/* $OpenBSD: frontend.c,v 1.80 2023/12/14 09:59:27 claudio Exp $ */ /* * Copyright (c) 2018 Florian Obser * Copyright (c) 2005 Claudio Jeker * Copyright (c) 2004 Esben Norby * Copyright (c) 2003, 2004 Henning Brauer * * 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 #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "libunbound/config.h" #include "libunbound/sldns/pkthdr.h" #include "libunbound/sldns/sbuffer.h" #include "libunbound/sldns/str2wire.h" #include "libunbound/sldns/wire2str.h" #include "libunbound/util/alloc.h" #include "libunbound/util/net_help.h" #include "libunbound/util/regional.h" #include "libunbound/util/data/dname.h" #include "libunbound/util/data/msgencode.h" #include "libunbound/util/data/msgparse.h" #include "libunbound/util/data/msgreply.h" #include "log.h" #include "unwind.h" #include "frontend.h" #include "control.h" #include "dns64_synth.h" #define MINIMUM(a, b) (((a) < (b)) ? (a) : (b)) #define ROUTE_SOCKET_BUF_SIZE 16384 /* * size of a resource record with name a two octed pointer to qname * 2 octets pointer to qname * 2 octets TYPE * 2 octets CLASS * 4 octets TTL * 2 octets RDLENGTH */ #define COMPRESSED_RR_SIZE 12 #define MINIMIZE_ANSWER 1 #define FD_RESERVE 5 #define TCP_TIMEOUT 15 #define DEFAULT_TCP_SIZE 512 struct udp_ev { struct event ev; uint8_t query[65536]; struct msghdr rcvmhdr; struct iovec rcviov[1]; struct sockaddr_storage from; } udp4ev, udp6ev; struct tcp_accept_ev { struct event ev; struct event pause; } tcp4ev, tcp6ev; struct pending_query { TAILQ_ENTRY(pending_query) entry; struct sockaddr_storage from; struct sldns_buffer *qbuf; struct sldns_buffer *abuf; struct regional *region; struct query_info qinfo; struct edns_data edns; struct event ev; /* for tcp */ struct event resp_ev; /* for tcp */ struct event tmo_ev; /* for tcp */ uint64_t imsg_id; uint16_t id; uint16_t flags; int fd; int tcp; int dns64_synthesize; }; TAILQ_HEAD(, pending_query) pending_queries; struct bl_node { RB_ENTRY(bl_node) entry; char *domain; }; __dead void frontend_shutdown(void); void frontend_sig_handler(int, short, void *); void frontend_startup(void); void udp_receive(int, short, void *); void handle_query(struct pending_query *); void free_pending_query(struct pending_query *); void tcp_accept(int, short, void *); int accept_reserve(int, struct sockaddr *, socklen_t *); void accept_paused(int, short, void *); void tcp_request(int, short, void *); void tcp_response(int, short, void *); void tcp_timeout(int, short, void *); int check_query(sldns_buffer*); void noerror_answer(struct pending_query *); void synthesize_dns64_answer(struct pending_query *); void resend_dns64_query(struct pending_query *); void chaos_answer(struct pending_query *); void error_answer(struct pending_query *, int rcode); void send_answer(struct pending_query *); void route_receive(int, short, void *); void handle_route_message(struct rt_msghdr *, struct sockaddr **); void get_rtaddrs(int, struct sockaddr *, struct sockaddr **); void rtmget_default(void); struct pending_query *find_pending_query(uint64_t); void parse_trust_anchor(struct trust_anchor_head *, int); void send_trust_anchors(struct trust_anchor_head *); void write_trust_anchors(struct trust_anchor_head *, int); void parse_blocklist(int); int bl_cmp(struct bl_node *, struct bl_node *); void free_bl(void); int pending_query_cnt(void); void check_available_af(void); struct uw_conf *frontend_conf; static struct imsgev *iev_main; static struct imsgev *iev_resolver; struct event ev_route; int udp4sock = -1, udp6sock = -1; int tcp4sock = -1, tcp6sock = -1; int ta_fd = -1; static struct trust_anchor_head trust_anchors, new_trust_anchors; RB_HEAD(bl_tree, bl_node) bl_head = RB_INITIALIZER(&bl_head); RB_PROTOTYPE(bl_tree, bl_node, entry, bl_cmp) RB_GENERATE(bl_tree, bl_node, entry, bl_cmp) struct dns64_prefix *dns64_prefixes; int dns64_prefix_count; void frontend_sig_handler(int sig, short event, void *bula) { /* * Normal signal handler rules don't apply because libevent * decouples for us. */ switch (sig) { case SIGINT: case SIGTERM: frontend_shutdown(); default: fatalx("unexpected signal"); } } void frontend(int debug, int verbose) { struct event ev_sigint, ev_sigterm; struct passwd *pw; frontend_conf = config_new_empty(); log_init(debug, LOG_DAEMON); log_setverbose(verbose); if ((pw = getpwnam(UNWIND_USER)) == NULL) fatal("getpwnam"); if (chroot(pw->pw_dir) == -1) fatal("chroot"); if (chdir("/") == -1) fatal("chdir(\"/\")"); setproctitle("%s", "frontend"); log_procinit("frontend"); if (setgroups(1, &pw->pw_gid) || setresgid(pw->pw_gid, pw->pw_gid, pw->pw_gid) || setresuid(pw->pw_uid, pw->pw_uid, pw->pw_uid)) fatal("can't drop privileges"); if (pledge("stdio dns unix recvfd", NULL) == -1) fatal("pledge"); event_init(); /* Setup signal handler. */ signal_set(&ev_sigint, SIGINT, frontend_sig_handler, NULL); signal_set(&ev_sigterm, SIGTERM, frontend_sig_handler, NULL); signal_add(&ev_sigint, NULL); signal_add(&ev_sigterm, NULL); signal(SIGPIPE, SIG_IGN); signal(SIGHUP, SIG_IGN); /* Setup pipe and event handler to the parent process. */ if (iev_main != NULL) fatal("iev_main"); if ((iev_main = malloc(sizeof(struct imsgev))) == NULL) fatal(NULL); imsg_init(&iev_main->ibuf, 3); iev_main->handler = frontend_dispatch_main; iev_main->events = EV_READ; event_set(&iev_main->ev, iev_main->ibuf.fd, iev_main->events, iev_main->handler, iev_main); event_add(&iev_main->ev, NULL); udp4ev.rcviov[0].iov_base = (caddr_t)udp4ev.query; udp4ev.rcviov[0].iov_len = sizeof(udp4ev.query); udp4ev.rcvmhdr.msg_name = (caddr_t)&udp4ev.from; udp4ev.rcvmhdr.msg_namelen = sizeof(udp4ev.from); udp4ev.rcvmhdr.msg_iov = udp4ev.rcviov; udp4ev.rcvmhdr.msg_iovlen = 1; udp6ev.rcviov[0].iov_base = (caddr_t)udp6ev.query; udp6ev.rcviov[0].iov_len = sizeof(udp6ev.query); udp6ev.rcvmhdr.msg_name = (caddr_t)&udp6ev.from; udp6ev.rcvmhdr.msg_namelen = sizeof(udp6ev.from); udp6ev.rcvmhdr.msg_iov = udp6ev.rcviov; udp6ev.rcvmhdr.msg_iovlen = 1; TAILQ_INIT(&pending_queries); TAILQ_INIT(&trust_anchors); TAILQ_INIT(&new_trust_anchors); add_new_ta(&trust_anchors, KSK2017); event_dispatch(); frontend_shutdown(); } __dead void frontend_shutdown(void) { /* Close pipes. */ msgbuf_write(&iev_resolver->ibuf.w); msgbuf_clear(&iev_resolver->ibuf.w); close(iev_resolver->ibuf.fd); msgbuf_write(&iev_main->ibuf.w); msgbuf_clear(&iev_main->ibuf.w); close(iev_main->ibuf.fd); config_clear(frontend_conf); free(iev_resolver); free(iev_main); log_info("frontend exiting"); exit(0); } int frontend_imsg_compose_main(int type, pid_t pid, void *data, uint16_t datalen) { return (imsg_compose_event(iev_main, type, 0, pid, -1, data, datalen)); } int frontend_imsg_compose_resolver(int type, pid_t pid, void *data, uint16_t datalen) { return (imsg_compose_event(iev_resolver, type, 0, pid, -1, data, datalen)); } void frontend_dispatch_main(int fd, short event, void *bula) { static struct uw_conf *nconf; struct imsg imsg; struct imsgev *iev = bula; struct imsgbuf *ibuf = &iev->ibuf; int n, shut = 0; if (event & EV_READ) { if ((n = imsg_read(ibuf)) == -1 && errno != EAGAIN) fatal("imsg_read error"); if (n == 0) /* Connection closed. */ shut = 1; } if (event & EV_WRITE) { if ((n = msgbuf_write(&ibuf->w)) == -1 && errno != EAGAIN) fatal("msgbuf_write"); if (n == 0) /* Connection closed. */ shut = 1; } for (;;) { if ((n = imsg_get(ibuf, &imsg)) == -1) fatal("%s: imsg_get error", __func__); if (n == 0) /* No more messages. */ break; switch (imsg.hdr.type) { case IMSG_SOCKET_IPC_RESOLVER: /* * Setup pipe and event handler to the resolver * process. */ if (iev_resolver) { fatalx("%s: received unexpected imsg fd " "to frontend", __func__); break; } if ((fd = imsg_get_fd(&imsg)) == -1) { fatalx("%s: expected to receive imsg fd to " "frontend but didn't receive any", __func__); break; } if (iev_resolver != NULL) fatal("iev_resolver"); iev_resolver = malloc(sizeof(struct imsgev)); if (iev_resolver == NULL) fatal(NULL); imsg_init(&iev_resolver->ibuf, fd); iev_resolver->handler = frontend_dispatch_resolver; iev_resolver->events = EV_READ; event_set(&iev_resolver->ev, iev_resolver->ibuf.fd, iev_resolver->events, iev_resolver->handler, iev_resolver); event_add(&iev_resolver->ev, NULL); break; case IMSG_RECONF_CONF: case IMSG_RECONF_BLOCKLIST_FILE: case IMSG_RECONF_FORWARDER: case IMSG_RECONF_DOT_FORWARDER: case IMSG_RECONF_FORCE: imsg_receive_config(&imsg, &nconf); break; case IMSG_RECONF_END: if (nconf == NULL) fatalx("%s: IMSG_RECONF_END without " "IMSG_RECONF_CONF", __func__); merge_config(frontend_conf, nconf); if (frontend_conf->blocklist_file == NULL) free_bl(); nconf = NULL; break; case IMSG_UDP6SOCK: if (udp6sock != -1) fatalx("%s: received unexpected udp6sock", __func__); if ((udp6sock = imsg_get_fd(&imsg)) == -1) fatalx("%s: expected to receive imsg " "UDP6 fd but didn't receive any", __func__); event_set(&udp6ev.ev, udp6sock, EV_READ | EV_PERSIST, udp_receive, &udp6ev); event_add(&udp6ev.ev, NULL); break; case IMSG_UDP4SOCK: if (udp4sock != -1) fatalx("%s: received unexpected udp4sock", __func__); if ((udp4sock = imsg_get_fd(&imsg)) == -1) fatalx("%s: expected to receive imsg " "UDP4 fd but didn't receive any", __func__); event_set(&udp4ev.ev, udp4sock, EV_READ | EV_PERSIST, udp_receive, &udp4ev); event_add(&udp4ev.ev, NULL); break; case IMSG_TCP4SOCK: if (tcp4sock != -1) fatalx("%s: received unexpected tcp4sock", __func__); if ((tcp4sock = imsg_get_fd(&imsg)) == -1) fatalx("%s: expected to receive imsg " "TCP4 fd but didn't receive any", __func__); event_set(&tcp4ev.ev, tcp4sock, EV_READ | EV_PERSIST, tcp_accept, &tcp4ev); event_add(&tcp4ev.ev, NULL); evtimer_set(&tcp4ev.pause, accept_paused, &tcp4ev); break; case IMSG_TCP6SOCK: if (tcp6sock != -1) fatalx("%s: received unexpected tcp6sock", __func__); if ((tcp6sock = imsg_get_fd(&imsg)) == -1) fatalx("%s: expected to receive imsg " "TCP6 fd but didn't receive any", __func__); event_set(&tcp6ev.ev, tcp6sock, EV_READ | EV_PERSIST, tcp_accept, &tcp6ev); event_add(&tcp6ev.ev, NULL); evtimer_set(&tcp6ev.pause, accept_paused, &tcp6ev); break; case IMSG_ROUTESOCK: { static int routesock = -1; if (routesock != -1) fatalx("%s: received unexpected routesock", __func__); if ((fd = imsg_get_fd(&imsg)) == -1) fatalx("%s: expected to receive imsg " "routesocket fd but didn't receive any", __func__); routesock = fd; event_set(&ev_route, fd, EV_READ | EV_PERSIST, route_receive, NULL); break; } case IMSG_STARTUP: frontend_startup(); break; case IMSG_CONTROLFD: if ((fd = imsg_get_fd(&imsg)) == -1) fatalx("%s: expected to receive imsg control " "fd but didn't receive any", __func__); /* Listen on control socket. */ control_listen(fd); break; case IMSG_TAFD: if ((ta_fd = imsg_get_fd(&imsg)) != -1) parse_trust_anchor(&trust_anchors, ta_fd); if (!TAILQ_EMPTY(&trust_anchors)) send_trust_anchors(&trust_anchors); break; case IMSG_BLFD: if ((fd = imsg_get_fd(&imsg)) == -1) fatalx("%s: expected to receive imsg block " "list fd but didn't receive any", __func__); parse_blocklist(fd); break; default: log_debug("%s: error handling imsg %d", __func__, imsg.hdr.type); break; } imsg_free(&imsg); } if (!shut) imsg_event_add(iev); else { /* This pipe is dead. Remove its event handler. */ event_del(&iev->ev); event_loopexit(NULL); } } void frontend_dispatch_resolver(int fd, short event, void *bula) { static struct dns64_prefix *new_dns64_prefixes = NULL; static int new_dns64_prefix_count = 0; static int new_dns64_prefix_pos = 0; struct pending_query *pq; struct imsgev *iev = bula; struct imsgbuf *ibuf = &iev->ibuf; struct imsg imsg; int n, shut = 0, chg; if (event & EV_READ) { if ((n = imsg_read(ibuf)) == -1 && errno != EAGAIN) fatal("imsg_read error"); if (n == 0) /* Connection closed. */ shut = 1; } if (event & EV_WRITE) { if ((n = msgbuf_write(&ibuf->w)) == -1 && errno != EAGAIN) fatal("msgbuf_write"); if (n == 0) /* Connection closed. */ shut = 1; } for (;;) { if ((n = imsg_get(ibuf, &imsg)) == -1) fatal("%s: imsg_get error", __func__); if (n == 0) /* No more messages. */ break; switch (imsg.hdr.type) { case IMSG_ANSWER: { struct answer_header *answer_header; int data_len; uint8_t *data; if (IMSG_DATA_SIZE(imsg) < sizeof(*answer_header)) fatalx("%s: IMSG_ANSWER wrong length: " "%lu", __func__, IMSG_DATA_SIZE(imsg)); answer_header = (struct answer_header *)imsg.data; data = (uint8_t *)imsg.data + sizeof(*answer_header); if (answer_header->answer_len > UINT16_MAX) fatalx("%s: IMSG_ANSWER answer too big: %d", __func__, answer_header->answer_len); data_len = IMSG_DATA_SIZE(imsg) - sizeof(*answer_header); if ((pq = find_pending_query(answer_header->id)) == NULL) { log_warnx("%s: cannot find pending query %llu", __func__, answer_header->id); break; } if (answer_header->srvfail) { error_answer(pq, LDNS_RCODE_SERVFAIL); send_answer(pq); break; } if (answer_header->bogus && !(pq->flags & BIT_CD)) { error_answer(pq, LDNS_RCODE_SERVFAIL); send_answer(pq); break; } if (sldns_buffer_position(pq->abuf) == 0 && !sldns_buffer_set_capacity(pq->abuf, answer_header->answer_len)) { error_answer(pq, LDNS_RCODE_SERVFAIL); send_answer(pq); break; } if (sldns_buffer_position(pq->abuf) + data_len > sldns_buffer_capacity(pq->abuf)) fatalx("%s: IMSG_ANSWER answer too big: %d", __func__, data_len); sldns_buffer_write(pq->abuf, data, data_len); if (sldns_buffer_position(pq->abuf) == sldns_buffer_capacity(pq->abuf)) { sldns_buffer_flip(pq->abuf); if (pq->dns64_synthesize) { synthesize_dns64_answer(pq); send_answer(pq); } else { noerror_answer(pq); if (pq->dns64_synthesize) /* we did not find a answer */ resend_dns64_query(pq); else send_answer(pq); } } break; } case IMSG_CTL_RESOLVER_INFO: case IMSG_CTL_AUTOCONF_RESOLVER_INFO: case IMSG_CTL_MEM_INFO: case IMSG_CTL_END: control_imsg_relay(&imsg); break; case IMSG_NEW_TA: /* make sure this is a string */ ((char *)imsg.data)[IMSG_DATA_SIZE(imsg) - 1] = '\0'; add_new_ta(&new_trust_anchors, imsg.data); break; case IMSG_NEW_TAS_ABORT: free_tas(&new_trust_anchors); break; case IMSG_NEW_TAS_DONE: chg = merge_tas(&new_trust_anchors, &trust_anchors); if (chg) send_trust_anchors(&trust_anchors); /* * always write trust anchors, the modify date on * the file is an indication when we made progress */ if (ta_fd != -1) write_trust_anchors(&trust_anchors, ta_fd); break; case IMSG_NEW_DNS64_PREFIXES_START: if (IMSG_DATA_SIZE(imsg) != sizeof(new_dns64_prefix_count)) fatalx("%s: IMSG_NEW_DNS64_PREFIXES_START " "wrong length: %lu", __func__, IMSG_DATA_SIZE(imsg)); memcpy(&new_dns64_prefix_count, imsg.data, sizeof(new_dns64_prefix_count)); free(new_dns64_prefixes); new_dns64_prefixes = NULL; if (new_dns64_prefix_count > 0) new_dns64_prefixes = calloc(new_dns64_prefix_count, sizeof(struct dns64_prefix)); new_dns64_prefix_pos = 0; break; case IMSG_NEW_DNS64_PREFIX: { if (IMSG_DATA_SIZE(imsg) != sizeof(struct dns64_prefix)) fatalx("%s: IMSG_NEW_DNS64_PREFIX wrong " "length: %lu", __func__, IMSG_DATA_SIZE(imsg)); if (new_dns64_prefixes == NULL) break; if (new_dns64_prefix_pos >= new_dns64_prefix_count) fatalx("%s: IMSG_NEW_DNS64_PREFIX: too many " "prefixes", __func__); memcpy(&new_dns64_prefixes[new_dns64_prefix_pos++], imsg.data, sizeof(struct dns64_prefix)); break; } case IMSG_NEW_DNS64_PREFIXES_DONE: free(dns64_prefixes); dns64_prefixes = new_dns64_prefixes; dns64_prefix_count = new_dns64_prefix_count; new_dns64_prefixes = NULL; break; default: log_debug("%s: error handling imsg %d", __func__, imsg.hdr.type); break; } imsg_free(&imsg); } if (!shut) imsg_event_add(iev); else { /* This pipe is dead. Remove its event handler. */ event_del(&iev->ev); event_loopexit(NULL); } } void frontend_startup(void) { if (!event_initialized(&ev_route)) fatalx("%s: did not receive a route socket from the main " "process", __func__); event_add(&ev_route, NULL); frontend_imsg_compose_main(IMSG_STARTUP_DONE, 0, NULL, 0); check_available_af(); } void free_pending_query(struct pending_query *pq) { if (!pq) return; TAILQ_REMOVE(&pending_queries, pq, entry); regional_destroy(pq->region); sldns_buffer_free(pq->qbuf); sldns_buffer_free(pq->abuf); if (pq->tcp) { if (event_initialized(&pq->ev)) event_del(&pq->ev); if (event_initialized(&pq->resp_ev)) event_del(&pq->resp_ev); if (event_initialized(&pq->tmo_ev)) event_del(&pq->tmo_ev); if (pq->fd != -1) close(pq->fd); } free(pq); } void udp_receive(int fd, short events, void *arg) { struct udp_ev *udpev = (struct udp_ev *)arg; struct pending_query *pq = NULL; ssize_t len; if ((len = recvmsg(fd, &udpev->rcvmhdr, 0)) == -1) { log_warn("recvmsg"); return; } if ((pq = calloc(1, sizeof(*pq))) == NULL) { log_warn(NULL); return; } do { arc4random_buf(&pq->imsg_id, sizeof(pq->imsg_id)); } while(find_pending_query(pq->imsg_id) != NULL); TAILQ_INSERT_TAIL(&pending_queries, pq, entry); pq->from = udpev->from; pq->fd = fd; pq->qbuf = sldns_buffer_new(len); pq->abuf = sldns_buffer_new(len); /* make sure we can send errors */ pq->region = regional_create(); if (!pq->qbuf || !pq->abuf || !pq->region) { log_warnx("out of memory"); free_pending_query(pq); return; } sldns_buffer_write(pq->qbuf, udpev->query, len); sldns_buffer_flip(pq->qbuf); handle_query(pq); } void handle_query(struct pending_query *pq) { struct query_imsg query_imsg; struct bl_node find; int rcode; char *str; char dname[LDNS_MAX_DOMAINLEN + 1]; char qclass_buf[16]; char qtype_buf[16]; if (log_getverbose() & OPT_VERBOSE2 && (str = sldns_wire2str_pkt(sldns_buffer_begin(pq->qbuf), sldns_buffer_limit(pq->qbuf))) != NULL) { log_debug("from: %s\n%s", ip_port((struct sockaddr *) &pq->from), str); free(str); } if (sldns_buffer_remaining(pq->qbuf) < LDNS_HEADER_SIZE) { log_warnx("bad query: too short, dropped"); goto drop; } pq->id = sldns_buffer_read_u16_at(pq->qbuf, 0); pq->flags = sldns_buffer_read_u16_at(pq->qbuf, 2); if (!query_info_parse(&pq->qinfo, pq->qbuf)) { log_warnx("query_info_parse failed"); goto drop; } rcode = check_query(pq->qbuf); switch (rcode) { case LDNS_RCODE_NOERROR: break; case -1: goto drop; default: error_answer(pq, rcode); goto send_answer; } rcode = parse_edns_from_query_pkt(pq->qbuf, &pq->edns, NULL, NULL, NULL, 0, pq->region); if (rcode != LDNS_RCODE_NOERROR) { error_answer(pq, rcode); goto send_answer; } if (!dname_valid(pq->qinfo.qname, pq->qinfo.qname_len)) { error_answer(pq, LDNS_RCODE_FORMERR); goto send_answer; } dname_str(pq->qinfo.qname, dname); sldns_wire2str_class_buf(pq->qinfo.qclass, qclass_buf, sizeof(qclass_buf)); sldns_wire2str_type_buf(pq->qinfo.qtype, qtype_buf, sizeof(qtype_buf)); log_debug("%s: %s %s %s ?", ip_port((struct sockaddr *)&pq->from), dname, qclass_buf, qtype_buf); find.domain = dname; if (RB_FIND(bl_tree, &bl_head, &find) != NULL) { if (frontend_conf->blocklist_log) log_info("blocking %s", dname); error_answer(pq, LDNS_RCODE_REFUSED); goto send_answer; } if (pq->qinfo.qtype == LDNS_RR_TYPE_AXFR || pq->qinfo.qtype == LDNS_RR_TYPE_IXFR) { error_answer(pq, LDNS_RCODE_REFUSED); goto send_answer; } if(pq->qinfo.qtype == LDNS_RR_TYPE_OPT || pq->qinfo.qtype == LDNS_RR_TYPE_TSIG || pq->qinfo.qtype == LDNS_RR_TYPE_TKEY || pq->qinfo.qtype == LDNS_RR_TYPE_MAILA || pq->qinfo.qtype == LDNS_RR_TYPE_MAILB || (pq->qinfo.qtype >= 128 && pq->qinfo.qtype <= 248)) { error_answer(pq, LDNS_RCODE_FORMERR); goto send_answer; } if (pq->qinfo.qclass == LDNS_RR_CLASS_CH) { if (strcasecmp(dname, "version.server.") == 0 || strcasecmp(dname, "version.bind.") == 0) { chaos_answer(pq); } else error_answer(pq, LDNS_RCODE_REFUSED); goto send_answer; } if (strlcpy(query_imsg.qname, dname, sizeof(query_imsg.qname)) >= sizeof(query_imsg.qname)) { log_warnx("qname too long"); error_answer(pq, LDNS_RCODE_FORMERR); goto send_answer; } query_imsg.id = pq->imsg_id; query_imsg.t = pq->qinfo.qtype; query_imsg.c = pq->qinfo.qclass; if (frontend_imsg_compose_resolver(IMSG_QUERY, 0, &query_imsg, sizeof(query_imsg)) == -1) { error_answer(pq, LDNS_RCODE_SERVFAIL); goto send_answer; } return; send_answer: send_answer(pq); return; drop: free_pending_query(pq); } void noerror_answer(struct pending_query *pq) { struct query_info skip, qinfo; struct reply_info *rinfo = NULL; struct alloc_cache alloc; struct edns_data edns; struct ub_packed_rrset_key *an_rrset = NULL; struct packed_rrset_data *an_rrset_data = NULL; alloc_init(&alloc, NULL, 0); memset(&qinfo, 0, sizeof(qinfo)); /* read past query section, no memory is allocated */ if (!query_info_parse(&skip, pq->abuf)) goto srvfail; if (reply_info_parse(pq->abuf, &alloc, &qinfo, &rinfo, pq->region, &edns) != 0) goto srvfail; if ((an_rrset = reply_find_answer_rrset(&qinfo, rinfo)) != NULL) an_rrset_data = (struct packed_rrset_data*)an_rrset->entry.data; /* reply_info_parse() allocates memory */ query_info_clear(&qinfo); /* XXX check that there a no AAAA records in answer section? */ if ((an_rrset_data == NULL || an_rrset_data->count == 0) && !pq->dns64_synthesize && pq->qinfo.qtype == LDNS_RR_TYPE_AAAA && pq->qinfo.qclass == LDNS_RR_CLASS_IN && dns64_prefix_count > 0) { pq->dns64_synthesize = 1; return; } sldns_buffer_clear(pq->abuf); if (reply_info_encode(&pq->qinfo, rinfo, htons(pq->id), rinfo->flags, pq->abuf, 0, pq->region, pq->tcp ? UINT16_MAX : pq->edns.udp_size, pq->edns.bits & EDNS_DO, MINIMIZE_ANSWER) == 0) goto srvfail; reply_info_parsedelete(rinfo, &alloc); alloc_clear(&alloc); return; srvfail: reply_info_parsedelete(rinfo, &alloc); alloc_clear(&alloc); error_answer(pq, LDNS_RCODE_SERVFAIL); } void synthesize_dns64_answer(struct pending_query *pq) { struct query_info skip, qinfo; struct reply_info *rinfo = NULL, *synth_rinfo = NULL; struct alloc_cache alloc; struct edns_data edns; size_t i; pq->dns64_synthesize = 0; alloc_init(&alloc, NULL, 0); memset(&qinfo, 0, sizeof(qinfo)); /* read past query section, no memory is allocated */ if (!query_info_parse(&skip, pq->abuf)) goto srvfail; if (reply_info_parse(pq->abuf, &alloc, &qinfo, &rinfo, pq->region, &edns) != 0) goto srvfail; /* reply_info_parse() allocates memory */ query_info_clear(&qinfo); synth_rinfo = construct_reply_info_base(pq->region, rinfo->flags, rinfo->qdcount, rinfo->ttl, rinfo->prefetch_ttl, rinfo->serve_expired_ttl, rinfo->an_numrrsets, rinfo->ns_numrrsets, rinfo->ar_numrrsets, rinfo->rrset_count, rinfo->security, rinfo->reason_bogus); if (!synth_rinfo) goto srvfail; if(!reply_info_alloc_rrset_keys(synth_rinfo, NULL, pq->region)) goto srvfail; for (i = 0; i < synth_rinfo->rrset_count; i++) { struct ub_packed_rrset_key *src_rrset_key, *dst_rrset_key; struct packed_rrset_data *src_rrset_data; struct packed_rrset_data *dst_rrset_data; src_rrset_key = rinfo->rrsets[i]; src_rrset_data = (struct packed_rrset_data *)src_rrset_key->entry.data; dst_rrset_key = synth_rinfo->rrsets[i]; dst_rrset_key->id = src_rrset_key->id; dst_rrset_key->rk = src_rrset_key->rk; if (i < rinfo->an_numrrsets && src_rrset_key->rk.type == htons(LDNS_RR_TYPE_A)) { dns64_synth_aaaa_data(src_rrset_key, src_rrset_data, dst_rrset_key, &dst_rrset_data, pq->region); if (dst_rrset_data == NULL) goto srvfail; } else { dst_rrset_key->entry.hash = src_rrset_key->entry.hash; dst_rrset_key->rk.dname = regional_alloc_init( pq->region, src_rrset_key->rk.dname, src_rrset_key->rk.dname_len); if (dst_rrset_key->rk.dname == NULL) goto srvfail; dst_rrset_data = regional_alloc_init(pq->region, src_rrset_data, packed_rrset_sizeof(src_rrset_data)); if (dst_rrset_data == NULL) goto srvfail; } packed_rrset_ptr_fixup(dst_rrset_data); dst_rrset_key->entry.data = dst_rrset_data; } if (!sldns_buffer_set_capacity(pq->abuf, pq->tcp ? UINT16_MAX : pq->edns.udp_size)) goto srvfail; sldns_buffer_clear(pq->abuf); if (reply_info_encode(&pq->qinfo, synth_rinfo, htons(pq->id), synth_rinfo->flags, pq->abuf, 0, pq->region, pq->tcp ? UINT16_MAX : pq->edns.udp_size, pq->edns.bits & EDNS_DO, MINIMIZE_ANSWER) == 0) goto srvfail; reply_info_parsedelete(rinfo, &alloc); alloc_clear(&alloc); return; srvfail: reply_info_parsedelete(rinfo, &alloc); alloc_clear(&alloc); error_answer(pq, LDNS_RCODE_SERVFAIL); } void resend_dns64_query(struct pending_query *opq) { struct pending_query *pq; struct query_imsg query_imsg; int rcode; char dname[LDNS_MAX_DOMAINLEN + 1]; if ((pq = calloc(1, sizeof(*pq))) == NULL) { log_warn(NULL); return; } do { arc4random_buf(&pq->imsg_id, sizeof(pq->imsg_id)); } while(find_pending_query(pq->imsg_id) != NULL); TAILQ_INSERT_TAIL(&pending_queries, pq, entry); pq->from = opq->from; pq->fd = opq->fd; opq->fd = -1; pq->tcp = opq->tcp; pq->qbuf = sldns_buffer_new(sldns_buffer_capacity(opq->qbuf)); pq->abuf = sldns_buffer_new(sldns_buffer_capacity(opq->abuf)); pq->region = regional_create(); if (!pq->qbuf || !pq->abuf || !pq->region) { log_warnx("out of memory"); free_pending_query(pq); free_pending_query(opq); return; } sldns_buffer_rewind(opq->qbuf); sldns_buffer_write(pq->qbuf, sldns_buffer_current(opq->qbuf), sldns_buffer_remaining(opq->qbuf)); sldns_buffer_flip(pq->qbuf); if (pq->tcp) { struct timeval timeout = {TCP_TIMEOUT, 0}; event_set(&pq->ev, pq->fd, EV_READ | EV_PERSIST, tcp_request, pq); event_set(&pq->resp_ev, pq->fd, EV_WRITE | EV_PERSIST, tcp_response, pq); evtimer_set(&pq->tmo_ev, tcp_timeout, pq); evtimer_add(&pq->tmo_ev, &timeout); } if (sldns_buffer_remaining(pq->qbuf) < LDNS_HEADER_SIZE) { log_warnx("bad query: too short, dropped"); goto drop; } pq->id = sldns_buffer_read_u16_at(pq->qbuf, 0); pq->flags = sldns_buffer_read_u16_at(pq->qbuf, 2); if (!query_info_parse(&pq->qinfo, pq->qbuf)) { log_warnx("query_info_parse failed"); goto drop; } rcode = parse_edns_from_query_pkt(pq->qbuf, &pq->edns, NULL, NULL, NULL, 0, pq->region); if (rcode != LDNS_RCODE_NOERROR) { error_answer(pq, rcode); goto send_answer; } dname_str(pq->qinfo.qname, dname); strlcpy(query_imsg.qname, dname, sizeof(query_imsg.qname)); query_imsg.id = pq->imsg_id; query_imsg.t = LDNS_RR_TYPE_A; query_imsg.c = pq->qinfo.qclass; pq->dns64_synthesize = 