mirror of
https://gitlab.com/apparmor/apparmor.git
synced 2025-03-04 00:14:44 +01:00

The mapping of AA_CONT_MATCH was being dropped resulting in the
tcp tests failing because they would only match up to the first conditional
match check in the layout.
Bug: https://gitlab.com/apparmor/apparmor/-/issues/462
Fixes: e29f5ce5f
("parser: if extended perms are supported by the kernel build a permstable")
Signed-off-by: John Johansen <john.johansen@canonical.com>
1018 lines
28 KiB
C++
1018 lines
28 KiB
C++
/*
|
|
* Copyright (c) 2014
|
|
* Canonical, Ltd. (All rights reserved)
|
|
*
|
|
* This program is free software; you can redistribute it and/or
|
|
* modify it under the terms of version 2 of the GNU General Public
|
|
* License published by the Free Software Foundation.
|
|
*
|
|
* This program 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, contact Novell, Inc. or Canonical
|
|
* Ltd.
|
|
*/
|
|
|
|
#include <iomanip>
|
|
#include <string>
|
|
#include <sstream>
|
|
#include <map>
|
|
#include <arpa/inet.h>
|
|
|
|
#include "lib.h"
|
|
#include "parser.h"
|
|
#include "profile.h"
|
|
#include "network.h"
|
|
|
|
#define ALL_TYPES 0x43e
|
|
|
|
int parse_net_perms(const char *str_mode, perm32_t *mode, int fail)
|
|
{
|
|
return parse_X_perms("net", AA_VALID_NET_PERMS, str_mode, mode, fail);
|
|
}
|
|
|
|
/* Bleah C++ doesn't have non-trivial designated initializers so we just
|
|
* have to make sure these are in order. This means we are more brittle
|
|
* but there isn't much we can do.
|
|
*/
|
|
struct sock_type_map {
|
|
const char *name;
|
|
int value;
|
|
};
|
|
|
|
struct sock_type_map sock_types[] = {
|
|
{ "none", 0 },
|
|
{ "stream", SOCK_STREAM },
|
|
{ "dgram", SOCK_DGRAM },
|
|
{ "raw", SOCK_RAW },
|
|
{ "rdm", SOCK_RDM },
|
|
{ "seqpacket", SOCK_SEQPACKET },
|
|
{ "dccp", SOCK_DCCP },
|
|
{ "invalid", -1 },
|
|
{ "invalid", -1 },
|
|
{ "invalid", -1 },
|
|
{ "packet", SOCK_PACKET },
|
|
{ NULL, -1 },
|
|
/*
|
|
* See comment above
|
|
*/
|
|
};
|
|
|
|
int net_find_type_val(const char *type)
|
|
{
|
|
int i;
|
|
for (i = 0; sock_types[i].name; i++) {
|
|
if (strcmp(sock_types[i].name, type) == 0)
|
|
return sock_types[i].value;
|
|
}
|
|
|
|
return -1;
|
|
}
|
|
|
|
const char *net_find_type_name(int type)
|
|
{
|
|
int i;
|
|
for (i = 0; sock_types[i].name; i++) {
|
|
if (sock_types[i].value == type)
|
|
return sock_types[i].name;
|
|
}
|
|
|
|
return NULL;
|
|
}
|
|
|
|
|
|
/* FIXME: currently just treating as a bit mask this will have to change
|
|
* set up a table of mappings, there can be several mappings for a
|
|
* given match.
|
|
* currently the mapping does not set the protocol for stream/dgram to
|
|
* anything other than 0.
|
|
* network inet tcp -> network inet stream 0 instead of
|
|
* network inet raw tcp.
|
|
* some entries are just provided for completeness at this time
|
|
*/
|
|
/* values stolen from /etc/protocols - needs to change */
|
|
#define RAW_TCP 6
|
|
#define RAW_UDP 17
|
|
#define RAW_ICMP 1
|
|
#define RAW_ICMPv6 58
|
|
|
|
/* used by af_name.h to auto generate table entries for "name", AF_NAME
|
|
* pair */
|
|
#define AA_GEN_NET_ENT(name, AF) \
|
|
{name, AF, "stream", SOCK_STREAM, "", 0xffffff}, \
|
|
{name, AF, "dgram", SOCK_DGRAM, "", 0xffffff}, \
|
|
{name, AF, "seqpacket", SOCK_SEQPACKET, "", 0xffffff}, \
|
|
{name, AF, "rdm", SOCK_RDM, "", 0xffffff}, \
|
|
{name, AF, "raw", SOCK_RAW, "", 0xffffff}, \
|
|
{name, AF, "packet", SOCK_PACKET, "", 0xffffff},
|
|
/*FIXME: missing {name, AF, "dccp", SOCK_DCCP, "", 0xfffffff}, */
|
|
|
|
static struct network_tuple network_mappings[] = {
|
|
/* basic types */
|
|
#include "af_names.h"
|
|
/* FIXME: af_names.h is missing AF_LLC, AF_TIPC */
|
|
/* mapped types */
|
|
{"inet", AF_INET, "raw", SOCK_RAW,
|
|
"tcp", 1 << RAW_TCP},
|
|
{"inet", AF_INET, "raw", SOCK_RAW,
|
|
"udp", 1 << RAW_UDP},
|
|
{"inet", AF_INET, "raw", SOCK_RAW,
|
|
"icmp", 1 << RAW_ICMP},
|
|
{"inet", AF_INET, "tcp", SOCK_STREAM,
|
|
"", 0xffffffff}, /* should we give raw tcp too? */
|
|
{"inet", AF_INET, "udp", SOCK_DGRAM,
|
|
"", 0xffffffff}, /* should these be open masks? */
|
|
{"inet", AF_INET, "icmp", SOCK_RAW,
|
|
"", 1 << RAW_ICMP},
|
|
{"inet6", AF_INET6, "tcp", SOCK_STREAM,
|
|
"", 0xffffffff},
|
|
{"inet6", AF_INET6, "udp", SOCK_DGRAM,
|
|
"", 0xffffffff},
|
|
/* what do we do with icmp on inet6?
