]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * lxc: linux Container library | |
3 | * | |
4 | * (C) Copyright IBM Corp. 2007, 2008 | |
5 | * | |
6 | * Authors: | |
7 | * Daniel Lezcano <dlezcano at fr.ibm.com> | |
8 | * | |
9 | * This library is free software; you can redistribute it and/or | |
10 | * modify it under the terms of the GNU Lesser General Public | |
11 | * License as published by the Free Software Foundation; either | |
12 | * version 2.1 of the License, or (at your option) any later version. | |
13 | * | |
14 | * This library is distributed in the hope that it will be useful, | |
15 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
16 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
17 | * Lesser General Public License for more details. | |
18 | * | |
19 | * You should have received a copy of the GNU Lesser General Public | |
20 | * License along with this library; if not, write to the Free Software | |
21 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
22 | */ | |
23 | #define _GNU_SOURCE | |
24 | #include <stdio.h> | |
25 | #undef _GNU_SOURCE | |
26 | #include <stdlib.h> | |
27 | #include <stdarg.h> | |
28 | #include <errno.h> | |
29 | #include <string.h> | |
30 | #include <dirent.h> | |
31 | #include <unistd.h> | |
32 | #include <sys/wait.h> | |
33 | #include <sys/syscall.h> | |
34 | ||
35 | #if HAVE_PTY_H | |
36 | #include <pty.h> | |
37 | #else | |
38 | #include <../include/openpty.h> | |
39 | #endif | |
40 | ||
41 | #include <linux/loop.h> | |
42 | ||
43 | #include <sys/types.h> | |
44 | #include <sys/utsname.h> | |
45 | #include <sys/param.h> | |
46 | #include <sys/stat.h> | |
47 | #include <sys/socket.h> | |
48 | #include <sys/mount.h> | |
49 | #include <sys/mman.h> | |
50 | #include <sys/prctl.h> | |
51 | ||
52 | #include <arpa/inet.h> | |
53 | #include <fcntl.h> | |
54 | #include <netinet/in.h> | |
55 | #include <net/if.h> | |
56 | #include <libgen.h> | |
57 | ||
58 | #include "network.h" | |
59 | #include "error.h" | |
60 | #include "parse.h" | |
61 | #include "config.h" | |
62 | #include "utils.h" | |
63 | #include "conf.h" | |
64 | #include "log.h" | |
65 | #include "lxc.h" /* for lxc_cgroup_set() */ | |
66 | #include "caps.h" /* for lxc_caps_last_cap() */ | |
67 | ||
68 | #if HAVE_APPARMOR | |
69 | #include <apparmor.h> | |
70 | #endif | |
71 | ||
72 | #if HAVE_SYS_CAPABILITY_H | |
73 | #include <sys/capability.h> | |
74 | #endif | |
75 | ||
76 | #if HAVE_SYS_PERSONALITY_H | |
77 | #include <sys/personality.h> | |
78 | #endif | |
79 | ||
80 | #if IS_BIONIC | |
81 | #include <../include/lxcmntent.h> | |
82 | #else | |
83 | #include <mntent.h> | |
84 | #endif | |
85 | ||
86 | #include "lxcseccomp.h" | |
87 | ||
88 | lxc_log_define(lxc_conf, lxc); | |
89 | ||
90 | #define MAXHWLEN 18 | |
91 | #define MAXINDEXLEN 20 | |
92 | #define MAXMTULEN 16 | |
93 | #define MAXLINELEN 128 | |
94 | ||
95 | #ifndef MS_DIRSYNC | |
96 | #define MS_DIRSYNC 128 | |
97 | #endif | |
98 | ||
99 | #ifndef MS_REC | |
100 | #define MS_REC 16384 | |
101 | #endif | |
102 | ||
103 | #ifndef MNT_DETACH | |
104 | #define MNT_DETACH 2 | |
105 | #endif | |
106 | ||
107 | #ifndef MS_SLAVE | |
108 | #define MS_SLAVE (1<<19) | |
109 | #endif | |
110 | ||
111 | #ifndef MS_RELATIME | |
112 | #define MS_RELATIME (1 << 21) | |
113 | #endif | |
114 | ||
115 | #ifndef MS_STRICTATIME | |
116 | #define MS_STRICTATIME (1 << 24) | |
117 | #endif | |
118 | ||
119 | #if HAVE_SYS_CAPABILITY_H | |
120 | #ifndef CAP_SETFCAP | |
121 | #define CAP_SETFCAP 31 | |
122 | #endif | |
123 | ||
124 | #ifndef CAP_MAC_OVERRIDE | |
125 | #define CAP_MAC_OVERRIDE 32 | |
126 | #endif | |
127 | ||
128 | #ifndef CAP_MAC_ADMIN | |
129 | #define CAP_MAC_ADMIN 33 | |
130 | #endif | |
131 | #endif | |
132 | ||
133 | #ifndef PR_CAPBSET_DROP | |
134 | #define PR_CAPBSET_DROP 24 | |
135 | #endif | |
136 | ||
137 | #ifndef LO_FLAGS_AUTOCLEAR | |
138 | #define LO_FLAGS_AUTOCLEAR 4 | |
139 | #endif | |
140 | ||
141 | /* Define pivot_root() if missing from the C library */ | |
142 | #ifndef HAVE_PIVOT_ROOT | |
143 | static int pivot_root(const char * new_root, const char * put_old) | |
144 | { | |
145 | #ifdef __NR_pivot_root | |
146 | return syscall(__NR_pivot_root, new_root, put_old); | |
147 | #else | |
148 | errno = ENOSYS; | |
149 | return -1; | |
150 | #endif | |
151 | } | |
152 | #else | |
153 | extern int pivot_root(const char * new_root, const char * put_old); | |
154 | #endif | |
155 | ||
156 | /* Define sethostname() if missing from the C library */ | |
157 | #ifndef HAVE_SETHOSTNAME | |
158 | static int sethostname(const char * name, size_t len) | |
159 | { | |
160 | #ifdef __NR_sethostname | |
161 | return syscall(__NR_sethostname, name, len); | |
162 | #else | |
163 | errno = ENOSYS; | |
164 | return -1; | |
165 | #endif | |
166 | } | |
167 | #endif | |
168 | ||
169 | /* Define __S_ISTYPE if missing from the C library */ | |
170 | #ifndef __S_ISTYPE | |
171 | #define __S_ISTYPE(mode, mask) (((mode) & S_IFMT) == (mask)) | |
172 | #endif | |
173 | ||
174 | char *lxchook_names[NUM_LXC_HOOKS] = { | |
175 | "pre-start", "pre-mount", "mount", "autodev", "start", "post-stop" }; | |
176 | ||
177 | typedef int (*instanciate_cb)(struct lxc_handler *, struct lxc_netdev *); | |
178 | ||
179 | struct mount_opt { | |
180 | char *name; | |
181 | int clear; | |
182 | int flag; | |
183 | }; | |
184 | ||
185 | struct caps_opt { | |
186 | char *name; | |
187 | int value; | |
188 | }; | |
189 | ||
190 | static int instanciate_veth(struct lxc_handler *, struct lxc_netdev *); | |
191 | static int instanciate_macvlan(struct lxc_handler *, struct lxc_netdev *); | |
192 | static int instanciate_vlan(struct lxc_handler *, struct lxc_netdev *); | |
193 | static int instanciate_phys(struct lxc_handler *, struct lxc_netdev *); | |
194 | static int instanciate_empty(struct lxc_handler *, struct lxc_netdev *); | |
195 | ||
196 | static instanciate_cb netdev_conf[LXC_NET_MAXCONFTYPE + 1] = { | |
197 | [LXC_NET_VETH] = instanciate_veth, | |
198 | [LXC_NET_MACVLAN] = instanciate_macvlan, | |
199 | [LXC_NET_VLAN] = instanciate_vlan, | |
200 | [LXC_NET_PHYS] = instanciate_phys, | |
201 | [LXC_NET_EMPTY] = instanciate_empty, | |
202 | }; | |
203 | ||
204 | static int shutdown_veth(struct lxc_handler *, struct lxc_netdev *); | |
205 | static int shutdown_macvlan(struct lxc_handler *, struct lxc_netdev *); | |
206 | static int shutdown_vlan(struct lxc_handler *, struct lxc_netdev *); | |
207 | static int shutdown_phys(struct lxc_handler *, struct lxc_netdev *); | |
208 | static int shutdown_empty(struct lxc_handler *, struct lxc_netdev *); | |
209 | ||
210 | static instanciate_cb netdev_deconf[LXC_NET_MAXCONFTYPE + 1] = { | |
211 | [LXC_NET_VETH] = shutdown_veth, | |
212 | [LXC_NET_MACVLAN] = shutdown_macvlan, | |
213 | [LXC_NET_VLAN] = shutdown_vlan, | |
214 | [LXC_NET_PHYS] = shutdown_phys, | |
215 | [LXC_NET_EMPTY] = shutdown_empty, | |
216 | }; | |
217 | ||
218 | static struct mount_opt mount_opt[] = { | |
219 | { "defaults", 0, 0 }, | |
220 | { "ro", 0, MS_RDONLY }, | |
221 | { "rw", 1, MS_RDONLY }, | |
222 | { "suid", 1, MS_NOSUID }, | |
223 | { "nosuid", 0, MS_NOSUID }, | |
224 | { "dev", 1, MS_NODEV }, | |
225 | { "nodev", 0, MS_NODEV }, | |
226 | { "exec", 1, MS_NOEXEC }, | |
227 | { "noexec", 0, MS_NOEXEC }, | |
228 | { "sync", 0, MS_SYNCHRONOUS }, | |
229 | { "async", 1, MS_SYNCHRONOUS }, | |
230 | { "dirsync", 0, MS_DIRSYNC }, | |
231 | { "remount", 0, MS_REMOUNT }, | |
232 | { "mand", 0, MS_MANDLOCK }, | |
233 | { "nomand", 1, MS_MANDLOCK }, | |
234 | { "atime", 1, MS_NOATIME }, | |
235 | { "noatime", 0, MS_NOATIME }, | |
236 | { "diratime", 1, MS_NODIRATIME }, | |
237 | { "nodiratime", 0, MS_NODIRATIME }, | |
238 | { "bind", 0, MS_BIND }, | |
239 | { "rbind", 0, MS_BIND|MS_REC }, | |
240 | { "relatime", 0, MS_RELATIME }, | |
241 | { "norelatime", 1, MS_RELATIME }, | |
242 | { "strictatime", 0, MS_STRICTATIME }, | |
243 | { "nostrictatime", 1, MS_STRICTATIME }, | |
244 | { NULL, 0, 0 }, | |
245 | }; | |
246 | ||
247 | #if HAVE_SYS_CAPABILITY_H | |
248 | static struct caps_opt caps_opt[] = { | |
249 | { "chown", CAP_CHOWN }, | |
250 | { "dac_override", CAP_DAC_OVERRIDE }, | |
251 | { "dac_read_search", CAP_DAC_READ_SEARCH }, | |
252 | { "fowner", CAP_FOWNER }, | |
253 | { "fsetid", CAP_FSETID }, | |
254 | { "kill", CAP_KILL }, | |
255 | { "setgid", CAP_SETGID }, | |
256 | { "setuid", CAP_SETUID }, | |
257 | { "setpcap", CAP_SETPCAP }, | |
258 | { "linux_immutable", CAP_LINUX_IMMUTABLE }, | |
259 | { "net_bind_service", CAP_NET_BIND_SERVICE }, | |
260 | { "net_broadcast", CAP_NET_BROADCAST }, | |
261 | { "net_admin", CAP_NET_ADMIN }, | |
262 | { "net_raw", CAP_NET_RAW }, | |
263 | { "ipc_lock", CAP_IPC_LOCK }, | |
264 | { "ipc_owner", CAP_IPC_OWNER }, | |
265 | { "sys_module", CAP_SYS_MODULE }, | |
266 | { "sys_rawio", CAP_SYS_RAWIO }, | |
267 | { "sys_chroot", CAP_SYS_CHROOT }, | |
268 | { "sys_ptrace", CAP_SYS_PTRACE }, | |
269 | { "sys_pacct", CAP_SYS_PACCT }, | |
270 | { "sys_admin", CAP_SYS_ADMIN }, | |
271 | { "sys_boot", CAP_SYS_BOOT }, | |
272 | { "sys_nice", CAP_SYS_NICE }, | |
273 | { "sys_resource", CAP_SYS_RESOURCE }, | |
274 | { "sys_time", CAP_SYS_TIME }, | |
275 | { "sys_tty_config", CAP_SYS_TTY_CONFIG }, | |
276 | { "mknod", CAP_MKNOD }, | |
277 | { "lease", CAP_LEASE }, | |
278 | #ifdef CAP_AUDIT_WRITE | |
279 | { "audit_write", CAP_AUDIT_WRITE }, | |
280 | #endif | |
281 | #ifdef CAP_AUDIT_CONTROL | |
