blob: 91ad0b63ddc4dfc299df9b869418045d2ada073c [file] [log] [blame]
Willy Tarreau59f98392012-07-06 14:13:49 +02001/*
2 * Connection management functions
3 *
4 * Copyright 2000-2012 Willy Tarreau <w@1wt.eu>
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 *
11 */
12
Willy Tarreaue1e4a612012-10-05 00:10:55 +020013#include <errno.h>
14
Willy Tarreau59f98392012-07-06 14:13:49 +020015#include <common/compat.h>
16#include <common/config.h>
17
Willy Tarreauc5788912012-08-24 18:12:41 +020018#include <proto/connection.h>
Willy Tarreaudd2f85e2012-09-02 22:34:23 +020019#include <proto/fd.h>
Willy Tarreau5f1504f2012-10-04 23:55:57 +020020#include <proto/frontend.h>
Willy Tarreau2da156f2012-07-23 15:07:23 +020021#include <proto/proto_tcp.h>
Willy Tarreau2542b532012-08-31 16:01:23 +020022#include <proto/session.h>
Willy Tarreau2c6be842012-07-06 17:12:34 +020023#include <proto/stream_interface.h>
Willy Tarreau59f98392012-07-06 14:13:49 +020024
Emeric Brun46591952012-05-18 15:47:34 +020025#ifdef USE_OPENSSL
26#include <proto/ssl_sock.h>
27#endif
28
Willy Tarreauf2943dc2012-10-26 20:10:28 +020029struct pool_head *pool2_connection;
30
31/* perform minimal intializations, report 0 in case of error, 1 if OK. */
32int init_connection()
33{
34 pool2_connection = create_pool("connection", sizeof (struct connection), MEM_F_SHARED);
35 return pool2_connection != NULL;
36}
37
Willy Tarreau59f98392012-07-06 14:13:49 +020038/* I/O callback for fd-based connections. It calls the read/write handlers
Willy Tarreauafad0e02012-08-09 14:45:22 +020039 * provided by the connection's sock_ops, which must be valid. It returns 0.
Willy Tarreau59f98392012-07-06 14:13:49 +020040 */
41int conn_fd_handler(int fd)
42{
Willy Tarreau80184712012-07-06 14:54:49 +020043 struct connection *conn = fdtab[fd].owner;
Willy Tarreau9e272bf2012-10-03 21:04:48 +020044 unsigned int flags;
Willy Tarreau59f98392012-07-06 14:13:49 +020045
Willy Tarreauc76ae332012-07-12 15:32:13 +020046 if (unlikely(!conn))
Willy Tarreau2542b532012-08-31 16:01:23 +020047 return 0;
Willy Tarreau59f98392012-07-06 14:13:49 +020048
Willy Tarreau7d281492012-12-16 19:19:13 +010049 conn_refresh_polling_flags(conn);
50 flags = conn->flags & ~CO_FL_ERROR; /* ensure to call the wake handler upon error */
Willy Tarreaud29a0662012-12-10 16:33:38 +010051
Willy Tarreauc76ae332012-07-12 15:32:13 +020052 process_handshake:
Willy Tarreauf9dabec2012-08-17 17:33:53 +020053 /* The handshake callbacks are called in sequence. If either of them is
54 * missing something, it must enable the required polling at the socket
55 * layer of the connection. Polling state is not guaranteed when entering
56 * these handlers, so any handshake handler which does not complete its
Willy Tarreaud6e999b2013-11-25 08:41:15 +010057 * work must explicitly disable events it's not interested in. Error
58 * handling is also performed here in order to reduce the number of tests
59 * around.
Willy Tarreauf9dabec2012-08-17 17:33:53 +020060 */
Willy Tarreaud6e999b2013-11-25 08:41:15 +010061 while (unlikely(conn->flags & (CO_FL_HANDSHAKE | CO_FL_ERROR))) {
Willy Tarreau310987a2014-01-22 19:46:33 +010062 if (unlikely(conn->flags & CO_FL_ERROR))
Willy Tarreau2c6be842012-07-06 17:12:34 +020063 goto leave;
Willy Tarreau59f98392012-07-06 14:13:49 +020064
Willy Tarreau22cda212012-08-31 17:43:29 +020065 if (conn->flags & CO_FL_ACCEPT_PROXY)
66 if (!conn_recv_proxy(conn, CO_FL_ACCEPT_PROXY))
67 goto leave;
68
Willy Tarreau57cd3e42013-10-24 22:01:26 +020069 if (conn->flags & CO_FL_SEND_PROXY)
70 if (!conn_si_send_proxy(conn, CO_FL_SEND_PROXY))
Willy Tarreau5f1504f2012-10-04 23:55:57 +020071 goto leave;
Emeric Brun46591952012-05-18 15:47:34 +020072#ifdef USE_OPENSSL
73 if (conn->flags & CO_FL_SSL_WAIT_HS)
74 if (!ssl_sock_handshake(conn, CO_FL_SSL_WAIT_HS))
75 goto leave;
76#endif
Willy Tarreauc76ae332012-07-12 15:32:13 +020077 }
78
Willy Tarreauf9dabec2012-08-17 17:33:53 +020079 /* Once we're purely in the data phase, we disable handshake polling */
80 if (!(conn->flags & CO_FL_POLL_SOCK))
81 __conn_sock_stop_both(conn);
Willy Tarreauc76ae332012-07-12 15:32:13 +020082
Willy Tarreau071e1372012-10-03 01:39:48 +020083 /* The data layer might not be ready yet (eg: when using embryonic
84 * sessions). If we're about to move data, we must initialize it first.