|
|
{"inet6", AF_INET, "icmp", SOCK_RAW, 0},
|
|
{"inet6", AF_INET, "icmpv6", SOCK_RAW, 0},
|
|
*/
|
|
/* terminate */
|
|
{NULL, 0, NULL, 0, NULL, 0}
|
|
};
|
|
|
|
/* The apparmor kernel patches up until 2.6.38 didn't handle networking
|
|
* tables with sizes > AF_MAX correctly. This could happen when the
|
|
* parser was built against newer kernel headers and then used to load
|
|
* policy on an older kernel. This could happen during upgrades or
|
|
* in multi-kernel boot systems.
|
|
*
|
|
* Try to detect the running kernel version and use that to determine
|
|
* AF_MAX
|
|
*/
|
|
#define PROC_VERSION "/proc/sys/kernel/osrelease"
|
|
static size_t kernel_af_max(void) {
|
|
char buffer[32];
|
|
int major;
|
|
autoclose int fd = -1;
|
|
int res;
|
|
|
|
if (!net_af_max_override) {
|
|
return 0;
|
|
}
|
|
/* the override parameter is specifying the max value */
|
|
if (net_af_max_override > 0)
|
|
return net_af_max_override;
|
|
|
|
fd = open(PROC_VERSION, O_RDONLY);
|
|
if (fd == -1)
|
|
/* fall back to default provided during build */
|
|
return 0;
|
|
res = read(fd, &buffer, sizeof(buffer) - 1);
|
|
if (res <= 0)
|
|
return 0;
|
|
buffer[res] = '\0';
|
|
res = sscanf(buffer, "2.6.%d", &major);
|
|
if (res != 1)
|
|
return 0;
|
|
|
|
switch(major) {
|
|
case 24:
|
|
case 25:
|
|
case 26:
|
|
return 34;
|
|
case 27:
|
|
return 35;
|
|
case 28:
|
|
case 29:
|
|
case 30:
|
|
return 36;
|
|
case 31:
|
|
case 32:
|
|
case 33:
|
|
case 34:
|
|
case 35:
|
|
return 37;
|
|
case 36:
|
|
case 37:
|
|
return 38;
|
|
/* kernels .38 and later should handle this correctly so no
|
|
* static mapping needed
|
|
*/
|
|
default:
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
/* Yuck. We grab AF_* values to define above from linux/socket.h because
|
|
* they are more accurate than sys/socket.h for what the kernel actually
|
|
* supports. However, we can't just include linux/socket.h directly,
|
|
* because the AF_* definitions are protected with an ifdef KERNEL
|
|
* wrapper, but we don't want to define that because that can cause
|
|
* other redefinitions from glibc. However, because the kernel may have
|
|
* more definitions than glibc, we need make sure AF_MAX reflects this,
|
|
* hence the wrapping function.
