]>
Commit | Line | Data |
---|---|---|
0ad19a3f | 1 | /* |
2 | * lxc: linux Container library | |
3 | * | |
4 | * (C) Copyright IBM Corp. 2007, 2008 | |
5 | * | |
6 | * Authors: | |
9afe19d6 | 7 | * Daniel Lezcano <daniel.lezcano at free.fr> |
e51d4895 | 8 | * Dwight Engen <dwight.engen@oracle.com> |
0ad19a3f | 9 | * |
10 | * This library is free software; you can redistribute it and/or | |
11 | * modify it under the terms of the GNU Lesser General Public | |
12 | * License as published by the Free Software Foundation; either | |
13 | * version 2.1 of the License, or (at your option) any later version. | |
14 | * | |
15 | * This library is distributed in the hope that it will be useful, | |
16 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
18 | * Lesser General Public License for more details. | |
19 | * | |
20 | * You should have received a copy of the GNU Lesser General Public | |
21 | * License along with this library; if not, write to the Free Software | |
250b1eec | 22 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
0ad19a3f | 23 | */ |
e51d4895 | 24 | |
0ad19a3f | 25 | #include <stdio.h> |
26 | #include <errno.h> | |
27 | #include <unistd.h> | |
28 | #include <string.h> | |
29 | #include <stdlib.h> | |
aae93dd3 | 30 | #include <stddef.h> |
0ad19a3f | 31 | #include <fcntl.h> |
b45c7011 DE |
32 | #include <inttypes.h> |
33 | #include <stdint.h> | |
0ad19a3f | 34 | #include <sys/types.h> |
35 | #include <sys/stat.h> | |
36 | #include <sys/param.h> | |
0ad19a3f | 37 | #include <sys/socket.h> |
e51d4895 | 38 | #include <sys/wait.h> |
0ad19a3f | 39 | #include <netinet/in.h> |
40 | #include <net/if.h> | |
b113348e | 41 | |
e2bcd7db | 42 | #include "error.h" |
31c53c2e | 43 | #include "af_unix.h" |
f2363e38 ÇO |
44 | #include "log.h" |
45 | #include "lxclock.h" | |
46 | #include "state.h" | |
47 | #include "monitor.h" | |
48 | #include "utils.h" | |
36eb9bde CLG |
49 | |
50 | lxc_log_define(lxc_monitor, lxc); | |
0ad19a3f | 51 | |
e51d4895 | 52 | /* routines used by monitor publishers (containers) */ |
9e60f51d DE |
53 | int lxc_monitor_fifo_name(const char *lxcpath, char *fifo_path, size_t fifo_path_sz, |
54 | int do_mkdirp) | |
55 | { | |
56 | int ret; | |
57 | const char *rundir; | |
58 | ||
59 | rundir = get_rundir(); | |
97a696c6 SG |
60 | if (!rundir) |
61 | return -1; | |
62 | ||
9e60f51d DE |
63 | if (do_mkdirp) { |
64 | ret = snprintf(fifo_path, fifo_path_sz, "%s/lxc/%s", rundir, lxcpath); | |
65 | if (ret < 0 || ret >= fifo_path_sz) { | |
66 | ERROR("rundir/lxcpath (%s/%s) too long for monitor fifo", rundir, lxcpath); | |
67 | return -1; | |
68 | } | |
9e60f51d | 69 | ret = mkdir_p(fifo_path, 0755); |
9e60f51d DE |
70 | if (ret < 0) { |
71 | ERROR("unable to create monitor fifo dir %s", fifo_path); | |
72 | return ret; | |
73 | } | |