282 | { "audit_control", CAP_AUDIT_CONTROL }, | |
283 | #endif | |
284 | { "setfcap", CAP_SETFCAP }, | |
285 | { "mac_override", CAP_MAC_OVERRIDE }, | |
286 | { "mac_admin", CAP_MAC_ADMIN }, | |
287 | #ifdef CAP_SYSLOG | |
288 | { "syslog", CAP_SYSLOG }, | |
289 | #endif | |
290 | #ifdef CAP_WAKE_ALARM | |
291 | { "wake_alarm", CAP_WAKE_ALARM }, | |
292 | #endif | |
293 | }; | |
294 | #else | |
295 | static struct caps_opt caps_opt[] = {}; | |
296 | #endif | |
297 | ||
298 | static int run_buffer(char *buffer) | |
299 | { | |
300 | FILE *f; | |
301 | char *output; | |
302 | ||
303 | f = popen(buffer, "r"); | |
304 | if (!f) { | |
305 | SYSERROR("popen failed"); | |
306 | return -1; | |
307 | } | |
308 | ||
309 | output = malloc(LXC_LOG_BUFFER_SIZE); | |
310 | if (!output) { | |
311 | ERROR("failed to allocate memory for script output"); | |
312 | return -1; | |
313 | } | |
314 | ||
315 | while(fgets(output, LXC_LOG_BUFFER_SIZE, f)) | |
316 | DEBUG("script output: %s", output); | |
317 | ||
318 | free(output); | |
319 | ||
320 | if (pclose(f) == -1) { | |
321 | SYSERROR("Script exited on error"); | |
322 | return -1; | |
323 | } | |
324 | ||
325 | return 0; | |
326 | } | |
327 | ||
328 | static int run_script(const char *name, const char *section, | |
329 | const char *script, ...) | |
330 | { | |
331 | int ret; | |
332 | char *buffer, *p; | |
333 | size_t size = 0; | |
334 | va_list ap; | |
335 | ||
336 | INFO("Executing script '%s' for container '%s', config section '%s'", | |
337 | script, name, section); | |
338 | ||
339 | va_start(ap, script); | |
340 | while ((p = va_arg(ap, char *))) | |
341 | size += strlen(p) + 1; | |
342 | va_end(ap); | |
343 | ||
344 | size += strlen(script); | |
345 | size += strlen(name); | |
346 | size += strlen(section); | |
347 | size += 3; | |
348 | ||
349 | if (size > INT_MAX) | |
350 | return -1; | |
351 | ||
352 | buffer = alloca(size); | |
353 | if (!buffer) { | |
354 | ERROR("failed to allocate memory"); | |
355 | return -1; | |
356 | } | |
357 | ||
358 | ret = snprintf(buffer, size, "%s %s %s", script, name, section); | |
359 | if (ret < 0 || ret >= size) { | |
360 | ERROR("Script name too long"); | |
361 | free(buffer); | |
362 | return -1; | |
363 | } | |
364 | ||
365 | va_start(ap, script); | |
366 | while ((p = va_arg(ap, char *))) { | |
367 | int len = size-ret; | |
368 | int rc; | |
369 | rc = snprintf(buffer + ret, len, " %s", p); | |
370 | if (rc < 0 || rc >= len) { | |
371 | free(buffer); | |
372 | ERROR("Script args too long"); | |
373 | return -1; | |
374 | } | |
375 | ret += rc; | |
376 | } | |
377 | va_end(ap); | |
378 | ||
379 | return run_buffer(buffer); | |
380 | } | |
381 | ||
382 | static int find_fstype_cb(char* buffer, void *data) | |
383 | { | |
384 | struct cbarg { | |
385 | const char *rootfs; | |
386 | const char *target; | |
387 | int mntopt; | |
388 | } *cbarg = data; | |
389 | ||
390 | char *fstype; | |
391 | ||
392 | /* we don't try 'nodev' entries */ | |
393 | if (strstr(buffer, "nodev")) | |
394 | return 0; | |
395 | ||
396 | fstype = buffer; | |
397 | fstype += lxc_char_left_gc(fstype, strlen(fstype)); | |
398 | fstype[lxc_char_right_gc(fstype, strlen(fstype))] = '\0'; | |
399 | ||
400 | DEBUG("trying to mount '%s'->'%s' with fstype '%s'", | |
401 | cbarg->rootfs, cbarg->target, fstype); | |
402 | ||
403 | if (mount(cbarg->rootfs, cbarg->target, fstype, cbarg->mntopt, NULL)) { | |
404 | DEBUG("mount failed with error: %s", strerror(errno)); | |
405 | return 0; | |
406 | } | |
407 | ||
408 | INFO("mounted '%s' on '%s', with fstype '%s'", | |
409 | cbarg->rootfs, cbarg->target, fstype); | |
410 | ||
411 | return 1; | |
412 | } | |
413 | ||
414 | static int mount_unknow_fs(const char *rootfs, const char *target, int mntopt) | |
415 | { | |
416 | int i; | |
417 | ||
418 | struct cbarg { | |
419 | const char *rootfs; | |
420 | const char *target; | |
421 | int mntopt; | |
422 | } cbarg = { | |
423 | .rootfs = rootfs, | |
424 | .target = target, | |
425 | .mntopt = mntopt, | |
426 | }; | |
427 | ||
428 | /* | |
429 | * find the filesystem type with brute force: | |
430 | * first we check with /etc/filesystems, in case the modules | |
431 | * are auto-loaded and fall back to the supported kernel fs | |
432 | */ | |
433 | char *fsfile[] = { | |
434 | "/etc/filesystems", | |
435 | "/proc/filesystems", | |
436 | }; | |
437 | ||
438 | for (i = 0; i < sizeof(fsfile)/sizeof(fsfile[0]); i++) { | |
439 | ||
440 | int ret; | |
441 | ||
442 | if (access(fsfile[i], F_OK)) | |
443 | continue; | |
444 | ||
445 | ret = lxc_file_for_each_line(fsfile[i], find_fstype_cb, &cbarg); | |
446 | if (ret < 0) { | |
447 | ERROR("failed to parse '%s'", fsfile[i]); | |
448 | return -1; | |
449 | } | |
450 | ||
451 | if (ret) | |
452 | return 0; | |
453 | } | |
454 | ||
455 | ERROR("failed to determine fs type for '%s'", rootfs); | |
456 | return -1; | |
457 | } | |
458 | ||
459 | static int mount_rootfs_dir(const char *rootfs, const char *target) | |
460 | { | |
461 | return mount(rootfs, target, "none", MS_BIND | MS_REC, NULL); | |
462 | } | |
463 | ||
464 | static int setup_lodev(const char *rootfs, int fd, struct loop_info64 *loinfo) | |
465 | { | |
466 | int rfd; | |
467 | int ret = -1; | |
468 | ||
469 | rfd = open(rootfs, O_RDWR); | |
470 | if (rfd < 0) { | |
471 | SYSERROR("failed to open '%s'", rootfs); | |
472 | return -1; | |
473 | } | |
474 | ||
475 | memset(loinfo, 0, sizeof(*loinfo)); | |
476 | ||
477 | loinfo->lo_flags = LO_FLAGS_AUTOCLEAR; | |
478 | ||
479 | if (ioctl(fd, LOOP_SET_FD, rfd)) { | |
480 | SYSERROR("failed to LOOP_SET_FD"); | |
481 | goto out; | |
482 | } | |
483 | ||
484 | if (ioctl(fd, LOOP_SET_STATUS64, loinfo)) { | |
485 | SYSERROR("failed to LOOP_SET_STATUS64"); | |
486 | goto out; | |
487 | } | |
488 | ||
489 | ret = 0; | |
490 | out: | |
491 | close(rfd); | |
492 | ||
493 | return ret; | |
494 | } | |
495 | ||
496 | static int mount_rootfs_file(const char *rootfs, const char *target) | |
497 | { | |
498 | struct dirent dirent, *direntp; | |
499 | struct loop_info64 loinfo; | |
500 | int ret = -1, fd = -1, rc; | |
501 | DIR *dir; | |
502 | char path[MAXPATHLEN]; | |
503 | ||
504 | dir = opendir("/dev"); | |
505 | if (!dir) { | |
506 | SYSERROR("failed to open '/dev'"); | |
507 | return -1; | |
508 | } | |
509 | ||
510 | while (!readdir_r(dir, &dirent, &direntp)) { | |
511 | ||
512 | if (!direntp) | |
513 | break; | |
514 | ||
515 | if (!strcmp(direntp->d_name, ".")) | |
516 | continue; | |
517 | ||
518 | if (!strcmp(direntp->d_name, "..")) | |
519 | continue; | |
520 | ||
521 | if (strncmp(direntp->d_name, "loop", 4)) | |
522 | continue; | |
523 | ||
524 | rc = snprintf(path, MAXPATHLEN, "/dev/%s", direntp->d_name); | |
525 | if (rc < 0 || rc >= MAXPATHLEN) | |
526 | continue; | |
527 | ||
528 | fd = open(path, O_RDWR); | |
529 | if (fd < 0) | |
530 | continue; | |
531 | ||
532 | if (ioctl(fd, LOOP_GET_STATUS64, &loinfo) == 0) { | |
533 | close(fd); | |
534 | continue; | |
535 | } | |
536 | ||
537 | if (errno != ENXIO) { | |
538 | WARN("unexpected error for ioctl on '%s': %m", | |
539 | direntp->d_name); | |
540 | continue; | |
541 | } | |
542 | ||
543 | DEBUG("found '%s' free lodev", path); | |
544 | ||
545 | ret = setup_lodev(rootfs, fd, &loinfo); | |
546 | if (!ret) | |
547 | ret = mount_unknow_fs(path, target, 0); | |
548 | close(fd); | |
549 | ||
550 | break; | |
551 | } | |
552 | ||
553 | if (closedir(dir)) | |
554 | WARN("failed to close directory"); | |
555 | ||
556 | return ret; | |
557 | } | |
558 | ||
559 | static int mount_rootfs_block(const char *rootfs, const char *target) | |
560 | { | |
561 | return mount_unknow_fs(rootfs, target, 0); | |
562 | } | |
563 | ||
564 | /* | |
565 | * pin_rootfs | |
566 | * if rootfs is a directory, then open ${rootfs}.hold for writing for the | |
567 | * duration of the container run, to prevent the container from marking the | |
568 | * underlying fs readonly on shutdown. | |
569 | * return -1 on error. | |
570 | * return -2 if nothing needed to be pinned. | |
571 | * return an open fd (>=0) if we pinned it. | |
572 | */ | |
573 | int pin_rootfs(const char *rootfs) | |
574 | { | |
575 | char absrootfs[MAXPATHLEN]; | |
576 | char absrootfspin[MAXPATHLEN]; | |
577 | struct stat s; | |
578 | int ret, fd; | |
579 | ||
580 | if (rootfs == NULL || strlen(rootfs) == 0) | |
581 | return 0; | |
582 | ||
583 | if (!realpath(rootfs, absrootfs)) { | |
584 | SYSERROR("failed to get real path for '%s'", rootfs); | |
585 | return -1; | |
586 | } | |
587 | ||
588 | if (access(absrootfs, F_OK)) { | |
589 | SYSERROR("'%s' is not accessible", absrootfs); | |
590 | return -1; | |
591 | } | |
592 | ||
593 | if (stat(absrootfs, &s)) { | |
594 | SYSERROR("failed to stat '%s'", absrootfs); | |
595 | return -1; | |
596 | } | |
597 | ||
598 | if (!S_ISDIR(s.st_mode)) | |
599 | return -2; | |
600 | ||
601 | ret = snprintf(absrootfspin, MAXPATHLEN, "%s%s", absrootfs, ".hold"); | |
602 | if (ret >= MAXPATHLEN) { | |
603 | SYSERROR("pathname too long for rootfs hold file"); | |
604 | return -1; | |
605 | } | |
606 | ||
607 | fd = open(absrootfspin, O_CREAT | O_RDWR, S_IWUSR|S_IRUSR); | |
608 | INFO("opened %s as fd %d\n", absrootfspin, fd); | |
609 | return fd; | |
610 | } | |
611 | ||
612 | static int mount_rootfs(const char *rootfs, const char *target) | |
613 | { | |
614 | char absrootfs[MAXPATHLEN]; | |
615 | struct stat s; | |
616 | int i; | |
617 | ||