|
|
*/
|
|
size_t get_af_max() {
|
|
size_t af_max;
|
|
/* HACK: declare that version without "create" had a static AF_MAX */
|
|
if (!perms_create && !net_af_max_override)
|
|
net_af_max_override = -1;
|
|
|
|
#if AA_AF_MAX > AF_MAX
|
|
af_max = AA_AF_MAX;
|
|
#else
|
|
af_max = AF_MAX;
|
|
#endif
|
|
|
|
/* HACK: some kernels didn't handle network tables from parsers
|
|
* compiled against newer kernel headers as they are larger than
|
|
* the running kernel expected. If net_override is defined check
|
|
* to see if there is a static max specified for that kernel
|
|
*/
|
|
if (net_af_max_override) {
|
|
size_t max = kernel_af_max();
|
|
if (max && max < af_max)
|
|
return max;
|
|
}
|
|
|
|
return af_max;
|
|
}
|
|
|
|
const char *net_find_af_name(unsigned int af)
|
|
{
|
|
size_t i;
|
|
|
|
if (af < 0 || af > get_af_max())
|
|
return NULL;
|
|
|
|
for (i = 0; i < sizeof(network_mappings) / sizeof(*network_mappings); i++) {
|
|
if (network_mappings[i].family == af)
|
|
return network_mappings[i].family_name;
|
|
}
|
|
|
|
return NULL;
|
|
}
|
|
|
|
const char *net_find_protocol_name(unsigned int protocol)
|
|
{
|
|
size_t i;
|
|
|
|
for (i = 0; i < sizeof(network_mappings) / sizeof(*network_mappings); i++) {
|
|
if (network_mappings[i].protocol == protocol) {
|
|
return network_mappings[i].protocol_name;
|
|
}
|
|
}
|
|
|
|
return NULL;
|
|
}
|
|
|
|
const struct network_tuple *net_find_mapping(const struct network_tuple *map,
|
|
const char *family,
|
|
const char *type,
|
|
const char *protocol)
|
|
{
|
|
if (!map)
|
|
map = network_mappings;
|
|
else
|
|
/* assumes it points to last entry returned */
|
|
map++;
|
|
|
|
for (; map->family_name; map++) {
|
|
if (family) {
|
|
PDEBUG("Checking family %s\n", map->family_name);
|
|
if (strcmp(family, map->family_name) != 0)
|
|
continue;
|
|
PDEBUG("Found family %s\n", family);
|
|
}
|
|
if (type) {
|
|
PDEBUG("Checking type %s\n", map->type_name);
|
|
if (strcmp(type, map->type_name) != 0)
|
|
continue;
|
|
PDEBUG("Found type %s\n", type);
|
|
}
|
|
if (protocol) {
|
|
/* allows the proto to be the "type", ie. tcp implies
|
|
* stream */
|
|
if (!type) {
|
|
PDEBUG("Checking protocol type %s\n", map->type_name);
|
|
if (strcmp(protocol, map->type_name) == 0)
|
|
goto match;
|
|
}
|
|
PDEBUG("Checking type %s protocol %s\n", map->type_name, map->protocol_name);
|
|
if (strcmp(protocol, map->protocol_name) != 0)
|
|
continue;
|
|
/* fixme should we allow specifying protocol by #
|
|
* without needing the protocol mapping? */
|
|
}
|
|
|
|
/* if we get this far we have a match */
|
|
match:
|
|
return map;
|
|
}
|
|
|
|
return NULL;
|
|
}
|
|
|
|
bool parse_ipv4_address(const char *input, struct ip_address *result)
|
|
{
|
|
struct in_addr addr;
|
|
if (inet_pton(AF_INET, input, &addr) == 1) {
|
|
result->family = AF_INET;
|
|
result->address.address_v4 = addr.s_addr;
|
|
return true;
|
|
}
|
|
return false;
|
|
}
|
|
|
|
bool parse_ipv6_address(const char *input, struct ip_address *result)
|
|
{
|
|
struct in6_addr addr;
|
|
if (inet_pton(AF_INET6, input, &addr) == 1) {
|
|
result->family = AF_INET6;
|
|
memcpy(result->address.address_v6, addr.s6_addr, 16);
|
|
return true;
|
|
}
|
|
return false;
|
|
}
|
|
|
|
bool parse_ip(const char *ip, struct ip_address *result)
|
|
{
|
|
return parse_ipv6_address(ip, result) ||
|
|
parse_ipv4_address(ip, result);
|
|
}
|
|
|
|
bool parse_port_number(const char *port_entry, uint16_t *port) {
|
|
char *eptr;
|
|
unsigned long port_tmp = strtoul(port_entry, &eptr, 10);
|
|
|
|
if (port_entry != eptr && *eptr == '\0' &&
|
|
port_tmp <= UINT16_MAX) {
|
|
*port = port_tmp;
|
|
return true;
|
|
}
|
|
return false;
|
|
}
|
|
|
|
bool parse_range(const char *range, uint16_t *from, uint16_t *to)
|
|
{
|
|
char *range_tmp = strdup(range);
|
|
char *dash = strchr(range_tmp, '-');
|
|
bool ret = false;
|
|
if (dash == NULL)
|
|
goto out;
|
|
*dash = '\0';
|
|
|
|
if (parse_port_number(range_tmp, from)) {
|
|
if (parse_port_number(dash + 1, to)) {
|
|
if (*from > *to) {
|
|
goto out;
|
|
}
|
|
ret = true;
|
|
goto out;
|
|
}
|
|
}
|
|
|
|
