hammer2 - Implement automatic reconnect
[dragonfly.git] / sbin / hammer2 / cmd_service.c
CommitLineData
9ab15106
MD
1/*
2 * Copyright (c) 2011-2012 The DragonFly Project. All rights reserved.
3 *
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@dragonflybsd.org>
6 * by Venkatesh Srinivas <vsrinivas@dragonflybsd.org>
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 *
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in
16 * the documentation and/or other materials provided with the
17 * distribution.
18 * 3. Neither the name of The DragonFly Project nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific, prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
25 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
26 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
27 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
28 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
29 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
30 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
31 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
32 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
34 */
35
36#include "hammer2.h"
37
62efe6ec 38static void *master_accept(void *data);
29ead430
MD
39static void master_auth_signal(hammer2_router_t *router);
40static void master_auth_rxmsg(hammer2_msg_t *msg);
41static void master_link_signal(hammer2_router_t *router);
42static void master_link_rxmsg(hammer2_msg_t *msg);
e1648a68 43static void master_reconnect(const char *mntpt);
9ab15106
MD
44
45/*
46 * Start-up the master listener daemon for the machine.
47 *
48 * The master listener serves as a rendezvous point in the cluster, accepting
49 * connections, performing registrations and authentications, maintaining
50 * the spanning tree, and keeping track of message state so disconnects can
51 * be handled properly.
52 *
53 * Once authenticated only low-level messaging protocols (which includes
54 * tracking persistent messages) are handled by this daemon. This daemon
55 * does not run the higher level quorum or locking protocols.
56 *
57 * This daemon can also be told to maintain connections to other nodes,
58 * forming a messaging backbone, which in turn allows PFS's (if desired) to
59 * simply connect to the master daemon via localhost if desired.
60 * Backbones are specified via /etc/hammer2.conf.
61 */
62int
62efe6ec 63cmd_service(void)
9ab15106
MD
64{
65 struct sockaddr_in lsin;
66 int on;
67 int lfd;
68
69 /*
70 * Acquire socket and set options
71 */
72 if ((lfd = socket(AF_INET, SOCK_STREAM, 0)) < 0) {
62efe6ec 73 fprintf(stderr, "master_listen: socket(): %s\n",
9ab15106
MD
74 strerror(errno));
75 return 1;
76 }
77 on = 1;
78 setsockopt(lfd, SOL_SOCKET, SO_REUSEADDR, &on, sizeof(on));
79
80 /*
81 * Setup listen port and try to bind. If the bind fails we assume
82 * that a master listener process is already running and silently
83 * fail.
84 */
85 bzero(&lsin, sizeof(lsin));
86 lsin.sin_family = AF_INET;
87 lsin.sin_addr.s_addr = INADDR_ANY;
88 lsin.sin_port = htons(HAMMER2_LISTEN_PORT);
89 if (bind(lfd, (struct sockaddr *)&lsin, sizeof(lsin)) < 0) {
90 close(lfd);
9b8b748f
MD
91 if (QuietOpt == 0) {
92 fprintf(stderr,
93 "master listen: daemon already running\n");
94 }
9ab15106
MD
95 return 0;
96 }
9b8b748f
MD
97 if (QuietOpt == 0)
98 fprintf(stderr, "master listen: startup\n");
9ab15106
MD
99 listen(lfd, 50);
100
101 /*
102 * Fork and disconnect the controlling terminal and parent process,
103 * executing the specified function as a pthread.
104 *
105 * Returns to the original process which can then continue running.
106 * In debug mode this call will create the pthread without forking
107 * and set NormalExit to 0, instead of fork.
108 */
62efe6ec 109 hammer2_demon(master_accept, (void *)(intptr_t)lfd);
9ab15106
MD
110 if (NormalExit)
111 close(lfd);
112 return 0;
113}
114
115/*
116 * Master listen/accept thread. Accept connections on the master socket,
117 * starting a pthread for each one.