618 | typedef int (*rootfs_cb)(const char *, const char *); | |
619 | ||
620 | struct rootfs_type { | |
621 | int type; | |
622 | rootfs_cb cb; | |
623 | } rtfs_type[] = { | |
624 | { S_IFDIR, mount_rootfs_dir }, | |
625 | { S_IFBLK, mount_rootfs_block }, | |
626 | { S_IFREG, mount_rootfs_file }, | |
627 | }; | |
628 | ||
629 | if (!realpath(rootfs, absrootfs)) { | |
630 | SYSERROR("failed to get real path for '%s'", rootfs); | |
631 | return -1; | |
632 | } | |
633 | ||
634 | if (access(absrootfs, F_OK)) { | |
635 | SYSERROR("'%s' is not accessible", absrootfs); | |
636 | return -1; | |
637 | } | |
638 | ||
639 | if (stat(absrootfs, &s)) { | |
640 | SYSERROR("failed to stat '%s'", absrootfs); | |
641 | return -1; | |
642 | } | |
643 | ||
644 | for (i = 0; i < sizeof(rtfs_type)/sizeof(rtfs_type[0]); i++) { | |
645 | ||
646 | if (!__S_ISTYPE(s.st_mode, rtfs_type[i].type)) | |
647 | continue; | |
648 | ||
649 | return rtfs_type[i].cb(absrootfs, target); | |
650 | } | |
651 | ||
652 | ERROR("unsupported rootfs type for '%s'", absrootfs); | |
653 | return -1; | |
654 | } | |
655 | ||
656 | static int setup_utsname(struct utsname *utsname) | |
657 | { | |
658 | if (!utsname) | |
659 | return 0; | |
660 | ||
661 | if (sethostname(utsname->nodename, strlen(utsname->nodename))) { | |
662 | SYSERROR("failed to set the hostname to '%s'", utsname->nodename); | |
663 | return -1; | |
664 | } | |
665 | ||
666 | INFO("'%s' hostname has been setup", utsname->nodename); | |
667 | ||
668 | return 0; | |
669 | } | |
670 | ||
671 | static int setup_tty(const struct lxc_rootfs *rootfs, | |
672 | const struct lxc_tty_info *tty_info, char *ttydir) | |
673 | { | |
674 | char path[MAXPATHLEN], lxcpath[MAXPATHLEN]; | |
675 | int i, ret; | |
676 | ||
677 | if (!rootfs->path) | |
678 | return 0; | |
679 | ||
680 | for (i = 0; i < tty_info->nbtty; i++) { | |
681 | ||
682 | struct lxc_pty_info *pty_info = &tty_info->pty_info[i]; | |
683 | ||
684 | ret = snprintf(path, sizeof(path), "%s/dev/tty%d", | |
685 | rootfs->mount, i + 1); | |
686 | if (ret >= sizeof(path)) { | |
687 | ERROR("pathname too long for ttys"); | |
688 | return -1; | |
689 | } | |
690 | if (ttydir) { | |
691 | /* create dev/lxc/tty%d" */ | |
692 | ret = snprintf(lxcpath, sizeof(lxcpath), "%s/dev/%s/tty%d", | |
693 | rootfs->mount, ttydir, i + 1); | |
694 | if (ret >= sizeof(lxcpath)) { | |
695 | ERROR("pathname too long for ttys"); | |
696 | return -1; | |
697 | } | |
698 | ret = creat(lxcpath, 0660); | |
699 | if (ret==-1 && errno != EEXIST) { | |
700 | SYSERROR("error creating %s\n", lxcpath); | |
701 | return -1; | |
702 | } | |
703 | close(ret); | |
704 | ret = unlink(path); | |
705 | if (ret && errno != ENOENT) { | |
706 | SYSERROR("error unlinking %s\n", path); | |
707 | return -1; | |
708 | } | |
709 | ||
710 | if (mount(pty_info->name, lxcpath, "none", MS_BIND, 0)) { | |
711 | WARN("failed to mount '%s'->'%s'", | |
712 | pty_info->name, path); | |
713 | continue; | |
714 | } | |
715 | ||
716 | ret = snprintf(lxcpath, sizeof(lxcpath), "%s/tty%d", ttydir, i+1); | |
717 | if (ret >= sizeof(lxcpath)) { | |
718 | ERROR("tty pathname too long"); | |
719 | return -1; | |
720 | } | |
721 | ret = symlink(lxcpath, path); | |
722 | if (ret) { | |
723 | SYSERROR("failed to create symlink for tty %d\n", i+1); | |
724 | return -1; | |
725 | } | |
726 | } else { | |
727 | /* If we populated /dev, then we need to create /dev/ttyN */ | |
728 | if (access(path, F_OK)) { | |
729 | ret = creat(path, 0660); | |
730 | if (ret==-1) { | |
731 | SYSERROR("error creating %s\n", path); | |
732 | /* this isn't fatal, continue */ | |
733 | } else | |
734 | close(ret); | |
735 | } | |
736 | if (mount(pty_info->name, path, "none", MS_BIND, 0)) { | |
737 | WARN("failed to mount '%s'->'%s'", | |
738 | pty_info->name, path); | |
739 | continue; | |
740 | } | |
741 | } | |
742 | } | |
743 | ||
744 | INFO("%d tty(s) has been setup", tty_info->nbtty); | |
745 | ||
746 | return 0; | |
747 | } | |
748 | ||
749 | static int setup_rootfs_pivot_root_cb(char *buffer, void *data) | |
750 | { | |
751 | struct lxc_list *mountlist, *listentry, *iterator; | |
752 | char *pivotdir, *mountpoint, *mountentry; | |
753 | int found; | |
754 | void **cbparm; | |
755 | ||
756 | mountentry = buffer; | |
757 | cbparm = (void **)data; | |
758 | ||
759 | mountlist = cbparm[0]; | |
760 | pivotdir = cbparm[1]; | |
761 | ||
762 | /* parse entry, first field is mountname, ignore */ | |
763 | mountpoint = strtok(mountentry, " "); | |
764 | if (!mountpoint) | |
765 | return -1; | |
766 | ||
767 | /* second field is mountpoint */ | |
768 | mountpoint = strtok(NULL, " "); | |
769 | if (!mountpoint) | |
770 | return -1; | |
771 | ||
772 | /* only consider mountpoints below old root fs */ | |
773 | if (strncmp(mountpoint, pivotdir, strlen(pivotdir))) | |
774 | return 0; | |
775 | ||
776 | /* filter duplicate mountpoints */ | |
777 | found = 0; | |
778 | lxc_list_for_each(iterator, mountlist) { | |
779 | if (!strcmp(iterator->elem, mountpoint)) { | |
780 | found = 1; | |
781 | break; | |
782 | } | |
783 | } | |
784 | if (found) | |
785 | return 0; | |
786 | ||
787 | /* add entry to list */ | |
788 | listentry = malloc(sizeof(*listentry)); | |
789 | if (!listentry) { | |
790 | SYSERROR("malloc for mountpoint listentry failed"); | |
791 | return -1; | |
792 | } | |
793 | ||
794 | listentry->elem = strdup(mountpoint); | |
795 | if (!listentry->elem) { | |
796 | SYSERROR("strdup failed"); | |
797 | return -1; | |
798 | } | |
799 | lxc_list_add_tail(mountlist, listentry); | |
800 | ||
801 | return 0; | |
802 | } | |
803 | ||
804 | static int umount_oldrootfs(const char *oldrootfs) | |
805 | { | |
806 | char path[MAXPATHLEN]; | |
807 | void *cbparm[2]; | |
808 | struct lxc_list mountlist, *iterator, *next; | |
809 | int ok, still_mounted, last_still_mounted; | |
810 | int rc; | |
811 | ||
812 | /* read and parse /proc/mounts in old root fs */ | |
813 | lxc_list_init(&mountlist); | |
814 | ||
815 | /* oldrootfs is on the top tree directory now */ | |
816 | rc = snprintf(path, sizeof(path), "/%s", oldrootfs); | |
817 | if (rc >= sizeof(path)) { | |
818 | ERROR("rootfs name too long"); | |
819 | return -1; | |
820 | } | |
821 | cbparm[0] = &mountlist; | |
822 | ||
823 | cbparm[1] = strdup(path); | |
824 | if (!cbparm[1]) { | |
825 | SYSERROR("strdup failed"); | |
826 | return -1; | |
827 | } | |
828 | ||
829 | rc = snprintf(path, sizeof(path), "%s/proc/mounts", oldrootfs); | |
830 | if (rc >= sizeof(path)) { | |
831 | ERROR("container proc/mounts name too long"); | |
832 | return -1; | |
833 | } | |
834 | ||
835 | ok = lxc_file_for_each_line(path, | |
836 | setup_rootfs_pivot_root_cb, &cbparm); | |
837 | if (ok < 0) { | |
838 | SYSERROR("failed to read or parse mount list '%s'", path); | |
839 | return -1; | |
840 | } | |
841 | ||
842 | /* umount filesystems until none left or list no longer shrinks */ | |
843 | still_mounted = 0; | |
844 | do { | |
845 | last_still_mounted = still_mounted; | |
846 | still_mounted = 0; | |
847 | ||
848 | lxc_list_for_each_safe(iterator, &mountlist, next) { | |
849 | ||
850 | /* umount normally */ | |
851 | if (!umount(iterator->elem)) { | |
852 | DEBUG("umounted '%s'", (char *)iterator->elem); | |
853 | lxc_list_del(iterator); | |
854 | continue; | |
855 | } | |
856 | ||
857 | still_mounted++; | |
858 | } | |
859 | ||
860 | } while (still_mounted > 0 && still_mounted != last_still_mounted); | |
861 | ||
862 | ||
863 | lxc_list_for_each(iterator, &mountlist) { | |
864 | ||
865 | /* let's try a lazy umount */ | |
866 | if (!umount2(iterator->elem, MNT_DETACH)) { | |
867 | INFO("lazy unmount of '%s'", (char *)iterator->elem); | |
868 | continue; | |
869 | } | |
870 | ||
871 | /* be more brutal (nfs) */ | |
872 | if (!umount2(iterator->elem, MNT_FORCE)) { | |
873 | INFO("forced unmount of '%s'", (char *)iterator->elem); | |
874 | continue; | |
875 | } | |
876 | ||
877 | WARN("failed to unmount '%s'", (char *)iterator->elem); | |
878 | } | |
879 | ||
880 | return 0; | |
881 | } | |
882 | ||
883 | static int setup_rootfs_pivot_root(const char *rootfs, const char *pivotdir) | |
884 | { | |
885 | char path[MAXPATHLEN]; | |
886 | int remove_pivotdir = 0; | |
887 | int rc; | |
888 | ||
889 | /* change into new root fs */ | |
890 | if (chdir(rootfs)) { | |
891 | SYSERROR("can't chdir to new rootfs '%s'", rootfs); | |
892 | return -1; | |
893 | } | |
894 | ||
895 | if (!pivotdir) | |
896 | pivotdir = "lxc_putold"; | |
897 | ||
898 | /* compute the full path to pivotdir under rootfs */ | |
899 | rc = snprintf(path, sizeof(path), "%s/%s", rootfs, pivotdir); | |
900 | if (rc >= sizeof(path)) { | |
901 | ERROR("pivot dir name too long"); | |
902 | return -1; | |
903 | } | |
904 | ||
905 | if (access(path, F_OK)) { | |
906 | ||
907 | if (mkdir_p(path, 0755)) { | |
908 | SYSERROR("failed to create pivotdir '%s'", path); | |
909 | return -1; | |
910 | } | |
911 | ||
912 | remove_pivotdir = 1; | |
913 | DEBUG("created '%s' directory", path); | |
914 | } | |
915 | ||
916 | DEBUG("mountpoint for old rootfs is '%s'", path); | |
917 | ||
918 | /* pivot_root into our new root fs */ | |
919 | if (pivot_root(".", path)) { | |
920 | SYSERROR("pivot_root syscall failed"); | |
921 | return -1; | |
922 | } | |
923 | ||
924 | if (chdir("/")) { | |
925 | SYSERROR("can't chdir to / after pivot_root"); | |
926 | return -1; | |
927 | } | |
928 | ||
929 | DEBUG("pivot_root syscall to '%s' successful", rootfs); | |
