Willy Tarreau | 59f9839 | 2012-07-06 14:13:49 +0200 | [diff] [blame] | 1 | /* |
| 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 Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 13 | #include <errno.h> |
| 14 | |
Willy Tarreau | 59f9839 | 2012-07-06 14:13:49 +0200 | [diff] [blame] | 15 | #include <common/compat.h> |
| 16 | #include <common/config.h> |
KOVACS Krisztian | b3e54fe | 2014-11-17 15:11:45 +0100 | [diff] [blame] | 17 | #include <common/namespace.h> |
Willy Tarreau | 59f9839 | 2012-07-06 14:13:49 +0200 | [diff] [blame] | 18 | |
Willy Tarreau | c578891 | 2012-08-24 18:12:41 +0200 | [diff] [blame] | 19 | #include <proto/connection.h> |
Willy Tarreau | dd2f85e | 2012-09-02 22:34:23 +0200 | [diff] [blame] | 20 | #include <proto/fd.h> |
Willy Tarreau | 5f1504f | 2012-10-04 23:55:57 +0200 | [diff] [blame] | 21 | #include <proto/frontend.h> |
Willy Tarreau | 2da156f | 2012-07-23 15:07:23 +0200 | [diff] [blame] | 22 | #include <proto/proto_tcp.h> |
Willy Tarreau | 2c6be84 | 2012-07-06 17:12:34 +0200 | [diff] [blame] | 23 | #include <proto/stream_interface.h> |
Emeric Brun | 4f60301 | 2017-01-05 15:11:44 +0100 | [diff] [blame] | 24 | #include <proto/sample.h> |
Willy Tarreau | 59f9839 | 2012-07-06 14:13:49 +0200 | [diff] [blame] | 25 | |
Emeric Brun | 4659195 | 2012-05-18 15:47:34 +0200 | [diff] [blame] | 26 | #ifdef USE_OPENSSL |
| 27 | #include <proto/ssl_sock.h> |
| 28 | #endif |
| 29 | |
Willy Tarreau | f2943dc | 2012-10-26 20:10:28 +0200 | [diff] [blame] | 30 | struct pool_head *pool2_connection; |
Willy Tarreau | 13e1410 | 2016-12-22 20:25:26 +0100 | [diff] [blame] | 31 | struct xprt_ops *registered_xprt[XPRT_ENTRIES] = { NULL, }; |
Willy Tarreau | f2943dc | 2012-10-26 20:10:28 +0200 | [diff] [blame] | 32 | |
Willy Tarreau | 2386be6 | 2017-09-21 19:40:52 +0200 | [diff] [blame^] | 33 | /* List head of all known muxes for ALPN */ |
| 34 | struct alpn_mux_list alpn_mux_list = { |
| 35 | .list = LIST_HEAD_INIT(alpn_mux_list.list) |
| 36 | }; |
| 37 | |
Willy Tarreau | f2943dc | 2012-10-26 20:10:28 +0200 | [diff] [blame] | 38 | /* perform minimal intializations, report 0 in case of error, 1 if OK. */ |
| 39 | int init_connection() |
| 40 | { |
| 41 | pool2_connection = create_pool("connection", sizeof (struct connection), MEM_F_SHARED); |
| 42 | return pool2_connection != NULL; |
| 43 | } |
| 44 | |
Willy Tarreau | 59f9839 | 2012-07-06 14:13:49 +0200 | [diff] [blame] | 45 | /* I/O callback for fd-based connections. It calls the read/write handlers |
Willy Tarreau | 7a798e5 | 2016-04-14 11:13:20 +0200 | [diff] [blame] | 46 | * provided by the connection's sock_ops, which must be valid. |
Willy Tarreau | 59f9839 | 2012-07-06 14:13:49 +0200 | [diff] [blame] | 47 | */ |
Willy Tarreau | 7a798e5 | 2016-04-14 11:13:20 +0200 | [diff] [blame] | 48 | void conn_fd_handler(int fd) |
Willy Tarreau | 59f9839 | 2012-07-06 14:13:49 +0200 | [diff] [blame] | 49 | { |
Willy Tarreau | 8018471 | 2012-07-06 14:54:49 +0200 | [diff] [blame] | 50 | struct connection *conn = fdtab[fd].owner; |
Willy Tarreau | 9e272bf | 2012-10-03 21:04:48 +0200 | [diff] [blame] | 51 | unsigned int flags; |
Willy Tarreau | 59f9839 | 2012-07-06 14:13:49 +0200 | [diff] [blame] | 52 | |
Willy Tarreau | c76ae33 | 2012-07-12 15:32:13 +0200 | [diff] [blame] | 53 | if (unlikely(!conn)) |
Willy Tarreau | 7a798e5 | 2016-04-14 11:13:20 +0200 | [diff] [blame] | 54 | return; |
Willy Tarreau | 59f9839 | 2012-07-06 14:13:49 +0200 | [diff] [blame] | 55 | |
Willy Tarreau | 7d28149 | 2012-12-16 19:19:13 +0100 | [diff] [blame] | 56 | conn_refresh_polling_flags(conn); |
Willy Tarreau | 916e12d | 2017-10-25 09:22:43 +0200 | [diff] [blame] | 57 | conn->flags |= CO_FL_WILL_UPDATE; |
| 58 | |
Willy Tarreau | 7d28149 | 2012-12-16 19:19:13 +0100 | [diff] [blame] | 59 | flags = conn->flags & ~CO_FL_ERROR; /* ensure to call the wake handler upon error */ |
Willy Tarreau | d29a066 | 2012-12-10 16:33:38 +0100 | [diff] [blame] | 60 | |
Willy Tarreau | c76ae33 | 2012-07-12 15:32:13 +0200 | [diff] [blame] | 61 | process_handshake: |
Willy Tarreau | f9dabec | 2012-08-17 17:33:53 +0200 | [diff] [blame] | 62 | /* The handshake callbacks are called in sequence. If either of them is |
| 63 | * missing something, it must enable the required polling at the socket |
| 64 | * layer of the connection. Polling state is not guaranteed when entering |
| 65 | * these handlers, so any handshake handler which does not complete its |
Willy Tarreau | d6e999b | 2013-11-25 08:41:15 +0100 | [diff] [blame] | 66 | * work must explicitly disable events it's not interested in. Error |
| 67 | * handling is also performed here in order to reduce the number of tests |
| 68 | * around. |
Willy Tarreau | f9dabec | 2012-08-17 17:33:53 +0200 | [diff] [blame] | 69 | */ |
Willy Tarreau | d6e999b | 2013-11-25 08:41:15 +0100 | [diff] [blame] | 70 | while (unlikely(conn->flags & (CO_FL_HANDSHAKE | CO_FL_ERROR))) { |
Willy Tarreau | 310987a | 2014-01-22 19:46:33 +0100 | [diff] [blame] | 71 | if (unlikely(conn->flags & CO_FL_ERROR)) |
Willy Tarreau | 2c6be84 | 2012-07-06 17:12:34 +0200 | [diff] [blame] | 72 | goto leave; |
Willy Tarreau | 59f9839 | 2012-07-06 14:13:49 +0200 | [diff] [blame] | 73 | |
Bertrand Jacquin | 93b227d | 2016-06-04 15:11:10 +0100 | [diff] [blame] | 74 | if (conn->flags & CO_FL_ACCEPT_CIP) |
| 75 | if (!conn_recv_netscaler_cip(conn, CO_FL_ACCEPT_CIP)) |
| 76 | goto leave; |
| 77 | |
Willy Tarreau | 22cda21 | 2012-08-31 17:43:29 +0200 | [diff] [blame] | 78 | if (conn->flags & CO_FL_ACCEPT_PROXY) |
| 79 | if (!conn_recv_proxy(conn, CO_FL_ACCEPT_PROXY)) |
| 80 | goto leave; |
| 81 | |
Willy Tarreau | 57cd3e4 | 2013-10-24 22:01:26 +0200 | [diff] [blame] | 82 | if (conn->flags & CO_FL_SEND_PROXY) |
| 83 | if (!conn_si_send_proxy(conn, CO_FL_SEND_PROXY)) |
Willy Tarreau | 5f1504f | 2012-10-04 23:55:57 +0200 | [diff] [blame] | 84 | goto leave; |
Emeric Brun | 4659195 | 2012-05-18 15:47:34 +0200 | [diff] [blame] | 85 | #ifdef USE_OPENSSL |
| 86 | if (conn->flags & CO_FL_SSL_WAIT_HS) |
| 87 | if (!ssl_sock_handshake(conn, CO_FL_SSL_WAIT_HS)) |
| 88 | goto leave; |
| 89 | #endif |
Willy Tarreau | c76ae33 | 2012-07-12 15:32:13 +0200 | [diff] [blame] | 90 | } |
| 91 | |
Willy Tarreau | f9dabec | 2012-08-17 17:33:53 +0200 | [diff] [blame] | 92 | /* Once we're purely in the data phase, we disable handshake polling */ |
| 93 | if (!(conn->flags & CO_FL_POLL_SOCK)) |
| 94 | __conn_sock_stop_both(conn); |
Willy Tarreau | c76ae33 | 2012-07-12 15:32:13 +0200 | [diff] [blame] | 95 | |
Willy Tarreau | 8e3c6ce | 2017-08-28 15:46:01 +0200 | [diff] [blame] | 96 | /* The connection owner might want to be notified about an end of |
| 97 | * handshake indicating the connection is ready, before we proceed with |
| 98 | * any data exchange. The callback may fail and cause the connection to |
| 99 | * be destroyed, thus we must not use it anymore and should immediately |
| 100 | * leave instead. The caller must immediately unregister itself once |
| 101 | * called. |
Willy Tarreau | 2542b53 | 2012-08-31 16:01:23 +0200 | [diff] [blame] | 102 | */ |
Willy Tarreau | 8e3c6ce | 2017-08-28 15:46:01 +0200 | [diff] [blame] | 103 | if (conn->xprt_done_cb && conn->xprt_done_cb(conn) < 0) |
Willy Tarreau | 7a798e5 | 2016-04-14 11:13:20 +0200 | [diff] [blame] | 104 | return; |
Willy Tarreau | 2542b53 | 2012-08-31 16:01:23 +0200 | [diff] [blame] | 105 | |
Willy Tarreau | 57ec32f | 2017-04-11 19:59:33 +0200 | [diff] [blame] | 106 | if (conn->xprt && fd_send_ready(fd) && |
Olivier Houchard | 1a0545f | 2017-09-13 18:30:23 +0200 | [diff] [blame] | 107 | ((conn->flags & (CO_FL_XPRT_WR_ENA|CO_FL_ERROR|CO_FL_HANDSHAKE)) == CO_FL_XPRT_WR_ENA)) { |
Willy Tarreau | 3c0cc49 | 2017-03-19 07:54:28 +0100 | [diff] [blame] | 108 | /* force reporting of activity by clearing the previous flags : |
| 109 | * we'll have at least ERROR or CONNECTED at the end of an I/O, |
| 110 | * both of which will be detected below. |
