_table(nl_table_t *tl, u_int type, const char *fname)
|
|
{
|
|
char *buf = NULL;
|
|
int l = 0;
|
|
FILE *fp;
|
|
size_t n;
|
|
|
|
fp = fopen(fname, "r");
|
|
if (fp == NULL) {
|
|
err(EXIT_FAILURE, "open '%s'", fname);
|
|
}
|
|
while (l++, getline(&buf, &n, fp) != -1) {
|
|
fam_addr_mask_t fam;
|
|
int alen;
|
|
|
|
if (*buf == '\n' || *buf == '#') {
|
|
continue;
|
|
}
|
|
|
|
if (!npfctl_parse_cidr(buf, &fam, &alen)) {
|
|
errx(EXIT_FAILURE,
|
|
"%s:%d: invalid table entry", fname, l);
|
|
}
|
|
if (type == NPF_TABLE_HASH && fam.fam_mask != NPF_NO_NETMASK) {
|
|
errx(EXIT_FAILURE,
|
|
"%s:%d: mask used with the hash table", fname, l);
|
|
}
|
|
|
|
/* Create and add a table entry. */
|
|
npf_table_add_entry(tl, fam.fam_family,
|
|
&fam.fam_addr, fam.fam_mask);
|
|
}
|
|
if (buf != NULL) {
|
|
free(buf);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* npfctl_build_table: create an NPF table, add to the configuration and,
|
|
* if required, fill with contents from a file.
|
|
*/
|
|
void
|
|
npfctl_build_table(const char *tid, u_int type, const char *fname)
|
|
{
|
|
nl_table_t *tl;
|
|
u_int id;
|
|
|
|
id = atoi(tid);
|
|
tl = npf_table_create(id, type);
|
|
assert(tl != NULL);
|
|
|
|
if (npf_table_insert(npf_conf, tl)) {
|
|
errx(EXIT_FAILURE, "table '%d' is already defined\n", id);
|
|
}
|
|
|
|
if (fname) {
|
|
npfctl_fill_table(tl, type, fname);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* npfctl_build_alg: create an NPF application level gatewayl and add it
|
|
* to the configuration.
|
|
*/
|
|
void
|
|
npfctl_build_alg(const char *al_name)
|
|
{
|
|
if (_npf_alg_load(npf_conf, al_name) != 0) {
|
|
errx(EXIT_FAILURE, "ALG '%s' already loaded", al_name);
|
|
}
|
|
}
|