summaryrefslogtreecommitdiffstats
path: root/src/resolve/resolved-dns-stream.c
blob: e57af6622180ad5e5f55968cc173ea75be3d2ecf (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
/* SPDX-License-Identifier: LGPL-2.1-or-later */

#include <linux/if_arp.h>
#include <netinet/tcp.h>
#include <unistd.h>

#include "alloc-util.h"
#include "fd-util.h"
#include "iovec-util.h"
#include "macro.h"
#include "missing_network.h"
#include "resolved-dns-stream.h"
#include "resolved-manager.h"

#define DNS_STREAMS_MAX 128

#define DNS_QUERIES_PER_STREAM 32

static void dns_stream_stop(DnsStream *s) {
        assert(s);

        s->io_event_source = sd_event_source_disable_unref(s->io_event_source);
        s->timeout_event_source = sd_event_source_disable_unref(s->timeout_event_source);
        s->fd = safe_close(s->fd);

        /* Disconnect us from the server object if we are now not usable anymore */
        dns_stream_detach(s);
}

static int dns_stream_update_io(DnsStream *s) {
        uint32_t f = 0;

        assert(s);

        if (s->write_packet && s->n_written < sizeof(s->write_size) + s->write_packet->size)
                f |= EPOLLOUT;
        else if (!ordered_set_isempty(s->write_queue)) {
                dns_packet_unref(s->write_packet);
                s->write_packet = ordered_set_steal_first(s->write_queue);
                s->write_size = htobe16(s->write_packet->size);
                s->n_written = 0;
                f |= EPOLLOUT;
        }

        /* Let's read a packet if we haven't queued any yet. Except if we already hit a limit of parallel
         * queries for this connection. */
        if ((!s->read_packet || s->n_read < sizeof(s->read_size) + s->read_packet->size) &&
                set_size(s->queries) < DNS_QUERIES_PER_STREAM)
                f |= EPOLLIN;

        s->requested_events = f;

#if ENABLE_DNS_OVER_TLS
        /* For handshake and clean closing purposes, TLS can override requested events */
        if (s->dnstls_events != 0)
                f = s->dnstls_events;
#endif

        return sd_event_source_set_io_events(s->io_event_source, f);
}

static int dns_stream_complete(DnsStream *s, int error) {
        _cleanup_(dns_stream_unrefp) _unused_ DnsStream *ref = dns_stream_ref(s); /* Protect stream while we process it */

        assert(s);
        assert(error >= 0);

        /* Error is > 0 when the connection failed for some reason in the network stack. It's == 0 if we sent
         * and received exactly one packet each (in the LLMNR client case). */

#if ENABLE_DNS_OVER_TLS
        if (s->encrypted) {
                int r;

                r = dnstls_stream_shutdown(s, error);
                if (r != -EAGAIN)
                        dns_stream_stop(s);
        } else
#endif
                dns_stream_stop(s);

        dns_stream_detach(s);

        if (s->complete)
                s->complete(s, error);
        else /* the default action if no completion function is set is to close the stream */
                dns_stream_unref(s);

        return 0;
}

static int dns_stream_identify(DnsStream *s) {
        CMSG_BUFFER_TYPE(CMSG_SPACE(MAXSIZE(struct in_pktinfo, struct in6_pktinfo))
                         + CMSG_SPACE(int) + /* for the TTL */
                         + EXTRA_CMSG_SPACE /* kernel appears to require extra space */) control;
        struct msghdr mh = {};
        struct cmsghdr *cmsg;
        socklen_t sl;
        int r;

        assert(s);

        if (s->identified)
                return 0;

        /* Query the local side */
        s->local_salen = sizeof(s->local);
        r = getsockname(s->fd, &s->local.sa, &s->local_salen);
        if (r < 0)
                return -errno;
        if (s->local.sa.sa_family == AF_INET6 && s->ifindex <= 0)
                s->ifindex = s->local.in6.sin6_scope_id;

        /* Query the remote side */
        s->peer_salen = sizeof(s->peer);
        r = getpeername(s->fd, &s->peer.sa, &s->peer_salen);
        if (r < 0)
                return -errno;
        if (s->peer.sa.sa_family == AF_INET6 && s->ifindex <= 0)
                s->ifindex = s->peer.in6.sin6_scope_id;

        /* Check consistency */
        assert(s->peer.sa.sa_family == s->local.sa.sa_family);
        assert(IN_SET(s->peer.sa.sa_family, AF_INET, AF_INET6));