930 | ||
931 | /* we switch from absolute path to relative path */ | |
932 | if (umount_oldrootfs(pivotdir)) | |
933 | return -1; | |
934 | ||
935 | /* remove temporary mount point, we don't consider the removing | |
936 | * as fatal */ | |
937 | if (remove_pivotdir && rmdir(pivotdir)) | |
938 | WARN("can't remove mountpoint '%s': %m", pivotdir); | |
939 | ||
940 | return 0; | |
941 | } | |
942 | ||
943 | /* | |
944 | * Do we want to add options for max size of /dev and a file to | |
945 | * specify which devices to create? | |
946 | */ | |
947 | static int mount_autodev(char *root) | |
948 | { | |
949 | int ret; | |
950 | char path[MAXPATHLEN]; | |
951 | ||
952 | INFO("Mounting /dev under %s\n", root); | |
953 | ret = snprintf(path, MAXPATHLEN, "%s/dev", root); | |
954 | if (ret < 0 || ret > MAXPATHLEN) | |
955 | return -1; | |
956 | ret = mount("none", path, "tmpfs", 0, "size=100000"); | |
957 | if (ret) { | |
958 | SYSERROR("Failed to mount /dev at %s\n", root); | |
959 | return -1; | |
960 | } | |
961 | ret = snprintf(path, MAXPATHLEN, "%s/dev/pts", root); | |
962 | if (ret < 0 || ret >= MAXPATHLEN) | |
963 | return -1; | |
964 | ret = mkdir(path, S_IRWXU | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH); | |
965 | if (ret) { | |
966 | SYSERROR("Failed to create /dev/pts in container"); | |
967 | return -1; | |
968 | } | |
969 | ||
970 | INFO("Mounted /dev under %s\n", root); | |
971 | return 0; | |
972 | } | |
973 | ||
974 | struct lxc_devs { | |
975 | char *name; | |
976 | mode_t mode; | |
977 | int maj; | |
978 | int min; | |
979 | }; | |
980 | ||
981 | struct lxc_devs lxc_devs[] = { | |
982 | { "null", S_IFCHR | S_IRWXU | S_IRWXG | S_IRWXO, 1, 3 }, | |
983 | { "zero", S_IFCHR | S_IRWXU | S_IRWXG | S_IRWXO, 1, 5 }, | |
984 | { "full", S_IFCHR | S_IRWXU | S_IRWXG | S_IRWXO, 1, 7 }, | |
985 | { "urandom", S_IFCHR | S_IRWXU | S_IRWXG | S_IRWXO, 1, 9 }, | |
986 | { "random", S_IFCHR | S_IRWXU | S_IRWXG | S_IRWXO, 1, 8 }, | |
987 | { "tty", S_IFCHR | S_IRWXU | S_IRWXG | S_IRWXO, 5, 0 }, | |
988 | { "console", S_IFCHR | S_IRUSR | S_IWUSR, 5, 1 }, | |
989 | }; | |
990 | ||
991 | static int setup_autodev(char *root) | |
992 | { | |
993 | int ret; | |
994 | struct lxc_devs *d; | |
995 | char path[MAXPATHLEN]; | |
996 | int i; | |
997 | mode_t cmask; | |
998 | ||
999 | INFO("Creating initial consoles under %s/dev\n", root); | |
1000 | ||
1001 | ret = snprintf(path, MAXPATHLEN, "%s/dev", root); | |
1002 | if (ret < 0 || ret >= MAXPATHLEN) { | |
1003 | ERROR("Error calculating container /dev location"); | |
1004 | return -1; | |
1005 | } | |
1006 | ||
1007 | INFO("Populating /dev under %s\n", root); | |
1008 | cmask = umask(S_IXUSR | S_IXGRP | S_IXOTH); | |
1009 | for (i = 0; i < sizeof(lxc_devs) / sizeof(lxc_devs[0]); i++) { | |
1010 | d = &lxc_devs[i]; | |
1011 | ret = snprintf(path, MAXPATHLEN, "%s/dev/%s", root, d->name); | |
1012 | if (ret < 0 || ret >= MAXPATHLEN) | |
1013 | return -1; | |
1014 | ret = mknod(path, d->mode, makedev(d->maj, d->min)); | |
1015 | if (ret && errno != EEXIST) { | |
1016 | SYSERROR("Error creating %s\n", d->name); | |
1017 | return -1; | |
1018 | } | |
1019 | } | |
1020 | umask(cmask); | |
1021 | ||
1022 | INFO("Populated /dev under %s\n", root); | |
1023 | return 0; | |
1024 | } | |
1025 | ||
1026 | /* | |
1027 | * Detect whether / is mounted MS_SHARED. The only way I know of to | |
1028 | * check that is through /proc/self/mountinfo. | |
1029 | * I'm only checking for /. If the container rootfs or mount location | |
1030 | * is MS_SHARED, but not '/', then you're out of luck - figuring that | |
1031 | * out would be too much work to be worth it. | |
1032 | */ | |
1033 | #define LINELEN 4096 | |
1034 | int detect_shared_rootfs(void) | |
1035 | { | |
1036 | char buf[LINELEN], *p; | |
1037 | FILE *f; | |
1038 | int i; | |
1039 | char *p2; | |
1040 | ||
1041 | f = fopen("/proc/self/mountinfo", "r"); | |
1042 | if (!f) | |
1043 | return 0; | |
1044 | while ((p = fgets(buf, LINELEN, f))) { | |
1045 | INFO("looking at .%s.", p); | |
1046 | for (p = buf, i=0; p && i < 4; i++) | |
1047 | p = index(p+1, ' '); | |
1048 | if (!p) | |
1049 | continue; | |
1050 | p2 = index(p+1, ' '); | |
1051 | if (!p2) | |
1052 | continue; | |
1053 | *p2 = '\0'; | |
1054 | INFO("now p is .%s.", p); | |
1055 | if (strcmp(p+1, "/") == 0) { | |
1056 | // this is '/'. is it shared? | |
1057 | p = index(p2+1, ' '); | |
1058 | if (strstr(p, "shared:")) | |
1059 | return 1; | |
1060 | } | |
1061 | } | |
1062 | fclose(f); | |
1063 | return 0; | |
1064 | } | |
1065 | ||
1066 | /* | |
1067 | * I'll forgive you for asking whether all of this is needed :) The | |
1068 | * answer is yes. | |
1069 | * pivot_root will fail if the new root, the put_old dir, or the parent | |
1070 | * of current->fs->root are MS_SHARED. (parent of current->fs_root may | |
1071 | * or may not be current->fs_root - if we assumed it always was, we could | |
1072 | * just mount --make-rslave /). So, | |
1073 | * 1. mount a tiny tmpfs to be parent of current->fs->root. | |
1074 | * 2. make that MS_SLAVE | |
1075 | * 3. make a 'root' directory under that | |
1076 | * 4. mount --rbind / under the $tinyroot/root. | |
1077 | * 5. make that rslave | |
1078 | * 6. chdir and chroot into $tinyroot/root | |
1079 | * 7. $tinyroot will be unmounted by our parent in start.c | |
1080 | */ | |
1081 | static int chroot_into_slave(struct lxc_conf *conf) | |
1082 | { | |
1083 | char path[MAXPATHLEN]; | |
1084 | const char *destpath = conf->rootfs.mount; | |
1085 | int ret; | |
1086 | ||
1087 | if (mount(destpath, destpath, NULL, MS_BIND, 0)) { | |
1088 | SYSERROR("failed to mount %s bind", destpath); | |
1089 | return -1; | |
1090 | } | |
1091 | if (mount("", destpath, NULL, MS_SLAVE, 0)) { | |
1092 | SYSERROR("failed to make %s slave", destpath); | |
1093 | return -1; | |
1094 | } | |
1095 | if (mount("none", destpath, "tmpfs", 0, "size=10000")) { | |
1096 | SYSERROR("Failed to mount tmpfs / at %s", destpath); | |
1097 | return -1; | |
1098 | } | |
1099 | ret = snprintf(path, MAXPATHLEN, "%s/root", destpath); | |
1100 | if (ret < 0 || ret >= MAXPATHLEN) { | |
1101 | ERROR("out of memory making root path"); | |
1102 | return -1; | |
1103 | } | |
1104 | if (mkdir(path, S_IRWXU | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH)) { | |
1105 | SYSERROR("Failed to create /dev/pts in container"); | |
1106 | return -1; | |
1107 | } | |
1108 | if (mount("/", path, NULL, MS_BIND|MS_REC, 0)) { | |
1109 | SYSERROR("Failed to rbind mount / to %s", path); | |
1110 | return -1; | |
1111 | } | |
1112 | if (mount("", destpath, NULL, MS_SLAVE|MS_REC, 0)) { | |
1113 | SYSERROR("Failed to make tmp-/ at %s rslave", path); | |
1114 | return -1; | |
1115 | } | |
1116 | if (chdir(path)) { | |
1117 | SYSERROR("Failed to chdir into tmp-/"); | |
1118 | return -1; | |
1119 | } | |
1120 | if (chroot(path)) { | |
1121 | SYSERROR("Failed to chroot into tmp-/"); | |
1122 | return -1; | |
1123 | } | |
1124 | INFO("Chrooted into tmp-/ at %s\n", path); | |
1125 | return 0; | |
1126 | } | |
1127 | ||
1128 | static int setup_rootfs(struct lxc_conf *conf) | |
1129 | { | |
1130 | const struct lxc_rootfs *rootfs = &conf->rootfs; | |
1131 | ||
1132 | if (!rootfs->path) | |
1133 | return 0; | |
1134 | ||
1135 | if (access(rootfs->mount, F_OK)) { | |
1136 | SYSERROR("failed to access to '%s', check it is present", | |
1137 | rootfs->mount); | |
1138 | return -1; | |
1139 | } | |
1140 | ||
1141 | if (detect_shared_rootfs()) { | |
1142 | if (chroot_into_slave(conf)) { | |
1143 | ERROR("Failed to chroot into slave /"); | |
1144 | return -1; | |
1145 | } | |
1146 | } | |
1147 | ||
1148 | if (mount_rootfs(rootfs->path, rootfs->mount)) { | |
1149 | ERROR("failed to mount rootfs"); | |
1150 | return -1; | |
1151 | } | |
1152 | ||
1153 | DEBUG("mounted '%s' on '%s'", rootfs->path, rootfs->mount); | |
1154 | ||
1155 | return 0; | |
1156 | } | |
1157 | ||
1158 | int setup_pivot_root(const struct lxc_rootfs *rootfs) | |
1159 | { | |
1160 | if (!rootfs->path) | |
1161 | return 0; | |
1162 | ||
1163 | if (setup_rootfs_pivot_root(rootfs->mount, rootfs->pivot)) { | |
1164 | ERROR("failed to setup pivot root"); | |
1165 | return -1; | |
1166 | } | |
1167 | ||
1168 | return 0; | |
1169 | } | |
1170 | ||
1171 | static int setup_pts(int pts) | |
1172 | { | |
1173 | char target[PATH_MAX]; | |
1174 | ||
1175 | if (!pts) | |
1176 | return 0; | |
1177 | ||
1178 | if (!access("/dev/pts/ptmx", F_OK) && umount("/dev/pts")) { | |
1179 | SYSERROR("failed to umount 'dev/pts'"); | |
1180 | return -1; | |
1181 | } | |
1182 | ||
1183 | if (mount("devpts", "/dev/pts", "devpts", MS_MGC_VAL, | |
1184 | "newinstance,ptmxmode=0666")) { | |
1185 | SYSERROR("failed to mount a new instance of '/dev/pts'"); | |
1186 | return -1; | |
1187 | } | |
1188 | ||
1189 | if (access("/dev/ptmx", F_OK)) { | |
1190 | if (!symlink("/dev/pts/ptmx", "/dev/ptmx")) | |
1191 | goto out; | |
1192 | SYSERROR("failed to symlink '/dev/pts/ptmx'->'/dev/ptmx'"); | |
1193 | return -1; | |
1194 | } | |
1195 | ||
1196 | if (realpath("/dev/ptmx", target) && !strcmp(target, "/dev/pts/ptmx")) | |
1197 | goto out; | |
1198 | ||
1199 | /* fallback here, /dev/pts/ptmx exists just mount bind */ | |
1200 | if (mount("/dev/pts/ptmx", "/dev/ptmx", "none", MS_BIND, 0)) { | |
1201 | SYSERROR("mount failed '/dev/pts/ptmx'->'/dev/ptmx'"); | |
1202 | return -1; | |
1203 | } | |
1204 | ||
1205 | INFO("created new pts instance"); | |