1; if (frontend_imsg_compose_resolver(IMSG_QUERY, 0, &query_imsg, sizeof(query_imsg)) == -1) { error_answer(pq, LDNS_RCODE_SERVFAIL); goto send_answer; } free_pending_query(opq); return; send_answer: free_pending_query(opq); send_answer(pq); return; drop: free_pending_query(opq); free_pending_query(pq); } void chaos_answer(struct pending_query *pq) { size_t len; const char *name = "unwind"; len = strlen(name); if (!sldns_buffer_set_capacity(pq->abuf, sldns_buffer_capacity(pq->qbuf) + COMPRESSED_RR_SIZE + 1 + len)) { error_answer(pq, LDNS_RCODE_SERVFAIL); return; } sldns_buffer_copy(pq->abuf, pq->qbuf); sldns_buffer_clear(pq->abuf); sldns_buffer_skip(pq->abuf, sizeof(uint16_t)); /* skip id */ sldns_buffer_write_u16(pq->abuf, 0); /* clear flags */ LDNS_QR_SET(sldns_buffer_begin(pq->abuf)); LDNS_RA_SET(sldns_buffer_begin(pq->abuf)); if (LDNS_RD_WIRE(sldns_buffer_begin(pq->qbuf))) LDNS_RD_SET(sldns_buffer_begin(pq->abuf)); if (LDNS_CD_WIRE(sldns_buffer_begin(pq->qbuf))) LDNS_CD_SET(sldns_buffer_begin(pq->abuf)); LDNS_RCODE_SET(sldns_buffer_begin(pq->abuf), LDNS_RCODE_NOERROR); sldns_buffer_write_u16(pq->abuf, 1); /* qdcount */ sldns_buffer_write_u16(pq->abuf, 1); /* ancount */ sldns_buffer_write_u16(pq->abuf, 0); /* nscount */ sldns_buffer_write_u16(pq->abuf, 0); /* arcount */ (void)query_dname_len(pq->abuf); /* skip qname */ sldns_buffer_skip(pq->abuf, sizeof(uint16_t)); /* skip qtype */ sldns_buffer_skip(pq->abuf, sizeof(uint16_t)); /* skip qclass */ sldns_buffer_write_u16(pq->abuf, 0xc00c); /* ptr to query */ sldns_buffer_write_u16(pq->abuf, LDNS_RR_TYPE_TXT); sldns_buffer_write_u16(pq->abuf, LDNS_RR_CLASS_CH); sldns_buffer_write_u32(pq->abuf, 0); /* TTL */ sldns_buffer_write_u16(pq->abuf, 1 + len); /* RDLENGTH */ sldns_buffer_write_u8(pq->abuf, len); /* length octed */ sldns_buffer_write(pq->abuf, name, len); sldns_buffer_flip(pq->abuf); } void error_answer(struct pending_query *pq, int rcode) { sldns_buffer_clear(pq->abuf); error_encode(pq->abuf, rcode, &pq->qinfo, htons(pq->id), pq->flags, pq->edns.edns_present ? &pq->edns : NULL); } int check_query(sldns_buffer* pkt) { if(sldns_buffer_limit(pkt) < LDNS_HEADER_SIZE) { log_warnx("bad query: too short, dropped"); return -1; } if(LDNS_QR_WIRE(sldns_buffer_begin(pkt))) { log_warnx("bad query: QR set, dropped"); return -1; } if(LDNS_TC_WIRE(sldns_buffer_begin(pkt))) { LDNS_TC_CLR(sldns_buffer_begin(pkt)); log_warnx("bad query: TC set"); return (LDNS_RCODE_FORMERR); } if(!(LDNS_RD_WIRE(sldns_buffer_begin(pkt)))) { log_warnx("bad query: RD not set"); return (LDNS_RCODE_REFUSED); } if(LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_QUERY) { log_warnx("bad query: unknown opcode %d", LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt))); return (LDNS_RCODE_NOTIMPL); } if (LDNS_QDCOUNT(sldns_buffer_begin(pkt)) != 1 && LDNS_ANCOUNT(sldns_buffer_begin(pkt))!= 0 && LDNS_NSCOUNT(sldns_buffer_begin(pkt))!= 0 && LDNS_ARCOUNT(sldns_buffer_begin(pkt)) > 1) { log_warnx("bad query: qdcount: %d, ancount: %d " "nscount: %d, arcount: %d", LDNS_QDCOUNT(sldns_buffer_begin(pkt)), LDNS_ANCOUNT(sldns_buffer_begin(pkt)), LDNS_NSCOUNT(sldns_buffer_begin(pkt)), LDNS_ARCOUNT(sldns_buffer_begin(pkt))); return (LDNS_RCODE_FORMERR); } return (LDNS_RCODE_NOERROR); } void send_answer(struct pending_query *pq) { char *str; if (log_getverbose() & OPT_VERBOSE2 && (str = sldns_wire2str_pkt(sldns_buffer_begin(pq->abuf), sldns_buffer_limit(pq->abuf))) != NULL) { log_debug("from: %s\n%s", ip_port((struct sockaddr *) &pq->from), str); free(str); } if (!pq->tcp) { if(sendto(pq->fd, sldns_buffer_current(pq->abuf), sldns_buffer_remaining(pq->abuf), 0, (struct sockaddr *)&pq->from, pq->from.ss_len) == -1) log_warn("sendto"); free_pending_query(pq); } else { struct sldns_buffer *tmp; tmp = sldns_buffer_new(sldns_buffer_limit(pq->abuf) + 2); if (!tmp) { free_pending_query(pq); return; } sldns_buffer_write_u16(tmp, sldns_buffer_limit(pq->abuf)); sldns_buffer_write(tmp, sldns_buffer_current(pq->abuf), sldns_buffer_remaining(pq->abuf)); sldns_buffer_flip(tmp); sldns_buffer_free(pq->abuf); pq->abuf = tmp; event_add(&pq->resp_ev, NULL); } } char* ip_port(struct sockaddr *sa) { static char hbuf[NI_MAXHOST], buf[NI_MAXHOST]; if (getnameinfo(sa, sa->sa_len, hbuf, sizeof(hbuf), NULL, 0, NI_NUMERICHOST) != 0) { snprintf(buf, sizeof(buf), "%s", "(unknown)"); return buf; } if (sa->sa_family == AF_INET6) snprintf(buf, sizeof(buf), "[%s]:%d", hbuf, ntohs( ((struct sockaddr_in6 *)sa)->sin6_port)); if (sa->sa_family == AF_INET) snprintf(buf, sizeof(buf), "[%s]:%d", hbuf, ntohs( ((struct sockaddr_in *)sa)->sin_port)); return buf; } struct pending_query* find_pending_query(uint64_t id) { struct pending_query *pq; TAILQ_FOREACH(pq, &pending_queries, entry) if (pq->imsg_id == id) return pq; return NULL; } void route_receive(int fd, short events, void *arg) { static uint8_t *buf; struct rt_msghdr *rtm; struct sockaddr *sa, *rti_info[RTAX_MAX]; ssize_t n; if (buf == NULL) { buf = malloc(ROUTE_SOCKET_BUF_SIZE); if (buf == NULL) fatal("malloc"); } rtm = (struct rt_msghdr *)buf; if ((n = read(fd, buf, ROUTE_SOCKET_BUF_SIZE)) == -1) { if (errno == EAGAIN || errno == EINTR) return; log_warn("dispatch_rtmsg: read error"); return; } if (n == 0) fatal("routing socket closed"); if (n < (ssize_t)sizeof(rtm->rtm_msglen) || n < rtm->rtm_msglen) { log_warnx("partial rtm of %zd in buffer", n); return; } if (rtm->rtm_version != RTM_VERSION) return; sa = (struct sockaddr *)(buf + rtm->rtm_hdrlen); get_rtaddrs(rtm->rtm_addrs, sa, rti_info); handle_route_message(rtm, rti_info); } #define ROUNDUP(a) \ ((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long)) void get_rtaddrs(int addrs, struct sockaddr *sa, struct sockaddr **rti_info) { int i; for (i = 0; i < RTAX_MAX; i++) { if (addrs & (1 << i)) { rti_info[i] = sa; sa = (struct sockaddr *)((char *)(sa) + ROUNDUP(sa->sa_len)); } else rti_info[i] = NULL; } } void handle_route_message(struct rt_msghdr *rtm, struct sockaddr **rti_info) { struct imsg_rdns_proposal rdns_proposal; struct sockaddr_rtdns *rtdns; struct if_announcemsghdr *ifan; switch (rtm->rtm_type) { case RTM_IFANNOUNCE: ifan = (struct if_announcemsghdr *)rtm; if (ifan->ifan_what == IFAN_ARRIVAL) break; rdns_proposal.if_index = ifan->ifan_index; rdns_proposal.src = 0; rdns_proposal.rtdns.sr_family = AF_INET; rdns_proposal.rtdns.sr_len = offsetof(struct sockaddr_rtdns, sr_dns); frontend_imsg_compose_resolver(IMSG_REPLACE_DNS, 0, &rdns_proposal, sizeof(rdns_proposal)); break; case RTM_IFINFO: frontend_imsg_compose_resolver(IMSG_NETWORK_CHANGED, 0, NULL, 0); break; case RTM_PROPOSAL: if (!(rtm->rtm_addrs & RTA_DNS)) break; rtdns = (struct sockaddr_rtdns*)rti_info[RTAX_DNS]; rdns_proposal.if_index = rtm->rtm_index; rdns_proposal.src = rtm->rtm_priority; memcpy(&rdns_proposal.rtdns, rtdns, sizeof(rdns_proposal.rtdns)); frontend_imsg_compose_resolver(IMSG_REPLACE_DNS, 0, &rdns_proposal, sizeof(rdns_proposal)); break; case RTM_NEWADDR: case RTM_DELADDR: case RTM_DESYNC: check_available_af(); break; default: break; } } void add_new_ta(struct trust_anchor_head *tah, char *val) { struct trust_anchor *ta, *i; int cmp; if ((ta = malloc(sizeof(*ta))) == NULL) fatal("%s", __func__); if ((ta->ta = strdup(val)) == NULL) fatal("%s", __func__); /* keep the list sorted to prevent churn if the order changes in DNS */ TAILQ_FOREACH(i, tah, entry) { cmp = strcmp(i->ta, ta->ta); if ( cmp == 0) { /* duplicate */ free(ta->ta); free(ta); return; } else if (cmp > 0) { TAILQ_INSERT_BEFORE(i, ta, entry); return; } } TAILQ_INSERT_TAIL(tah, ta, entry); } void free_tas(struct trust_anchor_head *tah) { struct trust_anchor *ta; while ((ta = TAILQ_FIRST(tah))) { TAILQ_REMOVE(tah, ta, entry); free(ta->ta); free(ta); } } int merge_tas(struct trust_anchor_head *newh, struct trust_anchor_head *oldh) { struct trust_anchor *i, *j; int chg = 0; j = TAILQ_FIRST(oldh); TAILQ_FOREACH(i, newh, entry) { if (j == NULL || strcmp(i->ta, j->ta) != 0) { chg = 1; break; } j = TAILQ_NEXT(j, entry); } if (j != NULL) chg = 1; if (chg) { free_tas(oldh); TAILQ_CONCAT(oldh, newh, entry); } else { free_tas(newh); } return (chg); } void parse_trust_anchor(struct trust_anchor_head *tah, int fd) { size_t len, dname_len; ssize_t n, sz; uint8_t rr[LDNS_RR_BUF_SIZE]; char *str, *p, buf[512], *line; sz = 0; str = NULL; while ((n = read(fd, buf, sizeof(buf))) > 0) { p = recallocarray(str, sz, sz + n, 1); if (p == NULL) { log_warn("%s", __func__); goto out; } str = p; memcpy(str + sz, buf, n); sz += n; } if (n == -1) { log_warn("%s", __func__); goto out; } /* make it a string */ p = recallocarray(str, sz, sz + 1, 1); if (p == NULL) { log_warn("%s", __func__); goto out; } str = p; sz++; len = sizeof(rr); while ((line = strsep(&p, "\n")) != NULL) { if (sldns_str2wire_rr_buf(line, rr, &len, &dname_len, ROOT_DNSKEY_TTL, NULL, 0, NULL, 0) != 0) continue; if (sldns_wirerr_get_type(rr, len, dname_len) == LDNS_RR_TYPE_DNSKEY) add_new_ta(tah, line); } out: free(str); return; } void send_trust_anchors(struct trust_anchor_head *tah) { struct trust_anchor *ta; TAILQ_FOREACH(ta, tah, entry) frontend_imsg_compose_resolver(IMSG_NEW_TA, 0, ta->ta, strlen(ta->ta) + 1); frontend_imsg_compose_resolver(IMSG_NEW_TAS_DONE, 0, NULL, 0); } void write_trust_anchors(struct trust_anchor_head *tah, int fd) { struct trust_anchor *ta; size_t len = 0; ssize_t n; char *str; if (lseek(fd, 0, SEEK_SET) == -1) { log_warn("%s", __func__); goto out; } TAILQ_FOREACH(ta, tah, entry) { if ((n = asprintf(&str, "%s\n", ta->ta)) == -1) { log_warn("%s", __func__); len = 0; goto out; } len += n; if (write(fd, str, n) != n) { log_warn("%s", __func__); free(str); len = 0; goto out; } free(str); } out: ftruncate(fd, len); fsync(fd); } void parse_blocklist(int fd) { FILE *f; struct bl_node *bl_node; char *line = NULL; size_t linesize = 0; ssize_t linelen; if((f = fdopen(fd, "r")) == NULL) { log_warn("cannot read block list"); close(fd); return; } free_bl(); while ((linelen = getline(&line, &linesize, f)) != -1) { if (line[linelen - 1] == '\n') { if (linelen >= 2 && line[linelen - 2] != '.') line[linelen - 1] = '.'; else line[linelen - 1] = '\0'; } bl_node = malloc(sizeof *bl_node); if (bl_node == NULL) fatal("%s: malloc", __func__); if ((bl_node->domain = strdup(line)) == NULL) fatal("%s: strdup", __func__); if (RB_INSERT(bl_tree, &bl_head, bl_node) != NULL) { log_warnx("duplicate blocked domain \"%s\"", line); free(bl_node->domain); free(bl_node); } } free(line); if (ferror(f)) log_warn("getline"); fclose(f); } int bl_cmp(struct bl_node *e1, struct bl_node *e2) { return (strcasecmp(e1->domain, e2->domain)); } void free_bl(void) { struct bl_node *n, *nxt; RB_FOREACH_SAFE(n, bl_tree, &bl_head, nxt) { RB_REMOVE(bl_tree, &bl_head, n); free(n->domain); free(n); } } int pending_query_cnt(void) { struct pending_query *e; int cnt = 0; TAILQ_FOREACH(e, &pending_queries, entry) cnt++; return cnt; } void accept_paused(int fd, short events, void *arg) { struct tcp_accept_ev *tcpev = arg; event_add(&tcpev->ev, NULL); } int accept_reserve(int sockfd, struct sockaddr *addr, socklen_t *addrlen) { if (getdtablecount() + FD_RESERVE >= getdtablesize()) { log_debug("%s: inflight fds exceeded", __func__); errno = EMFILE; return -1; } return accept4(sockfd, addr, addrlen, SOCK_NONBLOCK | SOCK_CLOEXEC); } void tcp_accept(int fd, short events, void *arg) { static struct timeval timeout = {TCP_TIMEOUT, 0}; static struct timeval backoff = {1, 0}; struct pending_query *pq; struct tcp_accept_ev *tcpev; struct sockaddr_storage ss; socklen_t len; int s; tcpev = arg; len = sizeof(ss); if ((s = accept_reserve(fd, (struct sockaddr *)&ss, &len)) == -1) { switch (errno) { case EINTR: case EWOULDBLOCK: case ECONNABORTED: return; case EMFILE: case ENFILE: event_del(&tcpev->ev); evtimer_add(&tcpev->pause, &backoff); return; default: fatal("accept"); } } if ((pq = calloc(1, sizeof(*pq))) == NULL) { log_warn(NULL); close(s); return; } do { arc4random_buf(&pq->imsg_id, sizeof(pq->imsg_id)); } while(find_pending_query(pq->imsg_id) != NULL); TAILQ_INSERT_TAIL(&pending_queries, pq, entry); pq->from = ss; pq->fd = s; pq->tcp = 1; pq->qbuf = sldns_buffer_new(DEFAULT_TCP_SIZE); pq->region = regional_create(); if (!pq->qbuf || !pq->region) { free_pending_query(pq); return; } event_set(&pq->ev, s, EV_READ | EV_PERSIST, tcp_request, pq); event_add(&pq->ev, NULL); event_set(&pq->resp_ev, s, EV_WRITE | EV_PERSIST, tcp_response, pq); evtimer_set(&pq->tmo_ev, tcp_timeout, pq); evtimer_add(&pq->tmo_ev, &timeout); } void tcp_request(int fd, short events, void *arg) { struct pending_query *pq; ssize_t n; pq = arg; n = read(fd, sldns_buffer_current(pq->qbuf), sldns_buffer_remaining(pq->qbuf)); switch (n) { case -1: switch (errno) { case EINTR: case EAGAIN: return; default: goto fail; } break; case 0: log_debug("closed connection"); goto fail; default: break; } sldns_buffer_skip(pq->qbuf, n); if (sldns_buffer_position(pq->qbuf) >= 2 && !pq->abuf) { struct sldns_buffer *tmp; size_t rem; uint16_t len; sldns_buffer_flip(pq->qbuf); len = sldns_buffer_read_u16(pq->qbuf); tmp = sldns_buffer_new(len); if (tmp == NULL) goto fail; pq->abuf = sldns_buffer_new(len); if (pq->abuf == NULL) { sldns_buffer_free(tmp); goto fail; } rem = sldns_buffer_remaining(pq->qbuf); sldns_buffer_write(tmp, sldns_buffer_current(pq->qbuf), MINIMUM(len, rem)); sldns_buffer_free(pq->qbuf); pq->qbuf = tmp; } if (sldns_buffer_remaining(pq->qbuf) == 0) { sldns_buffer_flip(pq->qbuf); shutdown(fd, SHUT_RD); event_del(&pq->ev); handle_query(pq); } return; fail: free_pending_query(pq); } void tcp_response(int fd, short events, void *arg) { struct pending_query *pq; ssize_t n; pq = arg; n = write(fd, sldns_buffer_current(pq->abuf), sldns_buffer_remaining(pq->abuf)); if (n == -1) { if (errno == EAGAIN || errno == EINTR) return; free_pending_query(pq); return; } sldns_buffer_skip(pq->abuf, n); if (sldns_buffer_remaining(pq->abuf) == 0) free_pending_query(pq); } void tcp_timeout(int fd, short events, void *arg) { free_pending_query(arg); } void check_available_af(void) { static int available_af = HAVE_IPV4 | HAVE_IPV6; static int rtable = -1; struct ifaddrs *ifap, *ifa; struct if_data *ifa_data; struct sockaddr_in *sin4; struct sockaddr_in6 *sin6; int new_available_af = 0, ifa_rtable = -1; if (rtable == -1) rtable = getrtable(); if (getifaddrs(&ifap) != 0) { log_warn("getifaddrs"); return; } for (ifa = ifap; ifa != NULL; ifa = ifa->ifa_next) { if (ifa->ifa_addr == NULL) continue; switch(ifa->ifa_addr->sa_family) { case AF_LINK: /* AF_LINK comes before inet / inet6 on an interface */ ifa_data = (struct if_data *)ifa->ifa_data; ifa_rtable = ifa_data->ifi_rdomain; break; case AF_INET: if (ifa_rtable != rtable) continue; sin4 = (struct sockaddr_in *)ifa->ifa_addr; if ((ntohl(sin4->sin_addr.s_addr) >> 24) == IN_LOOPBACKNET) continue; new_available_af |= HAVE_IPV4; break; case AF_INET6: if (ifa_rtable != rtable) continue; sin6 = (struct sockaddr_in6 *)ifa->ifa_addr; if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) || IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) || IN6_IS_ADDR_MC_LINKLOCAL(&sin6->sin6_addr) || IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr)) continue; new_available_af |= HAVE_IPV6; break; default: break; } if (new_available_af == (HAVE_IPV4 | HAVE_IPV6)) break; } freeifaddrs(ifap); if (new_available_af != available_af) { available_af = new_available_af; frontend_imsg_compose_resolver(IMSG_CHANGE_AFS, 0, &available_af, sizeof(available_af)); } }