85 * The function may fail and cause the connection to be destroyed, thus
Willy Tarreau2542b532012-08-31 16:01:23 +020086 * we must not use it anymore and should immediately leave instead.
87 */
Willy Tarreau071e1372012-10-03 01:39:48 +020088 if ((conn->flags & CO_FL_INIT_DATA) && conn->data->init(conn) < 0)
Willy Tarreau2542b532012-08-31 16:01:23 +020089 return 0;
90
Willy Tarreau153c3ca2012-10-22 22:47:55 +020091 /* The data transfer starts here and stops on error and handshakes. Note
92 * that we must absolutely test conn->xprt at each step in case it suddenly
93 * changes due to a quick unexpected close().
94 */
Willy Tarreaud8375892014-01-21 11:01:08 +010095 if (conn->xprt && fd_recv_ready(fd) &&
96 ((conn->flags & (CO_FL_DATA_RD_ENA|CO_FL_WAIT_ROOM|CO_FL_ERROR|CO_FL_HANDSHAKE)) == CO_FL_DATA_RD_ENA)) {
Willy Tarreau3b5bc662012-10-05 21:29:37 +020097 /* force detection of a flag change : it's impossible to have both
98 * CONNECTED and WAIT_CONN so we're certain to trigger a change.
Willy Tarreau9e272bf2012-10-03 21:04:48 +020099 */
Willy Tarreau3b5bc662012-10-05 21:29:37 +0200100 flags = CO_FL_WAIT_L4_CONN | CO_FL_CONNECTED;
Willy Tarreau74beec32012-10-03 00:41:04 +0200101 conn->data->recv(conn);
Willy Tarreau9e272bf2012-10-03 21:04:48 +0200102 }
Willy Tarreau59f98392012-07-06 14:13:49 +0200103
Willy Tarreaud8375892014-01-21 11:01:08 +0100104 if (conn->xprt && fd_send_ready(fd) &&
105 ((conn->flags & (CO_FL_DATA_WR_ENA|CO_FL_WAIT_DATA|CO_FL_ERROR|CO_FL_HANDSHAKE)) == CO_FL_DATA_WR_ENA)) {
Willy Tarreau3b5bc662012-10-05 21:29:37 +0200106 /* force detection of a flag change : it's impossible to have both
107 * CONNECTED and WAIT_CONN so we're certain to trigger a change.
Willy Tarreau9e272bf2012-10-03 21:04:48 +0200108 */
Willy Tarreau3b5bc662012-10-05 21:29:37 +0200109 flags = CO_FL_WAIT_L4_CONN | CO_FL_CONNECTED;
Willy Tarreau74beec32012-10-03 00:41:04 +0200110 conn->data->send(conn);
Willy Tarreau9e272bf2012-10-03 21:04:48 +0200111 }
Willy Tarreau2da156f2012-07-23 15:07:23 +0200112
Willy Tarreauc76ae332012-07-12 15:32:13 +0200113 /* It may happen during the data phase that a handshake is
114 * enabled again (eg: SSL)
115 */
Willy Tarreaud6e999b2013-11-25 08:41:15 +0100116 if (unlikely(conn->flags & (CO_FL_HANDSHAKE | CO_FL_ERROR)))
Willy Tarreauc76ae332012-07-12 15:32:13 +0200117 goto process_handshake;
118
Willy Tarreaufd803bb2014-01-20 15:13:07 +0100119 if (unlikely(conn->flags & CO_FL_WAIT_L4_CONN)) {
Willy Tarreauf8deb0c2012-09-01 17:59:22 +0200120 /* still waiting for a connection to establish and nothing was
121 * attempted yet to probe the connection. Then let's retry the
122 * connect().
