summaryrefslogtreecommitdiffstats
path: root/src/net_tcp.c
blob: 61676fc23e82c677e5b6f4c0b0e63d0506df55ad (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
/*
 * Copyright 2019-2022 OARC, Inc.
 * Copyright 2017-2018 Akamai Technologies
 * Copyright 2006-2016 Nominum, Inc.
 * All rights reserved.
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

#include "config.h"

#include "net.h"

#include "log.h"
#include "strerror.h"
#include "os.h"
#include "util.h"

#include <errno.h>
#include <fcntl.h>
#include <unistd.h>
#include <string.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include <ck_pr.h>

#define self ((struct perf__tcp_socket*)sock)

/*

About state sync between sending and receiving thread

Two variables in the TCP socket struct are used with libck to atomically
sync states between thread w.r.t connect/reconnect events.

  sock->fd is controlled by the sending thread (ST)
  self->recv_need_reconn is controlled by receiving thread (RT)

On connect/reconnect ST will open a new socket and atomically store it
into sock->fd.

When RT is trying to receive it will atomically load sock->fd and store it
in self->recvfd. Before storing it, it will compare it to what already in
self->recvfd and if it differ then a connect/reconnect event happend and RT
will reset receiving state and buffers.

If RT detects a disconnection it will atomically store self->recvfd into
self->recv_need_reconn to signal to ST that it needs to reconnect.

ST will load and check self->recv_need_reconn before sending and when
checking socket readiness, if its the same as sock->fd then it will start
reconnecting.

*/

struct perf__tcp_socket {
    struct perf_net_socket base;

    char   recvbuf[TCP_RECV_BUF_SIZE], sendbuf[TCP_SEND_BUF_SIZE];
    size_t at, sending;
    bool   is_ready, need_reconnect, have_more, is_sending;

    perf_sockaddr_t server, local;
    size_t          bufsize;

    int recvfd;
    int recv_need_reconn;

    uint16_t qid;

    uint64_t            conn_ts;
    perf_socket_event_t conn_event, conning_event;

    size_t       num_queries_per_conn, nqpc_timeout;
    unsigned int nqpc_sent, nqpc_recv;
    uint64_t     nqpc_ts;
};

static int perf__tcp_connect(struct perf_net_socket* sock)
{
    int fd;

    self->is_ready  = true;
    self->nqpc_sent = 0;
    ck_pr_store_uint(&self->nqpc_recv, 0);
    self->nqpc_ts = 0;

    fd = socket(self->server.sa.sa.sa_family, SOCK_STREAM, 0);
    if (fd == -1) {
        char __s[256];
        perf_log_fatal("socket: %s", perf_strerror_r(errno, __s, sizeof(__s)));
    }

    if (self->server.sa.sa.sa_family == AF_INET6) {
        int on = 1;

        if (setsockopt(fd, IPPROTO_IPV6, IPV6_V6ONLY, &on, sizeof(on)) == -1) {
            perf_log_warning("setsockopt(IPV6_V6ONLY) failed");
        }
    }

    if (bind(fd, &self->local.sa.sa, self->local.length) == -1) {
        char __s[256];
        perf_log_fatal("bind: %s", perf_strerror_r(errno, __s, sizeof(__s)));
    }

    if (self->bufsize) {
        int ret = setsockopt(fd, SOL_SOCKET, SO_RCVBUF,
            &self->bufsize, sizeof(self->bufsize));
        if (ret < 0)
            perf_log_warning("setsockbuf(SO_RCVBUF) failed");

        ret = setsockopt(fd, SOL_SOCKET, SO_SNDBUF,
            &self->bufsize, sizeof(self->bufsize));
        if (ret < 0)
            perf_log_warning("setsockbuf(SO_SNDBUF) failed");
    }

    int flags = fcntl(fd, F_GETFL, 0);
    if (flags < 0)
        perf_log_fatal("fcntl(F_GETFL)");
    int ret = fcntl(fd, F_SETFL, flags | O_NONBLOCK);
    if (ret < 0)
        perf_log_fatal("fcntl(F_SETFL)");

    self->conn_ts = perf_get_time();
    if (sock->event) {
        sock->event(sock, self->conning_event, self->conn_ts);
        self->conning_event = perf_socket_event_reconnecting;
    }
    if (connect(fd, &self->server.sa.sa, self->server.length)) {
        if (errno == EINPROGRESS) {
            self->is_ready = false;
        } else {
            char __s[256];
            perf_log_fatal("connect() failed: %s", perf_strerror_r(errno, __s, sizeof(__s)));
        }
    }