74 | } | |
75 | ret = snprintf(fifo_path, fifo_path_sz, "%s/lxc/%s/monitor-fifo", rundir, lxcpath); | |
76 | if (ret < 0 || ret >= fifo_path_sz) { | |
77 | ERROR("rundir/lxcpath (%s/%s) too long for monitor fifo", rundir, lxcpath); | |
78 | return -1; | |
79 | } | |
80 | return 0; | |
81 | } | |
82 | ||
e51d4895 | 83 | static void lxc_monitor_fifo_send(struct lxc_msg *msg, const char *lxcpath) |
0ad19a3f | 84 | { |
e51d4895 DE |
85 | int fd,ret; |
86 | char fifo_path[PATH_MAX]; | |
87 | ||
88 | BUILD_BUG_ON(sizeof(*msg) > PIPE_BUF); /* write not guaranteed atomic */ | |
9e60f51d DE |
89 | |
90 | ret = lxc_monitor_fifo_name(lxcpath, fifo_path, sizeof(fifo_path), 0); | |
91 | if (ret < 0) | |
9123e471 | 92 | return; |
80f41298 | 93 | |
e51d4895 DE |
94 | fd = open(fifo_path, O_WRONLY); |
95 | if (fd < 0) { | |
96 | /* it is normal for this open to fail when there is no monitor | |
97 | * running, so we don't log it | |
98 | */ | |
31c53c2e | 99 | return; |
e51d4895 | 100 | } |
0ad19a3f | 101 | |
e51d4895 DE |
102 | ret = write(fd, msg, sizeof(*msg)); |
103 | if (ret != sizeof(*msg)) { | |
e8b9ac8f | 104 | close(fd); |
e51d4895 DE |
105 | SYSERROR("failed to write monitor fifo %s", fifo_path); |
106 | return; | |
107 | } | |
0ad19a3f | 108 | |
109 | close(fd); | |
110 | } | |
111 | ||
9123e471 | 112 | void lxc_monitor_send_state(const char *name, lxc_state_t state, const char *lxcpath) |
eae6543d | 113 | { |
114 | struct lxc_msg msg = { .type = lxc_msg_state, | |
115 | .value = state }; | |
80f41298 | 116 | strncpy(msg.name, name, sizeof(msg.name)); |
f3bc28bd | 117 | msg.name[sizeof(msg.name) - 1] = 0; |
eae6543d | 118 | |
e51d4895 | 119 | lxc_monitor_fifo_send(&msg, lxcpath); |
0ad19a3f | 120 | } |
121 | ||
e51d4895 DE |
122 | |
123 | /* routines used by monitor subscribers (lxc-monitor) */ | |
124 | int lxc_monitor_close(int fd) | |
0ad19a3f | 125 | { |
dd1d77f9 | 126 | return close(fd); |
e51d4895 DE |
127 | } |
128 | ||
129 | int lxc_monitor_sock_name(const char *lxcpath, struct sockaddr_un *addr) { | |
130 | size_t len; | |
131 | int ret; | |
b45c7011 DE |
132 | char *sockname = &addr->sun_path[1]; |
133 | char path[PATH_MAX+18]; | |
134 | uint64_t hash; | |
e51d4895 | 135 | |
b45c7011 DE |
136 | /* addr.sun_path is only 108 bytes, so we hash the full name and |
137 | * then append as much of the name as we can fit. | |
9123e471 | 138 | */ |
e51d4895 DE |
139 | memset(addr, 0, sizeof(*addr)); |
140 | addr->sun_family = AF_UNIX; | |
141 | len = sizeof(addr->sun_path) - 1; | |
b45c7011 DE |
142 | ret = snprintf(path, sizeof(path), "lxc/%s/monitor-sock", lxcpath); |
143 | if (ret < 0 || ret >= sizeof(path)) { | |
144 | ERROR("lxcpath %s too long for monitor unix socket", lxcpath); | |
9e60f51d DE |
145 | return -1; |
146 | } | |
9e60f51d | 147 | |