Willy Tarreau2da156f2012-07-23 15:07:23 +0200123 */
Willy Tarreau239d7182012-07-23 18:53:03 +0200124 if (!tcp_connect_probe(conn))
Willy Tarreauafad0e02012-08-09 14:45:22 +0200125 goto leave;
Willy Tarreau2da156f2012-07-23 15:07:23 +0200126 }
127
Willy Tarreau2c6be842012-07-06 17:12:34 +0200128 leave:
Willy Tarreau2396c1c2012-10-03 21:12:16 +0200129 /* The wake callback may be used to process a critical error and abort the
130 * connection. If so, we don't want to go further as the connection will
131 * have been released and the FD destroyed.
132 */
133 if ((conn->flags & CO_FL_WAKE_DATA) &&
134 ((conn->flags ^ flags) & CO_FL_CONN_STATE) &&
135 conn->data->wake(conn) < 0)
136 return 0;
Willy Tarreaufd31e532012-07-23 18:24:25 +0200137
Willy Tarreau8f8c92f2012-07-23 19:45:44 +0200138 /* Last check, verify if the connection just established */
Willy Tarreauc76ae332012-07-12 15:32:13 +0200139 if (unlikely(!(conn->flags & (CO_FL_WAIT_L4_CONN | CO_FL_WAIT_L6_CONN | CO_FL_CONNECTED))))
Willy Tarreau8f8c92f2012-07-23 19:45:44 +0200140 conn->flags |= CO_FL_CONNECTED;
141
Willy Tarreau61ace1b2012-07-23 12:14:26 +0200142 /* remove the events before leaving */
Willy Tarreau26d7cfc2012-12-07 00:09:43 +0100143 fdtab[fd].ev &= FD_POLL_STICKY;
Willy Tarreauf9dabec2012-08-17 17:33:53 +0200144
145 /* commit polling changes */
146 conn_cond_update_polling(conn);
Willy Tarreauafad0e02012-08-09 14:45:22 +0200147 return 0;
Willy Tarreau59f98392012-07-06 14:13:49 +0200148}
Willy Tarreaub5e2cbd2012-08-17 11:55:04 +0200149
Willy Tarreaue9dfa792012-09-01 17:26:16 +0200150/* Update polling on connection <c>'s file descriptor depending on its current
151 * state as reported in the connection's CO_FL_CURR_* flags, reports of EAGAIN
152 * in CO_FL_WAIT_*, and the data layer expectations indicated by CO_FL_DATA_*.
153 * The connection flags are updated with the new flags at the end of the
Willy Tarreau0ffde2c2012-10-04 22:21:15 +0200154 * operation. Polling is totally disabled if an error was reported.
Willy Tarreaub5e2cbd2012-08-17 11:55:04 +0200155 */
Willy Tarreaue9dfa792012-09-01 17:26:16 +0200156void conn_update_data_polling(struct connection *c)
Willy Tarreaub5e2cbd2012-08-17 11:55:04 +0200157{
Willy Tarreaue9dfa792012-09-01 17:26:16 +0200158 unsigned int f = c->flags;
Willy Tarreaub5e2cbd2012-08-17 11:55:04 +0200159
Willy Tarreau3c728722014-01-23 13:50:42 +0100160 if (!conn_ctrl_ready(c))
Willy Tarreauf79c8172013-10-21 16:30:56 +0200161 return;
162
Willy Tarreaub5e2cbd2012-08-17 11:55:04 +0200163 /* update read status if needed */
Willy Tarreau310987a2014-01-22 19:46:33 +0100164 if (unlikely((f & (CO_FL_CURR_RD_ENA|CO_FL_DATA_RD_ENA)) == CO_FL_DATA_RD_ENA)) {
Willy Tarreauf817e9f2014-01-10 16:58:45 +0100165 fd_want_recv(c->t.sock.fd);
Willy Tarreauc8dd77f2012-11-05 17:52:26 +0100166 f |= CO_FL_CURR_RD_ENA;
Willy Tarreaue9dfa792012-09-01 17:26:16 +0200167 }
Willy Tarreauc8dd77f2012-11-05 17:52:26 +0100168 else if (unlikely((f & (CO_FL_CURR_RD_ENA|CO_FL_DATA_RD_ENA)) == CO_FL_CURR_RD_ENA)) {
169 fd_stop_recv(c->t.sock.fd);
170 f &= ~CO_FL_CURR_RD_ENA;
Willy Tarreaue9dfa792012-09-01 17:26:16 +0200171 }
Willy Tarreaub5e2cbd2012-08-17 11:55:04 +0200172
173 /* update write status if needed */
Willy Tarreau310987a2014-01-22 19:46:33 +0100174 if (unlikely((f & (CO_FL_CURR_WR_ENA|CO_FL_DATA_WR_ENA)) == CO_FL_DATA_WR_ENA)) {
Willy Tarreaub5e2cbd2012-08-17 11:55:04 +0200175 fd_want_send(c->t.sock.fd);
Willy Tarreauc8dd77f2012-11-05 17:52:26 +0100176 f |= CO_FL_CURR_WR_ENA;
177 }
178 else if (unlikely((f & (CO_FL_CURR_WR_ENA|CO_FL_DATA_WR_ENA)) == CO_FL_CURR_WR_ENA)) {
179 fd_stop_send(c->t.sock.fd);
180 f &= ~CO_FL_CURR_WR_ENA;
Willy Tarreaue9dfa792012-09-01 17:26:16 +0200181 }
Willy Tarreau310987a2014-01-22 19:46:33 +0100182 c->flags = f;
Willy Tarreaue9dfa792012-09-01 17:26:16 +0200183}
184
185/* Update polling on connection <c>'s file descriptor depending on its current
186 * state as reported in the connection's CO_FL_CURR_* flags, reports of EAGAIN
187 * in CO_FL_WAIT_*, and the sock layer expectations indicated by CO_FL_SOCK_*.