1206 | ||
1207 | out: | |
1208 | return 0; | |
1209 | } | |
1210 | ||
1211 | static int setup_personality(int persona) | |
1212 | { | |
1213 | #if HAVE_SYS_PERSONALITY_H | |
1214 | if (persona == -1) | |
1215 | return 0; | |
1216 | ||
1217 | if (personality(persona) < 0) { | |
1218 | SYSERROR("failed to set personality to '0x%x'", persona); | |
1219 | return -1; | |
1220 | } | |
1221 | ||
1222 | INFO("set personality to '0x%x'", persona); | |
1223 | #endif | |
1224 | ||
1225 | return 0; | |
1226 | } | |
1227 | ||
1228 | static int setup_dev_console(const struct lxc_rootfs *rootfs, | |
1229 | const struct lxc_console *console) | |
1230 | { | |
1231 | char path[MAXPATHLEN]; | |
1232 | struct stat s; | |
1233 | int ret; | |
1234 | ||
1235 | ret = snprintf(path, sizeof(path), "%s/dev/console", rootfs->mount); | |
1236 | if (ret >= sizeof(path)) { | |
1237 | ERROR("console path too long\n"); | |
1238 | return -1; | |
1239 | } | |
1240 | ||
1241 | if (access(path, F_OK)) { | |
1242 | WARN("rootfs specified but no console found at '%s'", path); | |
1243 | return 0; | |
1244 | } | |
1245 | ||
1246 | if (console->peer == -1) { | |
1247 | INFO("no console output required"); | |
1248 | return 0; | |
1249 | } | |
1250 | ||
1251 | if (stat(path, &s)) { | |
1252 | SYSERROR("failed to stat '%s'", path); | |
1253 | return -1; | |
1254 | } | |
1255 | ||
1256 | if (chmod(console->name, s.st_mode)) { | |
1257 | SYSERROR("failed to set mode '0%o' to '%s'", | |
1258 | s.st_mode, console->name); | |
1259 | return -1; | |
1260 | } | |
1261 | ||
1262 | if (mount(console->name, path, "none", MS_BIND, 0)) { | |
1263 | ERROR("failed to mount '%s' on '%s'", console->name, path); | |
1264 | return -1; | |
1265 | } | |
1266 | ||
1267 | INFO("console has been setup"); | |
1268 | return 0; | |
1269 | } | |
1270 | ||
1271 | static int setup_ttydir_console(const struct lxc_rootfs *rootfs, | |
1272 | const struct lxc_console *console, | |
1273 | char *ttydir) | |
1274 | { | |
1275 | char path[MAXPATHLEN], lxcpath[MAXPATHLEN]; | |
1276 | int ret; | |
1277 | ||
1278 | /* create rootfs/dev/<ttydir> directory */ | |
1279 | ret = snprintf(path, sizeof(path), "%s/dev/%s", rootfs->mount, | |
1280 | ttydir); | |
1281 | if (ret >= sizeof(path)) | |
1282 | return -1; | |
1283 | ret = mkdir(path, 0755); | |
1284 | if (ret && errno != EEXIST) { | |
1285 | SYSERROR("failed with errno %d to create %s\n", errno, path); | |
1286 | return -1; | |
1287 | } | |
1288 | INFO("created %s\n", path); | |
1289 | ||
1290 | ret = snprintf(lxcpath, sizeof(lxcpath), "%s/dev/%s/console", | |
1291 | rootfs->mount, ttydir); | |
1292 | if (ret >= sizeof(lxcpath)) { | |
1293 | ERROR("console path too long\n"); | |
1294 | return -1; | |
1295 | } | |
1296 | ||
1297 | snprintf(path, sizeof(path), "%s/dev/console", rootfs->mount); | |
1298 | ret = unlink(path); | |
1299 | if (ret && errno != ENOENT) { | |
1300 | SYSERROR("error unlinking %s\n", path); | |
1301 | return -1; | |
1302 | } | |
1303 | ||
1304 | ret = creat(lxcpath, 0660); | |
1305 | if (ret==-1 && errno != EEXIST) { | |
1306 | SYSERROR("error %d creating %s\n", errno, lxcpath); | |
1307 | return -1; | |
1308 | } | |
1309 | close(ret); | |
1310 | ||
1311 | if (console->peer == -1) { | |
1312 | INFO("no console output required"); | |
1313 | return 0; | |
1314 | } | |
1315 | ||
1316 | if (mount(console->name, lxcpath, "none", MS_BIND, 0)) { | |
1317 | ERROR("failed to mount '%s' on '%s'", console->name, lxcpath); | |
1318 | return -1; | |
1319 | } | |
1320 | ||
1321 | /* create symlink from rootfs/dev/console to 'lxc/console' */ | |
1322 | ret = snprintf(lxcpath, sizeof(lxcpath), "%s/console", ttydir); | |
1323 | if (ret >= sizeof(lxcpath)) { | |
1324 | ERROR("lxc/console path too long"); | |
1325 | return -1; | |
1326 | } | |
1327 | ret = symlink(lxcpath, path); | |
1328 | if (ret) { | |
1329 | SYSERROR("failed to create symlink for console"); | |
1330 | return -1; | |
1331 | } | |
1332 | ||
1333 | INFO("console has been setup on %s", lxcpath); | |
1334 | ||
1335 | return 0; | |
1336 | } | |
1337 | ||
1338 | static int setup_console(const struct lxc_rootfs *rootfs, | |
1339 | const struct lxc_console *console, | |
1340 | char *ttydir) | |
1341 | { | |
1342 | /* We don't have a rootfs, /dev/console will be shared */ | |
1343 | if (!rootfs->path) | |
1344 | return 0; | |
1345 | if (!ttydir) | |
1346 | return setup_dev_console(rootfs, console); | |
1347 | ||
1348 | return setup_ttydir_console(rootfs, console, ttydir); | |
1349 | } | |
1350 | ||
1351 | static int setup_kmsg(const struct lxc_rootfs *rootfs, | |
1352 | const struct lxc_console *console) | |
1353 | { | |
1354 | char kpath[MAXPATHLEN]; | |
1355 | int ret; | |
1356 | ||
1357 | if (!rootfs->path) | |
1358 | return 0; | |
1359 | ret = snprintf(kpath, sizeof(kpath), "%s/dev/kmsg", rootfs->mount); | |
1360 | if (ret < 0 || ret >= sizeof(kpath)) | |
1361 | return -1; | |
1362 | ||
1363 | ret = unlink(kpath); | |
1364 | if (ret && errno != ENOENT) { | |
1365 | SYSERROR("error unlinking %s\n", kpath); | |
1366 | return -1; | |
1367 | } | |
1368 | ||
1369 | ret = symlink("console", kpath); | |
1370 | if (ret) { | |
1371 | SYSERROR("failed to create symlink for kmsg"); | |
1372 | return -1; | |
1373 | } | |
1374 | ||
1375 | return 0; | |
1376 | } | |
1377 | ||
1378 | int setup_cgroup(const char *name, struct lxc_list *cgroups) | |
1379 | { | |
1380 | struct lxc_list *iterator; | |
1381 | struct lxc_cgroup *cg; | |
1382 | int ret = -1; | |
1383 | ||
1384 | if (lxc_list_empty(cgroups)) | |
1385 | return 0; | |
1386 | ||
1387 | lxc_list_for_each(iterator, cgroups) { | |
1388 | ||
1389 | cg = iterator->elem; | |
1390 | ||
1391 | if (lxc_cgroup_set(name, cg->subsystem, cg->value)) | |
1392 | goto out; | |
1393 | ||
1394 | DEBUG("cgroup '%s' set to '%s'", cg->subsystem, cg->value); | |
1395 | } | |
1396 | ||
1397 | ret = 0; | |
1398 | INFO("cgroup has been setup"); | |
1399 | out: | |
1400 | return ret; | |
1401 | } | |
1402 | ||
1403 | static void parse_mntopt(char *opt, unsigned long *flags, char **data) | |
1404 | { | |
1405 | struct mount_opt *mo; | |
1406 | ||
1407 | /* If opt is found in mount_opt, set or clear flags. | |
1408 | * Otherwise append it to data. */ | |
1409 | ||
1410 | for (mo = &mount_opt[0]; mo->name != NULL; mo++) { | |
1411 | if (!strncmp(opt, mo->name, strlen(mo->name))) { | |
1412 | if (mo->clear) | |
1413 | *flags &= ~mo->flag; | |
1414 | else | |
1415 | *flags |= mo->flag; | |
1416 | return; | |
1417 | } | |
1418 | } | |
1419 | ||
1420 | if (strlen(*data)) | |
1421 | strcat(*data, ","); | |
1422 | strcat(*data, opt); | |
1423 | } | |
1424 | ||
1425 | static int parse_mntopts(const char *mntopts, unsigned long *mntflags, | |
1426 | char **mntdata) | |
1427 | { | |
1428 | char *s, *data; | |
1429 | char *p, *saveptr = NULL; | |
1430 | ||
1431 | *mntdata = NULL; | |
1432 | *mntflags = 0L; | |
1433 | ||
1434 | if (!mntopts) | |
1435 | return 0; | |
1436 | ||
1437 | s = strdup(mntopts); | |
1438 | if (!s) { | |
1439 | SYSERROR("failed to allocate memory"); | |
1440 | return -1; | |
1441 | } | |
1442 | ||
1443 | data = malloc(strlen(s) + 1); | |
1444 | if (!data) { | |
1445 | SYSERROR("failed to allocate memory"); | |
1446 | free(s); | |
1447 | return -1; | |
1448 | } | |
1449 | *data = 0; | |
1450 | ||
1451 | for (p = strtok_r(s, ",", &saveptr); p != NULL; | |
1452 | p = strtok_r(NULL, ",", &saveptr)) | |
1453 | parse_mntopt(p, mntflags, &data); | |
1454 | ||
1455 | if (*data) | |
1456 | *mntdata = data; | |
1457 | else | |
1458 | free(data); | |
1459 | free(s); | |
1460 | ||
1461 | return 0; | |
1462 | } | |
1463 | ||
1464 | static int mount_entry(const char *fsname, const char *target, | |
1465 | const char *fstype, unsigned long mountflags, | |
1466 | const char *data) | |
1467 | { | |
1468 | if (mount(fsname, target, fstype, mountflags & ~MS_REMOUNT, data)) { | |
1469 | SYSERROR("failed to mount '%s' on '%s'", fsname, target); | |
1470 | return -1; | |
1471 | } | |
1472 | ||
1473 | if ((mountflags & MS_REMOUNT) || (mountflags & MS_BIND)) { | |
1474 | ||
1475 | DEBUG("remounting %s on %s to respect bind or remount options", | |
1476 | fsname, target); | |
1477 | ||
1478 | if (mount(fsname, target, fstype, | |
1479 | mountflags | MS_REMOUNT, data)) { | |
1480 | SYSERROR("failed to mount '%s' on '%s'", | |
1481 | fsname, target); | |
1482 | return -1; | |
1483 | } | |
1484 | } | |
1485 | ||
1486 | DEBUG("mounted '%s' on '%s', type '%s'", fsname, target, fstype); | |
1487 | ||
1488 | return 0; | |
1489 | } | |
1490 | ||
1491 | static inline int mount_entry_on_systemfs(struct mntent *mntent) | |
1492 | { | |
1493 | unsigned long mntflags; | |
1494 | char *mntdata; | |
1495 | int ret; | |
1496 | ||
1497 | if (parse_mntopts(mntent->mnt_opts, &mntflags, &mntdata) < 0) { | |
1498 | ERROR("failed to parse mount option '%s'", mntent->mnt_opts); | |
1499 | return -1; | |
1500 | } | |
1501 | ||
1502 | ret = mount_entry(mntent->mnt_fsname, mntent->mnt_dir, | |
1503 | mntent->mnt_type, mntflags, mntdata); | |
1504 | ||
1505 | free(mntdata); | |
1506 | ||
1507 | return ret; | |
1508 | } | |
1509 | ||
1510 | static int mount_entry_on_absolute_rootfs(struct mntent *mntent, | |
1511 | const struct lxc_rootfs *rootfs, | |
1512 | const char *lxc_name) | |
1513 | { | |
1514 | char *aux; | |
1515 | char path[MAXPATHLEN]; | |
1516 | unsigned long mntflags; | |
1517 | char *mntdata; | |
1518 | int r, ret = 0, offset; | |