Willy Tarreau | 9e272bf | 2012-10-03 21:04:48 +0200 | [diff] [blame] | 111 | */ |
Willy Tarreau | 3c0cc49 | 2017-03-19 07:54:28 +0100 | [diff] [blame] | 112 | flags = 0; |
Willy Tarreau | 53a4766 | 2017-08-28 10:53:00 +0200 | [diff] [blame] | 113 | conn->mux->send(conn); |
Willy Tarreau | 9e272bf | 2012-10-03 21:04:48 +0200 | [diff] [blame] | 114 | } |
Willy Tarreau | 59f9839 | 2012-07-06 14:13:49 +0200 | [diff] [blame] | 115 | |
Willy Tarreau | 57ec32f | 2017-04-11 19:59:33 +0200 | [diff] [blame] | 116 | /* The data transfer starts here and stops on error and handshakes. Note |
| 117 | * that we must absolutely test conn->xprt at each step in case it suddenly |
| 118 | * changes due to a quick unexpected close(). |
| 119 | */ |
| 120 | if (conn->xprt && fd_recv_ready(fd) && |
Olivier Houchard | 1a0545f | 2017-09-13 18:30:23 +0200 | [diff] [blame] | 121 | ((conn->flags & (CO_FL_XPRT_RD_ENA|CO_FL_WAIT_ROOM|CO_FL_ERROR|CO_FL_HANDSHAKE)) == CO_FL_XPRT_RD_ENA)) { |
Willy Tarreau | 3c0cc49 | 2017-03-19 07:54:28 +0100 | [diff] [blame] | 122 | /* force reporting of activity by clearing the previous flags : |
| 123 | * we'll have at least ERROR or CONNECTED at the end of an I/O, |
| 124 | * both of which will be detected below. |
Willy Tarreau | 9e272bf | 2012-10-03 21:04:48 +0200 | [diff] [blame] | 125 | */ |
Willy Tarreau | 3c0cc49 | 2017-03-19 07:54:28 +0100 | [diff] [blame] | 126 | flags = 0; |
Willy Tarreau | 53a4766 | 2017-08-28 10:53:00 +0200 | [diff] [blame] | 127 | conn->mux->recv(conn); |
Willy Tarreau | 9e272bf | 2012-10-03 21:04:48 +0200 | [diff] [blame] | 128 | } |
Willy Tarreau | 2da156f | 2012-07-23 15:07:23 +0200 | [diff] [blame] | 129 | |
Willy Tarreau | c76ae33 | 2012-07-12 15:32:13 +0200 | [diff] [blame] | 130 | /* It may happen during the data phase that a handshake is |
| 131 | * enabled again (eg: SSL) |
| 132 | */ |
Willy Tarreau | d6e999b | 2013-11-25 08:41:15 +0100 | [diff] [blame] | 133 | if (unlikely(conn->flags & (CO_FL_HANDSHAKE | CO_FL_ERROR))) |
Willy Tarreau | c76ae33 | 2012-07-12 15:32:13 +0200 | [diff] [blame] | 134 | goto process_handshake; |
| 135 | |
Willy Tarreau | fd803bb | 2014-01-20 15:13:07 +0100 | [diff] [blame] | 136 | if (unlikely(conn->flags & CO_FL_WAIT_L4_CONN)) { |
Willy Tarreau | f8deb0c | 2012-09-01 17:59:22 +0200 | [diff] [blame] | 137 | /* still waiting for a connection to establish and nothing was |
| 138 | * attempted yet to probe the connection. Then let's retry the |
| 139 | * connect(). |
Willy Tarreau | 2da156f | 2012-07-23 15:07:23 +0200 | [diff] [blame] | 140 | */ |
Willy Tarreau | 239d718 | 2012-07-23 18:53:03 +0200 | [diff] [blame] | 141 | if (!tcp_connect_probe(conn)) |
Willy Tarreau | afad0e0 | 2012-08-09 14:45:22 +0200 | [diff] [blame] | 142 | goto leave; |
Willy Tarreau | 2da156f | 2012-07-23 15:07:23 +0200 | [diff] [blame] | 143 | } |
Willy Tarreau | 2c6be84 | 2012-07-06 17:12:34 +0200 | [diff] [blame] | 144 | leave: |
Willy Tarreau | 3c0cc49 | 2017-03-19 07:54:28 +0100 | [diff] [blame] | 145 | /* Verify if the connection just established. */ |
Willy Tarreau | 7bf3fa3 | 2017-03-14 20:19:29 +0100 | [diff] [blame] | 146 | if (unlikely(!(conn->flags & (CO_FL_WAIT_L4_CONN | CO_FL_WAIT_L6_CONN | CO_FL_CONNECTED)))) |
| 147 | conn->flags |= CO_FL_CONNECTED; |
| 148 | |
Willy Tarreau | 8e3c6ce | 2017-08-28 15:46:01 +0200 | [diff] [blame] | 149 | /* The connection owner might want to be notified about failures to |
| 150 | * complete the handshake. The callback may fail and cause the |
| 151 | * connection to be destroyed, thus we must not use it anymore and |
| 152 | * should immediately leave instead. The caller must immediately |
| 153 | * unregister itself once called. |
| 154 | */ |
| 155 | if (((conn->flags ^ flags) & CO_FL_NOTIFY_DONE) && |
| 156 | conn->xprt_done_cb && conn->xprt_done_cb(conn) < 0) |
| 157 | return; |
| 158 | |
Willy Tarreau | 3c0cc49 | 2017-03-19 07:54:28 +0100 | [diff] [blame] | 159 | /* The wake callback is normally used to notify the data layer about |
| 160 | * data layer activity (successful send/recv), connection establishment, |
| 161 | * shutdown and fatal errors. We need to consider the following |
| 162 | * situations to wake up the data layer : |
| 163 | * - change among the CO_FL_NOTIFY_DATA flags : |
| 164 | * {DATA,SOCK}_{RD,WR}_SH, ERROR, |
| 165 | * - absence of any of {L4,L6}_CONN and CONNECTED, indicating the |
| 166 | * end of handshake and transition to CONNECTED |
| 167 | * - raise of CONNECTED with HANDSHAKE down |
| 168 | * - end of HANDSHAKE with CONNECTED set |
| 169 | * - regular data layer activity |
| 170 | * |
| 171 | * Note that the wake callback is allowed to release the connection and |
| 172 | * the fd (and return < 0 in this case). |
Willy Tarreau | 2396c1c | 2012-10-03 21:12:16 +0200 | [diff] [blame] | 173 | */ |
Willy Tarreau | 9fa1ee6 | 2017-03-18 15:39:57 +0100 | [diff] [blame] | 174 | if ((((conn->flags ^ flags) & CO_FL_NOTIFY_DATA) || |
Willy Tarreau | 3c0cc49 | 2017-03-19 07:54:28 +0100 | [diff] [blame] | 175 | ((flags & (CO_FL_CONNECTED|CO_FL_HANDSHAKE)) != CO_FL_CONNECTED && |
| 176 | (conn->flags & (CO_FL_CONNECTED|CO_FL_HANDSHAKE)) == CO_FL_CONNECTED)) && |
Willy Tarreau | 53a4766 | 2017-08-28 10:53:00 +0200 | [diff] [blame] | 177 | conn->mux->wake(conn) < 0) |
Willy Tarreau | 7a798e5 | 2016-04-14 11:13:20 +0200 | [diff] [blame] | 178 | return; |
Willy Tarreau | fd31e53 | 2012-07-23 18:24:25 +0200 | [diff] [blame] | 179 | |
Willy Tarreau | 61ace1b | 2012-07-23 12:14:26 +0200 | [diff] [blame] | 180 | /* remove the events before leaving */ |
Willy Tarreau | 26d7cfc | 2012-12-07 00:09:43 +0100 | [diff] [blame] | 181 | fdtab[fd].ev &= FD_POLL_STICKY; |
Willy Tarreau | f9dabec | 2012-08-17 17:33:53 +0200 | [diff] [blame] | 182 | |
| 183 | /* commit polling changes */ |
Willy Tarreau | 916e12d | 2017-10-25 09:22:43 +0200 | [diff] [blame] | 184 | conn->flags &= ~CO_FL_WILL_UPDATE; |
Willy Tarreau | f9dabec | 2012-08-17 17:33:53 +0200 | [diff] [blame] | 185 | conn_cond_update_polling(conn); |
Willy Tarreau | 7a798e5 | 2016-04-14 11:13:20 +0200 | [diff] [blame] | 186 | return; |
Willy Tarreau | 59f9839 | 2012-07-06 14:13:49 +0200 | [diff] [blame] | 187 | } |
Willy Tarreau | b5e2cbd | 2012-08-17 11:55:04 +0200 | [diff] [blame] | 188 | |
Willy Tarreau | e9dfa79 | 2012-09-01 17:26:16 +0200 | [diff] [blame] | 189 | /* Update polling on connection <c>'s file descriptor depending on its current |
| 190 | * state as reported in the connection's CO_FL_CURR_* flags, reports of EAGAIN |
Olivier Houchard | 1a0545f | 2017-09-13 18:30:23 +0200 | [diff] [blame] | 191 | * in CO_FL_WAIT_*, and the data layer expectations indicated by CO_FL_XPRT_*. |
Willy Tarreau | e9dfa79 | 2012-09-01 17:26:16 +0200 | [diff] [blame] | 192 | * The connection flags are updated with the new flags at the end of the |
Willy Tarreau | 0ffde2c | 2012-10-04 22:21:15 +0200 | [diff] [blame] | 193 | * operation. Polling is totally disabled if an error was reported. |
Willy Tarreau | b5e2cbd | 2012-08-17 11:55:04 +0200 | [diff] [blame] | 194 | */ |
Olivier Houchard | 1a0545f | 2017-09-13 18:30:23 +0200 | [diff] [blame] | 195 | void conn_update_xprt_polling(struct connection *c) |
Willy Tarreau | b5e2cbd | 2012-08-17 11:55:04 +0200 | [diff] [blame] | 196 | { |
Willy Tarreau | e9dfa79 | 2012-09-01 17:26:16 +0200 | [diff] [blame] | 197 | unsigned int f = c->flags; |
Willy Tarreau | b5e2cbd | 2012-08-17 11:55:04 +0200 | [diff] [blame] | 198 | |
Willy Tarreau | 3c72872 | 2014-01-23 13:50:42 +0100 | [diff] [blame] | 199 | if (!conn_ctrl_ready(c)) |
Willy Tarreau | f79c817 | 2013-10-21 16:30:56 +0200 | [diff] [blame] | 200 | return; |
| 201 | |
Willy Tarreau | b5e2cbd | 2012-08-17 11:55:04 +0200 | [diff] [blame] | 202 | /* update read status if needed */ |
Olivier Houchard | 1a0545f | 2017-09-13 18:30:23 +0200 | [diff] [blame] | 203 | if (unlikely((f & (CO_FL_CURR_RD_ENA|CO_FL_XPRT_RD_ENA)) == CO_FL_XPRT_RD_ENA)) { |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 204 | fd_want_recv(c->handle.fd); |