188 * The connection flags are updated with the new flags at the end of the
Willy Tarreau0ffde2c2012-10-04 22:21:15 +0200189 * operation. Polling is totally disabled if an error was reported.
Willy Tarreaue9dfa792012-09-01 17:26:16 +0200190 */
191void conn_update_sock_polling(struct connection *c)
192{
193 unsigned int f = c->flags;
Willy Tarreaub5e2cbd2012-08-17 11:55:04 +0200194
Willy Tarreau3c728722014-01-23 13:50:42 +0100195 if (!conn_ctrl_ready(c))
Willy Tarreauf79c8172013-10-21 16:30:56 +0200196 return;
197
Willy Tarreaue9dfa792012-09-01 17:26:16 +0200198 /* update read status if needed */
Willy Tarreau310987a2014-01-22 19:46:33 +0100199 if (unlikely((f & (CO_FL_CURR_RD_ENA|CO_FL_SOCK_RD_ENA)) == CO_FL_SOCK_RD_ENA)) {
Willy Tarreauf817e9f2014-01-10 16:58:45 +0100200 fd_want_recv(c->t.sock.fd);
Willy Tarreauc8dd77f2012-11-05 17:52:26 +0100201 f |= CO_FL_CURR_RD_ENA;
Willy Tarreaue9dfa792012-09-01 17:26:16 +0200202 }
Willy Tarreauc8dd77f2012-11-05 17:52:26 +0100203 else if (unlikely((f & (CO_FL_CURR_RD_ENA|CO_FL_SOCK_RD_ENA)) == CO_FL_CURR_RD_ENA)) {
204 fd_stop_recv(c->t.sock.fd);
205 f &= ~CO_FL_CURR_RD_ENA;
Willy Tarreaue9dfa792012-09-01 17:26:16 +0200206 }
207
208 /* update write status if needed */
Willy Tarreau310987a2014-01-22 19:46:33 +0100209 if (unlikely((f & (CO_FL_CURR_WR_ENA|CO_FL_SOCK_WR_ENA)) == CO_FL_SOCK_WR_ENA)) {
Willy Tarreauf817e9f2014-01-10 16:58:45 +0100210 fd_want_send(c->t.sock.fd);
Willy Tarreauc8dd77f2012-11-05 17:52:26 +0100211 f |= CO_FL_CURR_WR_ENA;
Willy Tarreaue9dfa792012-09-01 17:26:16 +0200212 }
Willy Tarreauc8dd77f2012-11-05 17:52:26 +0100213 else if (unlikely((f & (CO_FL_CURR_WR_ENA|CO_FL_SOCK_WR_ENA)) == CO_FL_CURR_WR_ENA)) {
214 fd_stop_send(c->t.sock.fd);
215 f &= ~CO_FL_CURR_WR_ENA;
Willy Tarreaue9dfa792012-09-01 17:26:16 +0200216 }
Willy Tarreau310987a2014-01-22 19:46:33 +0100217 c->flags = f;
Willy Tarreaub5e2cbd2012-08-17 11:55:04 +0200218}
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200219
220/* This handshake handler waits a PROXY protocol header at the beginning of the
221 * raw data stream. The header looks like this :
222 *
223 * "PROXY" <SP> PROTO <SP> SRC3 <SP> DST3 <SP> SRC4 <SP> <DST4> "\r\n"
224 *
225 * There must be exactly one space between each field. Fields are :
226 * - PROTO : layer 4 protocol, which must be "TCP4" or "TCP6".
227 * - SRC3 : layer 3 (eg: IP) source address in standard text form
228 * - DST3 : layer 3 (eg: IP) destination address in standard text form
229 * - SRC4 : layer 4 (eg: TCP port) source address in standard text form
230 * - DST4 : layer 4 (eg: TCP port) destination address in standard text form
231 *
232 * This line MUST be at the beginning of the buffer and MUST NOT wrap.