b45c7011 DE |
148 | hash = fnv_64a_buf(path, ret, FNV1A_64_INIT); |
149 | ret = snprintf(sockname, len, "lxc/%016" PRIx64 "/%s", hash, lxcpath); | |
150 | if (ret < 0) | |
9123e471 | 151 | return -1; |
b45c7011 DE |
152 | sockname[sizeof(addr->sun_path)-2] = '\0'; |
153 | INFO("using monitor sock name %s", sockname); | |
e51d4895 DE |
154 | return 0; |
155 | } | |
0ad19a3f | 156 | |
e51d4895 DE |
157 | int lxc_monitor_open(const char *lxcpath) |
158 | { | |
159 | struct sockaddr_un addr; | |
160 | int fd,ret; | |
161 | int retry,backoff_ms[] = {10, 50, 100}; | |
aae93dd3 | 162 | size_t len; |
e51d4895 DE |
163 | |
164 | if (lxc_monitor_sock_name(lxcpath, &addr) < 0) | |
165 | return -1; | |
166 | ||
167 | fd = socket(PF_UNIX, SOCK_STREAM, 0); | |
2c396e12 MN |
168 | if (fd < 0) { |
169 | ERROR("socket : %s", strerror(errno)); | |
31c53c2e | 170 | return -1; |
2c396e12 | 171 | } |
0ad19a3f | 172 | |
aae93dd3 ÇO |
173 | len = strlen(&addr.sun_path[1]) + 1; |
174 | if (len >= sizeof(addr.sun_path) - 1) { | |
175 | ret = -1; | |
176 | errno = ENAMETOOLONG; | |
177 | goto err1; | |
178 | } | |
179 | ||
e51d4895 | 180 | for (retry = 0; retry < sizeof(backoff_ms)/sizeof(backoff_ms[0]); retry++) { |
aae93dd3 | 181 | ret = connect(fd, (struct sockaddr *)&addr, offsetof(struct sockaddr_un, sun_path) + len); |
e51d4895 DE |
182 | if (ret == 0 || errno != ECONNREFUSED) |
183 | break; | |
184 | ERROR("connect : backing off %d", backoff_ms[retry]); | |
185 | usleep(backoff_ms[retry] * 1000); | |
0ad19a3f | 186 | } |
187 | ||
e51d4895 DE |
188 | if (ret < 0) { |
189 | ERROR("connect : %s", strerror(errno)); | |
190 | goto err1; | |
191 | } | |
0ad19a3f | 192 | return fd; |
e51d4895 DE |
193 | err1: |
194 | close(fd); | |
195 | return ret; | |
0ad19a3f | 196 | } |
197 | ||
8d06bd13 DE |
198 | int lxc_monitor_read_fdset(fd_set *rfds, int nfds, struct lxc_msg *msg, |
199 | int timeout) | |
0ad19a3f | 200 | { |
8d06bd13 DE |
201 | struct timeval tval,*tv = NULL; |
202 | int ret,i; | |
72d0e1cb SG |
203 | |
204 | if (timeout != -1) { | |
8d06bd13 DE |
205 | tv = &tval; |
206 | tv->tv_sec = timeout; | |
207 | tv->tv_usec = 0; | |
72d0e1cb | 208 | } |
0ad19a3f | 209 | |
8d06bd13 DE |
210 | ret = select(nfds, rfds, NULL, NULL, tv); |
211 | if (ret == -1) | |
75b1e198 | 212 | return -1; |
8d06bd13 DE |
213 | else if (ret == 0) |
214 | return -2; // timed out | |
215 | ||
216 | /* only read from the first ready fd, the others will remain ready | |
217 | * for when this routine is called again | |
218 | */ | |
219 | for (i = 0; i < nfds; i++) { | |
220 | if (FD_ISSET(i, rfds)) { | |
221 | ret = recv(i, msg, sizeof(*msg), 0); | |
222 | if (ret <= 0) { | |
223 | SYSERROR("client failed to recv (monitord died?) %s", | |
224 | strerror(errno)); | |
225 | return -1; | |
226 | } | |