        /* Query connection meta information */
        sl = sizeof(control);
        if (s->peer.sa.sa_family == AF_INET) {
                r = getsockopt(s->fd, IPPROTO_IP, IP_PKTOPTIONS, &control, &sl);
                if (r < 0)
                        return -errno;
        } else if (s->peer.sa.sa_family == AF_INET6) {

                r = getsockopt(s->fd, IPPROTO_IPV6, IPV6_2292PKTOPTIONS, &control, &sl);
                if (r < 0)
                        return -errno;
        } else
                return -EAFNOSUPPORT;

        mh.msg_control = &control;
        mh.msg_controllen = sl;

        CMSG_FOREACH(cmsg, &mh) {

                if (cmsg->cmsg_level == IPPROTO_IPV6) {
                        assert(s->peer.sa.sa_family == AF_INET6);

                        switch (cmsg->cmsg_type) {

                        case IPV6_PKTINFO: {
                                struct in6_pktinfo *i = CMSG_TYPED_DATA(cmsg, struct in6_pktinfo);

                                if (s->ifindex <= 0)
                                        s->ifindex = i->ipi6_ifindex;
                                break;
                        }

                        case IPV6_HOPLIMIT:
                                s->ttl = *CMSG_TYPED_DATA(cmsg, int);
                                break;
                        }

                } else if (cmsg->cmsg_level == IPPROTO_IP) {
                        assert(s->peer.sa.sa_family == AF_INET);

                        switch (cmsg->cmsg_type) {

                        case IP_PKTINFO: {
                                struct in_pktinfo *i = CMSG_TYPED_DATA(cmsg, struct in_pktinfo);

                                if (s->ifindex <= 0)
                                        s->ifindex = i->ipi_ifindex;
                                break;
                        }

                        case IP_TTL:
                                s->ttl = *CMSG_TYPED_DATA(cmsg, int);
                                break;
                        }
                }
        }

        /* The Linux kernel sets the interface index to the loopback
         * device if the connection came from the local host since it
         * avoids the routing table in such a case. Let's unset the
         * interface index in such a case. */
        if (s->ifindex == LOOPBACK_IFINDEX)
                s->ifindex = 0;

        /* If we don't know the interface index still, we look for the
         * first local interface with a matching address. Yuck! */
        if (s->ifindex <= 0)
                s->ifindex = manager_find_ifindex(s->manager, s->local.sa.sa_family, sockaddr_in_addr(&s->local.sa));

        if (s->protocol == DNS_PROTOCOL_LLMNR && s->ifindex > 0) {
                /* Make sure all packets for this connection are sent on the same interface */
                r = socket_set_unicast_if(s->fd, s->local.sa.sa_family, s->ifindex);
                if (r < 0)
                        log_debug_errno(r, "Failed to invoke IP_UNICAST_IF/IPV6_UNICAST_IF: %m");
        }

        s->identified = true;

        return 0;
}

ssize_t dns_stream_writev(DnsStream *s, const struct iovec *iov, size_t iovcnt, int flags) {
        ssize_t m;
        int r;

        assert(s);
        assert(iov);

#if ENABLE_DNS_OVER_TLS
        if (s->encrypted && !(flags & DNS_STREAM_WRITE_TLS_DATA))
                return dnstls_stream_writev(s, iov, iovcnt);
#endif

        if (s->tfo_salen > 0) {
                struct msghdr hdr = {
                        .msg_iov = (struct iovec*) iov,
                        .msg_iovlen = iovcnt,
                        .msg_name = &s->tfo_address.sa,
                        .msg_namelen = s->tfo_salen
                };

                m = sendmsg(s->fd, &hdr, MSG_FASTOPEN);
                if (m < 0) {
                        if (ERRNO_IS_NOT_SUPPORTED(errno)) {
                                /* MSG_FASTOPEN not supported? Then try to connect() traditionally */
                                r = RET_NERRNO(connect(s->fd, &s->tfo_address.sa, s->tfo_salen));
                                s->tfo_salen = 0; /* connection is made */
                                if (r < 0 && r != -EINPROGRESS)
                                        return r;

                                return -EAGAIN; /* In case of EINPROGRESS, EAGAIN or success: return EAGAIN, so that caller calls us again */
                        }
                        if (errno == EINPROGRESS)
                                return -EAGAIN;