233 *
234 * The header line is small and in all cases smaller than the smallest normal
235 * TCP MSS. So it MUST always be delivered as one segment, which ensures we
236 * can safely use MSG_PEEK and avoid buffering.
237 *
238 * Once the data is fetched, the values are set in the connection's address
239 * fields, and data are removed from the socket's buffer. The function returns
240 * zero if it needs to wait for more data or if it fails, or 1 if it completed
241 * and removed itself.
242 */
243int conn_recv_proxy(struct connection *conn, int flag)
244{
245 char *line, *end;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200246
247 /* we might have been called just after an asynchronous shutr */
248 if (conn->flags & CO_FL_SOCK_RD_SH)
249 goto fail;
250
Willy Tarreau3c728722014-01-23 13:50:42 +0100251 if (!conn_ctrl_ready(conn))
Willy Tarreauf79c8172013-10-21 16:30:56 +0200252 goto fail;
253
Willy Tarreaufd803bb2014-01-20 15:13:07 +0100254 if (!fd_recv_ready(conn->t.sock.fd))
255 return 0;
256
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200257 do {
Willy Tarreau19d14ef2012-10-29 16:51:55 +0100258 trash.len = recv(conn->t.sock.fd, trash.str, trash.size, MSG_PEEK);
259 if (trash.len < 0) {
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200260 if (errno == EINTR)
261 continue;
262 if (errno == EAGAIN) {
Willy Tarreaue1f50c42014-01-22 20:02:06 +0100263 fd_cant_recv(conn->t.sock.fd);
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200264 return 0;
265 }
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100266 goto recv_abort;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200267 }
268 } while (0);
269
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100270 if (!trash.len) {
271 /* client shutdown */
272 conn->err_code = CO_ER_PRX_EMPTY;
273 goto fail;
274 }
275
Willy Tarreau19d14ef2012-10-29 16:51:55 +0100276 if (trash.len < 6)
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200277 goto missing;
278
Willy Tarreau19d14ef2012-10-29 16:51:55 +0100279 line = trash.str;
280 end = trash.str + trash.len;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200281
282 /* Decode a possible proxy request, fail early if it does not match */
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100283 if (strncmp(line, "PROXY ", 6) != 0) {
284 conn->err_code = CO_ER_PRX_NOT_HDR;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200285 goto fail;
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100286 }
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200287
288 line += 6;
Willy Tarreau19d14ef2012-10-29 16:51:55 +0100289 if (trash.len < 18) /* shortest possible line */
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200290 goto missing;
291
292 if (!memcmp(line, "TCP4 ", 5) != 0) {
293 u32 src3, dst3, sport, dport;
294
295 line += 5;
296
297 src3 = inetaddr_host_lim_ret(line, end, &line);
298 if (line == end)
299 goto missing;
300 if (*line++ != ' ')
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100301 goto bad_header;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200302
303 dst3 = inetaddr_host_lim_ret(line, end, &line);
304 if (line == end)
305 goto missing;
306 if (*line++ != ' ')
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100307 goto bad_header;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200308
309 sport = read_uint((const char **)&line, end);
310 if (line == end)
311 goto missing;
312 if (*line++ != ' ')
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100313 goto bad_header;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200314
315 dport = read_uint((const char **)&line, end);
316 if (line > end - 2)
317 goto missing;
318 if (*line++ != '\r')
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100319 goto bad_header;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200320 if (*line++ != '\n')
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100321 goto bad_header;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200322
323 /* update the session's addresses and mark them set */
324 ((struct sockaddr_in *)&conn->addr.from)->sin_family = AF_INET;
325 ((struct sockaddr_in *)&conn->addr.from)->sin_addr.s_addr = htonl(src3);
326 ((struct sockaddr_in *)&conn->addr.from)->sin_port = htons(sport);
327
328 ((struct sockaddr_in *)&conn->addr.to)->sin_family = AF_INET;
329 ((struct sockaddr_in *)&conn->addr.to)->sin_addr.s_addr = htonl(dst3);
330 ((struct sockaddr_in *)&conn->addr.to)->sin_port = htons(dport);
331 conn->flags |= CO_FL_ADDR_FROM_SET | CO_FL_ADDR_TO_SET;
332 }
333 else if (!memcmp(line, "TCP6 ", 5) != 0) {
334 u32 sport, dport;
335 char *src_s;
336 char *dst_s, *sport_s, *dport_s;
337 struct in6_addr src3, dst3;
338
339 line += 5;
340
341 src_s = line;
342 dst_s = sport_s = dport_s = NULL;
343 while (1) {
