diff options
Diffstat (limited to 'src/grp-resolve/systemd-resolved/resolved-dns-scope.c')
-rw-r--r-- | src/grp-resolve/systemd-resolved/resolved-dns-scope.c | 1044 |
1 files changed, 1044 insertions, 0 deletions
diff --git a/src/grp-resolve/systemd-resolved/resolved-dns-scope.c b/src/grp-resolve/systemd-resolved/resolved-dns-scope.c new file mode 100644 index 0000000000..9f20fd304a --- /dev/null +++ b/src/grp-resolve/systemd-resolved/resolved-dns-scope.c @@ -0,0 +1,1044 @@ +/*** + This file is part of systemd. + + Copyright 2014 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd 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 + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <netinet/tcp.h> + +#include "systemd-basic/af-list.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/hostname-util.h" +#include "systemd-basic/missing.h" +#include "systemd-basic/random-util.h" +#include "systemd-basic/socket-util.h" +#include "systemd-basic/strv.h" +#include "systemd-shared/dns-domain.h" + +#include "resolved-dns-scope.h" +#include "resolved-llmnr.h" +#include "resolved-mdns.h" + +#define MULTICAST_RATELIMIT_INTERVAL_USEC (1*USEC_PER_SEC) +#define MULTICAST_RATELIMIT_BURST 1000 + +/* After how much time to repeat LLMNR requests, see RFC 4795 Section 7 */ +#define MULTICAST_RESEND_TIMEOUT_MIN_USEC (100 * USEC_PER_MSEC) +#define MULTICAST_RESEND_TIMEOUT_MAX_USEC (1 * USEC_PER_SEC) + +int dns_scope_new(Manager *m, DnsScope **ret, Link *l, DnsProtocol protocol, int family) { + DnsScope *s; + + assert(m); + assert(ret); + + s = new0(DnsScope, 1); + if (!s) + return -ENOMEM; + + s->manager = m; + s->link = l; + s->protocol = protocol; + s->family = family; + s->resend_timeout = MULTICAST_RESEND_TIMEOUT_MIN_USEC; + + if (protocol == DNS_PROTOCOL_DNS) { + /* Copy DNSSEC mode from the link if it is set there, + * otherwise take the manager's DNSSEC mode. Note that + * we copy this only at scope creation time, and do + * not update it from the on, even if the setting + * changes. */ + + if (l) + s->dnssec_mode = link_get_dnssec_mode(l); + else + s->dnssec_mode = manager_get_dnssec_mode(m); + } else + s->dnssec_mode = DNSSEC_NO; + + LIST_PREPEND(scopes, m->dns_scopes, s); + + dns_scope_llmnr_membership(s, true); + dns_scope_mdns_membership(s, true); + + log_debug("New scope on link %s, protocol %s, family %s", l ? l->name : "*", dns_protocol_to_string(protocol), family == AF_UNSPEC ? "*" : af_to_name(family)); + + /* Enforce ratelimiting for the multicast protocols */ + RATELIMIT_INIT(s->ratelimit, MULTICAST_RATELIMIT_INTERVAL_USEC, MULTICAST_RATELIMIT_BURST); + + *ret = s; + return 0; +} + +static void dns_scope_abort_transactions(DnsScope *s) { + assert(s); + + while (s->transactions) { + DnsTransaction *t = s->transactions; + + /* Abort the transaction, but make sure it is not + * freed while we still look at it */ + + t->block_gc++; + if (DNS_TRANSACTION_IS_LIVE(t->state)) + dns_transaction_complete(t, DNS_TRANSACTION_ABORTED); + t->block_gc--; + + dns_transaction_free(t); + } +} + +DnsScope* dns_scope_free(DnsScope *s) { + DnsResourceRecord *rr; + + if (!s) + return NULL; + + log_debug("Removing scope on link %s, protocol %s, family %s", s->link ? s->link->name : "*", dns_protocol_to_string(s->protocol), s->family == AF_UNSPEC ? "*" : af_to_name(s->family)); + + dns_scope_llmnr_membership(s, false); + dns_scope_mdns_membership(s, false); + dns_scope_abort_transactions(s); + + while (s->query_candidates) + dns_query_candidate_free(s->query_candidates); + + hashmap_free(s->transactions_by_key); + + while ((rr = ordered_hashmap_steal_first(s->conflict_queue))) + dns_resource_record_unref(rr); + + ordered_hashmap_free(s->conflict_queue); + sd_event_source_unref(s->conflict_event_source); + + dns_cache_flush(&s->cache); + dns_zone_flush(&s->zone); + + LIST_REMOVE(scopes, s->manager->dns_scopes, s); + return mfree(s); +} + +DnsServer *dns_scope_get_dns_server(DnsScope *s) { + assert(s); + + if (s->protocol != DNS_PROTOCOL_DNS) + return NULL; + + if (s->link) + return link_get_dns_server(s->link); + else + return manager_get_dns_server(s->manager); +} + +void dns_scope_next_dns_server(DnsScope *s) { + assert(s); + + if (s->protocol != DNS_PROTOCOL_DNS) + return; + + if (s->link) + link_next_dns_server(s->link); + else + manager_next_dns_server(s->manager); +} + +void dns_scope_packet_received(DnsScope *s, usec_t rtt) { + assert(s); + + if (rtt <= s->max_rtt) + return; + + s->max_rtt = rtt; + s->resend_timeout = MIN(MAX(MULTICAST_RESEND_TIMEOUT_MIN_USEC, s->max_rtt * 2), MULTICAST_RESEND_TIMEOUT_MAX_USEC); +} + +void dns_scope_packet_lost(DnsScope *s, usec_t usec) { + assert(s); + + if (s->resend_timeout <= usec) + s->resend_timeout = MIN(s->resend_timeout * 2, MULTICAST_RESEND_TIMEOUT_MAX_USEC); +} + +static int dns_scope_emit_one(DnsScope *s, int fd, DnsPacket *p) { + union in_addr_union addr; + int ifindex = 0, r; + int family; + uint32_t mtu; + + assert(s); + assert(p); + assert(p->protocol == s->protocol); + + if (s->link) { + mtu = s->link->mtu; + ifindex = s->link->ifindex; + } else + mtu = manager_find_mtu(s->manager); + + switch (s->protocol) { + + case DNS_PROTOCOL_DNS: + assert(fd >= 0); + + if (DNS_PACKET_QDCOUNT(p) > 1) + return -EOPNOTSUPP; + + if (p->size > DNS_PACKET_UNICAST_SIZE_MAX) + return -EMSGSIZE; + + if (p->size + UDP_PACKET_HEADER_SIZE > mtu) + return -EMSGSIZE; + + r = manager_write(s->manager, fd, p); + if (r < 0) + return r; + + break; + + case DNS_PROTOCOL_LLMNR: + assert(fd < 0); + + if (DNS_PACKET_QDCOUNT(p) > 1) + return -EOPNOTSUPP; + + if (!ratelimit_test(&s->ratelimit)) + return -EBUSY; + + family = s->family; + + if (family == AF_INET) { + addr.in = LLMNR_MULTICAST_IPV4_ADDRESS; + fd = manager_llmnr_ipv4_udp_fd(s->manager); + } else if (family == AF_INET6) { + addr.in6 = LLMNR_MULTICAST_IPV6_ADDRESS; + fd = manager_llmnr_ipv6_udp_fd(s->manager); + } else + return -EAFNOSUPPORT; + if (fd < 0) + return fd; + + r = manager_send(s->manager, fd, ifindex, family, &addr, LLMNR_PORT, NULL, p); + if (r < 0) + return r; + + break; + + case DNS_PROTOCOL_MDNS: + assert(fd < 0); + + if (!ratelimit_test(&s->ratelimit)) + return -EBUSY; + + family = s->family; + + if (family == AF_INET) { + addr.in = MDNS_MULTICAST_IPV4_ADDRESS; + fd = manager_mdns_ipv4_fd(s->manager); + } else if (family == AF_INET6) { + addr.in6 = MDNS_MULTICAST_IPV6_ADDRESS; + fd = manager_mdns_ipv6_fd(s->manager); + } else + return -EAFNOSUPPORT; + if (fd < 0) + return fd; + + r = manager_send(s->manager, fd, ifindex, family, &addr, MDNS_PORT, NULL, p); + if (r < 0) + return r; + + break; + + default: + return -EAFNOSUPPORT; + } + + return 1; +} + +int dns_scope_emit_udp(DnsScope *s, int fd, DnsPacket *p) { + int r; + + assert(s); + assert(p); + assert(p->protocol == s->protocol); + assert((s->protocol == DNS_PROTOCOL_DNS) == (fd >= 0)); + + do { + /* If there are multiple linked packets, set the TC bit in all but the last of them */ + if (p->more) { + assert(p->protocol == DNS_PROTOCOL_MDNS); + dns_packet_set_flags(p, true, true); + } + + r = dns_scope_emit_one(s, fd, p); + if (r < 0) + return r; + + p = p->more; + } while (p); + + return 0; +} + +static int dns_scope_socket( + DnsScope *s, + int type, + int family, + const union in_addr_union *address, + DnsServer *server, + uint16_t port) { + + _cleanup_close_ int fd = -1; + union sockaddr_union sa = {}; + socklen_t salen; + static const int one = 1; + int ret, r, ifindex; + + assert(s); + + if (server) { + assert(family == AF_UNSPEC); + assert(!address); + + ifindex = dns_server_ifindex(server); + + sa.sa.sa_family = server->family; + if (server->family == AF_INET) { + sa.in.sin_port = htobe16(port); + sa.in.sin_addr = server->address.in; + salen = sizeof(sa.in); + } else if (server->family == AF_INET6) { + sa.in6.sin6_port = htobe16(port); + sa.in6.sin6_addr = server->address.in6; + sa.in6.sin6_scope_id = ifindex; + salen = sizeof(sa.in6); + } else + return -EAFNOSUPPORT; + } else { + assert(family != AF_UNSPEC); + assert(address); + + sa.sa.sa_family = family; + ifindex = s->link ? s->link->ifindex : 0; + + if (family == AF_INET) { + sa.in.sin_port = htobe16(port); + sa.in.sin_addr = address->in; + salen = sizeof(sa.in); + } else if (family == AF_INET6) { + sa.in6.sin6_port = htobe16(port); + sa.in6.sin6_addr = address->in6; + sa.in6.sin6_scope_id = ifindex; + salen = sizeof(sa.in6); + } else + return -EAFNOSUPPORT; + } + + fd = socket(sa.sa.sa_family, type|SOCK_CLOEXEC|SOCK_NONBLOCK, 0); + if (fd < 0) + return -errno; + + if (type == SOCK_STREAM) { + r = setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &one, sizeof(one)); + if (r < 0) + return -errno; + } + + if (s->link) { + be32_t ifindex_be = htobe32(ifindex); + + if (sa.sa.sa_family == AF_INET) { + r = setsockopt(fd, IPPROTO_IP, IP_UNICAST_IF, &ifindex_be, sizeof(ifindex_be)); + if (r < 0) + return -errno; + } else if (sa.sa.sa_family == AF_INET6) { + r = setsockopt(fd, IPPROTO_IPV6, IPV6_UNICAST_IF, &ifindex_be, sizeof(ifindex_be)); + if (r < 0) + return -errno; + } + } + + if (s->protocol == DNS_PROTOCOL_LLMNR) { + /* RFC 4795, section 2.5 requires the TTL to be set to 1 */ + + if (sa.sa.sa_family == AF_INET) { + r = setsockopt(fd, IPPROTO_IP, IP_TTL, &one, sizeof(one)); + if (r < 0) + return -errno; + } else if (sa.sa.sa_family == AF_INET6) { + r = setsockopt(fd, IPPROTO_IPV6, IPV6_UNICAST_HOPS, &one, sizeof(one)); + if (r < 0) + return -errno; + } + } + + r = connect(fd, &sa.sa, salen); + if (r < 0 && errno != EINPROGRESS) + return -errno; + + ret = fd; + fd = -1; + + return ret; +} + +int dns_scope_socket_udp(DnsScope *s, DnsServer *server, uint16_t port) { + return dns_scope_socket(s, SOCK_DGRAM, AF_UNSPEC, NULL, server, port); +} + +int dns_scope_socket_tcp(DnsScope *s, int family, const union in_addr_union *address, DnsServer *server, uint16_t port) { + return dns_scope_socket(s, SOCK_STREAM, family, address, server, port); +} + +DnsScopeMatch dns_scope_good_domain(DnsScope *s, int ifindex, uint64_t flags, const char *domain) { + DnsSearchDomain *d; + DnsServer *dns_server; + + assert(s); + assert(domain); + + /* Checks if the specified domain is something to look up on + * this