    return fd;
}

static ssize_t perf__tcp_recv(struct perf_net_socket* sock, void* buf, size_t len, int flags)
{
    ssize_t  n;
    uint16_t dnslen, dnslen2;

    int fd = ck_pr_load_int(&sock->fd);
    if (fd != self->recvfd) {
        /* reconnecting happened, reset buffers */
        self->have_more = false;
        self->at        = 0;
        self->recvfd    = fd;
    }

    if (!self->have_more) {
        n = recv(fd, self->recvbuf + self->at, TCP_RECV_BUF_SIZE - self->at, flags);
        if (!n) {
            // need reconnect
            ck_pr_store_int(&self->recv_need_reconn, fd);
            return 0;
        } else if (n < 0) {
            switch (errno) {
            case ECONNREFUSED:
            case ECONNRESET:
            case ENOTCONN:
                // need reconnect
                ck_pr_store_int(&self->recv_need_reconn, fd);
                errno = EAGAIN;
                break;
            default:
                break;
            }
            return n;
        }
        self->at += n;
        if (self->at < 3) {
            errno = EAGAIN;
            return -1;
        }
    }

    memcpy(&dnslen, self->recvbuf, 2);
    dnslen = ntohs(dnslen);
    if (self->at < dnslen + 2) {
        errno = EAGAIN;
        return -1;
    }
    memcpy(buf, self->recvbuf + 2, len < dnslen ? len : dnslen);
    memmove(self->recvbuf, self->recvbuf + 2 + dnslen, self->at - 2 - dnslen);
    self->at -= 2 + dnslen;
    if (self->num_queries_per_conn) {
        ck_pr_inc_uint(&self->nqpc_recv);
    }

    if (self->at > 2) {
        memcpy(&dnslen2, self->recvbuf, 2);
        dnslen2 = ntohs(dnslen2);
        if (self->at >= dnslen2 + 2) {
            self->have_more = true;
            return dnslen;
        }
    }

    self->have_more = false;
    return dnslen;
}

static ssize_t perf__tcp_sendto(struct perf_net_socket* sock, uint16_t qid, const void* buf, size_t len, int flags, const struct sockaddr* dest_addr, socklen_t addrlen)
{
    size_t send = len < TCP_SEND_BUF_SIZE - 2 ? len : (TCP_SEND_BUF_SIZE - 2);
    // TODO: We only send what we can send, because we can't continue sending
    uint16_t dnslen = htons(send);
    ssize_t  n;

    memcpy(self->sendbuf, &dnslen, 2);
    memcpy(self->sendbuf + 2, buf, send);
    self->qid = qid;

    int recv_need_reconn = ck_pr_load_int(&self->recv_need_reconn);
    if (recv_need_reconn == sock->fd) {
        self->need_reconnect = true;
        self->is_sending     = true;
        self->sending        = 0;
        errno                = EINPROGRESS;
        return -1;
    }

    n = sendto(sock->fd, self->sendbuf, send + 2, 0, 0, 0);

    if (n < 0) {
        switch (errno) {
        case ECONNREFUSED:
        case ECONNRESET:
        case ENOTCONN:
        case EPIPE:
            self->need_reconnect = true;
            self->is_sending     = true;
            self->sending        = 0;
            errno                = EINPROGRESS;
            return -1;
        default:
            break;
        }
        return -1;
    }

    if (n < send + 2) {
        self->is_sending = true;
        self->sending    = n;
        errno            = EINPROGRESS;
        return -1;
    }
    self->nqpc_sent++;

    return n - 2;
}

static int perf__tcp_close(struct perf_net_socket* sock)
{
    return close(sock->fd);
}

static int perf__tcp_sockeq(struct perf_net_socket* sock_a, struct perf_net_socket* sock_b)
{
    return sock_a->fd == sock_b->fd;
}

static int perf__tcp_sockready(struct perf_net_socket* sock, int pipe_fd, int64_t timeout)
{
    int recv_need_reconn = ck_pr_load_int(&self->recv_need_reconn);
    if (recv_need_reconn == sock->fd || self->need_reconnect) {
        int fd = perf__tcp_connect(sock), oldfd = ck_pr_load_int(&sock->fd);
        ck_pr_store_int(&sock->fd, fd);
        close(oldfd);
        self->need_reconnect = false;
    }

    if (self->is_ready) {
        if (self->is_sending) {
            uint16_t dnslen;
            ssize_t  n;