344 if (line > end - 2) {
345 goto missing;
346 }
347 else if (*line == '\r') {
348 *line = 0;
349 line++;
350 if (*line++ != '\n')
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100351 goto bad_header;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200352 break;
353 }
354
355 if (*line == ' ') {
356 *line = 0;
357 if (!dst_s)
358 dst_s = line + 1;
359 else if (!sport_s)
360 sport_s = line + 1;
361 else if (!dport_s)
362 dport_s = line + 1;
363 }
364 line++;
365 }
366
367 if (!dst_s || !sport_s || !dport_s)
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100368 goto bad_header;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200369
370 sport = read_uint((const char **)&sport_s,dport_s - 1);
371 if (*sport_s != 0)
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100372 goto bad_header;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200373
374 dport = read_uint((const char **)&dport_s,line - 2);
375 if (*dport_s != 0)
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100376 goto bad_header;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200377
378 if (inet_pton(AF_INET6, src_s, (void *)&src3) != 1)
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100379 goto bad_header;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200380
381 if (inet_pton(AF_INET6, dst_s, (void *)&dst3) != 1)
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100382 goto bad_header;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200383
384 /* update the session's addresses and mark them set */
385 ((struct sockaddr_in6 *)&conn->addr.from)->sin6_family = AF_INET6;
386 memcpy(&((struct sockaddr_in6 *)&conn->addr.from)->sin6_addr, &src3, sizeof(struct in6_addr));
387 ((struct sockaddr_in6 *)&conn->addr.from)->sin6_port = htons(sport);
388
389 ((struct sockaddr_in6 *)&conn->addr.to)->sin6_family = AF_INET6;
390 memcpy(&((struct sockaddr_in6 *)&conn->addr.to)->sin6_addr, &dst3, sizeof(struct in6_addr));
391 ((struct sockaddr_in6 *)&conn->addr.to)->sin6_port = htons(dport);
392 conn->flags |= CO_FL_ADDR_FROM_SET | CO_FL_ADDR_TO_SET;
393 }
394 else {
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100395 /* The protocol does not match something known (TCP4/TCP6) */
396 conn->err_code = CO_ER_PRX_BAD_PROTO;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200397 goto fail;
398 }
399
400 /* remove the PROXY line from the request. For this we re-read the
401 * exact line at once. If we don't get the exact same result, we
402 * fail.
403 */
Willy Tarreau19d14ef2012-10-29 16:51:55 +0100404 trash.len = line - trash.str;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200405 do {
Willy Tarreau19d14ef2012-10-29 16:51:55 +0100406 int len2 = recv(conn->t.sock.fd, trash.str, trash.len, 0);
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200407 if (len2 < 0 && errno == EINTR)
408 continue;
Willy Tarreau19d14ef2012-10-29 16:51:55 +0100409 if (len2 != trash.len)
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100410 goto recv_abort;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200411 } while (0);
412
413 conn->flags &= ~flag;
414 return 1;
415
416 missing:
417 /* Missing data. Since we're using MSG_PEEK, we can only poll again if
418 * we have not read anything. Otherwise we need to fail because we won't
419 * be able to poll anymore.
420 */
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100421 conn->err_code = CO_ER_PRX_TRUNCATED;
422 goto fail;
423
424 bad_header:
425 /* This is not a valid proxy protocol header */
426 conn->err_code = CO_ER_PRX_BAD_HDR;
427 goto fail;
428
429 recv_abort:
430 conn->err_code = CO_ER_PRX_ABORT;
Willy Tarreau26f4a042013-12-04 23:44:10 +0100431 conn->flags |= CO_FL_SOCK_RD_SH | CO_FL_SOCK_WR_SH;
Willy Tarreau8e3bf692012-12-03 15:41:18 +0100432 goto fail;
433
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200434 fail:
Willy Tarreaud486ef52012-12-10 17:03:52 +0100435 __conn_sock_stop_both(conn);
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200436 conn->flags |= CO_FL_ERROR;
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200437 return 0;
438}
439
David Safb76832014-05-08 23:42:08 -0400440int make_proxy_line(char *buf, int buf_len, struct server *srv, struct connection *remote)
441{
442 int ret = 0;
443
444 if (srv && (srv->pp_opts & SRV_PP_V2)) {
445 ret = make_proxy_line_v2(buf, buf_len, srv, remote);
446 }
447 else {
448 if (remote)
449 ret = make_proxy_line_v1(buf, buf_len, &remote->addr.from, &remote->addr.to);
450 else
451 ret = make_proxy_line_v1(buf, buf_len, NULL, NULL);
452 }
453
454 return ret;
455}
456
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200457/* Makes a PROXY protocol line from the two addresses. The output is sent to
458 * buffer <buf> for a maximum size of <buf_len> (including the trailing zero).