out:
|
|
free(range_tmp);
|
|
return ret;
|
|
}
|
|
|
|
bool network_rule::parse_port(ip_conds &entry)
|
|
{
|
|
entry.is_port = true;
|
|
if (parse_range(entry.sport, &entry.from_port, &entry.to_port))
|
|
return true;
|
|
if (parse_port_number(entry.sport, &entry.from_port)) {
|
|
/* if range is not used, from and to have the same value */
|
|
entry.to_port = entry.from_port;
|
|
return true;
|
|
}
|
|
return false;
|
|
}
|
|
|
|
bool network_rule::parse_address(ip_conds &entry)
|
|
{
|
|
if (strcmp(entry.sip, "none") == 0) {
|
|
entry.is_none = true;
|
|
return true;
|
|
}
|
|
entry.is_ip = true;
|
|
return parse_ip(entry.sip, &entry.ip);
|
|
}
|
|
|
|
void network_rule::move_conditionals(struct cond_entry *conds, ip_conds &ip_cond)
|
|
{
|
|
struct cond_entry *cond_ent;
|
|
|
|
list_for_each(conds, cond_ent) {
|
|
/* for now disallow keyword 'in' (list) */
|
|
if (!cond_ent->eq)
|
|
yyerror("keyword \"in\" is not allowed in network rules\n");
|
|
if (strcmp(cond_ent->name, "ip") == 0) {
|
|
move_conditional_value("network", &ip_cond.sip, cond_ent);
|
|
if (!parse_address(ip_cond))
|
|
yyerror("network invalid ip='%s'\n", ip_cond.sip);
|
|
} else if (strcmp(cond_ent->name, "port") == 0) {
|
|
move_conditional_value("network", &ip_cond.sport, cond_ent);
|
|
if (!parse_port(ip_cond))
|
|
yyerror("network invalid port='%s'\n", ip_cond.sport);
|
|
} else {
|
|
yyerror("invalid network rule conditional \"%s\"\n",
|
|
cond_ent->name);
|
|
}
|
|
}
|
|
}
|
|
|
|
void network_rule::set_netperm(unsigned int family, unsigned int type, unsigned int protocol)
|
|
{
|
|
if (type > SOCK_PACKET) {
|
|
/* setting mask instead of a bit */
|
|
network_perms[family].first |= type;
|
|
} else
|
|
network_perms[family].first |= 1 << type;
|
|
network_perms[family].second |= protocol;
|
|
}
|
|
|
|
network_rule::network_rule(perm32_t perms_p, struct cond_entry *conds,
|
|
struct cond_entry *peer_conds):
|
|
dedup_perms_rule_t(AA_CLASS_NETV8), label(NULL)
|
|
{
|
|
size_t family_index, i;
|
|
|
|
move_conditionals(conds, local);
|
|
move_conditionals(peer_conds, peer);
|
|
free_cond_list(conds);
|
|
free_cond_list(peer_conds);
|
|
|
|
if (has_local_conds() || has_peer_conds()) {
|
|
const char *family[] = { "inet", "inet6" };
|
|
for (i = 0; i < sizeof(family)/sizeof(family[0]); i++) {
|
|
const struct network_tuple *mapping = NULL;
|
|
while ((mapping = net_find_mapping(mapping, family[i], NULL, NULL))) {
|
|
network_map[mapping->family].push_back({ mapping->family, mapping->type, mapping->protocol });
|
|
set_netperm(mapping->family, mapping->type, mapping->protocol);
|
|
}
|
|
}
|
|
} else {
|
|
for (family_index = AF_UNSPEC; family_index < get_af_max(); family_index++) {
|
|
network_map[family_index].push_back({ family_index, 0xFFFFFFFF, 0xFFFFFFFF });
|
|
set_netperm(family_index, 0xFFFFFFFF, 0xFFFFFFFF);
|
|
}
|
|
}
|
|
|
|
|
|
|
|
if (perms_p) {
|
|
perms = perms_p;
|
|
if (perms & ~AA_VALID_NET_PERMS)
|
|
yyerror("perms contains invalid permissions for network rules\n");
|
|
else if ((perms & ~AA_PEER_NET_PERMS) && has_peer_conds())
|
|
yyerror("network 'create', 'shutdown', 'setattr', 'getattr', 'bind', 'listen', 'setopt', and/or 'getopt' accesses cannot be used with peer socket conditionals\n");
|
|
} else {
|
|
perms = AA_VALID_NET_PERMS;
|
|
}
|
|
}
|
|
|
|
network_rule::network_rule(perm32_t perms_p, const char *family, const char *type,
|
|
const char *protocol, struct cond_entry *conds,
|
|
struct cond_entry *peer_conds):
|
|
dedup_perms_rule_t(AA_CLASS_NETV8), label(NULL)
|
|
{
|
|
const struct network_tuple *mapping = NULL;
|
|
|
|
move_conditionals(conds, local);
|
|
move_conditionals(peer_conds, peer);
|
|
free_cond_list(conds);
|
|
free_cond_list(peer_conds);
|
|
|
|
while ((mapping = net_find_mapping(mapping, family, type, protocol))) {
|
|
/* if inet conds and family are specified, fail if
|
|
* family is not af_inet or af_inet6
|
|
*/
|
|
if ((has_local_conds() || has_peer_conds()) &&
|
|
mapping->family != AF_INET && mapping->family != AF_INET6) {
|
|
yyerror("network family does not support local or peer conditionals\n");