118 */
119static
120void *
62efe6ec 121master_accept(void *data)
9ab15106
MD
122{
123 struct sockaddr_in asin;
124 socklen_t alen;
125 pthread_t thread;
e1648a68 126 hammer2_master_service_info_t *info;
9ab15106
MD
127 int lfd = (int)(intptr_t)data;
128 int fd;
e1648a68
MD
129 int i;
130 int count;
131 struct statfs *mntbuf = NULL;
132 struct statvfs *mntvbuf = NULL;
9ab15106
MD
133
134 /*
135 * Nobody waits for us
136 */
137 setproctitle("hammer2 master listen");
138 pthread_detach(pthread_self());
139
140 /*
e1648a68
MD
141 * Scan existing hammer2 mounts and reconnect to them using
142 * HAMMER2IOC_RECLUSTER.
143 */
144 count = getmntvinfo(&mntbuf, &mntvbuf, MNT_NOWAIT);
145 for (i = 0; i < count; ++i) {
146 if (strcmp(mntbuf[i].f_fstypename, "hammer2") == 0)
147 master_reconnect(mntbuf[i].f_mntonname);
148 }
149
150 /*
9ab15106
MD
151 * Accept connections and create pthreads to handle them after
152 * validating the IP.
153 */
154 for (;;) {
155 alen = sizeof(asin);
156 fd = accept(lfd, (struct sockaddr *)&asin, &alen);
157 if (fd < 0) {
158 if (errno == EINTR)
159 continue;
160 break;
161 }
162 thread = NULL;
62efe6ec 163 fprintf(stderr, "master_accept: accept fd %d\n", fd);
e1648a68
MD
164 info = malloc(sizeof(*info));
165 bzero(info, sizeof(*info));
166 info->fd = fd;
167 info->detachme = 1;
168 pthread_create(&thread, NULL, master_service, info);
9ab15106
MD
169 }
170 return (NULL);
171}
172
173/*
e1648a68
MD
174 * Normally the mount program supplies a cluster communications
175 * descriptor to the hammer2 vfs on mount, but if you kill the service
176 * daemon and restart it that link will be lost.
177 *
178 * This procedure attempts to [re]connect to existing mounts when
179 * the service daemon is started up before going into its accept
180 * loop.
eae0d690
MD
181 *
182 * NOTE: A hammer2 mount point can only accomodate one connection at a time
183 * so this will disconnect any existing connection during the
184 * reconnect.
e1648a68
MD
185 */
186static
187void
188master_reconnect(const char *mntpt)
189{
190 struct hammer2_ioc_recluster recls;
191 hammer2_master_service_info_t *info;
192 pthread_t thread;
193 int fd;
194 int pipefds[2];
195
196 fd = open(mntpt, O_RDONLY);
197 if (fd < 0) {
198 fprintf(stderr, "reconnect %s: no access to mount\n", mntpt);
199 return;
200 }
201 if (pipe(pipefds) < 0) {
202 fprintf(stderr, "reconnect %s: pipe() failed\n", mntpt);
203 return;
204 }
205 bzero(&recls, sizeof(recls));
206 recls.fd = pipefds[0];
207 if (ioctl(fd, HAMMER2IOC_RECLUSTER, &recls) < 0) {
208 fprintf(stderr, "reconnect %s: ioctl failed\n", mntpt);
209 close(pipefds[0]);
210 close(pipefds[1]);
211 close(fd);
212 return;
213 }
214 close(pipefds[0]);
215
216 info = malloc(sizeof(*info));
217 bzero(info, sizeof(*info));
218 info->fd = pipefds[1];
219 info->detachme = 1;
220 pthread_create(&thread, NULL, master_service, info);
221}
222
223/*
9ab15106 224 * Service an accepted connection (runs as a pthread)
02454b3e
MD
225 *
226 * (also called from a couple of other places)
9ab15106 227 */
9ab15106 228void *
62efe6ec 229master_service(void *data)
9ab15106 230{
e1648a68 231 hammer2_master_service_info_t *info = data;
9ab15106 232 hammer2_iocom_t iocom;
9ab15106 233
e1648a68
MD
234 if (info->detachme)
235 pthread_detach(pthread_self());
236
237 hammer2_iocom_init(&iocom, info->fd, -1,
29ead430
MD
238 master_auth_signal,
239 master_auth_rxmsg,
240 NULL);
5903497c 241 hammer2_iocom_core(&iocom);
9ab15106
MD
242
243 fprintf(stderr,
244 "iocom on fd %d terminated error rx=%d, tx=%d\n",
e1648a68
MD
245 info->fd, iocom.ioq_rx.error, iocom.ioq_tx.error);
246 close(info->fd);
247 info->fd = -1; /* safety */
248 free(info);
9ab15106
MD
249
250 return (NULL);
251}
252
62efe6ec
MD
253/************************************************************************
254 * AUTHENTICATION *
255 ************************************************************************
256 *
1b195a98
MD
257 * Callback via hammer2_iocom_core().