459 * It returns the number of bytes composing this line (including the trailing
460 * LF), or zero in case of failure (eg: not enough space). It supports TCP4,
Willy Tarreau2e1401a2013-10-01 11:41:55 +0200461 * TCP6 and "UNKNOWN" formats. If any of <src> or <dst> is null, UNKNOWN is
462 * emitted as well.
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200463 */
David Safb76832014-05-08 23:42:08 -0400464int make_proxy_line_v1(char *buf, int buf_len, struct sockaddr_storage *src, struct sockaddr_storage *dst)
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200465{
466 int ret = 0;
467
Willy Tarreau2e1401a2013-10-01 11:41:55 +0200468 if (src && dst && src->ss_family == dst->ss_family && src->ss_family == AF_INET) {
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200469 ret = snprintf(buf + ret, buf_len - ret, "PROXY TCP4 ");
470 if (ret >= buf_len)
471 return 0;
472
473 /* IPv4 src */
474 if (!inet_ntop(src->ss_family, &((struct sockaddr_in *)src)->sin_addr, buf + ret, buf_len - ret))
475 return 0;
476
477 ret += strlen(buf + ret);
478 if (ret >= buf_len)
479 return 0;
480
481 buf[ret++] = ' ';
482
483 /* IPv4 dst */
484 if (!inet_ntop(dst->ss_family, &((struct sockaddr_in *)dst)->sin_addr, buf + ret, buf_len - ret))
485 return 0;
486
487 ret += strlen(buf + ret);
488 if (ret >= buf_len)
489 return 0;
490
491 /* source and destination ports */
492 ret += snprintf(buf + ret, buf_len - ret, " %u %u\r\n",
493 ntohs(((struct sockaddr_in *)src)->sin_port),
494 ntohs(((struct sockaddr_in *)dst)->sin_port));
495 if (ret >= buf_len)
496 return 0;
497 }
Willy Tarreau2e1401a2013-10-01 11:41:55 +0200498 else if (src && dst && src->ss_family == dst->ss_family && src->ss_family == AF_INET6) {
Willy Tarreaue1e4a612012-10-05 00:10:55 +0200499 ret = snprintf(buf + ret, buf_len - ret, "PROXY TCP6 ");
500 if (ret >= buf_len)
501 return 0;
502
503 /* IPv6 src */
504 if (!inet_ntop(src->ss_family, &((struct sockaddr_in6 *)src)->sin6_addr, buf + ret, buf_len - ret))
505 return 0;
506
507 ret += strlen(buf + ret);
508 if (ret >= buf_len)
509 return 0;
510
511 buf[ret++] = ' ';
512
513 /* IPv6 dst */
514 if (!inet_ntop(dst->ss_family, &((struct sockaddr_in6 *)dst)->sin6_addr, buf + ret, buf_len - ret))
515 return 0;
516
517 ret += strlen(buf + ret);
518 if (ret >= buf_len)
519 return 0;
520
521 /* source and destination ports */
522 ret += snprintf(buf + ret, buf_len - ret, " %u %u\r\n",
523 ntohs(((struct sockaddr_in6 *)src)->sin6_port),
524 ntohs(((struct sockaddr_in6 *)dst)->sin6_port));
525 if (ret >= buf_len)
526 return 0;
527 }
528 else {
529 /* unknown family combination */
530 ret = snprintf(buf, buf_len, "PROXY UNKNOWN\r\n");
531 if (ret >= buf_len)
532 return 0;
533 }
534 return ret;
535}
David Safb76832014-05-08 23:42:08 -0400536
537#ifdef USE_OPENSSL
538static int make_tlv(char *dest, int dest_len, char type, uint16_t length, char *value)
539{
540 struct tlv *tlv;
541
542 if (!dest || (length + sizeof(*tlv) > dest_len))
543 return 0;
544
545 tlv = (struct tlv *)dest;
546
547 tlv->type = type;
548 tlv->length_hi = length >> 8;
549 tlv->length_lo = length & 0x00ff;
550 memcpy(tlv->value, value, length);
551 return length + sizeof(*tlv);
552}
553#endif
554
555int make_proxy_line_v2(char *buf, int buf_len, struct server *srv, struct connection *remote)
556{