|
|
}
|
|
network_map[mapping->family].push_back({ mapping->family, mapping->type, mapping->protocol });
|
|
set_netperm(mapping->family, mapping->type, mapping->protocol);
|
|
}
|
|
|
|
if (type == NULL && network_map.empty()) {
|
|
while ((mapping = net_find_mapping(mapping, type, family, protocol))) {
|
|
/* if inet conds and type/protocol are
|
|
* specified, only add rules for af_inet and
|
|
* af_inet6
|
|
*/
|
|
if ((has_local_conds() || has_peer_conds()) &&
|
|
mapping->family != AF_INET && mapping->family != AF_INET6)
|
|
continue;
|
|
|
|
network_map[mapping->family].push_back({ mapping->family, mapping->type, mapping->protocol });
|
|
set_netperm(mapping->family, mapping->type, mapping->protocol);
|
|
}
|
|
}
|
|
|
|
if (network_map.empty())
|
|
yyerror(_("Invalid network entry."));
|
|
|
|
if (perms_p) {
|
|
perms = perms_p;
|
|
if (perms & ~AA_VALID_NET_PERMS)
|
|
yyerror("perms contains invalid permissions for network rules\n");
|
|
else if ((perms & ~AA_PEER_NET_PERMS) && has_peer_conds())
|
|
yyerror("network 'create', 'shutdown', 'setattr', 'getattr', 'bind', 'listen', 'setopt', and/or 'getopt' accesses cannot be used with peer socket conditionals\n");
|
|
} else {
|
|
perms = AA_VALID_NET_PERMS;
|
|
}
|
|
}
|
|
|
|
network_rule::network_rule(perm32_t perms_p, unsigned int family, unsigned int type):
|
|
dedup_perms_rule_t(AA_CLASS_NETV8), label(NULL)
|
|
{
|
|
network_map[family].push_back({ family, type, 0xFFFFFFFF });
|
|
set_netperm(family, type, 0xFFFFFFFF);
|
|
|
|
if (perms_p) {
|
|
perms = perms_p;
|
|
if (perms & ~AA_VALID_NET_PERMS)
|
|
yyerror("perms contains invalid permissions for network rules\n");
|
|
else if ((perms & ~AA_PEER_NET_PERMS) && has_peer_conds())
|
|
yyerror("network 'create', 'shutdown', 'setattr', 'getattr', 'bind', 'listen', 'setopt', and/or 'getopt' accesses cannot be used with peer socket conditionals\n");
|
|
} else {
|
|
perms = AA_VALID_NET_PERMS;
|
|
}
|
|
}
|
|
|
|
ostream &network_rule::dump(ostream &os)
|
|
{
|
|
class_rule_t::dump(os);
|
|
|
|
unsigned int count = sizeof(sock_types)/sizeof(sock_types[0]);
|
|
unsigned int mask = ~((1 << count) -1);
|
|
unsigned int j;
|
|
|
|
/* This can only be set by an unqualified network rule */
|
|
if (network_map.find(AF_UNSPEC) != network_map.end()) {
|
|
os << ",\n";
|
|
return os;
|
|
}
|
|
|
|
for (const auto& perm : network_perms) {
|
|
unsigned int family = perm.first;
|
|
unsigned int type = perm.second.first;
|
|
unsigned int protocol = perm.second.second;
|
|
|
|
const char *family_name = net_find_af_name(family);
|
|
if (family_name)
|
|
os << " " << family_name;
|
|
else
|
|
os << " #" << family;
|
|
|
|
/* All types/protocols */
|
|
if (type == 0xffffffff || type == ALL_TYPES)
|
|
continue;
|
|
|
|
printf(" {");
|
|
|
|
for (j = 0; j < count; j++) {
|
|
const char *type_name;
|
|
if (type & (1 << j)) {
|
|
type_name = sock_types[j].name;
|
|
if (type_name)
|
|
os << " " << type_name;
|
|
else
|
|
os << " #" << j;
|
|
}
|
|
}
|
|
if (type & mask)
|
|
os << " #" << std::hex << (type & mask);
|
|
|
|
printf(" }");
|
|
|
|
const char *protocol_name = net_find_protocol_name(protocol);
|
|
if (protocol_name)
|
|
os << " " << protocol_name;
|
|
else
|
|
os << " #" << protocol;
|
|
}
|
|
|
|
os << ",\n";
|
|
|
|
return os;
|
|
}
|
|
|
|
|
|
int network_rule::expand_variables(void)
|
|
{
|
|
return 0;
|
|
}
|
|
|
|
void network_rule::warn_once(const char *name)
|
|
{
|
|
rule_t::warn_once(name, "network rules not enforced");
|
|
}
|
|
|
|
std::string gen_ip_cond(const struct ip_address ip)
|
|
{
|
|
std::ostringstream oss;
|
|
int i;
|
|
if (ip.family == AF_INET) {
|
|
/* add a byte containing the size of the following ip */
|
|
oss << "\\x" << std::setfill('0') << std::setw(2) << std::hex << IPV4_SIZE;
|
|
|
|
u8 *byte = (u8 *) &ip.address.address_v4; /* in network byte order */
|
|
for (i = 0; i < 4; i++)
|
|
oss << "\\x" << std::setfill('0') << std::setw(2) << std::hex << static_cast<unsigned int>(byte[i]);
|
|
} else {
|
|
/* add a byte containing the size of the following ip */
|
|