scope. Note that this accepts non-qualified hostnames, + * i.e. those without any search path prefixed yet. */ + + if (ifindex != 0 && (!s->link || s->link->ifindex != ifindex)) + return DNS_SCOPE_NO; + + if ((SD_RESOLVED_FLAGS_MAKE(s->protocol, s->family, 0) & flags) == 0) + return DNS_SCOPE_NO; + + /* Never resolve any loopback hostname or IP address via DNS, + * LLMNR or mDNS. Instead, always rely on synthesized RRs for + * these. */ + if (is_localhost(domain) || + dns_name_endswith(domain, "127.in-addr.arpa") > 0 || + dns_name_equal(domain, "1.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.ip6.arpa") > 0) + return DNS_SCOPE_NO; + + /* Never respond to some of the domains listed in RFC6303 */ + if (dns_name_endswith(domain, "0.in-addr.arpa") > 0 || + dns_name_equal(domain, "255.255.255.255.in-addr.arpa") > 0 || + dns_name_equal(domain, "0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.ip6.arpa") > 0) + return DNS_SCOPE_NO; + + /* Never respond to some of the domains listed in RFC6761 */ + if (dns_name_endswith(domain, "invalid") > 0) + return DNS_SCOPE_NO; + + /* Always honour search domains for routing queries. Note that + * we return DNS_SCOPE_YES here, rather than just + * DNS_SCOPE_MAYBE, which means wildcard scopes won't be + * considered anymore. */ + LIST_FOREACH(domains, d, dns_scope_get_search_domains(s)) + if (dns_name_endswith(domain, d->name) > 0) + return DNS_SCOPE_YES; + + /* If the DNS server has route-only domains, don't send other requests + * to it. This would be a privacy violation, will most probably fail + * anyway, and adds unnecessary load. */ + dns_server = dns_scope_get_dns_server(s); + if (dns_server && dns_server_limited_domains(dns_server)) + return DNS_SCOPE_NO; + + switch (s->protocol) { + + case DNS_PROTOCOL_DNS: + + /* Exclude link-local IP ranges */ + if (dns_name_endswith(domain, "254.169.in-addr.arpa") == 0 && + dns_name_endswith(domain, "8.e.f.ip6.arpa") == 0 && + dns_name_endswith(domain, "9.e.f.ip6.arpa") == 0 && + dns_name_endswith(domain, "a.e.f.ip6.arpa") == 0 && + dns_name_endswith(domain, "b.e.f.ip6.arpa") == 0 && + /* If networks use .local in their private setups, they are supposed to also add .local to their search + * domains, which we already checked above. Otherwise, we consider .local specific to mDNS and won't + * send such queries ordinary DNS servers. */ + dns_name_endswith(domain, "local") == 0) + return DNS_SCOPE_MAYBE; + + return DNS_SCOPE_NO; + + case DNS_PROTOCOL_MDNS: + if ((s->family == AF_INET && dns_name_endswith(domain, "in-addr.arpa") > 0) || + (s->family == AF_INET6 && dns_name_endswith(domain, "ip6.arpa") > 0) || + (dns_name_endswith(domain, "local") > 0 && /* only resolve names ending in .local via mDNS */ + dns_name_equal(domain, "local") == 0 && /* but not the single-label "local" name itself */ + manager_is_own_hostname(s->manager, domain) <= 0)) /* never resolve the local hostname via mDNS */ + return DNS_SCOPE_MAYBE; + + return DNS_SCOPE_NO; + + case DNS_PROTOCOL_LLMNR: + if ((s->family == AF_INET && dns_name_endswith(domain, "in-addr.arpa") > 0) || + (s->family == AF_INET6 && dns_name_endswith(domain, "ip6.arpa") > 0) || + (dns_name_is_single_label(domain) && /* only resolve single label names via LLMNR */ + !is_gateway_hostname(domain) && /* don't resolve "gateway" with LLMNR, let nss-myhostname handle this */ + manager_is_own_hostname(s->manager, domain) <= 