Willy Tarreau | c8dd77f | 2012-11-05 17:52:26 +0100 | [diff] [blame] | 205 | f |= CO_FL_CURR_RD_ENA; |
Willy Tarreau | e9dfa79 | 2012-09-01 17:26:16 +0200 | [diff] [blame] | 206 | } |
Olivier Houchard | 1a0545f | 2017-09-13 18:30:23 +0200 | [diff] [blame] | 207 | else if (unlikely((f & (CO_FL_CURR_RD_ENA|CO_FL_XPRT_RD_ENA)) == CO_FL_CURR_RD_ENA)) { |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 208 | fd_stop_recv(c->handle.fd); |
Willy Tarreau | c8dd77f | 2012-11-05 17:52:26 +0100 | [diff] [blame] | 209 | f &= ~CO_FL_CURR_RD_ENA; |
Willy Tarreau | e9dfa79 | 2012-09-01 17:26:16 +0200 | [diff] [blame] | 210 | } |
Willy Tarreau | b5e2cbd | 2012-08-17 11:55:04 +0200 | [diff] [blame] | 211 | |
| 212 | /* update write status if needed */ |
Olivier Houchard | 1a0545f | 2017-09-13 18:30:23 +0200 | [diff] [blame] | 213 | if (unlikely((f & (CO_FL_CURR_WR_ENA|CO_FL_XPRT_WR_ENA)) == CO_FL_XPRT_WR_ENA)) { |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 214 | fd_want_send(c->handle.fd); |
Willy Tarreau | c8dd77f | 2012-11-05 17:52:26 +0100 | [diff] [blame] | 215 | f |= CO_FL_CURR_WR_ENA; |
| 216 | } |
Olivier Houchard | 1a0545f | 2017-09-13 18:30:23 +0200 | [diff] [blame] | 217 | else if (unlikely((f & (CO_FL_CURR_WR_ENA|CO_FL_XPRT_WR_ENA)) == CO_FL_CURR_WR_ENA)) { |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 218 | fd_stop_send(c->handle.fd); |
Willy Tarreau | c8dd77f | 2012-11-05 17:52:26 +0100 | [diff] [blame] | 219 | f &= ~CO_FL_CURR_WR_ENA; |
Willy Tarreau | e9dfa79 | 2012-09-01 17:26:16 +0200 | [diff] [blame] | 220 | } |
Willy Tarreau | 310987a | 2014-01-22 19:46:33 +0100 | [diff] [blame] | 221 | c->flags = f; |
Willy Tarreau | e9dfa79 | 2012-09-01 17:26:16 +0200 | [diff] [blame] | 222 | } |
| 223 | |
| 224 | /* Update polling on connection <c>'s file descriptor depending on its current |
| 225 | * state as reported in the connection's CO_FL_CURR_* flags, reports of EAGAIN |
| 226 | * in CO_FL_WAIT_*, and the sock layer expectations indicated by CO_FL_SOCK_*. |
| 227 | * The connection flags are updated with the new flags at the end of the |
Willy Tarreau | 0ffde2c | 2012-10-04 22:21:15 +0200 | [diff] [blame] | 228 | * operation. Polling is totally disabled if an error was reported. |
Willy Tarreau | e9dfa79 | 2012-09-01 17:26:16 +0200 | [diff] [blame] | 229 | */ |
| 230 | void conn_update_sock_polling(struct connection *c) |
| 231 | { |
| 232 | unsigned int f = c->flags; |
Willy Tarreau | b5e2cbd | 2012-08-17 11:55:04 +0200 | [diff] [blame] | 233 | |
Willy Tarreau | 3c72872 | 2014-01-23 13:50:42 +0100 | [diff] [blame] | 234 | if (!conn_ctrl_ready(c)) |
Willy Tarreau | f79c817 | 2013-10-21 16:30:56 +0200 | [diff] [blame] | 235 | return; |
| 236 | |
Willy Tarreau | e9dfa79 | 2012-09-01 17:26:16 +0200 | [diff] [blame] | 237 | /* update read status if needed */ |
Willy Tarreau | 310987a | 2014-01-22 19:46:33 +0100 | [diff] [blame] | 238 | if (unlikely((f & (CO_FL_CURR_RD_ENA|CO_FL_SOCK_RD_ENA)) == CO_FL_SOCK_RD_ENA)) { |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 239 | fd_want_recv(c->handle.fd); |
Willy Tarreau | c8dd77f | 2012-11-05 17:52:26 +0100 | [diff] [blame] | 240 | f |= CO_FL_CURR_RD_ENA; |
Willy Tarreau | e9dfa79 | 2012-09-01 17:26:16 +0200 | [diff] [blame] | 241 | } |
Willy Tarreau | c8dd77f | 2012-11-05 17:52:26 +0100 | [diff] [blame] | 242 | else if (unlikely((f & (CO_FL_CURR_RD_ENA|CO_FL_SOCK_RD_ENA)) == CO_FL_CURR_RD_ENA)) { |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 243 | fd_stop_recv(c->handle.fd); |
Willy Tarreau | c8dd77f | 2012-11-05 17:52:26 +0100 | [diff] [blame] | 244 | f &= ~CO_FL_CURR_RD_ENA; |
Willy Tarreau | e9dfa79 | 2012-09-01 17:26:16 +0200 | [diff] [blame] | 245 | } |
| 246 | |
| 247 | /* update write status if needed */ |
Willy Tarreau | 310987a | 2014-01-22 19:46:33 +0100 | [diff] [blame] | 248 | if (unlikely((f & (CO_FL_CURR_WR_ENA|CO_FL_SOCK_WR_ENA)) == CO_FL_SOCK_WR_ENA)) { |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 249 | fd_want_send(c->handle.fd); |
Willy Tarreau | c8dd77f | 2012-11-05 17:52:26 +0100 | [diff] [blame] | 250 | f |= CO_FL_CURR_WR_ENA; |
Willy Tarreau | e9dfa79 | 2012-09-01 17:26:16 +0200 | [diff] [blame] | 251 | } |
Willy Tarreau | c8dd77f | 2012-11-05 17:52:26 +0100 | [diff] [blame] | 252 | else if (unlikely((f & (CO_FL_CURR_WR_ENA|CO_FL_SOCK_WR_ENA)) == CO_FL_CURR_WR_ENA)) { |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 253 | fd_stop_send(c->handle.fd); |
Willy Tarreau | c8dd77f | 2012-11-05 17:52:26 +0100 | [diff] [blame] | 254 | f &= ~CO_FL_CURR_WR_ENA; |
Willy Tarreau | e9dfa79 | 2012-09-01 17:26:16 +0200 | [diff] [blame] | 255 | } |
Willy Tarreau | 310987a | 2014-01-22 19:46:33 +0100 | [diff] [blame] | 256 | c->flags = f; |
Willy Tarreau | b5e2cbd | 2012-08-17 11:55:04 +0200 | [diff] [blame] | 257 | } |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 258 | |
Willy Tarreau | ff3e648 | 2015-03-12 23:56:52 +0100 | [diff] [blame] | 259 | /* Send a message over an established connection. It makes use of send() and |
| 260 | * returns the same return code and errno. If the socket layer is not ready yet |
| 261 | * then -1 is returned and ENOTSOCK is set into errno. If the fd is not marked |
| 262 | * as ready, or if EAGAIN or ENOTCONN is returned, then we return 0. It returns |
| 263 | * EMSGSIZE if called with a zero length message. The purpose is to simplify |
| 264 | * some rare attempts to directly write on the socket from above the connection |
| 265 | * (typically send_proxy). In case of EAGAIN, the fd is marked as "cant_send". |
| 266 | * It automatically retries on EINTR. Other errors cause the connection to be |
| 267 | * marked as in error state. It takes similar arguments as send() except the |
| 268 | * first one which is the connection instead of the file descriptor. Note, |
| 269 | * MSG_DONTWAIT and MSG_NOSIGNAL are forced on the flags. |
| 270 | */ |
| 271 | int conn_sock_send(struct connection *conn, const void *buf, int len, int flags) |
| 272 | { |
| 273 | int ret; |
| 274 | |
| 275 | ret = -1; |
| 276 | errno = ENOTSOCK; |
| 277 | |
| 278 | if (conn->flags & CO_FL_SOCK_WR_SH) |
| 279 | goto fail; |
| 280 | |
| 281 | if (!conn_ctrl_ready(conn)) |
| 282 | goto fail; |
| 283 | |
| 284 | errno = EMSGSIZE; |
| 285 | if (!len) |
| 286 | goto fail; |
| 287 | |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 288 | if (!fd_send_ready(conn->handle.fd)) |
Willy Tarreau | ff3e648 | 2015-03-12 23:56:52 +0100 | [diff] [blame] | 289 | goto wait; |
| 290 | |
| 291 | do { |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 292 | ret = send(conn->handle.fd, buf, len, flags | MSG_DONTWAIT | MSG_NOSIGNAL); |
Willy Tarreau | ff3e648 | 2015-03-12 23:56:52 +0100 | [diff] [blame] | 293 | } while (ret < 0 && errno == EINTR); |
| 294 | |
| 295 | |
| 296 | if (ret > 0) |
| 297 | return ret; |
| 298 | |
| 299 | if (ret == 0 || errno == EAGAIN || errno == ENOTCONN) { |
| 300 | wait: |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 301 | fd_cant_send(conn->handle.fd); |
Willy Tarreau | ff3e648 | 2015-03-12 23:56:52 +0100 | [diff] [blame] | 302 | return 0; |
| 303 | } |
| 304 | fail: |
| 305 | conn->flags |= CO_FL_SOCK_RD_SH | CO_FL_SOCK_WR_SH | CO_FL_ERROR; |
| 306 | return ret; |
| 307 | } |
| 308 | |
Willy Tarreau | d85c485 | 2015-03-13 00:40:28 +0100 | [diff] [blame] | 309 | /* Drains possibly pending incoming data on the file descriptor attached to the |
| 310 | * connection and update the connection's flags accordingly. This is used to |
| 311 | * know whether we need to disable lingering on close. Returns non-zero if it |
| 312 | * is safe to close without disabling lingering, otherwise zero. The SOCK_RD_SH |
| 313 | * flag may also be updated if the incoming shutdown was reported by the drain() |
| 314 | * function. |
| 315 | */ |
| 316 | int conn_sock_drain(struct connection *conn) |
| 317 | { |
| 318 | if (!conn_ctrl_ready(conn)) |
| 319 | return 1; |
| 320 | |