oss << "\\x" << std::setfill('0') << std::setw(2) << std::hex << IPV6_SIZE;
|
|
for (i = 0; i < 16; ++i)
|
|
oss << "\\x" << std::setfill('0') << std::setw(2) << std::hex << static_cast<unsigned int>(ip.address.address_v6[i]);
|
|
}
|
|
return oss.str();
|
|
}
|
|
|
|
std::string gen_port_cond(uint16_t port)
|
|
{
|
|
std::ostringstream oss;
|
|
if (port > 0) {
|
|
oss << "\\x" << std::setfill('0') << std::setw(2) << std::hex << ((port & 0xff00) >> 8);
|
|
oss << "\\x" << std::setfill('0') << std::setw(2) << std::hex << (port & 0xff);
|
|
} else {
|
|
oss << "..";
|
|
}
|
|
return oss.str();
|
|
}
|
|
|
|
std::list<std::ostringstream> gen_all_ip_options(std::ostringstream &oss) {
|
|
|
|
std::list<std::ostringstream> all_streams;
|
|
std::ostringstream none, ipv4, ipv6;
|
|
int i;
|
|
none << oss.str();
|
|
ipv4 << oss.str();
|
|
ipv6 << oss.str();
|
|
|
|
none << "\\x" << std::setfill('0') << std::setw(2) << std::hex << NONE_SIZE;
|
|
|
|
/* add a byte containing the size of the following ip */
|
|
ipv4 << "\\x" << std::setfill('0') << std::setw(2) << std::hex << IPV4_SIZE;
|
|
for (i = 0; i < 4; i++)
|
|
ipv4 << ".";
|
|
|
|
/* add a byte containing the size of the following ip */
|
|
ipv6 << "\\x" << std::setfill('0') << std::setw(2) << std::hex << IPV6_SIZE;
|
|
for (i = 0; i < 16; ++i)
|
|
ipv6 << ".";
|
|
|
|
all_streams.push_back(std::move(none));
|
|
all_streams.push_back(std::move(ipv4));
|
|
all_streams.push_back(std::move(ipv6));
|
|
|
|
return all_streams;
|
|
}
|
|
|
|
std::list<std::ostringstream> copy_streams_list(std::list<std::ostringstream> &streams)
|
|
{
|
|
std::list<std::ostringstream> streams_copy;
|
|
for (auto &oss : streams) {
|
|
std::ostringstream oss_copy(oss.str());
|
|
streams_copy.push_back(std::move(oss_copy));
|
|
}
|
|
return streams_copy;
|
|
}
|
|
|
|
bool network_rule::gen_ip_conds(Profile &prof, std::list<std::ostringstream> &streams, ip_conds &entry, bool is_peer, uint16_t port, bool is_port, bool is_cmd)
|
|
{
|
|
std::string buf;
|
|
perm32_t cond_perms;
|
|
std::list<std::ostringstream> ip_streams;
|
|
|
|
for (auto &oss : streams) {
|
|
if (is_port && !(entry.is_ip && entry.is_none)) {
|
|
/* encode port type (privileged - 1, remote - 2, unprivileged - 0) */
|
|
if (!is_peer && perms & AA_NET_BIND && port < IPPORT_RESERVED)
|
|
oss << "\\x01";
|
|
else if (is_peer)
|
|
oss << "\\x02";
|
|
else
|
|
oss << "\\x00";
|
|
|
|
oss << gen_port_cond(port);
|
|
} else {
|
|
/* port type + port number */
|
|
oss << "...";
|
|
}
|
|
}
|
|
|
|
ip_streams = std::move(streams);
|
|
streams.clear();
|
|
|
|
for (auto &oss : ip_streams) {
|
|
if (entry.is_ip) {
|
|
oss << gen_ip_cond(entry.ip);
|
|
streams.push_back(std::move(oss));
|
|
} else if (entry.is_none) {
|
|
oss << "\\x" << std::setfill('0') << std::setw(2) << std::hex << NONE_SIZE;
|
|
streams.push_back(std::move(oss));
|
|
} else {
|
|
streams.splice(streams.end(), gen_all_ip_options(oss));
|
|
}
|
|
}
|
|
|
|
cond_perms = map_perms(perms);
|
|
if (!is_cmd && (label || is_peer))
|
|
cond_perms = AA_COMPAT_CONT_MATCH;
|
|
|
|
for (auto &oss : streams) {
|
|
oss << "\\x00"; /* null transition */
|
|
|
|
buf = oss.str();
|
|
/* AA_CONT_MATCH mapping (cond_perms) only applies to perms, not audit */
|
|
if (!prof.policy.rules->add_rule(buf.c_str(), priority,
|
|
rule_mode, cond_perms,
|
|
dedup_perms_rule_t::audit == AUDIT_FORCE ? map_perms(perms) : 0,
|
|
parseopts))
|
|
return false;
|
|
|
|
if (label || is_peer) {
|
|
if (!is_peer)
|
|
cond_perms = map_perms(perms);
|
|
|
|
oss << default_match_pattern; /* label - not used for now */
|
|
oss << "\\x00"; /* null transition */
|
|
|
|
buf = oss.str();
|
|
if (!prof.policy.rules->add_rule(buf.c_str(), priority,
|
|
rule_mode, cond_perms,
|
|
dedup_perms_rule_t::audit == AUDIT_FORCE ? map_perms(perms) : 0,
|
|
parseopts))
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
bool network_rule::gen_net_rule(Profile &prof, u16 family, unsigned int type_mask, unsigned int protocol) {
|
|
std::ostringstream buffer;
|
|
std::string buf;
|
|
|
|
buffer << "\\x" << std::setfill('0') << std::setw(2) << std::hex << AA_CLASS_NETV8;