227 | return ret; | |
228 | } | |
0ad19a3f | 229 | } |
8d06bd13 DE |
230 | SYSERROR("no ready fd found?"); |
231 | return -1; | |
232 | } | |
233 | ||
234 | int lxc_monitor_read_timeout(int fd, struct lxc_msg *msg, int timeout) | |
235 | { | |
236 | fd_set rfds; | |
237 | ||
238 | FD_ZERO(&rfds); | |
239 | FD_SET(fd, &rfds); | |
240 | ||
241 | return lxc_monitor_read_fdset(&rfds, fd+1, msg, timeout); | |
0ad19a3f | 242 | } |
243 | ||
72d0e1cb SG |
244 | int lxc_monitor_read(int fd, struct lxc_msg *msg) |
245 | { | |
246 | return lxc_monitor_read_timeout(fd, msg, -1); | |
247 | } | |
248 | ||
e51d4895 | 249 | |
45e854dc | 250 | #define LXC_MONITORD_PATH LIBEXECDIR "/lxc/lxc-monitord" |
e51d4895 DE |
251 | |
252 | /* used to spawn a monitord either on startup of a daemon container, or when | |
253 | * lxc-monitor starts | |
254 | */ | |
255 | int lxc_monitord_spawn(const char *lxcpath) | |
0ad19a3f | 256 | { |
e51d4895 DE |
257 | pid_t pid1,pid2; |
258 | int pipefd[2]; | |
259 | char pipefd_str[11]; | |
260 | ||
261 | char * const args[] = { | |
45e854dc | 262 | LXC_MONITORD_PATH, |
e51d4895 DE |
263 | (char *)lxcpath, |
264 | pipefd_str, | |
265 | NULL, | |
266 | }; | |
267 | ||
268 | /* double fork to avoid zombies when monitord exits */ | |
269 | pid1 = fork(); | |
270 | if (pid1 < 0) { | |
271 | SYSERROR("failed to fork"); | |
272 | return -1; | |
273 | } | |
274 | ||
275 | if (pid1) { | |
f2bbe86d DE |
276 | if (waitpid(pid1, NULL, 0) != pid1) |
277 | return -1; | |
e51d4895 DE |
278 | return 0; |
279 | } | |
280 | ||
281 | if (pipe(pipefd) < 0) { | |
282 | SYSERROR("failed to create pipe"); | |
283 | exit(EXIT_FAILURE); | |
284 | } | |
285 | ||
286 | pid2 = fork(); | |
287 | if (pid2 < 0) { | |
288 | SYSERROR("failed to fork"); | |
289 | exit(EXIT_FAILURE); | |
290 | } | |
291 | if (pid2) { | |
292 | char c; | |
293 | /* wait for daemon to create socket */ | |
294 | close(pipefd[1]); | |
295 | /* sync with child, we're ignoring the return from read | |
296 | * because regardless if it works or not, either way we've | |
297 | * synced with the child process. the if-empty-statement | |
298 | * construct is to quiet the warn-unused-result warning. | |
299 | */ | |
8f47bc3f SG |
300 | if (read(pipefd[0], &c, 1)) |
301 | ; | |
e51d4895 DE |
302 | close(pipefd[0]); |
303 | exit(EXIT_SUCCESS); | |
304 | } | |
305 | ||
e51d4895 DE |
306 | if (setsid() < 0) { |
307 | SYSERROR("failed to setsid"); | |
308 | exit(EXIT_FAILURE); | |
309 | } | |
310 | close(0); | |
311 | close(1); | |
312 | close(2); | |
313 | open("/dev/null", O_RDONLY); | |
314 | open("/dev/null", O_RDWR); | |
315 | open("/dev/null", O_RDWR); | |
316 | close(pipefd[0]); | |
317 | sprintf(pipefd_str, "%d", pipefd[1]); | |
318 | execvp(args[0], args); | |
319 | exit(EXIT_FAILURE); | |
0ad19a3f | 320 | } |