| 321 | if (conn->flags & (CO_FL_ERROR | CO_FL_SOCK_RD_SH)) |
| 322 | return 1; |
| 323 | |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 324 | if (fdtab[conn->handle.fd].ev & (FD_POLL_ERR|FD_POLL_HUP)) { |
| 325 | fdtab[conn->handle.fd].linger_risk = 0; |
Willy Tarreau | d85c485 | 2015-03-13 00:40:28 +0100 | [diff] [blame] | 326 | } |
| 327 | else { |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 328 | if (!fd_recv_ready(conn->handle.fd)) |
Willy Tarreau | d85c485 | 2015-03-13 00:40:28 +0100 | [diff] [blame] | 329 | return 0; |
| 330 | |
| 331 | /* disable draining if we were called and have no drain function */ |
| 332 | if (!conn->ctrl->drain) { |
Olivier Houchard | 1a0545f | 2017-09-13 18:30:23 +0200 | [diff] [blame] | 333 | __conn_xprt_stop_recv(conn); |
Willy Tarreau | d85c485 | 2015-03-13 00:40:28 +0100 | [diff] [blame] | 334 | return 0; |
| 335 | } |
| 336 | |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 337 | if (conn->ctrl->drain(conn->handle.fd) <= 0) |
Willy Tarreau | d85c485 | 2015-03-13 00:40:28 +0100 | [diff] [blame] | 338 | return 0; |
| 339 | } |
| 340 | |
| 341 | conn->flags |= CO_FL_SOCK_RD_SH; |
| 342 | return 1; |
| 343 | } |
| 344 | |
KOVACS Krisztian | b3e54fe | 2014-11-17 15:11:45 +0100 | [diff] [blame] | 345 | /* |
| 346 | * Get data length from tlv |
| 347 | */ |
| 348 | static int get_tlv_length(const struct tlv *src) |
| 349 | { |
| 350 | return (src->length_hi << 8) | src->length_lo; |
| 351 | } |
| 352 | |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 353 | /* This handshake handler waits a PROXY protocol header at the beginning of the |
| 354 | * raw data stream. The header looks like this : |
| 355 | * |
| 356 | * "PROXY" <SP> PROTO <SP> SRC3 <SP> DST3 <SP> SRC4 <SP> <DST4> "\r\n" |
| 357 | * |
| 358 | * There must be exactly one space between each field. Fields are : |
| 359 | * - PROTO : layer 4 protocol, which must be "TCP4" or "TCP6". |
| 360 | * - SRC3 : layer 3 (eg: IP) source address in standard text form |
| 361 | * - DST3 : layer 3 (eg: IP) destination address in standard text form |
| 362 | * - SRC4 : layer 4 (eg: TCP port) source address in standard text form |
| 363 | * - DST4 : layer 4 (eg: TCP port) destination address in standard text form |
| 364 | * |
| 365 | * This line MUST be at the beginning of the buffer and MUST NOT wrap. |
| 366 | * |
| 367 | * The header line is small and in all cases smaller than the smallest normal |
| 368 | * TCP MSS. So it MUST always be delivered as one segment, which ensures we |
| 369 | * can safely use MSG_PEEK and avoid buffering. |
| 370 | * |
| 371 | * Once the data is fetched, the values are set in the connection's address |
| 372 | * fields, and data are removed from the socket's buffer. The function returns |
| 373 | * zero if it needs to wait for more data or if it fails, or 1 if it completed |
| 374 | * and removed itself. |
| 375 | */ |
| 376 | int conn_recv_proxy(struct connection *conn, int flag) |
| 377 | { |
| 378 | char *line, *end; |
Willy Tarreau | 7799267 | 2014-06-14 11:06:17 +0200 | [diff] [blame] | 379 | struct proxy_hdr_v2 *hdr_v2; |
| 380 | const char v2sig[] = PP2_SIGNATURE; |
KOVACS Krisztian | b3e54fe | 2014-11-17 15:11:45 +0100 | [diff] [blame] | 381 | int tlv_length = 0; |
KOVACS Krisztian | 7209c20 | 2015-07-03 14:09:10 +0200 | [diff] [blame] | 382 | int tlv_offset = 0; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 383 | |
| 384 | /* we might have been called just after an asynchronous shutr */ |
| 385 | if (conn->flags & CO_FL_SOCK_RD_SH) |
| 386 | goto fail; |
| 387 | |
Willy Tarreau | 3c72872 | 2014-01-23 13:50:42 +0100 | [diff] [blame] | 388 | if (!conn_ctrl_ready(conn)) |
Willy Tarreau | f79c817 | 2013-10-21 16:30:56 +0200 | [diff] [blame] | 389 | goto fail; |
| 390 | |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 391 | if (!fd_recv_ready(conn->handle.fd)) |
Willy Tarreau | fd803bb | 2014-01-20 15:13:07 +0100 | [diff] [blame] | 392 | return 0; |
| 393 | |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 394 | do { |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 395 | trash.len = recv(conn->handle.fd, trash.str, trash.size, MSG_PEEK); |
Willy Tarreau | 19d14ef | 2012-10-29 16:51:55 +0100 | [diff] [blame] | 396 | if (trash.len < 0) { |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 397 | if (errno == EINTR) |
| 398 | continue; |
| 399 | if (errno == EAGAIN) { |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 400 | fd_cant_recv(conn->handle.fd); |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 401 | return 0; |
| 402 | } |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 403 | goto recv_abort; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 404 | } |
| 405 | } while (0); |
| 406 | |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 407 | if (!trash.len) { |
| 408 | /* client shutdown */ |
| 409 | conn->err_code = CO_ER_PRX_EMPTY; |
| 410 | goto fail; |
| 411 | } |
| 412 | |
Willy Tarreau | 19d14ef | 2012-10-29 16:51:55 +0100 | [diff] [blame] | 413 | if (trash.len < 6) |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 414 | goto missing; |
| 415 | |
Willy Tarreau | 19d14ef | 2012-10-29 16:51:55 +0100 | [diff] [blame] | 416 | line = trash.str; |
| 417 | end = trash.str + trash.len; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 418 | |
| 419 | /* Decode a possible proxy request, fail early if it does not match */ |
Willy Tarreau | 7799267 | 2014-06-14 11:06:17 +0200 | [diff] [blame] | 420 | if (strncmp(line, "PROXY ", 6) != 0) |
| 421 | goto not_v1; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 422 | |
| 423 | line += 6; |
Willy Tarreau | 4c20d29 | 2014-06-14 11:41:36 +0200 | [diff] [blame] | 424 | if (trash.len < 9) /* shortest possible line */ |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 425 | goto missing; |
| 426 | |
David CARLIER | 42ff05e | 2016-03-24 09:22:36 +0000 | [diff] [blame] | 427 | if (memcmp(line, "TCP4 ", 5) == 0) { |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 428 | u32 src3, dst3, sport, dport; |
| 429 | |
| 430 | line += 5; |
| 431 | |
| 432 | src3 = inetaddr_host_lim_ret(line, end, &line); |
| 433 | if (line == end) |
| 434 | goto missing; |
| 435 | if (*line++ != ' ') |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 436 | goto bad_header; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 437 | |
| 438 | dst3 = inetaddr_host_lim_ret(line, end, &line); |
| 439 | if (line == end) |
| 440 | goto missing; |
| 441 | if (*line++ != ' ') |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 442 | goto bad_header; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 443 | |
| 444 | sport = read_uint((const char **)&line, end); |
| 445 | if (line == end) |
| 446 | goto missing; |
| 447 | if (*line++ != ' ') |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 448 | goto bad_header; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 449 | |
| 450 | dport = read_uint((const char **)&line, end); |
| 451 | if (line > end - 2) |
| 452 | goto missing; |
| 453 | if (*line++ != '\r') |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 454 | goto bad_header; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 455 | if (*line++ != '\n') |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 456 | goto bad_header; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 457 | |
| 458 | /* update the session's addresses and mark them set */ |
| 459 | ((struct sockaddr_in *)&conn->addr.from)->sin_family = AF_INET; |
| 460 | ((struct sockaddr_in *)&conn->addr.from)->sin_addr.s_addr = htonl(src3); |
| 461 | ((struct sockaddr_in *)&conn->addr.from)->sin_port = htons(sport); |
| 462 | |
| 463 | ((struct sockaddr_in *)&conn->addr.to)->sin_family = AF_INET; |
| 464 | ((struct sockaddr_in *)&conn->addr.to)->sin_addr.s_addr = htonl(dst3); |
| 465 | ((struct sockaddr_in *)&conn->addr.to)->sin_port = htons(dport); |
| 466 | conn->flags |= CO_FL_ADDR_FROM_SET | CO_FL_ADDR_TO_SET; |