1519 | ||
1520 | if (parse_mntopts(mntent->mnt_opts, &mntflags, &mntdata) < 0) { | |
1521 | ERROR("failed to parse mount option '%s'", mntent->mnt_opts); | |
1522 | return -1; | |
1523 | } | |
1524 | ||
1525 | /* if rootfs->path is a blockdev path, allow container fstab to | |
1526 | * use $LXCPATH/CN/rootfs as the target prefix */ | |
1527 | r = snprintf(path, MAXPATHLEN, LXCPATH "/%s/rootfs", lxc_name); | |
1528 | if (r < 0 || r >= MAXPATHLEN) | |
1529 | goto skipvarlib; | |
1530 | ||
1531 | aux = strstr(mntent->mnt_dir, path); | |
1532 | if (aux) { | |
1533 | offset = strlen(path); | |
1534 | goto skipabs; | |
1535 | } | |
1536 | ||
1537 | skipvarlib: | |
1538 | aux = strstr(mntent->mnt_dir, rootfs->path); | |
1539 | if (!aux) { | |
1540 | WARN("ignoring mount point '%s'", mntent->mnt_dir); | |
1541 | goto out; | |
1542 | } | |
1543 | offset = strlen(rootfs->path); | |
1544 | ||
1545 | skipabs: | |
1546 | ||
1547 | r = snprintf(path, MAXPATHLEN, "%s/%s", rootfs->mount, | |
1548 | aux + offset); | |
1549 | if (r < 0 || r >= MAXPATHLEN) { | |
1550 | WARN("pathnme too long for '%s'", mntent->mnt_dir); | |
1551 | ret = -1; | |
1552 | goto out; | |
1553 | } | |
1554 | ||
1555 | ||
1556 | ret = mount_entry(mntent->mnt_fsname, path, mntent->mnt_type, | |
1557 | mntflags, mntdata); | |
1558 | ||
1559 | out: | |
1560 | free(mntdata); | |
1561 | return ret; | |
1562 | } | |
1563 | ||
1564 | static int mount_entry_on_relative_rootfs(struct mntent *mntent, | |
1565 | const char *rootfs) | |
1566 | { | |
1567 | char path[MAXPATHLEN]; | |
1568 | unsigned long mntflags; | |
1569 | char *mntdata; | |
1570 | int ret; | |
1571 | ||
1572 | if (parse_mntopts(mntent->mnt_opts, &mntflags, &mntdata) < 0) { | |
1573 | ERROR("failed to parse mount option '%s'", mntent->mnt_opts); | |
1574 | return -1; | |
1575 | } | |
1576 | ||
1577 | /* relative to root mount point */ | |
1578 | ret = snprintf(path, sizeof(path), "%s/%s", rootfs, mntent->mnt_dir); | |
1579 | if (ret >= sizeof(path)) { | |
1580 | ERROR("path name too long"); | |
1581 | return -1; | |
1582 | } | |
1583 | ||
1584 | ret = mount_entry(mntent->mnt_fsname, path, mntent->mnt_type, | |
1585 | mntflags, mntdata); | |
1586 | ||
1587 | free(mntdata); | |
1588 | ||
1589 | return ret; | |
1590 | } | |
1591 | ||
1592 | static int mount_file_entries(const struct lxc_rootfs *rootfs, FILE *file, | |
1593 | const char *lxc_name) | |
1594 | { | |
1595 | struct mntent *mntent; | |
1596 | int ret = -1; | |
1597 | ||
1598 | while ((mntent = getmntent(file))) { | |
1599 | ||
1600 | if (!rootfs->path) { | |
1601 | if (mount_entry_on_systemfs(mntent)) | |
1602 | goto out; | |
1603 | continue; | |
1604 | } | |
1605 | ||
1606 | /* We have a separate root, mounts are relative to it */ | |
1607 | if (mntent->mnt_dir[0] != '/') { | |
1608 | if (mount_entry_on_relative_rootfs(mntent, | |
1609 | rootfs->mount)) | |
1610 | goto out; | |
1611 | continue; | |
1612 | } | |
1613 | ||
1614 | if (mount_entry_on_absolute_rootfs(mntent, rootfs, lxc_name)) | |
1615 | goto out; | |
1616 | } | |
1617 | ||
1618 | ret = 0; | |
1619 | ||
1620 | INFO("mount points have been setup"); | |
1621 | out: | |
1622 | return ret; | |
1623 | } | |
1624 | ||
1625 | static int setup_mount(const struct lxc_rootfs *rootfs, const char *fstab, | |
1626 | const char *lxc_name) | |
1627 | { | |
1628 | FILE *file; | |
1629 | int ret; | |
1630 | ||
1631 | if (!fstab) | |
1632 | return 0; | |
1633 | ||
1634 | file = setmntent(fstab, "r"); | |
1635 | if (!file) { | |
1636 | SYSERROR("failed to use '%s'", fstab); | |
1637 | return -1; | |
1638 | } | |
1639 | ||
1640 | ret = mount_file_entries(rootfs, file, lxc_name); | |
1641 | ||
1642 | endmntent(file); | |
1643 | return ret; | |
1644 | } | |
1645 | ||
1646 | static int setup_mount_entries(const struct lxc_rootfs *rootfs, struct lxc_list *mount, | |
1647 | const char *lxc_name) | |
1648 | { | |
1649 | FILE *file; | |
1650 | struct lxc_list *iterator; | |
1651 | char *mount_entry; | |
1652 | int ret; | |
1653 | ||
1654 | file = tmpfile(); | |
1655 | if (!file) { | |
1656 | ERROR("tmpfile error: %m"); | |
1657 | return -1; | |
1658 | } | |
1659 | ||
1660 | lxc_list_for_each(iterator, mount) { | |
1661 | mount_entry = iterator->elem; | |
1662 | fprintf(file, "%s\n", mount_entry); | |
1663 | } | |
1664 | ||
1665 | rewind(file); | |
1666 | ||
1667 | ret = mount_file_entries(rootfs, file, lxc_name); | |
1668 | ||
1669 | fclose(file); | |
1670 | return ret; | |
1671 | } | |
1672 | ||
1673 | static int setup_caps(struct lxc_list *caps) | |
1674 | { | |
1675 | struct lxc_list *iterator; | |
1676 | char *drop_entry; | |
1677 | char *ptr; | |
1678 | int i, capid; | |
1679 | ||
1680 | lxc_list_for_each(iterator, caps) { | |
1681 | ||
1682 | drop_entry = iterator->elem; | |
1683 | ||
1684 | capid = -1; | |
1685 | ||
1686 | for (i = 0; i < sizeof(caps_opt)/sizeof(caps_opt[0]); i++) { | |
1687 | ||
1688 | if (strcmp(drop_entry, caps_opt[i].name)) | |
1689 | continue; | |
1690 | ||
1691 | capid = caps_opt[i].value; | |
1692 | break; | |
1693 | } | |
1694 | ||
1695 | if (capid < 0) { | |
1696 | /* try to see if it's numeric, so the user may specify | |
1697 | * capabilities that the running kernel knows about but | |
1698 | * we don't */ | |
1699 | capid = strtol(drop_entry, &ptr, 10); | |
1700 | if (!ptr || *ptr != '\0' || | |
1701 | capid == LONG_MIN || capid == LONG_MAX) | |
1702 | /* not a valid number */ | |
1703 | capid = -1; | |
1704 | else if (capid > lxc_caps_last_cap()) | |
1705 | /* we have a number but it's not a valid | |
1706 | * capability */ | |
1707 | capid = -1; | |
1708 | } | |
1709 | ||
1710 | if (capid < 0) { | |
1711 | ERROR("unknown capability %s", drop_entry); | |
1712 | return -1; | |
1713 | } | |
1714 | ||
1715 | DEBUG("drop capability '%s' (%d)", drop_entry, capid); | |
1716 | ||
1717 | if (prctl(PR_CAPBSET_DROP, capid, 0, 0, 0)) { | |
1718 | SYSERROR("failed to remove %s capability", drop_entry); | |
1719 | return -1; | |
1720 | } | |
1721 | ||
1722 | } | |
1723 | ||
1724 | DEBUG("capabilities has been setup"); | |
1725 | ||
1726 | return 0; | |
1727 | } | |
1728 | ||
1729 | static int setup_hw_addr(char *hwaddr, const char *ifname) | |
1730 | { | |
1731 | struct sockaddr sockaddr; | |
1732 | struct ifreq ifr; | |
1733 | int ret, fd; | |
1734 | ||
1735 | ret = lxc_convert_mac(hwaddr, &sockaddr); | |
1736 | if (ret) { | |
1737 | ERROR("mac address '%s' conversion failed : %s", | |
1738 | hwaddr, strerror(-ret)); | |
1739 | return -1; | |
1740 | } | |
1741 | ||
1742 | memcpy(ifr.ifr_name, ifname, IFNAMSIZ); | |
1743 | memcpy((char *) &ifr.ifr_hwaddr, (char *) &sockaddr, sizeof(sockaddr)); | |
1744 | ||
1745 | fd = socket(AF_INET, SOCK_DGRAM, 0); | |
1746 | if (fd < 0) { | |
1747 | ERROR("socket failure : %s", strerror(errno)); | |
1748 | return -1; | |
1749 | } | |
1750 | ||
1751 | ret = ioctl(fd, SIOCSIFHWADDR, &ifr); | |
1752 | close(fd); | |
1753 | if (ret) | |
1754 | ERROR("ioctl failure : %s", strerror(errno)); | |
1755 | ||
1756 | DEBUG("mac address '%s' on '%s' has been setup", hwaddr, ifname); | |
1757 | ||
1758 | return ret; | |
1759 | } | |
1760 | ||
1761 | static int setup_ipv4_addr(struct lxc_list *ip, int ifindex) | |
1762 | { | |
1763 | struct lxc_list *iterator; | |
1764 | struct lxc_inetdev *inetdev; | |
1765 | int err; | |
1766 | ||
1767 | lxc_list_for_each(iterator, ip) { | |
1768 | ||
1769 | inetdev = iterator->elem; | |
1770 | ||
1771 | err = lxc_ipv4_addr_add(ifindex, &inetdev->addr, | |
1772 | &inetdev->bcast, inetdev->prefix); | |
1773 | if (err) { | |
1774 | ERROR("failed to setup_ipv4_addr ifindex %d : %s", | |
1775 | ifindex, strerror(-err)); | |
1776 | return -1; | |
1777 | } | |
1778 | } | |
1779 | ||
1780 | return 0; | |
1781 | } | |
1782 | ||
1783 | static int setup_ipv6_addr(struct lxc_list *ip, int ifindex) | |
1784 | { | |
1785 | struct lxc_list *iterator; | |
1786 | struct lxc_inet6dev *inet6dev; | |
1787 | int err; | |
1788 | ||
1789 | lxc_list_for_each(iterator, ip) { | |
1790 | ||
1791 | inet6dev = iterator->elem; | |
1792 | ||
1793 | err = lxc_ipv6_addr_add(ifindex, &inet6dev->addr, | |
1794 | &inet6dev->mcast, &inet6dev->acast, | |
1795 | inet6dev->prefix); | |
1796 | if (err) { | |
1797 | ERROR("failed to setup_ipv6_addr ifindex %d : %s", | |
1798 | ifindex, strerror(-err)); | |
1799 | return -1; | |
1800 | } | |
1801 | } | |
1802 | ||
1803 | return 0; | |
1804 | } | |
1805 | ||
1806 | static int setup_netdev(struct lxc_netdev *netdev) | |
1807 | { | |
1808 | char ifname[IFNAMSIZ]; | |
1809 | char *current_ifname = ifname; | |
1810 | int err; | |
1811 | ||
1812 | /* empty network namespace */ | |
1813 | if (!netdev->ifindex) { | |
1814 | if (netdev->flags & IFF_UP) { | |
1815 | err = lxc_netdev_up("lo"); | |
1816 | if (err) { | |
1817 | ERROR("failed to set the loopback up : %s", | |
1818 | strerror(-err)); | |
1819 | return -1; | |
1820 | } | |
1821 | } | |
1822 | return 0; | |
1823 | } | |
1824 | ||
1825 | /* retrieve the name of the interface */ | |
1826 | if (!if_indextoname(netdev->ifindex, current_ifname)) { | |
1827 | ERROR("no interface corresponding to index '%d'", | |
1828 | netdev->ifindex); | |
1829 | return -1; | |
1830 | } | |
1831 | ||
1832 | /* default: let the system to choose one interface name */ | |
1833 | if (!netdev->name) | |
1834 | netdev->name = netdev->type == LXC_NET_PHYS ? | |
1835 | netdev->link : "eth%d"; | |
1836 | ||
1837 | /* rename the interface name */ | |