                        return -errno;
                } else
                        s->tfo_salen = 0; /* connection is made */
        } else {
                m = writev(s->fd, iov, iovcnt);
                if (m < 0)
                        return -errno;
        }

        return m;
}

static ssize_t dns_stream_read(DnsStream *s, void *buf, size_t count) {
        ssize_t ss;

#if ENABLE_DNS_OVER_TLS
        if (s->encrypted)
                ss = dnstls_stream_read(s, buf, count);
        else
#endif
        {
                ss = read(s->fd, buf, count);
                if (ss < 0)
                        return -errno;
        }

        return ss;
}

static int on_stream_timeout(sd_event_source *es, usec_t usec, void *userdata) {
        DnsStream *s = ASSERT_PTR(userdata);

        return dns_stream_complete(s, ETIMEDOUT);
}

static DnsPacket *dns_stream_take_read_packet(DnsStream *s) {
        assert(s);

        /* Note, dns_stream_update() should be called after this is called. When this is called, the
         * stream may be already full and the EPOLLIN flag is dropped from the stream IO event source.
         * Even this makes a room to read in the stream, this does not call dns_stream_update(), hence
         * EPOLLIN flag is not set automatically. So, to read further packets from the stream,
         * dns_stream_update() must be called explicitly. Currently, this is only called from
         * on_stream_io(), and there dns_stream_update() is called. */

        if (!s->read_packet)
                return NULL;

        if (s->n_read < sizeof(s->read_size))
                return NULL;

        if (s->n_read < sizeof(s->read_size) + be16toh(s->read_size))
                return NULL;

        s->n_read = 0;
        return TAKE_PTR(s->read_packet);
}

static int on_stream_io(sd_event_source *es, int fd, uint32_t revents, void *userdata) {
        _cleanup_(dns_stream_unrefp) DnsStream *s = dns_stream_ref(userdata); /* Protect stream while we process it */
        bool progressed = false;
        int r;

        assert(s);

#if ENABLE_DNS_OVER_TLS
        if (s->encrypted) {
                r = dnstls_stream_on_io(s, revents);
                if (r == DNSTLS_STREAM_CLOSED)
                        return 0;
                if (r == -EAGAIN)
                        return dns_stream_update_io(s);
                if (r < 0)
                        return dns_stream_complete(s, -r);

                r = dns_stream_update_io(s);
                if (r < 0)
                        return r;
        }
#endif

        /* only identify after connecting */
        if (s->tfo_salen == 0) {
                r = dns_stream_identify(s);
                if (r < 0)
                        return dns_stream_complete(s, -r);
        }

        if (revents & EPOLLERR) {
                socklen_t errlen = sizeof(r);
                if (getsockopt(s->fd, SOL_SOCKET, SO_ERROR, &r, &errlen) == 0)
                        return dns_stream_complete(s, r);
        }

        if ((revents & EPOLLOUT) &&
            s->write_packet &&
            s->n_written < sizeof(s->write_size) + s->write_packet->size) {

                struct iovec iov[] = {
                        IOVEC_MAKE(&s->write_size, sizeof(s->write_size)),
                        IOVEC_MAKE(DNS_PACKET_DATA(s->write_packet), s->write_packet->size),
                };

                iovec_increment(iov, ELEMENTSOF(iov), s->n_written);

                ssize_t ss = dns_stream_writev(s, iov, ELEMENTSOF(iov), 0);
                if (ss < 0) {
                        if (!ERRNO_IS_TRANSIENT(ss))
                                return dns_stream_complete(s, -ss);
                } else {
                        progressed = true;
                        s->n_written += ss;
                }

                /* Are we done? If so, disable the event source for EPOLLOUT */
                if (s->n_written >= sizeof(s->write_size) + s->write_packet->size) {
                        r = dns_stream_update_io(s);
                        if (r < 0)
                                return dns_stream_complete(s, -r);
                }
        }

        while ((revents & (EPOLLIN|EPOLLHUP|EPOLLRDHUP)) &&
               (!s->read_packet ||
                s->n_read < sizeof(s->read_size) + s->read_packet->size)) {

                if (s->n_read < sizeof(s->read_size)) {
                        ssize_t ss;