258 *
62efe6ec
MD
259 * Additional messaging-based authentication must occur before normal
260 * message operation. The connection has already been encrypted at
261 * this point.
262 */
29ead430 263static void master_auth_conn_rx(hammer2_msg_t *msg);
5903497c
MD
264
265static
266void
29ead430 267master_auth_signal(hammer2_router_t *router)
5903497c
MD
268{
269 hammer2_msg_t *msg;
270
271 /*
272 * Transmit LNK_CONN, enabling the SPAN protocol if both sides
273 * agree.
274 *
275 * XXX put additional authentication states here
276 */
29ead430
MD
277 msg = hammer2_msg_alloc(router, 0, HAMMER2_LNK_CONN |
278 HAMMER2_MSGF_CREATE,
279 master_auth_conn_rx, NULL);
5903497c 280 snprintf(msg->any.lnk_conn.label, sizeof(msg->any.lnk_conn.label), "*");
29ead430 281 hammer2_msg_write(msg);
5903497c 282
29ead430
MD
283 hammer2_router_restate(router,
284 master_link_signal,
285 master_link_rxmsg,
286 NULL);
5903497c
MD
287}
288
62efe6ec
MD
289static
290void
29ead430 291master_auth_conn_rx(hammer2_msg_t *msg)
62efe6ec 292{
5903497c 293 if (msg->any.head.cmd & HAMMER2_MSGF_DELETE)
29ead430 294 hammer2_msg_reply(msg, 0);
62efe6ec
MD
295}
296
297static
298void
29ead430 299master_auth_rxmsg(hammer2_msg_t *msg __unused)
62efe6ec 300{
62efe6ec
MD
301}
302
1b195a98
MD
303/************************************************************************
304 * POST-AUTHENTICATION SERVICE MSGS *
305 ************************************************************************
306 *
307 * Callback via hammer2_iocom_core().
9ab15106
MD
308 */
309static
310void
29ead430 311master_link_signal(hammer2_router_t *router)
9ab15106 312{
29ead430 313 hammer2_msg_lnk_signal(router);
9ab15106
MD
314}
315
9ab15106
MD
316static
317void
29ead430 318master_link_rxmsg(hammer2_msg_t *msg)
9ab15106 319{
5903497c
MD
320 hammer2_state_t *state;
321 uint32_t cmd;
322
323 /*
324 * If the message state has a function established we just
325 * call the function, otherwise we call the appropriate
326 * link-level protocol related to the original command and
327 * let it sort it out.
328 *
329 * Non-transactional one-off messages, on the otherhand,
330 * might have REPLY set.
331 */
332 state = msg->state;
81666e1b
MD
333 cmd = state ? state->msg->any.head.cmd : msg->any.head.cmd;
334
335 fprintf(stderr, "service-receive: %s\n", hammer2_msg_str(msg));
336
5903497c
MD
337 if (state && state->func) {
338 assert(state->func != NULL);
29ead430 339 state->func(msg);
5903497c
MD
340 } else {
341 switch(cmd & HAMMER2_MSGF_PROTOS) {
342 case HAMMER2_MSG_PROTO_LNK:
29ead430 343 hammer2_msg_lnk(msg);
5903497c
MD
344 break;
345 case HAMMER2_MSG_PROTO_DBG:
29ead430 346 hammer2_msg_dbg(msg);
5903497c
MD
347 break;
348 default:
29ead430 349 hammer2_msg_reply(msg, HAMMER2_MSG_ERR_NOSUPP);
5903497c
MD
350 break;
351 }
352 }
9ab15106 353}