| 467 | } |
David CARLIER | 42ff05e | 2016-03-24 09:22:36 +0000 | [diff] [blame] | 468 | else if (memcmp(line, "TCP6 ", 5) == 0) { |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 469 | u32 sport, dport; |
| 470 | char *src_s; |
| 471 | char *dst_s, *sport_s, *dport_s; |
| 472 | struct in6_addr src3, dst3; |
| 473 | |
| 474 | line += 5; |
| 475 | |
| 476 | src_s = line; |
| 477 | dst_s = sport_s = dport_s = NULL; |
| 478 | while (1) { |
| 479 | if (line > end - 2) { |
| 480 | goto missing; |
| 481 | } |
| 482 | else if (*line == '\r') { |
| 483 | *line = 0; |
| 484 | line++; |
| 485 | if (*line++ != '\n') |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 486 | goto bad_header; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 487 | break; |
| 488 | } |
| 489 | |
| 490 | if (*line == ' ') { |
| 491 | *line = 0; |
| 492 | if (!dst_s) |
| 493 | dst_s = line + 1; |
| 494 | else if (!sport_s) |
| 495 | sport_s = line + 1; |
| 496 | else if (!dport_s) |
| 497 | dport_s = line + 1; |
| 498 | } |
| 499 | line++; |
| 500 | } |
| 501 | |
| 502 | if (!dst_s || !sport_s || !dport_s) |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 503 | goto bad_header; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 504 | |
| 505 | sport = read_uint((const char **)&sport_s,dport_s - 1); |
| 506 | if (*sport_s != 0) |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 507 | goto bad_header; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 508 | |
| 509 | dport = read_uint((const char **)&dport_s,line - 2); |
| 510 | if (*dport_s != 0) |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 511 | goto bad_header; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 512 | |
| 513 | if (inet_pton(AF_INET6, src_s, (void *)&src3) != 1) |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 514 | goto bad_header; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 515 | |
| 516 | if (inet_pton(AF_INET6, dst_s, (void *)&dst3) != 1) |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 517 | goto bad_header; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 518 | |
| 519 | /* update the session's addresses and mark them set */ |
| 520 | ((struct sockaddr_in6 *)&conn->addr.from)->sin6_family = AF_INET6; |
| 521 | memcpy(&((struct sockaddr_in6 *)&conn->addr.from)->sin6_addr, &src3, sizeof(struct in6_addr)); |
| 522 | ((struct sockaddr_in6 *)&conn->addr.from)->sin6_port = htons(sport); |
| 523 | |
| 524 | ((struct sockaddr_in6 *)&conn->addr.to)->sin6_family = AF_INET6; |
| 525 | memcpy(&((struct sockaddr_in6 *)&conn->addr.to)->sin6_addr, &dst3, sizeof(struct in6_addr)); |
| 526 | ((struct sockaddr_in6 *)&conn->addr.to)->sin6_port = htons(dport); |
| 527 | conn->flags |= CO_FL_ADDR_FROM_SET | CO_FL_ADDR_TO_SET; |
| 528 | } |
Willy Tarreau | 4c20d29 | 2014-06-14 11:41:36 +0200 | [diff] [blame] | 529 | else if (memcmp(line, "UNKNOWN\r\n", 9) == 0) { |
| 530 | /* This can be a UNIX socket forwarded by an haproxy upstream */ |
| 531 | line += 9; |
| 532 | } |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 533 | else { |
Willy Tarreau | 4c20d29 | 2014-06-14 11:41:36 +0200 | [diff] [blame] | 534 | /* The protocol does not match something known (TCP4/TCP6/UNKNOWN) */ |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 535 | conn->err_code = CO_ER_PRX_BAD_PROTO; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 536 | goto fail; |
| 537 | } |
| 538 | |
Willy Tarreau | 7799267 | 2014-06-14 11:06:17 +0200 | [diff] [blame] | 539 | trash.len = line - trash.str; |
| 540 | goto eat_header; |
| 541 | |
| 542 | not_v1: |
| 543 | /* try PPv2 */ |
| 544 | if (trash.len < PP2_HEADER_LEN) |
| 545 | goto missing; |
| 546 | |
| 547 | hdr_v2 = (struct proxy_hdr_v2 *)trash.str; |
| 548 | |
| 549 | if (memcmp(hdr_v2->sig, v2sig, PP2_SIGNATURE_LEN) != 0 || |
| 550 | (hdr_v2->ver_cmd & PP2_VERSION_MASK) != PP2_VERSION) { |
| 551 | conn->err_code = CO_ER_PRX_NOT_HDR; |
| 552 | goto fail; |
| 553 | } |
| 554 | |
| 555 | if (trash.len < PP2_HEADER_LEN + ntohs(hdr_v2->len)) |
| 556 | goto missing; |
| 557 | |
| 558 | switch (hdr_v2->ver_cmd & PP2_CMD_MASK) { |
| 559 | case 0x01: /* PROXY command */ |
| 560 | switch (hdr_v2->fam) { |
| 561 | case 0x11: /* TCPv4 */ |
KOVACS Krisztian | efd3aa9 | 2014-11-19 10:53:20 +0100 | [diff] [blame] | 562 | if (ntohs(hdr_v2->len) < PP2_ADDR_LEN_INET) |
| 563 | goto bad_header; |
| 564 | |
Willy Tarreau | 7799267 | 2014-06-14 11:06:17 +0200 | [diff] [blame] | 565 | ((struct sockaddr_in *)&conn->addr.from)->sin_family = AF_INET; |
| 566 | ((struct sockaddr_in *)&conn->addr.from)->sin_addr.s_addr = hdr_v2->addr.ip4.src_addr; |
| 567 | ((struct sockaddr_in *)&conn->addr.from)->sin_port = hdr_v2->addr.ip4.src_port; |
| 568 | ((struct sockaddr_in *)&conn->addr.to)->sin_family = AF_INET; |
| 569 | ((struct sockaddr_in *)&conn->addr.to)->sin_addr.s_addr = hdr_v2->addr.ip4.dst_addr; |
| 570 | ((struct sockaddr_in *)&conn->addr.to)->sin_port = hdr_v2->addr.ip4.dst_port; |
| 571 | conn->flags |= CO_FL_ADDR_FROM_SET | CO_FL_ADDR_TO_SET; |
KOVACS Krisztian | 7209c20 | 2015-07-03 14:09:10 +0200 | [diff] [blame] | 572 | tlv_offset = PP2_HEADER_LEN + PP2_ADDR_LEN_INET; |
KOVACS Krisztian | b3e54fe | 2014-11-17 15:11:45 +0100 | [diff] [blame] | 573 | tlv_length = ntohs(hdr_v2->len) - PP2_ADDR_LEN_INET; |
Willy Tarreau | 7799267 | 2014-06-14 11:06:17 +0200 | [diff] [blame] | 574 | break; |
| 575 | case 0x21: /* TCPv6 */ |
KOVACS Krisztian | efd3aa9 | 2014-11-19 10:53:20 +0100 | [diff] [blame] | 576 | if (ntohs(hdr_v2->len) < PP2_ADDR_LEN_INET6) |
| 577 | goto bad_header; |
| 578 | |
Willy Tarreau | 7799267 | 2014-06-14 11:06:17 +0200 | [diff] [blame] | 579 | ((struct sockaddr_in6 *)&conn->addr.from)->sin6_family = AF_INET6; |
| 580 | memcpy(&((struct sockaddr_in6 *)&conn->addr.from)->sin6_addr, hdr_v2->addr.ip6.src_addr, 16); |
| 581 | ((struct sockaddr_in6 *)&conn->addr.from)->sin6_port = hdr_v2->addr.ip6.src_port; |
| 582 | ((struct sockaddr_in6 *)&conn->addr.to)->sin6_family = AF_INET6; |
| 583 | memcpy(&((struct sockaddr_in6 *)&conn->addr.to)->sin6_addr, hdr_v2->addr.ip6.dst_addr, 16); |
| 584 | ((struct sockaddr_in6 *)&conn->addr.to)->sin6_port = hdr_v2->addr.ip6.dst_port; |
| 585 | conn->flags |= CO_FL_ADDR_FROM_SET | CO_FL_ADDR_TO_SET; |
KOVACS Krisztian | 7209c20 | 2015-07-03 14:09:10 +0200 | [diff] [blame] | 586 | tlv_offset = PP2_HEADER_LEN + PP2_ADDR_LEN_INET6; |
KOVACS Krisztian | b3e54fe | 2014-11-17 15:11:45 +0100 | [diff] [blame] | 587 | tlv_length = ntohs(hdr_v2->len) - PP2_ADDR_LEN_INET6; |
Willy Tarreau | 7799267 | 2014-06-14 11:06:17 +0200 | [diff] [blame] | 588 | break; |
| 589 | } |
KOVACS Krisztian | b3e54fe | 2014-11-17 15:11:45 +0100 | [diff] [blame] | 590 | |
| 591 | /* TLV parsing */ |
| 592 | if (tlv_length > 0) { |
KOVACS Krisztian | b3e54fe | 2014-11-17 15:11:45 +0100 | [diff] [blame] | 593 | while (tlv_offset + TLV_HEADER_SIZE <= trash.len) { |
| 594 | const struct tlv *tlv_packet = (struct tlv *) &trash.str[tlv_offset]; |
| 595 | const int tlv_len = get_tlv_length(tlv_packet); |
| 596 | tlv_offset += tlv_len + TLV_HEADER_SIZE; |
| 597 | |
| 598 | switch (tlv_packet->type) { |
| 599 | #ifdef CONFIG_HAP_NS |
| 600 | case PP2_TYPE_NETNS: { |
| 601 | const struct netns_entry *ns; |
| 602 | ns = netns_store_lookup((char*)tlv_packet->value, tlv_len); |
| 603 | if (ns) |
| 604 | conn->proxy_netns = ns; |
| 605 | break; |
| 606 | } |
| 607 | #endif |
| 608 | default: |
| 609 | break; |
| 610 | } |
| 611 | } |
| 612 | } |
| 613 | |
Willy Tarreau | 7799267 | 2014-06-14 11:06:17 +0200 | [diff] [blame] | 614 | /* unsupported protocol, keep local connection address */ |
| 615 | break; |
| 616 | case 0x00: /* LOCAL command */ |
| 617 | /* keep local connection address for LOCAL */ |
| 618 | break; |
| 619 | default: |
| 620 | goto bad_header; /* not a supported command */ |
| 621 | } |
| 622 | |
| 623 | trash.len = PP2_HEADER_LEN + ntohs(hdr_v2->len); |
| 624 | goto eat_header; |
| 625 | |