                        ss = dns_stream_read(s, (uint8_t*) &s->read_size + s->n_read, sizeof(s->read_size) - s->n_read);
                        if (ss < 0) {
                                if (!ERRNO_IS_TRANSIENT(ss))
                                        return dns_stream_complete(s, -ss);
                                break;
                        } else if (ss == 0)
                                return dns_stream_complete(s, ECONNRESET);
                        else {
                                progressed = true;
                                s->n_read += ss;
                        }
                }

                if (s->n_read >= sizeof(s->read_size)) {

                        if (be16toh(s->read_size) < DNS_PACKET_HEADER_SIZE)
                                return dns_stream_complete(s, EBADMSG);

                        if (s->n_read < sizeof(s->read_size) + be16toh(s->read_size)) {
                                ssize_t ss;

                                if (!s->read_packet) {
                                        r = dns_packet_new(&s->read_packet, s->protocol, be16toh(s->read_size), DNS_PACKET_SIZE_MAX);
                                        if (r < 0)
                                                return dns_stream_complete(s, -r);

                                        s->read_packet->size = be16toh(s->read_size);
                                        s->read_packet->ipproto = IPPROTO_TCP;
                                        s->read_packet->family = s->peer.sa.sa_family;
                                        s->read_packet->ttl = s->ttl;
                                        s->read_packet->ifindex = s->ifindex;
                                        s->read_packet->timestamp = now(CLOCK_BOOTTIME);

                                        if (s->read_packet->family == AF_INET) {
                                                s->read_packet->sender.in = s->peer.in.sin_addr;
                                                s->read_packet->sender_port = be16toh(s->peer.in.sin_port);
                                                s->read_packet->destination.in = s->local.in.sin_addr;
                                                s->read_packet->destination_port = be16toh(s->local.in.sin_port);
                                        } else {
                                                assert(s->read_packet->family == AF_INET6);
                                                s->read_packet->sender.in6 = s->peer.in6.sin6_addr;
                                                s->read_packet->sender_port = be16toh(s->peer.in6.sin6_port);
                                                s->read_packet->destination.in6 = s->local.in6.sin6_addr;
                                                s->read_packet->destination_port = be16toh(s->local.in6.sin6_port);

                                                if (s->read_packet->ifindex == 0)
                                                        s->read_packet->ifindex = s->peer.in6.sin6_scope_id;
                                                if (s->read_packet->ifindex == 0)
                                                        s->read_packet->ifindex = s->local.in6.sin6_scope_id;
                                        }
                                }

                                ss = dns_stream_read(s,
                                          (uint8_t*) DNS_PACKET_DATA(s->read_packet) + s->n_read - sizeof(s->read_size),
                                          sizeof(s->read_size) + be16toh(s->read_size) - s->n_read);
                                if (ss < 0) {
                                        if (!ERRNO_IS_TRANSIENT(ss))
                                                return dns_stream_complete(s, -ss);
                                        break;
                                } else if (ss == 0)
                                        return dns_stream_complete(s, ECONNRESET);
                                else
                                        s->n_read += ss;
                        }

                        /* Are we done? If so, call the packet handler and re-enable EPOLLIN for the
                         * event source if necessary. */
                        _cleanup_(dns_packet_unrefp) DnsPacket *p = dns_stream_take_read_packet(s);
                        if (p) {
                                assert(s->on_packet);
                                r = s->on_packet(s, p);
                                if (r < 0)
                                        return r;

                                r = dns_stream_update_io(s);
                                if (r < 0)
                                        return dns_stream_complete(s, -r);

                                s->packet_received = true;

                                /* If we just disabled the read event, stop reading */
                                if (!FLAGS_SET(s->requested_events, EPOLLIN))
                                        break;
                        }
                }
        }

        /* Complete the stream if finished reading and writing one packet, and there's nothing
         * else left to write. */
        if (s->type == DNS_STREAM_LLMNR_SEND && s->packet_received &&
            !FLAGS_SET(s->requested_events, EPOLLOUT))
                return dns_stream_complete(s, 0);

        /* If we did something, let's restart the timeout event source */
        if (progressed && s->timeout_event_source) {
                r = sd_event_source_set_time_relative(s->timeout_event_source, DNS_STREAM_ESTABLISHED_TIMEOUT_USEC);
                if (r < 0)
                        log_warning_errno(r, "Couldn't restart TCP connection timeout, ignoring: %m");
        }

        return 0;
}

static DnsStream *dns_stream_free(DnsStream *s) {
        DnsPacket *p;

        assert(s);

        dns_stream_stop(s);

        if (s->manager) {
                LIST_REMOVE(streams, s->manager->dns_streams, s);
                s->manager->n_dns_streams[s->type]--;
        }