1838 | err = lxc_netdev_rename_by_name(ifname, netdev->name); | |
1839 | if (err) { | |
1840 | ERROR("failed to rename %s->%s : %s", ifname, netdev->name, | |
1841 | strerror(-err)); | |
1842 | return -1; | |
1843 | } | |
1844 | ||
1845 | /* Re-read the name of the interface because its name has changed | |
1846 | * and would be automatically allocated by the system | |
1847 | */ | |
1848 | if (!if_indextoname(netdev->ifindex, current_ifname)) { | |
1849 | ERROR("no interface corresponding to index '%d'", | |
1850 | netdev->ifindex); | |
1851 | return -1; | |
1852 | } | |
1853 | ||
1854 | /* set a mac address */ | |
1855 | if (netdev->hwaddr) { | |
1856 | if (setup_hw_addr(netdev->hwaddr, current_ifname)) { | |
1857 | ERROR("failed to setup hw address for '%s'", | |
1858 | current_ifname); | |
1859 | return -1; | |
1860 | } | |
1861 | } | |
1862 | ||
1863 | /* setup ipv4 addresses on the interface */ | |
1864 | if (setup_ipv4_addr(&netdev->ipv4, netdev->ifindex)) { | |
1865 | ERROR("failed to setup ip addresses for '%s'", | |
1866 | ifname); | |
1867 | return -1; | |
1868 | } | |
1869 | ||
1870 | /* setup ipv6 addresses on the interface */ | |
1871 | if (setup_ipv6_addr(&netdev->ipv6, netdev->ifindex)) { | |
1872 | ERROR("failed to setup ipv6 addresses for '%s'", | |
1873 | ifname); | |
1874 | return -1; | |
1875 | } | |
1876 | ||
1877 | /* set the network device up */ | |
1878 | if (netdev->flags & IFF_UP) { | |
1879 | int err; | |
1880 | ||
1881 | err = lxc_netdev_up(current_ifname); | |
1882 | if (err) { | |
1883 | ERROR("failed to set '%s' up : %s", current_ifname, | |
1884 | strerror(-err)); | |
1885 | return -1; | |
1886 | } | |
1887 | ||
1888 | /* the network is up, make the loopback up too */ | |
1889 | err = lxc_netdev_up("lo"); | |
1890 | if (err) { | |
1891 | ERROR("failed to set the loopback up : %s", | |
1892 | strerror(-err)); | |
1893 | return -1; | |
1894 | } | |
1895 | } | |
1896 | ||
1897 | /* We can only set up the default routes after bringing | |
1898 | * up the interface, sine bringing up the interface adds | |
1899 | * the link-local routes and we can't add a default | |
1900 | * route if the gateway is not reachable. */ | |
1901 | ||
1902 | /* setup ipv4 gateway on the interface */ | |
1903 | if (netdev->ipv4_gateway) { | |
1904 | if (!(netdev->flags & IFF_UP)) { | |
1905 | ERROR("Cannot add ipv4 gateway for %s when not bringing up the interface", ifname); | |
1906 | return -1; | |
1907 | } | |
1908 | ||
1909 | if (lxc_list_empty(&netdev->ipv4)) { | |
1910 | ERROR("Cannot add ipv4 gateway for %s when not assigning an address", ifname); | |
1911 | return -1; | |
1912 | } | |
1913 | ||
1914 | err = lxc_ipv4_gateway_add(netdev->ifindex, netdev->ipv4_gateway); | |
1915 | if (err) { | |
1916 | ERROR("failed to setup ipv4 gateway for '%s': %s", | |
1917 | ifname, strerror(-err)); | |
1918 | if (netdev->ipv4_gateway_auto) { | |
1919 | char buf[INET_ADDRSTRLEN]; | |
1920 | inet_ntop(AF_INET, netdev->ipv4_gateway, buf, sizeof(buf)); | |
1921 | ERROR("tried to set autodetected ipv4 gateway '%s'", buf); | |
1922 | } | |
1923 | return -1; | |
1924 | } | |
1925 | } | |
1926 | ||
1927 | /* setup ipv6 gateway on the interface */ | |
1928 | if (netdev->ipv6_gateway) { | |
1929 | if (!(netdev->flags & IFF_UP)) { | |
1930 | ERROR("Cannot add ipv6 gateway for %s when not bringing up the interface", ifname); | |
1931 | return -1; | |
1932 | } | |
1933 | ||
1934 | if (lxc_list_empty(&netdev->ipv6) && !IN6_IS_ADDR_LINKLOCAL(netdev->ipv6_gateway)) { | |
1935 | ERROR("Cannot add ipv6 gateway for %s when not assigning an address", ifname); | |
1936 | return -1; | |
1937 | } | |
1938 | ||
1939 | err = lxc_ipv6_gateway_add(netdev->ifindex, netdev->ipv6_gateway); | |
1940 | if (err) { | |
1941 | ERROR("failed to setup ipv6 gateway for '%s': %s", | |
1942 | ifname, strerror(-err)); | |
1943 | if (netdev->ipv6_gateway_auto) { | |
1944 | char buf[INET6_ADDRSTRLEN]; | |
1945 | inet_ntop(AF_INET6, netdev->ipv6_gateway, buf, sizeof(buf)); | |
1946 | ERROR("tried to set autodetected ipv6 gateway '%s'", buf); | |
1947 | } | |
1948 | return -1; | |
1949 | } | |
1950 | } | |
1951 | ||
1952 | DEBUG("'%s' has been setup", current_ifname); | |
1953 | ||
1954 | return 0; | |
1955 | } | |
1956 | ||
1957 | static int setup_network(struct lxc_list *network) | |
1958 | { | |
1959 | struct lxc_list *iterator; | |
1960 | struct lxc_netdev *netdev; | |
1961 | ||
1962 | lxc_list_for_each(iterator, network) { | |
1963 | ||
1964 | netdev = iterator->elem; | |
1965 | ||
1966 | if (setup_netdev(netdev)) { | |
1967 | ERROR("failed to setup netdev"); | |
1968 | return -1; | |
1969 | } | |
1970 | } | |
1971 | ||
1972 | if (!lxc_list_empty(network)) | |
1973 | INFO("network has been setup"); | |
1974 | ||
1975 | return 0; | |
1976 | } | |
1977 | ||
1978 | void lxc_rename_phys_nics_on_shutdown(struct lxc_conf *conf) | |
1979 | { | |
1980 | int i; | |
1981 | ||
1982 | INFO("running to reset %d nic names", conf->num_savednics); | |
1983 | for (i=0; i<conf->num_savednics; i++) { | |
1984 | struct saved_nic *s = &conf->saved_nics[i]; | |
1985 | INFO("resetting nic %d to %s\n", s->ifindex, s->orig_name); | |
1986 | lxc_netdev_rename_by_index(s->ifindex, s->orig_name); | |
1987 | free(s->orig_name); | |
1988 | } | |
1989 | conf->num_savednics = 0; | |
1990 | free(conf->saved_nics); | |
1991 | } | |
1992 | ||
1993 | static int setup_private_host_hw_addr(char *veth1) | |
1994 | { | |
1995 | struct ifreq ifr; | |
1996 | int err; | |
1997 | int sockfd; | |
1998 | ||
1999 | sockfd = socket(AF_INET, SOCK_DGRAM, 0); | |
2000 | if (sockfd < 0) | |
2001 | return -errno; | |
2002 | ||
2003 | snprintf((char *)ifr.ifr_name, IFNAMSIZ, "%s", veth1); | |
2004 | err = ioctl(sockfd, SIOCGIFHWADDR, &ifr); | |
2005 | if (err < 0) { | |
2006 | close(sockfd); | |
2007 | return -errno; | |
2008 | } | |
2009 | ||
2010 | ifr.ifr_hwaddr.sa_data[0] = 0xfe; | |
2011 | err = ioctl(sockfd, SIOCSIFHWADDR, &ifr); | |
2012 | close(sockfd); | |
2013 | if (err < 0) | |
2014 | return -errno; | |
2015 | ||
2016 | DEBUG("mac address of host interface '%s' changed to private " | |
2017 | "%02x:%02x:%02x:%02x:%02x:%02x", veth1, | |
2018 | ifr.ifr_hwaddr.sa_data[0] & 0xff, | |
2019 | ifr.ifr_hwaddr.sa_data[1] & 0xff, | |
2020 | ifr.ifr_hwaddr.sa_data[2] & 0xff, | |
2021 | ifr.ifr_hwaddr.sa_data[3] & 0xff, | |
2022 | ifr.ifr_hwaddr.sa_data[4] & 0xff, | |
2023 | ifr.ifr_hwaddr.sa_data[5] & 0xff); | |
2024 | ||
2025 | return 0; | |
2026 | } | |
2027 | ||
2028 | static char *default_rootfs_mount = LXCROOTFSMOUNT; | |
2029 | ||
2030 | struct lxc_conf *lxc_conf_init(void) | |
2031 | { | |
2032 | struct lxc_conf *new; | |
2033 | int i; | |
2034 | ||
2035 | new = malloc(sizeof(*new)); | |
2036 | if (!new) { | |
2037 | ERROR("lxc_conf_init : %m"); | |
2038 | return NULL; | |
2039 | } | |
2040 | memset(new, 0, sizeof(*new)); | |
2041 | ||
2042 | new->personality = -1; | |
2043 | new->console.log_path = NULL; | |
2044 | new->console.log_fd = -1; | |
2045 | new->console.path = NULL; | |
2046 | new->console.peer = -1; | |
2047 | new->console.master = -1; | |
2048 | new->console.slave = -1; | |
2049 | new->console.name[0] = '\0'; | |
2050 | new->maincmd_fd = -1; | |
2051 | new->rootfs.mount = default_rootfs_mount; | |
2052 | lxc_list_init(&new->cgroup); | |
2053 | lxc_list_init(&new->network); | |
2054 | lxc_list_init(&new->mount_list); | |
2055 | lxc_list_init(&new->caps); | |
2056 | for (i=0; i<NUM_LXC_HOOKS; i++) | |
2057 | lxc_list_init(&new->hooks[i]); | |
2058 | #if HAVE_APPARMOR | |
2059 | new->aa_profile = NULL; | |
2060 | #endif | |
2061 | #if HAVE_APPARMOR /* || HAVE_SMACK || HAVE_SELINUX */ | |
2062 | new->lsm_umount_proc = 0; | |
2063 | #endif | |
2064 | ||
2065 | return new; | |
2066 | } | |
2067 | ||
2068 | static int instanciate_veth(struct lxc_handler *handler, struct lxc_netdev *netdev) | |
2069 | { | |
2070 | char veth1buf[IFNAMSIZ], *veth1; | |
2071 | char veth2buf[IFNAMSIZ], *veth2; | |
2072 | int err; | |
2073 | ||
2074 | if (netdev->priv.veth_attr.pair) | |
2075 | veth1 = netdev->priv.veth_attr.pair; | |
2076 | else { | |
2077 | err = snprintf(veth1buf, sizeof(veth1buf), "vethXXXXXX"); | |
2078 | if (err >= sizeof(veth1buf)) { /* can't *really* happen, but... */ | |
2079 | ERROR("veth1 name too long"); | |
2080 | return -1; | |
2081 | } | |
2082 | veth1 = mktemp(veth1buf); | |
2083 | /* store away for deconf */ | |
2084 | memcpy(netdev->priv.veth_attr.veth1, veth1, IFNAMSIZ); | |
2085 | } | |
2086 | ||
2087 | snprintf(veth2buf, sizeof(veth2buf), "vethXXXXXX"); | |
2088 | veth2 = mktemp(veth2buf); | |
2089 | ||
2090 | if (!strlen(veth1) || !strlen(veth2)) { | |
2091 | ERROR("failed to allocate a temporary name"); | |
2092 | return -1; | |
2093 | } | |
2094 | ||
2095 | err = lxc_veth_create(veth1, veth2); | |
2096 | if (err) { | |
2097 | ERROR("failed to create %s-%s : %s", veth1, veth2, | |
2098 | strerror(-err)); | |
2099 | return -1; | |
2100 | } | |
2101 | ||
2102 | /* changing the high byte of the mac address to 0xfe, the bridge interface | |
2103 | * will always keep the host's mac address and not take the mac address | |
2104 | * of a container */ | |
2105 | err = setup_private_host_hw_addr(veth1); | |
2106 | if (err) { | |
2107 | ERROR("failed to change mac address of host interface '%s' : %s", | |
2108 | veth1, strerror(-err)); | |
2109 | goto out_delete; | |
2110 | } | |
2111 | ||
2112 | if (netdev->mtu) { | |