| 626 | eat_header: |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 627 | /* remove the PROXY line from the request. For this we re-read the |
| 628 | * exact line at once. If we don't get the exact same result, we |
| 629 | * fail. |
| 630 | */ |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 631 | do { |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 632 | int len2 = recv(conn->handle.fd, trash.str, trash.len, 0); |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 633 | if (len2 < 0 && errno == EINTR) |
| 634 | continue; |
Willy Tarreau | 19d14ef | 2012-10-29 16:51:55 +0100 | [diff] [blame] | 635 | if (len2 != trash.len) |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 636 | goto recv_abort; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 637 | } while (0); |
| 638 | |
| 639 | conn->flags &= ~flag; |
Emeric Brun | 4f60301 | 2017-01-05 15:11:44 +0100 | [diff] [blame] | 640 | conn->flags |= CO_FL_RCVD_PROXY; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 641 | return 1; |
| 642 | |
| 643 | missing: |
| 644 | /* Missing data. Since we're using MSG_PEEK, we can only poll again if |
| 645 | * we have not read anything. Otherwise we need to fail because we won't |
| 646 | * be able to poll anymore. |
| 647 | */ |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 648 | conn->err_code = CO_ER_PRX_TRUNCATED; |
| 649 | goto fail; |
| 650 | |
| 651 | bad_header: |
| 652 | /* This is not a valid proxy protocol header */ |
| 653 | conn->err_code = CO_ER_PRX_BAD_HDR; |
| 654 | goto fail; |
| 655 | |
| 656 | recv_abort: |
| 657 | conn->err_code = CO_ER_PRX_ABORT; |
Willy Tarreau | 26f4a04 | 2013-12-04 23:44:10 +0100 | [diff] [blame] | 658 | conn->flags |= CO_FL_SOCK_RD_SH | CO_FL_SOCK_WR_SH; |
Willy Tarreau | 8e3bf69 | 2012-12-03 15:41:18 +0100 | [diff] [blame] | 659 | goto fail; |
| 660 | |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 661 | fail: |
Willy Tarreau | d486ef5 | 2012-12-10 17:03:52 +0100 | [diff] [blame] | 662 | __conn_sock_stop_both(conn); |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 663 | conn->flags |= CO_FL_ERROR; |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 664 | return 0; |
| 665 | } |
| 666 | |
Bertrand Jacquin | 93b227d | 2016-06-04 15:11:10 +0100 | [diff] [blame] | 667 | /* This handshake handler waits a NetScaler Client IP insertion header |
| 668 | * at the beginning of the raw data stream. The header looks like this: |
| 669 | * |
| 670 | * 4 bytes: CIP magic number |
| 671 | * 4 bytes: Header length |
| 672 | * 20+ bytes: Header of the last IP packet sent by the client during |
| 673 | * TCP handshake. |
| 674 | * 20+ bytes: Header of the last TCP packet sent by the client during |
| 675 | * TCP handshake. |
| 676 | * |
| 677 | * This line MUST be at the beginning of the buffer and MUST NOT be |
| 678 | * fragmented. |
| 679 | * |
| 680 | * The header line is small and in all cases smaller than the smallest normal |
| 681 | * TCP MSS. So it MUST always be delivered as one segment, which ensures we |
| 682 | * can safely use MSG_PEEK and avoid buffering. |
| 683 | * |
| 684 | * Once the data is fetched, the values are set in the connection's address |
| 685 | * fields, and data are removed from the socket's buffer. The function returns |
| 686 | * zero if it needs to wait for more data or if it fails, or 1 if it completed |
| 687 | * and removed itself. |
| 688 | */ |
| 689 | int conn_recv_netscaler_cip(struct connection *conn, int flag) |
| 690 | { |
| 691 | char *line; |
| 692 | uint32_t cip_magic; |
| 693 | uint32_t cip_len; |
| 694 | uint8_t ip_v; |
| 695 | |
| 696 | /* we might have been called just after an asynchronous shutr */ |
| 697 | if (conn->flags & CO_FL_SOCK_RD_SH) |
| 698 | goto fail; |
| 699 | |
| 700 | if (!conn_ctrl_ready(conn)) |
| 701 | goto fail; |
| 702 | |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 703 | if (!fd_recv_ready(conn->handle.fd)) |
Bertrand Jacquin | 93b227d | 2016-06-04 15:11:10 +0100 | [diff] [blame] | 704 | return 0; |
| 705 | |
| 706 | do { |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 707 | trash.len = recv(conn->handle.fd, trash.str, trash.size, MSG_PEEK); |
Bertrand Jacquin | 93b227d | 2016-06-04 15:11:10 +0100 | [diff] [blame] | 708 | if (trash.len < 0) { |
| 709 | if (errno == EINTR) |
| 710 | continue; |
| 711 | if (errno == EAGAIN) { |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 712 | fd_cant_recv(conn->handle.fd); |
Bertrand Jacquin | 93b227d | 2016-06-04 15:11:10 +0100 | [diff] [blame] | 713 | return 0; |
| 714 | } |
| 715 | goto recv_abort; |
| 716 | } |
| 717 | } while (0); |
| 718 | |
| 719 | if (!trash.len) { |
| 720 | /* client shutdown */ |
| 721 | conn->err_code = CO_ER_CIP_EMPTY; |
| 722 | goto fail; |
| 723 | } |
| 724 | |
| 725 | /* Fail if buffer length is not large enough to contain |
| 726 | * CIP magic, CIP length */ |
| 727 | if (trash.len < 8) |
| 728 | goto missing; |
| 729 | |
| 730 | line = trash.str; |
| 731 | |
| 732 | cip_magic = ntohl(*(uint32_t *)line); |
| 733 | cip_len = ntohl(*(uint32_t *)(line+4)); |
| 734 | |
| 735 | /* Decode a possible NetScaler Client IP request, fail early if |
| 736 | * it does not match */ |
| 737 | if (cip_magic != objt_listener(conn->target)->bind_conf->ns_cip_magic) |
| 738 | goto bad_magic; |
| 739 | |
| 740 | /* Fail if buffer length is not large enough to contain |
| 741 | * CIP magic, CIP length, minimal IP header */ |
| 742 | if (trash.len < 28) |
| 743 | goto missing; |
| 744 | |
| 745 | line += 8; |
| 746 | |
| 747 | /* Get IP version from the first four bits */ |
| 748 | ip_v = (*line & 0xf0) >> 4; |
| 749 | |
| 750 | if (ip_v == 4) { |
| 751 | struct ip *hdr_ip4; |
David Carlier | 3015a2e | 2016-07-04 22:51:33 +0100 | [diff] [blame] | 752 | struct my_tcphdr *hdr_tcp; |
Bertrand Jacquin | 93b227d | 2016-06-04 15:11:10 +0100 | [diff] [blame] | 753 | |
| 754 | hdr_ip4 = (struct ip *)line; |
| 755 | |
| 756 | if (trash.len < (8 + ntohs(hdr_ip4->ip_len))) { |
| 757 | /* Fail if buffer length is not large enough to contain |
| 758 | * CIP magic, CIP length, IPv4 header */ |
| 759 | goto missing; |
| 760 | } else if (hdr_ip4->ip_p != IPPROTO_TCP) { |
| 761 | /* The protocol does not include a TCP header */ |
| 762 | conn->err_code = CO_ER_CIP_BAD_PROTO; |
| 763 | goto fail; |
| 764 | } else if (trash.len < (28 + ntohs(hdr_ip4->ip_len))) { |
| 765 | /* Fail if buffer length is not large enough to contain |
| 766 | * CIP magic, CIP length, IPv4 header, TCP header */ |
| 767 | goto missing; |
| 768 | } |
| 769 | |
David Carlier | 3015a2e | 2016-07-04 22:51:33 +0100 | [diff] [blame] | 770 | hdr_tcp = (struct my_tcphdr *)(line + (hdr_ip4->ip_hl * 4)); |
Bertrand Jacquin | 93b227d | 2016-06-04 15:11:10 +0100 | [diff] [blame] | 771 | |
| 772 | /* update the session's addresses and mark them set */ |
| 773 | ((struct sockaddr_in *)&conn->addr.from)->sin_family = AF_INET; |
| 774 | ((struct sockaddr_in *)&conn->addr.from)->sin_addr.s_addr = hdr_ip4->ip_src.s_addr; |
| 775 | ((struct sockaddr_in *)&conn->addr.from)->sin_port = hdr_tcp->source; |
| 776 | |
| 777 | ((struct sockaddr_in *)&conn->addr.to)->sin_family = AF_INET; |
| 778 | ((struct sockaddr_in *)&conn->addr.to)->sin_addr.s_addr = hdr_ip4->ip_dst.s_addr; |
| 779 | ((struct sockaddr_in *)&conn->addr.to)->sin_port = hdr_tcp->dest; |
| 780 | |
| 781 | conn->flags |= CO_FL_ADDR_FROM_SET | CO_FL_ADDR_TO_SET; |
| 782 | } |
| 783 | else if (ip_v == 6) { |
| 784 | struct ip6_hdr *hdr_ip6; |
David Carlier | 3015a2e | 2016-07-04 22:51:33 +0100 | [diff] [blame] | 785 | struct my_tcphdr *hdr_tcp; |
Bertrand Jacquin | 93b227d | 2016-06-04 15:11:10 +0100 | [diff] [blame] | 786 | |
| 787 | hdr_ip6 = (struct ip6_hdr *)line; |
| 788 | |
| 789 | if (trash.len < 28) { |
| 790 | /* Fail if buffer length is not large enough to contain |
| 791 | * CIP magic, CIP length, IPv6 header */ |
| 792 | goto missing; |
| 793 | } else if (hdr_ip6->ip6_nxt != IPPROTO_TCP) { |