|
|
buffer << "\\x" << std::setfill('0') << std::setw(2) << std::hex << ((family & 0xff00) >> 8);
|
|
buffer << "\\x" << std::setfill('0') << std::setw(2) << std::hex << (family & 0xff);
|
|
if (type_mask > 0xffff) {
|
|
buffer << "..";
|
|
} else {
|
|
buffer << "\\x" << std::setfill('0') << std::setw(2) << std::hex << ((type_mask & 0xff00) >> 8);
|
|
buffer << "\\x" << std::setfill('0') << std::setw(2) << std::hex << (type_mask & 0xff);
|
|
}
|
|
|
|
if (!features_supports_inet || (family != AF_INET && family != AF_INET6)) {
|
|
buf = buffer.str();
|
|
if (!prof.policy.rules->add_rule(buf.c_str(), priority,
|
|
rule_mode, map_perms(perms),
|
|
dedup_perms_rule_t::audit == AUDIT_FORCE ? map_perms(perms) : 0,
|
|
parseopts))
|
|
return false;
|
|
return true;
|
|
}
|
|
|
|
buf = buffer.str();
|
|
/* create perms need to be generated excluding the rest of the perms */
|
|
if (perms & AA_NET_CREATE) {
|
|
if (!prof.policy.rules->add_rule(buf.c_str(), priority,
|
|
rule_mode, map_perms(perms & AA_NET_CREATE) | (AA_CONT_MATCH << 1),
|
|
dedup_perms_rule_t::audit == AUDIT_FORCE ? map_perms(perms & AA_NET_CREATE) : 0,
|
|
parseopts))
|
|
return false;
|
|
}
|
|
|
|
/* encode protocol */
|
|
if (protocol > 0xffff) {
|
|
buffer << "..";
|
|
} else {
|
|
buffer << "\\x" << std::setfill('0') << std::setw(2) << std::hex << ((protocol & 0xff00) >> 8);
|
|
buffer << "\\x" << std::setfill('0') << std::setw(2) << std::hex << (protocol & 0xff);
|
|
}
|
|
|
|
if (perms & AA_PEER_NET_PERMS) {
|
|
for (int peer_port = peer.from_port; peer_port <= peer.to_port; peer_port++) {
|
|
std::list<std::ostringstream> streams;
|
|
std::ostringstream cmd_buffer;
|
|
|
|
cmd_buffer << buffer.str();
|
|
streams.push_back(std::move(cmd_buffer));
|
|
|
|
if (!gen_ip_conds(prof, streams, peer, true, peer_port, peer.is_port, false))
|
|
return false;
|
|
|
|
for (auto &oss : streams) {
|
|
oss << "\\x" << std::setfill('0') << std::setw(2) << std::hex << CMD_ADDR;
|
|
}
|
|
|
|
for (int local_port = local.from_port; local_port <= local.to_port; local_port++) {
|
|
std::list<std::ostringstream> localstreams;
|
|
|
|
for (auto &oss : streams) {
|
|
/* we need to copy streams because each local_port should be an unique entry */
|
|
std::ostringstream local_buffer;
|
|
local_buffer << oss.str();
|
|
localstreams.push_back(std::move(local_buffer));
|
|
}
|
|
|
|
if (!gen_ip_conds(prof, localstreams, local, false, local_port, local.is_port, true))
|
|
return false;
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
for (int local_port = local.from_port; local_port <= local.to_port; local_port++) {
|
|
std::list<std::ostringstream> streams;
|
|
std::ostringstream common_buffer;
|
|
|
|
common_buffer << buffer.str();
|
|
streams.push_back(std::move(common_buffer));
|
|
if (!gen_ip_conds(prof, streams, local, false, local_port, local.is_port, false))
|
|
return false;
|
|
|
|
if (perms & AA_NET_LISTEN) {
|
|
std::list<std::ostringstream> cmd_streams;
|
|
cmd_streams = copy_streams_list(streams);
|
|
|
|
for (auto &cmd_buffer : streams) {
|
|
std::ostringstream listen_buffer;
|
|
listen_buffer << cmd_buffer.str();
|
|
listen_buffer << "\\x" << std::setfill('0') << std::setw(2) << std::hex << CMD_LISTEN;
|
|
/* length of queue allowed - not used for now */
|
|
listen_buffer << "..";
|
|
buf = listen_buffer.str();
|
|
if (!prof.policy.rules->add_rule(buf.c_str(), priority,
|
|
rule_mode, map_perms(perms),
|
|
dedup_perms_rule_t::audit == AUDIT_FORCE ? map_perms(perms) : 0,
|
|
parseopts))
|
|
return false;
|
|
}
|
|
}
|
|
if (perms & AA_NET_OPT) {
|
|
std::list<std::ostringstream> cmd_streams;
|
|
cmd_streams = copy_streams_list(streams);
|
|
|
|
for (auto &cmd_buffer : streams) {
|
|
std::ostringstream opt_buffer;
|
|
opt_buffer << cmd_buffer.str();
|
|
opt_buffer << "\\x" << std::setfill('0') << std::setw(2) << std::hex << CMD_OPT;
|
|
/* level - not used for now */
|
|
opt_buffer << "..";
|
|
/* socket mapping - not used for now */
|
|
opt_buffer << "..";
|
|
buf = opt_buffer.str();
|
|
if (!prof.policy.rules->add_rule(buf.c_str(), priority,
|