#if ENABLE_DNS_OVER_TLS
        if (s->encrypted)
                dnstls_stream_free(s);
#endif

        ORDERED_SET_FOREACH(p, s->write_queue)
                dns_packet_unref(ordered_set_remove(s->write_queue, p));

        dns_packet_unref(s->write_packet);
        dns_packet_unref(s->read_packet);
        dns_server_unref(s->server);

        ordered_set_free(s->write_queue);

        return mfree(s);
}

DEFINE_TRIVIAL_REF_UNREF_FUNC(DnsStream, dns_stream, dns_stream_free);

int dns_stream_new(
                Manager *m,
                DnsStream **ret,
                DnsStreamType type,
                DnsProtocol protocol,
                int fd,
                const union sockaddr_union *tfo_address,
                int (on_packet)(DnsStream*, DnsPacket*),
                int (complete)(DnsStream*, int), /* optional */
                usec_t connect_timeout_usec) {

        _cleanup_(dns_stream_unrefp) DnsStream *s = NULL;
        int r;

        assert(m);
        assert(ret);
        assert(type >= 0);
        assert(type < _DNS_STREAM_TYPE_MAX);
        assert(protocol >= 0);
        assert(protocol < _DNS_PROTOCOL_MAX);
        assert(fd >= 0);
        assert(on_packet);

        if (m->n_dns_streams[type] > DNS_STREAMS_MAX)
                return -EBUSY;

        s = new(DnsStream, 1);
        if (!s)
                return -ENOMEM;

        *s = (DnsStream) {
                .n_ref = 1,
                .fd = -EBADF,
                .protocol = protocol,
                .type = type,
        };

        r = ordered_set_ensure_allocated(&s->write_queue, &dns_packet_hash_ops);
        if (r < 0)
                return r;

        r = sd_event_add_io(m->event, &s->io_event_source, fd, EPOLLIN, on_stream_io, s);
        if (r < 0)
                return r;

        (void) sd_event_source_set_description(s->io_event_source, "dns-stream-io");

        r = sd_event_add_time_relative(
                        m->event,
                        &s->timeout_event_source,
                        CLOCK_BOOTTIME,
                        connect_timeout_usec, 0,
                        on_stream_timeout, s);
        if (r < 0)
                return r;

        (void) sd_event_source_set_description(s->timeout_event_source, "dns-stream-timeout");

        LIST_PREPEND(streams, m->dns_streams, s);
        m->n_dns_streams[type]++;
        s->manager = m;

        s->fd = fd;
        s->on_packet = on_packet;
        s->complete = complete;

        if (tfo_address) {
                s->tfo_address = *tfo_address;
                s->tfo_salen = SOCKADDR_LEN(*tfo_address);
        }

        *ret = TAKE_PTR(s);

        return 0;
}

int dns_stream_write_packet(DnsStream *s, DnsPacket *p) {
        int r;

        assert(s);
        assert(p);

        r = ordered_set_put(s->write_queue, p);
        if (r < 0)
                return r;

        dns_packet_ref(p);

        return dns_stream_update_io(s);
}

void dns_stream_detach(DnsStream *s) {
        assert(s);

        if (!s->server)
                return;

        if (s->server->stream != s)
                return;

        dns_server_unref_stream(s->server);
}

DEFINE_PRIVATE_HASH_OPS_WITH_KEY_DESTRUCTOR(
                dns_stream_hash_ops,
                void,
                trivial_hash_func,
                trivial_compare_func,
                dns_stream_unref);

int dns_stream_disconnect_all(Manager *m) {
        _cleanup_(set_freep) Set *closed = NULL;
        int r;

        assert(m);

        /* Terminates all TCP connections (called after system suspend for example, to speed up recovery) */

        log_info("Closing all remaining TCP connections.");

        bool restart;
        do {
                restart = false;

                LIST_FOREACH(streams, s, m->dns_streams) {
                        r = set_ensure_put(&closed, &dns_stream_hash_ops, s);
                        if (r < 0)
                                return log_oom();
                        if (r > 0) {
                                /* Haven't seen this one before. Close it. */
                                dns_stream_ref(s);
                                (void) dns_stream_complete(s, ECONNRESET);

                                /* This might have a ripple effect, let's hence no look at the list further,
                                 * but scan from the beginning again */
                                restart = true;
                                break;
                        }
                }
        } while (restart);

        return 0;
}