0)) /* never resolve the local hostname via LLMNR */ + return DNS_SCOPE_MAYBE; + + return DNS_SCOPE_NO; + + default: + assert_not_reached("Unknown scope protocol"); + } +} + +bool dns_scope_good_key(DnsScope *s, const DnsResourceKey *key) { + int key_family; + + assert(s); + assert(key); + + /* Check if it makes sense to resolve the specified key on + * this scope. Note that this call assumes as fully qualified + * name, i.e. the search suffixes already appended. */ + + if (key->class != DNS_CLASS_IN) + return false; + + if (s->protocol == DNS_PROTOCOL_DNS) { + + /* On classic DNS, looking up non-address RRs is always + * fine. (Specifically, we want to permit looking up + * DNSKEY and DS records on the root and top-level + * domains.) */ + if (!dns_resource_key_is_address(key)) + return true; + + /* However, we refuse to look up A and AAAA RRs on the + * root and single-label domains, under the assumption + * that those should be resolved via LLMNR or search + * path only, and should not be leaked onto the + * internet. */ + return !(dns_name_is_single_label(dns_resource_key_name(key)) || + dns_name_is_root(dns_resource_key_name(key))); + } + + /* On mDNS and LLMNR, send A and AAAA queries only on the + * respective scopes */ + + key_family = dns_type_to_af(key->type); + if (key_family < 0) + return true; + + return key_family == s->family; +} + +static int dns_scope_multicast_membership(DnsScope *s, bool b, struct in_addr in, struct in6_addr in6) { + int fd; + + assert(s); + assert(s->link); + + if (s->family == AF_INET) { + struct ip_mreqn mreqn = { + .imr_multiaddr = in, + .imr_ifindex = s->link->ifindex, + }; + + fd = manager_llmnr_ipv4_udp_fd(s->manager); + if (fd < 0) + return fd; + + /* Always first try to drop membership before we add + * one. This is necessary on some devices, such as + * veth. */ + if (b) + (void) setsockopt(fd, IPPROTO_IP, IP_DROP_MEMBERSHIP, &mreqn, sizeof(mreqn)); + + if (setsockopt(fd, IPPROTO_IP, b ? IP_ADD_MEMBERSHIP : IP_DROP_MEMBERSHIP, &mreqn, sizeof(mreqn)) < 0) + return -errno; + + } else if (s->family == AF_INET6) { + struct ipv6_mreq mreq = { + .ipv6mr_multiaddr = in6, + .ipv6mr_interface = s->link->ifindex, + }; + + fd = manager_llmnr_ipv6_udp_fd(s->manager); + if (fd < 0) + return fd; + + if (b) + (void) setsockopt(fd, IPPROTO_IPV6, IPV6_DROP_MEMBERSHIP, &mreq, sizeof(mreq)); + + if (setsockopt(fd, IPPROTO_IPV6, b ? IPV6_ADD_MEMBERSHIP : IPV6_DROP_MEMBERSHIP, &mreq, sizeof(mreq)) < 0) + return -errno; + } else + return -EAFNOSUPPORT; + + return 0; +} + +int dns_scope_llmnr_membership(DnsScope *s, bool b) { + assert(s); + + if (s->protocol != DNS_PROTOCOL_LLMNR) + return 0; + + return dns_scope_multicast_membership(s, b, LLMNR_MULTICAST_IPV4_ADDRESS, LLMNR_MULTICAST_IPV6_ADDRESS); +} + +int dns_scope_mdns_membership(DnsScope *s, bool b) { + assert(s); + + if (s->protocol != DNS_PROTOCOL_MDNS) + return 0; + + return dns_scope_multicast_membership(s, b, MDNS_MULTICAST_IPV4_ADDRESS, MDNS_MULTICAST_IPV6_ADDRESS); +} + +static int dns_scope_make_reply_packet( + DnsScope *s, + uint16_t id, + int rcode, + DnsQuestion *q, + DnsAnswer *answer, + DnsAnswer *soa, + bool tentative, + DnsPacket **ret) { + + _cleanup_(dns_packet_unrefp) DnsPacket *p = NULL; + int r; + + assert(s); + assert(ret); + + if (dns_question_isempty(q) && + dns_answer_isempty(answer) && + dns_answer_isempty(soa)) + return -EINVAL; + + r = dns_packet_new(&p, s->protocol, 0); + if (r < 0) + return r; + + DNS_PACKET_HEADER(p)->id = id; + DNS_PACKET_HEADER(p)->flags = htobe16(DNS_PACKET_MAKE_FLAGS( + 1 /* qr */, + 0 /* opcode */, + 0 /* c */, + 0 /* tc */, + tentative, + 0 /* (ra) */, + 0 /* (ad) */, + 0 /* (cd) */, + rcode)); + + r = dns_packet_append_question(p, q); + if (r < 0) + return r; + DNS_PACKET_HEADER(p)->qdcount = htobe16(dns_question_size(q)); + + r = dns_packet_append_answer(p, answer); + if (r < 0) + return r; + DNS_PACKET_HEADER(p)->ancount = htobe16(dns_answer_size(answer)); + + r = dns_packet_append_answer(p, soa); + if (r < 0) + return r; + DNS_PACKET_HEADER(p)->arcount = htobe16(dns_answer_size(soa)); + + *ret = p; + p = NULL; + + return 0; +} + +static void dns_scope_verify_conflicts(DnsScope *s, DnsPacket *p) { + DnsResourceRecord *rr; + DnsResourceKey *key; + + assert(s); + assert(p); + + DNS_QUESTION_FOREACH(key, p->question) + dns_zone_verify_conflicts(&s->zone, key); + + DNS_ANSWER_FOREACH(rr, p->answer) + dns_zone_verify_conflicts(&s->zone, rr->key); +} + +void dns_scope_process_query(DnsScope *s, DnsStream *stream, DnsPacket *p) { + _cleanup_(dns_answer_unrefp) DnsAnswer *answer = NULL, *soa = NULL; + _cleanup_(dns_packet_unrefp) DnsPacket *reply = NULL; + DnsResourceKey *key = NULL; + bool tentative = false; + int r; + + assert(s); + assert(p); + + if (p->protocol != DNS_PROTOCOL_LLMNR) + return; + + if (p->ipproto == IPPROTO_UDP) { + /* Don't accept UDP queries directed to anything but + * the LLMNR multicast addresses. See RFC 4795, + * section 2.5. */ + + if (p->family == AF_INET && !in_addr_equal(AF_INET, &p->destination, (union in_addr_union*) &LLMNR_MULTICAST_IPV4_ADDRESS)) + return; + + if (p->family == AF_INET6 && !in_addr_equal(AF_INET6, &p->destination, (union in_addr_union*) &LLMNR_MULTICAST_IPV6_ADDRESS)) + return; + } + + r = dns_packet_extract(p); + if (r < 0) { + log_debug_errno(r, "Failed to extract resource records from incoming packet: %m"); + return; + } + + if (DNS_PACKET_LLMNR_C(p)) { + /* Somebody notified us about a possible conflict */ + dns_scope_verify_conflicts(s, p); + return; + } + + assert(dns_question_size(p->question) == 1); + key = p->question->keys[0]; + + r = dns_zone_lookup(&s->zone, key, 0, &answer, &soa, &tentative); + if (r < 0) { + log_debug_errno(r, "Failed to lookup key: %m"); + return; + } + if (r == 0) + return; + + if (answer) + dns_answer_order_by_scope(answer, in_addr_is_link_local(p->family, &p->sender) > 0); + + r = dns_scope_make_reply_packet(s, DNS_PACKET_ID(p), DNS_RCODE_SUCCESS, p->question, answer, soa, tentative, &reply); + if (r < 0) { + log_debug_errno(r, "Failed to build reply packet: %m"); + return; + } + + if (stream) { + r = dns_stream_write_packet(stream, reply); + if (r < 0) { + log_debug_errno(r, "Failed to enqueue reply packet: %m"); + return; + } + + /* Let's take an extra reference on this stream, so that it stays around after returning. The reference + * will be dangling until the stream is disconnected, and the default completion handler of the stream + * will then unref the stream and destroy it */ + if (DNS_STREAM_QUEUED(stream)) + dns_stream_ref(stream); + } else { + int fd; + + if (!ratelimit_test(&s->ratelimit)) + return; + + if (p->family == AF_INET) + fd = manager_llmnr_ipv4_udp_fd(s->manager); + else if (p->family == AF_INET6) + fd = manager_llmnr_ipv6_udp_fd(s->manager); + else { + log_debug("Unknown protocol"); + return; + } + if (fd < 0) { + log_debug_errno(fd, "Failed to get reply socket: %m"); + return; + } + + /* Note that we always immediately reply to all LLMNR + * requests, and do not wait any time, since we + * verified uniqueness for all records. Also see RFC + * 4795, Section 2.7 */ + + r = manager_send(s->manager, fd, p->ifindex, p->family, &p->sender, p->sender_port, NULL, reply); + if (r < 0) { + log_debug_errno(r, "Failed to send reply packet: %m"); + return; + } + } +} + +DnsTransaction *dns_scope_find_transaction(DnsScope *scope, DnsResourceKey *key, bool cache_ok) { + DnsTransaction *t; + + assert(scope); + assert(key); + + /* Try to find an ongoing transaction that is a equal to the + * specified question */ + t = hashmap_get(scope->transactions_by_key, key); + if (!t) + return NULL; + + /* Refuse reusing transactions that completed based on cached + * data instead of a real packet, if that's requested. */ + if (!cache_ok && + IN_SET(t->state, DNS_TRANSACTION_SUCCESS, DNS_TRANSACTION_RCODE_FAILURE) && + t->answer_source != DNS_TRANSACTION_NETWORK) + return NULL; + + return t; +} + +static int dns_scope_make_conflict_packet( + DnsScope *s, + DnsResourceRecord *rr, + DnsPacket **ret) { + + _cleanup_(dns_packet_unrefp) DnsPacket *p = NULL; + int r; + + assert(s); + assert(rr); + assert(ret); + + r = dns_packet_new(&p, s->protocol, 0); + if (r < 0) + return r; + + DNS_PACKET_HEADER(p)->flags = htobe16(DNS_PACKET_MAKE_FLAGS( + 0 /* qr */, + 0 /* opcode */, + 1 /* conflict */, + 0 /* tc */, + 0 /* t */, + 0 /* (ra) */, + 0 /* (ad) */, + 0 /* (cd) */, + 0)); + + /* For mDNS, the transaction ID should always be 0 */ + if (s->protocol != DNS_PROTOCOL_MDNS) + random_bytes(&DNS_PACKET_HEADER(p)->id, sizeof(uint16_t)); + + DNS_PACKET_HEADER(p)->qdcount = htobe16(1); + DNS_PACKET_HEADER(p)->arcount = htobe16(1); + + r = dns_packet_append_key(p, rr->key, NULL); + if (r < 0) + return r; + + r = dns_packet_append_rr(p, rr, NULL, NULL); + if (r < 0) + return r; + + *ret = p; + p = NULL; + + return 0; +} + +static int on_conflict_dispatch(sd_event_source *es, usec_t usec, void *userdata) { + DnsScope *scope = userdata; + int r; + + assert(es); + assert(scope); + + scope->conflict_event_source = sd_event_source_unref(scope->conflict_event_source); + + for (;;) { + _cleanup_(dns_resource_record_unrefp) DnsResourceRecord *rr = NULL; + _cleanup_(dns_packet_unrefp) DnsPacket *p = NULL; + + rr = ordered_hashmap_steal_first(scope->conflict_queue); + if (!rr) + break; + + r = dns_scope_make_conflict_packet(scope, rr, &p); + if (r < 0) { + log_error_errno(r, "Failed to make conflict packet: %m"); + return 0; + } + + r = dns_scope_emit_udp(scope, -1, p); + if (r < 0) + log_debug_errno(r, "Failed to send conflict packet: %m"); + } + + return 0; +} + +int dns_scope_notify_conflict(DnsScope *scope, DnsResourceRecord *rr) { + usec_t jitter; + int r; + + assert(scope); + assert(rr); + + /* We don't send these queries immediately. Instead, we queue + * them, and send them after some jitter delay. */ + r = ordered_hashmap_ensure_allocated(&scope->conflict_queue, &dns_resource_key_hash_ops); + if (r < 0) { + log_oom(); + return r; + } + + /* We only place one RR per key in the conflict + * messages, not all of them. That should be enough to + * indicate where there might be a conflict */ + r = ordered_hashmap_put(scope->conflict_queue, rr->key, rr); + if (r == -EEXIST || r == 0) + return 0; + if (r < 0) + return log_debug_errno(r, "Failed to queue conflicting RR: %m"); + + dns_resource_record_ref(rr); + + if (scope->conflict_event_source) + return 0; + + random_bytes(&jitter, sizeof(jitter)); + jitter %= LLMNR_JITTER_INTERVAL_USEC; + + r = sd_event_add_time(scope->manager->event, + &scope->conflict_event_source, + clock_boottime_or_monotonic(), + now(clock_boottime_or_monotonic()) + jitter, + LLMNR_JITTER_INTERVAL_USEC, + on_conflict_dispatch, scope); + if (r < 0) + return log_debug_errno(r, "Failed to add conflict dispatch event: %m"); + + (void) sd_event_source_set_description(scope->conflict_event_source, "scope-conflict"); + + return 0; +} + +void dns_scope_check_conflicts(DnsScope *scope, DnsPacket *p) { + unsigned i; + int r; + + assert(scope); + assert(p); + + if (p->protocol != DNS_PROTOCOL_LLMNR) + return; + + if (DNS_PACKET_RRCOUNT(p) <= 0) + return; + + if (DNS_PACKET_LLMNR_C(p) != 0) + return; + + if (DNS_PACKET_LLMNR_T(p) != 0) + return; + + if (manager_our_packet(scope->manager, p)) + return; + + r = dns_packet_extract(p); + if (r < 0) { + log_debug_errno(r, "Failed to extract packet: %m"); + return; + } + + log_debug("Checking for conflicts..."); + + for (i = 0; i < p->answer->n_rrs; i++) { + + /* Check for conflicts against the local zone. If we + * found one, we won't check any further */ + r = dns_zone_check_conflicts(&scope->zone, p->answer->items[i].rr); + if (r != 0) + continue; + + /* Check for conflicts against the local cache. If so, + * send out an advisory query, to inform everybody */ + r = dns_cache_check_conflicts(&scope->cache, p->answer->items[i].rr, p->family, &p->sender); + if (r <= 0) + continue; + + dns_scope_notify_conflict(scope, p->answer->items[i].rr); + } +} + +void dns_scope_dump(DnsScope *s, FILE *f) { + assert(s); + + if (!f) + f = stdout; + + fputs("[Scope protocol=", f); + fputs(dns_protocol_to_string(s->protocol), f); + + if (s->link) { + fputs(" interface=", f); + fputs(s->link->name, f); + } + + if (s->family != AF_UNSPEC) { + fputs(" family=", f); + fputs(af_to_name(s->family), f); + } + + fputs("]\n", f); + + if (!dns_zone_is_empty(&s->zone)) { + fputs("ZONE:\n", f); + dns_zone_dump(&s->zone, f); + } + + if (!dns_cache_is_empty(&s->cache)) { + fputs("CACHE:\n", f); + dns_cache_dump(&s->cache, f); + } +} + +DnsSearchDomain *dns_scope_get_search_domains(DnsScope *s) { + assert(s); + + if (s->protocol != DNS_PROTOCOL_DNS) + return NULL; + + if (s->link) + return s->link->search_domains; + + return s->manager->search_domains; +} + +bool dns_scope_name_needs_search_domain(DnsScope *s, const char *name) { + assert(s); + + if (s->protocol != DNS_PROTOCOL_DNS) + return false; + + return dns_name_is_single_label(name); +} + +bool dns_scope_network_good(DnsScope *s) { + /* Checks whether the network is in good state for lookups on this scope. For mDNS/LLMNR/Classic DNS scopes + * bound to links this is easy, as they don't even exist if the link isn't in a suitable state. For the global + * DNS scope we check whether there are any links that are up and have an address. */ + + if (s->link) + return true; + + return manager_routable(s->manager, AF_UNSPEC); +} + +int dns_scope_ifindex(DnsScope *s) { + assert(s); + + if (s->link) + return s->link->ifindex; + + return 0; +} |