|
rule_mode, map_perms(perms),
|
|
dedup_perms_rule_t::audit == AUDIT_FORCE ? map_perms(perms) : 0,
|
|
parseopts))
|
|
return false;
|
|
}
|
|
}
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
int network_rule::gen_policy_re(Profile &prof)
|
|
{
|
|
std::ostringstream buffer;
|
|
std::string buf;
|
|
|
|
if (!features_supports_networkv8) {
|
|
warn_once(prof.name);
|
|
return RULE_NOT_SUPPORTED;
|
|
}
|
|
|
|
for (const auto& perm : network_perms) {
|
|
unsigned int family = perm.first;
|
|
unsigned int type = perm.second.first;
|
|
unsigned int protocol = perm.second.second;
|
|
|
|
if (type > 0xffff) {
|
|
if (!gen_net_rule(prof, family, type, protocol))
|
|
goto fail;
|
|
} else {
|
|
int t;
|
|
/* generate rules for types that are set */
|
|
for (t = 0; t < 16; t++) {
|
|
if (type & (1 << t)) {
|
|
if (!gen_net_rule(prof, family, t, protocol))
|
|
goto fail;
|
|
}
|
|
}
|
|
}
|
|
|
|
}
|
|
return RULE_OK;
|
|
|
|
fail:
|
|
return RULE_ERROR;
|
|
|
|
}
|
|
|
|
/* initialize static members */
|
|
unsigned int *network_rule::allow = NULL;
|
|
unsigned int *network_rule::audit = NULL;
|
|
unsigned int *network_rule::deny = NULL;
|
|
unsigned int *network_rule::quiet = NULL;
|
|
|
|
bool network_rule::alloc_net_table()
|
|
{
|
|
if (allow)
|
|
return true;
|
|
allow = (unsigned int *) calloc(get_af_max(), sizeof(unsigned int));
|
|
audit = (unsigned int *) calloc(get_af_max(), sizeof(unsigned int));
|
|
deny = (unsigned int *) calloc(get_af_max(), sizeof(unsigned int));
|
|
quiet = (unsigned int *) calloc(get_af_max(), sizeof(unsigned int));
|
|
if (!allow || !audit || !deny || !quiet)
|
|
return false;
|
|
|
|
return true;
|
|
}
|
|
|
|
/* update is required because at the point of the creation of the
|
|
* network_rule object, we don't have owner, rule_mode, or audit
|
|
* set.
|
|
*/
|
|
void network_rule::update_compat_net(void)
|
|
{
|
|
if (!alloc_net_table())
|
|
yyerror(_("Memory allocation error."));
|
|
|
|
for (auto& nm: network_map) {
|
|
for (auto& entry : nm.second) {
|
|
if (entry.type > SOCK_PACKET) {
|
|
/* setting mask instead of a bit */
|
|
if (rule_mode == RULE_DENY) {
|
|
deny[entry.family] |= entry.type;
|
|
if (dedup_perms_rule_t::audit != AUDIT_FORCE)
|
|
quiet[entry.family] |= entry.type;
|
|
} else {
|
|
allow[entry.family] |= entry.type;
|
|
if (dedup_perms_rule_t::audit == AUDIT_FORCE)
|
|
audit[entry.family] |= entry.type;
|
|
}
|
|
} else {
|
|
if (rule_mode == RULE_DENY) {
|
|
deny[entry.family] |= 1 << entry.type;
|
|
if (dedup_perms_rule_t::audit != AUDIT_FORCE)
|
|
quiet[entry.family] |= 1 << entry.type;
|
|
} else {
|
|
allow[entry.family] |= 1 << entry.type;
|
|
if (dedup_perms_rule_t::audit == AUDIT_FORCE)
|
|
audit[entry.family] |= 1 << entry.type;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
static int cmp_ip_conds(ip_conds const &lhs, ip_conds const &rhs)
|
|
{
|
|
int res = null_strcmp(lhs.sip, rhs.sip);
|
|
if (res)
|
|
return res;
|
|
res = null_strcmp(lhs.sport, rhs.sport);
|
|
if (res)
|
|
return res;
|
|
return lhs.is_none - rhs.is_none;
|
|
}
|
|
|
|
static int cmp_network_map(std::unordered_map<unsigned int, std::pair<unsigned int, unsigned int>> lhs,
|
|
std::unordered_map<unsigned int, std::pair<unsigned int, unsigned int>> rhs)
|
|
{
|
|
int res;
|
|
size_t family_index;
|
|
for (family_index = AF_UNSPEC; family_index < get_af_max(); family_index++) {
|
|
res = lhs[family_index].first - rhs[family_index].first;
|
|
if (res)
|
|
return res;
|
|
res = lhs[family_index].second - rhs[family_index].second;
|
|
if (res)
|
|
return res;
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
int network_rule::cmp(rule_t const &rhs) const
|
|
{
|
|
int res = dedup_perms_rule_t::cmp(rhs);
|
|
if (res)
|
|
return res;
|
|
network_rule const &nrhs = rule_cast<network_rule const &>(rhs);
|
|
res = cmp_network_map(network_perms, nrhs.network_perms);
|
|
if (res)
|
|
return res;
|
|
res = cmp_ip_conds(local, nrhs.local);
|
|
if (res)
|
|
return res;
|
|
res = cmp_ip_conds(peer, nrhs.peer);
|
|
if (res)
|
|
return res;
|
|
return null_strcmp(label, nrhs.label);
|
|
};
|