557 const char pp2_signature[12] = {0x0D, 0x0A, 0x0D, 0x0A, 0x00, 0x0D, 0x0A, 0x51, 0x55, 0x49, 0x54, 0x0A};
558 int ret = 0;
559 struct proxy_hdr_v2 *hdr_p = (struct proxy_hdr_v2 *)buf;
560 union proxy_addr *addr_p = (union proxy_addr *)(buf + PP2_HEADER_LEN);
561 struct sockaddr_storage null_addr = {0};
562 struct sockaddr_storage *src = &null_addr;
563 struct sockaddr_storage *dst = &null_addr;
564#ifdef USE_OPENSSL
565 int tlv_len = 0;
566 char *value = NULL;
567 struct tlv_ssl *tlv;
568 int ssl_tlv_len = 0;
569#endif
570
571 if (buf_len < PP2_HEADER_LEN)
572 return 0;
573 memcpy(hdr_p->sig, pp2_signature, PP2_SIGNATURE_LEN);
574
575 if (remote) {
576 src = &remote->addr.from;
577 dst = &remote->addr.to;
578 }
579 if (src && dst && src->ss_family == dst->ss_family && src->ss_family == AF_INET) {
580 if (buf_len < PP2_HDR_LEN_INET)
581 return 0;
582 hdr_p->cmd = PP2_VERSION | PP2_CMD_PROXY;
583 hdr_p->fam = PP2_FAM_INET | PP2_TRANS_STREAM;
584 addr_p->ipv4_addr.src_addr = ((struct sockaddr_in *)src)->sin_addr.s_addr;
585 addr_p->ipv4_addr.dst_addr = ((struct sockaddr_in *)dst)->sin_addr.s_addr;
586 addr_p->ipv4_addr.src_port = ((struct sockaddr_in *)src)->sin_port;
587 addr_p->ipv4_addr.dst_port = ((struct sockaddr_in *)dst)->sin_port;
588 ret = PP2_HDR_LEN_INET;
589 }
590 else if (src && dst && src->ss_family == dst->ss_family && src->ss_family == AF_INET6) {
591 if (buf_len < PP2_HDR_LEN_INET6)
592 return 0;
593 hdr_p->cmd = PP2_VERSION | PP2_CMD_PROXY;
594 hdr_p->fam = PP2_FAM_INET6 | PP2_TRANS_STREAM;
595 memcpy(addr_p->ipv6_addr.src_addr, &((struct sockaddr_in6 *)src)->sin6_addr, 16);
596 memcpy(addr_p->ipv6_addr.dst_addr, &((struct sockaddr_in6 *)dst)->sin6_addr, 16);
597 addr_p->ipv6_addr.src_port = ((struct sockaddr_in6 *)src)->sin6_port;
598 addr_p->ipv6_addr.dst_port = ((struct sockaddr_in6 *)dst)->sin6_port;
599 ret = PP2_HDR_LEN_INET6;
600 }
601 else {
602 if (buf_len < PP2_HDR_LEN_UNSPEC)
603 return 0;
604 hdr_p->cmd = PP2_VERSION | PP2_CMD_LOCAL;
605 hdr_p->fam = PP2_FAM_UNSPEC | PP2_TRANS_UNSPEC;
606 ret = PP2_HDR_LEN_UNSPEC;
607 }
608
609#ifdef USE_OPENSSL
610 if (srv->pp_opts & SRV_PP_V2_SSL) {
611 if ((buf_len - ret) < sizeof(struct tlv_ssl))
612 return 0;
613 tlv = (struct tlv_ssl *)&buf[ret];
614 memset(tlv, 0, sizeof(struct tlv_ssl));
615 ssl_tlv_len += sizeof(struct tlv_ssl);
616 tlv->tlv.type = PP2_TYPE_SSL;
617 if (ssl_sock_is_ssl(remote)) {
618 tlv->client |= PP2_CLIENT_SSL;
619 value = ssl_sock_get_version(remote);
620 if (value) {
621 tlv_len = make_tlv(&buf[ret+ssl_tlv_len], (buf_len-ret-ssl_tlv_len), PP2_TYPE_SSL_VERSION, strlen(value), value);
622 ssl_tlv_len += tlv_len;
623 }
624 if (ssl_sock_get_cert_used(remote)) {
625 tlv->client |= PP2_CLIENT_CERT;
626 tlv->verify = htonl(ssl_sock_get_verify_result(remote));
627 }
628 if (srv->pp_opts & SRV_PP_V2_SSL_CN) {
629 value = ssl_sock_get_common_name(remote);
630 if (value) {
631 tlv_len = make_tlv(&buf[ret+ssl_tlv_len], (buf_len - ret - ssl_tlv_len), PP2_TYPE_SSL_CN, strlen(value), value);
632 ssl_tlv_len += tlv_len;
633 }
634 }
635 }
636 tlv->tlv.length_hi = (uint16_t)(ssl_tlv_len - sizeof(struct tlv)) >> 8;
637 tlv->tlv.length_lo = (uint16_t)(ssl_tlv_len - sizeof(struct tlv)) & 0x00ff;
638 ret += ssl_tlv_len;
639 }
640#endif
641
642 hdr_p->len = htons((uint16_t)(ret - PP2_HEADER_LEN));
643
644 return ret;
645}