            memcpy(&dnslen, self->sendbuf, 2);
            dnslen = ntohs(dnslen);
            n      = sendto(sock->fd, self->sendbuf + self->sending, dnslen + 2 - self->sending, 0, 0, 0);
            if (n < 0) {
                int fd = perf__tcp_connect(sock), oldfd = ck_pr_load_int(&sock->fd);
                ck_pr_store_int(&sock->fd, fd);
                close(oldfd);
                if (self->sending) {
                    self->sending    = 0;
                    self->is_sending = false;
                }
                goto conn_cont;
            }
            self->sending += n;
            if (self->sending < dnslen + 2) {
                return 0;
            }
            self->sending    = 0;
            self->is_sending = false;
            if (sock->sent) {
                sock->sent(sock, self->qid);
            }
            self->nqpc_sent++;
        }
        if (self->num_queries_per_conn && self->nqpc_sent >= self->num_queries_per_conn) {
            if (!self->nqpc_ts) {
                self->nqpc_ts = perf_get_time() + self->nqpc_timeout;
            }
            unsigned int r = ck_pr_load_uint(&self->nqpc_recv);
            if (r >= self->nqpc_sent || perf_get_time() > self->nqpc_ts) {
                self->need_reconnect = true;
            }
            return 0;
        }
        return 1;
    }

conn_cont:
    switch (perf_os_waituntilanywritable(&sock, 1, pipe_fd, timeout)) {
    case PERF_R_TIMEDOUT:
        return -1;
    case PERF_R_SUCCESS: {
        int       error = 0;
        socklen_t len   = (socklen_t)sizeof(error);

        getsockopt(sock->fd, SOL_SOCKET, SO_ERROR, (void*)&error, &len);
        if (error != 0) {
            if (error == EINPROGRESS
#if EWOULDBLOCK != EAGAIN
                || error == EWOULDBLOCK
#endif
                || error == EAGAIN) {
                return 0;
            }
            // unrecoverable error, reconnect
            self->need_reconnect = true;
            return 0;
        }
        self->is_ready = true;
        if (sock->event) {
            sock->event(sock, self->conn_event, perf_get_time() - self->conn_ts);
            self->conn_event = perf_socket_event_reconnected;
        }
        if (self->is_sending) {
            uint16_t dnslen;
            ssize_t  n;

            memcpy(&dnslen, self->sendbuf, 2);
            dnslen = ntohs(dnslen);
            n      = sendto(sock->fd, self->sendbuf + self->sending, dnslen + 2 - self->sending, 0, 0, 0);
            if (n < 0) {
                self->need_reconnect = true;
                return 0;
            }
            self->sending += n;
            if (self->sending < dnslen + 2) {
                return 0;
            }
            self->sending    = 0;
            self->is_sending = false;
            if (sock->sent) {
                sock->sent(sock, self->qid);
            }
            self->nqpc_sent++;
        }
        return 1;
    }
    default:
        break;
    }

    return -1;
}

static bool perf__tcp_have_more(struct perf_net_socket* sock)
{
    return self->have_more;
}

static void perf__tcp_num_queries_per_conn(struct perf_net_socket* sock, size_t num_queries_per_conn, size_t timeout)
{
    self->num_queries_per_conn = num_queries_per_conn;
    self->nqpc_timeout         = timeout;
}

struct perf_net_socket* perf_net_tcp_opensocket(const perf_sockaddr_t* server, const perf_sockaddr_t* local, size_t bufsize, void* data, perf_net_sent_cb_t sent, perf_net_event_cb_t event)
{
    struct perf__tcp_socket* tmp  = calloc(1, sizeof(struct perf__tcp_socket)); // clang scan-build
    struct perf_net_socket*  sock = (struct perf_net_socket*)tmp;

    if (!sock) {
        perf_log_fatal("perf_net_tcp_opensocket() out of memory");
        return 0; // needed for clang scan build
    }

    sock->recv      = perf__tcp_recv;
    sock->sendto    = perf__tcp_sendto;
    sock->close     = perf__tcp_close;
    sock->sockeq    = perf__tcp_sockeq;
    sock->sockready = perf__tcp_sockready;
    sock->have_more = perf__tcp_have_more;

    sock->num_queries_per_conn = perf__tcp_num_queries_per_conn;

    sock->data  = data;
    sock->sent  = sent;
    sock->event = event;

    self->server  = *server;
    self->local   = *local;
    self->bufsize = bufsize;
    if (self->bufsize > 0) {
        self->bufsize *= 1024;
    }
    self->conning_event = perf_socket_event_connecting;
    self->conn_event    = perf_socket_event_connected;

    sock->fd               = perf__tcp_connect(sock);
    self->recvfd           = sock->fd;
    self->recv_need_reconn = -1;

    return sock;
}