| 794 | /* The protocol does not include a TCP header */ |
| 795 | conn->err_code = CO_ER_CIP_BAD_PROTO; |
| 796 | goto fail; |
| 797 | } else if (trash.len < 48) { |
| 798 | /* Fail if buffer length is not large enough to contain |
| 799 | * CIP magic, CIP length, IPv6 header, TCP header */ |
| 800 | goto missing; |
| 801 | } |
| 802 | |
David Carlier | 3015a2e | 2016-07-04 22:51:33 +0100 | [diff] [blame] | 803 | hdr_tcp = (struct my_tcphdr *)(line + sizeof(struct ip6_hdr)); |
Bertrand Jacquin | 93b227d | 2016-06-04 15:11:10 +0100 | [diff] [blame] | 804 | |
| 805 | /* update the session's addresses and mark them set */ |
| 806 | ((struct sockaddr_in6 *)&conn->addr.from)->sin6_family = AF_INET6; |
| 807 | ((struct sockaddr_in6 *)&conn->addr.from)->sin6_addr = hdr_ip6->ip6_src; |
| 808 | ((struct sockaddr_in6 *)&conn->addr.from)->sin6_port = hdr_tcp->source; |
| 809 | |
| 810 | ((struct sockaddr_in6 *)&conn->addr.to)->sin6_family = AF_INET6; |
| 811 | ((struct sockaddr_in6 *)&conn->addr.to)->sin6_addr = hdr_ip6->ip6_dst; |
| 812 | ((struct sockaddr_in6 *)&conn->addr.to)->sin6_port = hdr_tcp->dest; |
| 813 | |
| 814 | conn->flags |= CO_FL_ADDR_FROM_SET | CO_FL_ADDR_TO_SET; |
| 815 | } |
| 816 | else { |
| 817 | /* The protocol does not match something known (IPv4/IPv6) */ |
| 818 | conn->err_code = CO_ER_CIP_BAD_PROTO; |
| 819 | goto fail; |
| 820 | } |
| 821 | |
| 822 | line += cip_len; |
| 823 | trash.len = line - trash.str; |
| 824 | |
| 825 | /* remove the NetScaler Client IP header from the request. For this |
| 826 | * we re-read the exact line at once. If we don't get the exact same |
| 827 | * result, we fail. |
| 828 | */ |
| 829 | do { |
Willy Tarreau | 585744b | 2017-08-24 14:31:19 +0200 | [diff] [blame] | 830 | int len2 = recv(conn->handle.fd, trash.str, trash.len, 0); |
Bertrand Jacquin | 93b227d | 2016-06-04 15:11:10 +0100 | [diff] [blame] | 831 | if (len2 < 0 && errno == EINTR) |
| 832 | continue; |
| 833 | if (len2 != trash.len) |
| 834 | goto recv_abort; |
| 835 | } while (0); |
| 836 | |
| 837 | conn->flags &= ~flag; |
| 838 | return 1; |
| 839 | |
| 840 | missing: |
| 841 | /* Missing data. Since we're using MSG_PEEK, we can only poll again if |
| 842 | * we have not read anything. Otherwise we need to fail because we won't |
| 843 | * be able to poll anymore. |
| 844 | */ |
| 845 | conn->err_code = CO_ER_CIP_TRUNCATED; |
| 846 | goto fail; |
| 847 | |
| 848 | bad_magic: |
| 849 | conn->err_code = CO_ER_CIP_BAD_MAGIC; |
| 850 | goto fail; |
| 851 | |
| 852 | recv_abort: |
| 853 | conn->err_code = CO_ER_CIP_ABORT; |
| 854 | conn->flags |= CO_FL_SOCK_RD_SH | CO_FL_SOCK_WR_SH; |
| 855 | goto fail; |
| 856 | |
| 857 | fail: |
| 858 | __conn_sock_stop_both(conn); |
| 859 | conn->flags |= CO_FL_ERROR; |
| 860 | return 0; |
| 861 | } |
| 862 | |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 863 | int make_proxy_line(char *buf, int buf_len, struct server *srv, struct connection *remote) |
| 864 | { |
| 865 | int ret = 0; |
| 866 | |
| 867 | if (srv && (srv->pp_opts & SRV_PP_V2)) { |
| 868 | ret = make_proxy_line_v2(buf, buf_len, srv, remote); |
| 869 | } |
| 870 | else { |
| 871 | if (remote) |
| 872 | ret = make_proxy_line_v1(buf, buf_len, &remote->addr.from, &remote->addr.to); |
| 873 | else |
| 874 | ret = make_proxy_line_v1(buf, buf_len, NULL, NULL); |
| 875 | } |
| 876 | |
| 877 | return ret; |
| 878 | } |
| 879 | |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 880 | /* Makes a PROXY protocol line from the two addresses. The output is sent to |
| 881 | * buffer <buf> for a maximum size of <buf_len> (including the trailing zero). |
| 882 | * It returns the number of bytes composing this line (including the trailing |
| 883 | * LF), or zero in case of failure (eg: not enough space). It supports TCP4, |
Willy Tarreau | 2e1401a | 2013-10-01 11:41:55 +0200 | [diff] [blame] | 884 | * TCP6 and "UNKNOWN" formats. If any of <src> or <dst> is null, UNKNOWN is |
| 885 | * emitted as well. |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 886 | */ |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 887 | int make_proxy_line_v1(char *buf, int buf_len, struct sockaddr_storage *src, struct sockaddr_storage *dst) |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 888 | { |
| 889 | int ret = 0; |
| 890 | |
Willy Tarreau | 2e1401a | 2013-10-01 11:41:55 +0200 | [diff] [blame] | 891 | if (src && dst && src->ss_family == dst->ss_family && src->ss_family == AF_INET) { |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 892 | ret = snprintf(buf + ret, buf_len - ret, "PROXY TCP4 "); |
| 893 | if (ret >= buf_len) |
| 894 | return 0; |
| 895 | |
| 896 | /* IPv4 src */ |
| 897 | if (!inet_ntop(src->ss_family, &((struct sockaddr_in *)src)->sin_addr, buf + ret, buf_len - ret)) |
| 898 | return 0; |
| 899 | |
| 900 | ret += strlen(buf + ret); |
| 901 | if (ret >= buf_len) |
| 902 | return 0; |
| 903 | |
| 904 | buf[ret++] = ' '; |
| 905 | |
| 906 | /* IPv4 dst */ |
| 907 | if (!inet_ntop(dst->ss_family, &((struct sockaddr_in *)dst)->sin_addr, buf + ret, buf_len - ret)) |
| 908 | return 0; |
| 909 | |
| 910 | ret += strlen(buf + ret); |
| 911 | if (ret >= buf_len) |
| 912 | return 0; |
| 913 | |
| 914 | /* source and destination ports */ |
| 915 | ret += snprintf(buf + ret, buf_len - ret, " %u %u\r\n", |
| 916 | ntohs(((struct sockaddr_in *)src)->sin_port), |
| 917 | ntohs(((struct sockaddr_in *)dst)->sin_port)); |
| 918 | if (ret >= buf_len) |
| 919 | return 0; |
| 920 | } |
Willy Tarreau | 2e1401a | 2013-10-01 11:41:55 +0200 | [diff] [blame] | 921 | else if (src && dst && src->ss_family == dst->ss_family && src->ss_family == AF_INET6) { |
Willy Tarreau | e1e4a61 | 2012-10-05 00:10:55 +0200 | [diff] [blame] | 922 | ret = snprintf(buf + ret, buf_len - ret, "PROXY TCP6 "); |
| 923 | if (ret >= buf_len) |
| 924 | return 0; |
| 925 | |
| 926 | /* IPv6 src */ |
| 927 | if (!inet_ntop(src->ss_family, &((struct sockaddr_in6 *)src)->sin6_addr, buf + ret, buf_len - ret)) |
| 928 | return 0; |
| 929 | |
| 930 | ret += strlen(buf + ret); |
| 931 | if (ret >= buf_len) |
| 932 | return 0; |
| 933 | |
| 934 | buf[ret++] = ' '; |
| 935 | |
| 936 | /* IPv6 dst */ |
| 937 | if (!inet_ntop(dst->ss_family, &((struct sockaddr_in6 *)dst)->sin6_addr, buf + ret, buf_len - ret)) |
| 938 | return 0; |
| 939 | |
| 940 | ret += strlen(buf + ret); |
| 941 | if (ret >= buf_len) |
| 942 | return 0; |
| 943 | |
| 944 | /* source and destination ports */ |
| 945 | ret += snprintf(buf + ret, buf_len - ret, " %u %u\r\n", |
| 946 | ntohs(((struct sockaddr_in6 *)src)->sin6_port), |
| 947 | ntohs(((struct sockaddr_in6 *)dst)->sin6_port)); |
| 948 | if (ret >= buf_len) |
| 949 | return 0; |
| 950 | } |
| 951 | else { |
| 952 | /* unknown family combination */ |
| 953 | ret = snprintf(buf, buf_len, "PROXY UNKNOWN\r\n"); |
| 954 | if (ret >= buf_len) |
| 955 | return 0; |
| 956 | } |
| 957 | return ret; |
| 958 | } |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 959 | |
KOVACS Krisztian | b3e54fe | 2014-11-17 15:11:45 +0100 | [diff] [blame] | 960 | static int make_tlv(char *dest, int dest_len, char type, uint16_t length, const char *value) |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 961 | { |
| 962 | struct tlv *tlv; |
| 963 | |
| 964 | if (!dest || (length + sizeof(*tlv) > dest_len)) |
| 965 | return 0; |
| 966 | |
| 967 | tlv = (struct tlv *)dest; |
| 968 | |
| 969 | tlv->type = type; |
| 970 | tlv->length_hi = length >> 8; |
| 971 | tlv->length_lo = length & 0x00ff; |
| 972 | memcpy(tlv->value, value, length); |
| 973 | return length + sizeof(*tlv); |
| 974 | } |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 975 | |
| 976 | int make_proxy_line_v2(char *buf, int buf_len, struct server *srv, struct connection *remote) |
| 977 | { |
Willy Tarreau | 8fccfa2 | 2014-06-14 08:28:06 +0200 | [diff] [blame] | 978 | const char pp2_signature[] = PP2_SIGNATURE; |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 979 | int ret = 0; |
Willy Tarreau | 8fccfa2 | 2014-06-14 08:28:06 +0200 | [diff] [blame] | 980 | struct proxy_hdr_v2 *hdr = (struct proxy_hdr_v2 *)buf; |
Vincent Bernat | 6e61589 | 2016-05-18 16:17:44 +0200 | [diff] [blame] | 981 | struct sockaddr_storage null_addr = { .ss_family = 0 }; |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 982 | struct sockaddr_storage *src = &null_addr; |
| 983 | struct sockaddr_storage *dst = &null_addr; |
Emmanuel Hocdet | 404d978 | 2017-10-24 10:55:14 +0200 | [diff] [blame] | 984 | const char *value; |
| 985 | int value_len; |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 986 | |
| 987 | if (buf_len < PP2_HEADER_LEN) |
| 988 | return 0; |
Willy Tarreau | 8fccfa2 | 2014-06-14 08:28:06 +0200 | [diff] [blame] | 989 | memcpy(hdr->sig, pp2_signature, PP2_SIGNATURE_LEN); |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 990 | |
| 991 | if (remote) { |
| 992 | src = &remote->addr.from; |
| 993 | dst = &remote->addr.to; |
| 994 | } |
KOVACS Krisztian | b3e54fe | 2014-11-17 15:11:45 +0100 | [diff] [blame] | 995 | |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 996 | if (src && dst && src->ss_family == dst->ss_family && src->ss_family == AF_INET) { |
| 997 | if (buf_len < PP2_HDR_LEN_INET) |
| 998 | return 0; |
Willy Tarreau | 8fccfa2 | 2014-06-14 08:28:06 +0200 | [diff] [blame] | 999 | hdr->ver_cmd = PP2_VERSION | PP2_CMD_PROXY; |
| 1000 | hdr->fam = PP2_FAM_INET | PP2_TRANS_STREAM; |
| 1001 | hdr->addr.ip4.src_addr = ((struct sockaddr_in *)src)->sin_addr.s_addr; |
| 1002 | hdr->addr.ip4.dst_addr = ((struct sockaddr_in *)dst)->sin_addr.s_addr; |
| 1003 | hdr->addr.ip4.src_port = ((struct sockaddr_in *)src)->sin_port; |
| 1004 | hdr->addr.ip4.dst_port = ((struct sockaddr_in *)dst)->sin_port; |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 1005 | ret = PP2_HDR_LEN_INET; |
| 1006 | } |
| 1007 | else if (src && dst && src->ss_family == dst->ss_family && src->ss_family == AF_INET6) { |
| 1008 | if (buf_len < PP2_HDR_LEN_INET6) |
| 1009 | return 0; |
Willy Tarreau | 8fccfa2 | 2014-06-14 08:28:06 +0200 | [diff] [blame] | 1010 | hdr->ver_cmd = PP2_VERSION | PP2_CMD_PROXY; |
| 1011 | hdr->fam = PP2_FAM_INET6 | PP2_TRANS_STREAM; |
| 1012 | memcpy(hdr->addr.ip6.src_addr, &((struct sockaddr_in6 *)src)->sin6_addr, 16); |
| 1013 | memcpy(hdr->addr.ip6.dst_addr, &((struct sockaddr_in6 *)dst)->sin6_addr, 16); |
| 1014 | hdr->addr.ip6.src_port = ((struct sockaddr_in6 *)src)->sin6_port; |
| 1015 | hdr->addr.ip6.dst_port = ((struct sockaddr_in6 *)dst)->sin6_port; |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 1016 | ret = PP2_HDR_LEN_INET6; |
| 1017 | } |
| 1018 | else { |
| 1019 | if (buf_len < PP2_HDR_LEN_UNSPEC) |
| 1020 | return 0; |
Willy Tarreau | 8fccfa2 | 2014-06-14 08:28:06 +0200 | [diff] [blame] | 1021 | hdr->ver_cmd = PP2_VERSION | PP2_CMD_LOCAL; |
| 1022 | hdr->fam = PP2_FAM_UNSPEC | PP2_TRANS_UNSPEC; |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 1023 | ret = PP2_HDR_LEN_UNSPEC; |
| 1024 | } |
| 1025 | |
Emmanuel Hocdet | 404d978 | 2017-10-24 10:55:14 +0200 | [diff] [blame] | 1026 | if (conn_get_alpn(remote, &value, &value_len)) { |
| 1027 | if ((buf_len - ret) < sizeof(struct tlv)) |
| 1028 | return 0; |
| 1029 | ret += make_tlv(&buf[ret], buf_len, PP2_TYPE_ALPN, value_len, value); |
| 1030 | } |
| 1031 | |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 1032 | #ifdef USE_OPENSSL |
| 1033 | if (srv->pp_opts & SRV_PP_V2_SSL) { |
Emmanuel Hocdet | 404d978 | 2017-10-24 10:55:14 +0200 | [diff] [blame] | 1034 | struct tlv_ssl *tlv; |
| 1035 | int ssl_tlv_len = 0; |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 1036 | if ((buf_len - ret) < sizeof(struct tlv_ssl)) |
| 1037 | return 0; |
| 1038 | tlv = (struct tlv_ssl *)&buf[ret]; |
| 1039 | memset(tlv, 0, sizeof(struct tlv_ssl)); |
| 1040 | ssl_tlv_len += sizeof(struct tlv_ssl); |
| 1041 | tlv->tlv.type = PP2_TYPE_SSL; |
| 1042 | if (ssl_sock_is_ssl(remote)) { |
| 1043 | tlv->client |= PP2_CLIENT_SSL; |
Emmanuel Hocdet | 01da571 | 2017-10-13 16:59:49 +0200 | [diff] [blame] | 1044 | value = ssl_sock_get_proto_version(remote); |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 1045 | if (value) { |
Emmanuel Hocdet | 58118b4 | 2017-10-13 12:15:28 +0200 | [diff] [blame] | 1046 | ssl_tlv_len += make_tlv(&buf[ret+ssl_tlv_len], (buf_len-ret-ssl_tlv_len), PP2_SUBTYPE_SSL_VERSION, strlen(value), value); |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 1047 | } |
Dave McCowan | 328fb58 | 2014-07-30 10:39:13 -0400 | [diff] [blame] | 1048 | if (ssl_sock_get_cert_used_sess(remote)) { |
| 1049 | tlv->client |= PP2_CLIENT_CERT_SESS; |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 1050 | tlv->verify = htonl(ssl_sock_get_verify_result(remote)); |
Dave McCowan | 328fb58 | 2014-07-30 10:39:13 -0400 | [diff] [blame] | 1051 | if (ssl_sock_get_cert_used_conn(remote)) |
| 1052 | tlv->client |= PP2_CLIENT_CERT_CONN; |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 1053 | } |
| 1054 | if (srv->pp_opts & SRV_PP_V2_SSL_CN) { |
Emmanuel Hocdet | 404d978 | 2017-10-24 10:55:14 +0200 | [diff] [blame] | 1055 | struct chunk *cn_trash = get_trash_chunk(); |
Willy Tarreau | 3b9a0c9 | 2014-07-19 06:37:33 +0200 | [diff] [blame] | 1056 | if (ssl_sock_get_remote_common_name(remote, cn_trash) > 0) { |
Emmanuel Hocdet | 58118b4 | 2017-10-13 12:15:28 +0200 | [diff] [blame] | 1057 | ssl_tlv_len += make_tlv(&buf[ret+ssl_tlv_len], (buf_len - ret - ssl_tlv_len), PP2_SUBTYPE_SSL_CN, cn_trash->len, cn_trash->str); |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 1058 | } |
| 1059 | } |
| 1060 | } |
| 1061 | tlv->tlv.length_hi = (uint16_t)(ssl_tlv_len - sizeof(struct tlv)) >> 8; |
| 1062 | tlv->tlv.length_lo = (uint16_t)(ssl_tlv_len - sizeof(struct tlv)) & 0x00ff; |
| 1063 | ret += ssl_tlv_len; |
| 1064 | } |
| 1065 | #endif |
| 1066 | |
KOVACS Krisztian | b3e54fe | 2014-11-17 15:11:45 +0100 | [diff] [blame] | 1067 | #ifdef CONFIG_HAP_NS |
| 1068 | if (remote && (remote->proxy_netns)) { |
| 1069 | if ((buf_len - ret) < sizeof(struct tlv)) |
| 1070 | return 0; |
| 1071 | ret += make_tlv(&buf[ret], buf_len, PP2_TYPE_NETNS, remote->proxy_netns->name_len, remote->proxy_netns->node.key); |
| 1072 | } |
| 1073 | #endif |
| 1074 | |
Willy Tarreau | 8fccfa2 | 2014-06-14 08:28:06 +0200 | [diff] [blame] | 1075 | hdr->len = htons((uint16_t)(ret - PP2_HEADER_LEN)); |
David S | afb7683 | 2014-05-08 23:42:08 -0400 | [diff] [blame] | 1076 | |
| 1077 | return ret; |
| 1078 | } |
Emeric Brun | 4f60301 | 2017-01-05 15:11:44 +0100 | [diff] [blame] | 1079 | |
| 1080 | /* fetch if the received connection used a PROXY protocol header */ |
| 1081 | int smp_fetch_fc_rcvd_proxy(const struct arg *args, struct sample *smp, const char *kw, void *private) |
| 1082 | { |
| 1083 | struct connection *conn; |
| 1084 | |
| 1085 | conn = objt_conn(smp->sess->origin); |
| 1086 | if (!conn) |
| 1087 | return 0; |
| 1088 | |
| 1089 | if (!(conn->flags & CO_FL_CONNECTED)) { |
| 1090 | smp->flags |= SMP_F_MAY_CHANGE; |
| 1091 | return 0; |
| 1092 | } |
| 1093 | |
| 1094 | smp->flags = 0; |
| 1095 | smp->data.type = SMP_T_BOOL; |
| 1096 | smp->data.u.sint = (conn->flags & CO_FL_RCVD_PROXY) ? 1 : 0; |
| 1097 | |
| 1098 | return 1; |
| 1099 | } |
| 1100 | |
| 1101 | /* Note: must not be declared <const> as its list will be overwritten. |
| 1102 | * Note: fetches that may return multiple types must be declared as the lowest |
| 1103 | * common denominator, the type that can be casted into all other ones. For |
| 1104 | * instance v4/v6 must be declared v4. |
| 1105 | */ |
| 1106 | static struct sample_fetch_kw_list sample_fetch_keywords = {ILH, { |
| 1107 | { "fc_rcvd_proxy", smp_fetch_fc_rcvd_proxy, 0, NULL, SMP_T_BOOL, SMP_USE_L4CLI }, |
| 1108 | { /* END */ }, |
| 1109 | }}; |
| 1110 | |
| 1111 | |
| 1112 | __attribute__((constructor)) |
| 1113 | static void __connection_init(void) |
| 1114 | { |
| 1115 | sample_register_fetches(&sample_fetch_keywords); |
| 1116 | } |