2113 | err = lxc_netdev_set_mtu(veth1, atoi(netdev->mtu)); | |
2114 | if (!err) | |
2115 | err = lxc_netdev_set_mtu(veth2, atoi(netdev->mtu)); | |
2116 | if (err) { | |
2117 | ERROR("failed to set mtu '%s' for %s-%s : %s", | |
2118 | netdev->mtu, veth1, veth2, strerror(-err)); | |
2119 | goto out_delete; | |
2120 | } | |
2121 | } | |
2122 | ||
2123 | if (netdev->link) { | |
2124 | err = lxc_bridge_attach(netdev->link, veth1); | |
2125 | if (err) { | |
2126 | ERROR("failed to attach '%s' to the bridge '%s' : %s", | |
2127 | veth1, netdev->link, strerror(-err)); | |
2128 | goto out_delete; | |
2129 | } | |
2130 | } | |
2131 | ||
2132 | netdev->ifindex = if_nametoindex(veth2); | |
2133 | if (!netdev->ifindex) { | |
2134 | ERROR("failed to retrieve the index for %s", veth2); | |
2135 | goto out_delete; | |
2136 | } | |
2137 | ||
2138 | err = lxc_netdev_up(veth1); | |
2139 | if (err) { | |
2140 | ERROR("failed to set %s up : %s", veth1, strerror(-err)); | |
2141 | goto out_delete; | |
2142 | } | |
2143 | ||
2144 | if (netdev->upscript) { | |
2145 | err = run_script(handler->name, "net", netdev->upscript, "up", | |
2146 | "veth", veth1, (char*) NULL); | |
2147 | if (err) | |
2148 | goto out_delete; | |
2149 | } | |
2150 | ||
2151 | DEBUG("instanciated veth '%s/%s', index is '%d'", | |
2152 | veth1, veth2, netdev->ifindex); | |
2153 | ||
2154 | return 0; | |
2155 | ||
2156 | out_delete: | |
2157 | lxc_netdev_delete_by_name(veth1); | |
2158 | return -1; | |
2159 | } | |
2160 | ||
2161 | static int shutdown_veth(struct lxc_handler *handler, struct lxc_netdev *netdev) | |
2162 | { | |
2163 | char *veth1; | |
2164 | int err; | |
2165 | ||
2166 | if (netdev->priv.veth_attr.pair) | |
2167 | veth1 = netdev->priv.veth_attr.pair; | |
2168 | else | |
2169 | veth1 = netdev->priv.veth_attr.veth1; | |
2170 | ||
2171 | if (netdev->downscript) { | |
2172 | err = run_script(handler->name, "net", netdev->downscript, | |
2173 | "down", "veth", veth1, (char*) NULL); | |
2174 | if (err) | |
2175 | return -1; | |
2176 | } | |
2177 | return 0; | |
2178 | } | |
2179 | ||
2180 | static int instanciate_macvlan(struct lxc_handler *handler, struct lxc_netdev *netdev) | |
2181 | { | |
2182 | char peerbuf[IFNAMSIZ], *peer; | |
2183 | int err; | |
2184 | ||
2185 | if (!netdev->link) { | |
2186 | ERROR("no link specified for macvlan netdev"); | |
2187 | return -1; | |
2188 | } | |
2189 | ||
2190 | err = snprintf(peerbuf, sizeof(peerbuf), "mcXXXXXX"); | |
2191 | if (err >= sizeof(peerbuf)) | |
2192 | return -1; | |
2193 | ||
2194 | peer = mktemp(peerbuf); | |
2195 | if (!strlen(peer)) { | |
2196 | ERROR("failed to make a temporary name"); | |
2197 | return -1; | |
2198 | } | |
2199 | ||
2200 | err = lxc_macvlan_create(netdev->link, peer, | |
2201 | netdev->priv.macvlan_attr.mode); | |
2202 | if (err) { | |
2203 | ERROR("failed to create macvlan interface '%s' on '%s' : %s", | |
2204 | peer, netdev->link, strerror(-err)); | |
2205 | return -1; | |
2206 | } | |
2207 | ||
2208 | netdev->ifindex = if_nametoindex(peer); | |
2209 | if (!netdev->ifindex) { | |
2210 | ERROR("failed to retrieve the index for %s", peer); | |
2211 | lxc_netdev_delete_by_name(peer); | |
2212 | return -1; | |
2213 | } | |
2214 | ||
2215 | if (netdev->upscript) { | |
2216 | err = run_script(handler->name, "net", netdev->upscript, "up", | |
2217 | "macvlan", netdev->link, (char*) NULL); | |
2218 | if (err) | |
2219 | return -1; | |
2220 | } | |
2221 | ||
2222 | DEBUG("instanciated macvlan '%s', index is '%d' and mode '%d'", | |
2223 | peer, netdev->ifindex, netdev->priv.macvlan_attr.mode); | |
2224 | ||
2225 | return 0; | |
2226 | } | |
2227 | ||
2228 | static int shutdown_macvlan(struct lxc_handler *handler, struct lxc_netdev *netdev) | |
2229 | { | |
2230 | int err; | |
2231 | ||
2232 | if (netdev->downscript) { | |
2233 | err = run_script(handler->name, "net", netdev->downscript, | |
2234 | "down", "macvlan", netdev->link, | |
2235 | (char*) NULL); | |
2236 | if (err) | |
2237 | return -1; | |
2238 | } | |
2239 | return 0; | |
2240 | } | |
2241 | ||
2242 | /* XXX: merge with instanciate_macvlan */ | |
2243 | static int instanciate_vlan(struct lxc_handler *handler, struct lxc_netdev *netdev) | |
2244 | { | |
2245 | char peer[IFNAMSIZ]; | |
2246 | int err; | |
2247 | ||
2248 | if (!netdev->link) { | |
2249 | ERROR("no link specified for vlan netdev"); | |
2250 | return -1; | |
2251 | } | |
2252 | ||
2253 | err = snprintf(peer, sizeof(peer), "vlan%d", netdev->priv.vlan_attr.vid); | |
2254 | if (err >= sizeof(peer)) { | |
2255 | ERROR("peer name too long"); | |
2256 | return -1; | |
2257 | } | |
2258 | ||
2259 | err = lxc_vlan_create(netdev->link, peer, netdev->priv.vlan_attr.vid); | |
2260 | if (err) { | |
2261 | ERROR("failed to create vlan interface '%s' on '%s' : %s", | |
2262 | peer, netdev->link, strerror(-err)); | |
2263 | return -1; | |
2264 | } | |
2265 | ||
2266 | netdev->ifindex = if_nametoindex(peer); | |
2267 | if (!netdev->ifindex) { | |
2268 | ERROR("failed to retrieve the ifindex for %s", peer); | |
2269 | lxc_netdev_delete_by_name(peer); | |
2270 | return -1; | |
2271 | } | |
2272 | ||
2273 | DEBUG("instanciated vlan '%s', ifindex is '%d'", " vlan1000", | |
2274 | netdev->ifindex); | |
2275 | ||
2276 | return 0; | |
2277 | } | |
2278 | ||
2279 | static int shutdown_vlan(struct lxc_handler *handler, struct lxc_netdev *netdev) | |
2280 | { | |
2281 | return 0; | |
2282 | } | |
2283 | ||
2284 | static int instanciate_phys(struct lxc_handler *handler, struct lxc_netdev *netdev) | |
2285 | { | |
2286 | if (!netdev->link) { | |
2287 | ERROR("no link specified for the physical interface"); | |
2288 | return -1; | |
2289 | } | |
2290 | ||
2291 | netdev->ifindex = if_nametoindex(netdev->link); | |
2292 | if (!netdev->ifindex) { | |
2293 | ERROR("failed to retrieve the index for %s", netdev->link); | |
2294 | return -1; | |
2295 | } | |
2296 | ||
2297 | if (netdev->upscript) { | |
2298 | int err; | |
2299 | err = run_script(handler->name, "net", netdev->upscript, | |
2300 | "up", "phys", netdev->link, (char*) NULL); | |
2301 | if (err) | |
2302 | return -1; | |
2303 | } | |
2304 | ||
2305 | return 0; | |
2306 | } | |
2307 | ||
2308 | static int shutdown_phys(struct lxc_handler *handler, struct lxc_netdev *netdev) | |
2309 | { | |
2310 | int err; | |
2311 | ||
2312 | if (netdev->downscript) { | |
2313 | err = run_script(handler->name, "net", netdev->downscript, | |
2314 | "down", "phys", netdev->link, (char*) NULL); | |
2315 | if (err) | |
2316 | return -1; | |
2317 | } | |
2318 | return 0; | |
2319 | } | |
2320 | ||
2321 | static int instanciate_empty(struct lxc_handler *handler, struct lxc_netdev *netdev) | |
2322 | { | |
2323 | netdev->ifindex = 0; | |
2324 | if (netdev->upscript) { | |
2325 | int err; | |
2326 | err = run_script(handler->name, "net", netdev->upscript, | |
2327 | "up", "empty", (char*) NULL); | |
2328 | if (err) | |
2329 | return -1; | |
2330 | } | |
2331 | return 0; | |
2332 | } | |
2333 | ||
2334 | static int shutdown_empty(struct lxc_handler *handler, struct lxc_netdev *netdev) | |
2335 | { | |
2336 | int err; | |
2337 | ||
2338 | if (netdev->downscript) { | |
2339 | err = run_script(handler->name, "net", netdev->downscript, | |
2340 | "down", "empty", (char*) NULL); | |
2341 | if (err) | |
2342 | return -1; | |
2343 | } | |
2344 | return 0; | |
2345 | } | |
2346 | ||
2347 | int lxc_create_network(struct lxc_handler *handler) | |
2348 | { | |
2349 | struct lxc_list *network = &handler->conf->network; | |
2350 | struct lxc_list *iterator; | |
2351 | struct lxc_netdev *netdev; | |
2352 | ||
2353 | lxc_list_for_each(iterator, network) { | |
2354 | ||
2355 | netdev = iterator->elem; | |
2356 | ||
2357 | if (netdev->type < 0 || netdev->type > LXC_NET_MAXCONFTYPE) { | |
2358 | ERROR("invalid network configuration type '%d'", | |
2359 | netdev->type); | |
2360 | return -1; | |
2361 | } | |
2362 | ||
2363 | if (netdev_conf[netdev->type](handler, netdev)) { | |
2364 | ERROR("failed to create netdev"); | |
2365 | return -1; | |
2366 | } | |
2367 | ||
2368 | } | |
2369 | ||
2370 | return 0; | |
2371 | } | |
2372 | ||
2373 | void lxc_delete_network(struct lxc_handler *handler) | |
2374 | { | |
2375 | struct lxc_list *network = &handler->conf->network; | |
2376 | struct lxc_list *iterator; | |
2377 | struct lxc_netdev *netdev; | |
2378 | ||
2379 | lxc_list_for_each(iterator, network) { | |
2380 | netdev = iterator->elem; | |
2381 | ||
2382 | if (netdev->ifindex != 0 && netdev->type == LXC_NET_PHYS) { | |
2383 | if (lxc_netdev_rename_by_index(netdev->ifindex, netdev->link)) | |
2384 | WARN("failed to rename to the initial name the " \ | |
2385 | "netdev '%s'", netdev->link); | |
2386 | continue; | |
2387 | } | |
2388 | ||
2389 | if (netdev_deconf[netdev->type](handler, netdev)) { | |
2390 | WARN("failed to destroy netdev"); | |
2391 | } | |
2392 | ||
2393 | /* Recent kernel remove the virtual interfaces when the network | |
2394 | * namespace is destroyed but in case we did not moved the | |
2395 | * interface to the network namespace, we have to destroy it | |
2396 | */ | |
2397 | if (netdev->ifindex != 0 && | |
2398 | lxc_netdev_delete_by_index(netdev->ifindex)) | |
2399 | WARN("failed to remove interface '%s'", netdev->name); | |
2400 | } | |
2401 | } | |
2402 | ||
2403 | int lxc_assign_network(struct lxc_list *network, pid_t pid) | |
2404 | { | |
2405 | struct lxc_list *iterator; | |
2406 | struct lxc_netdev *netdev; | |
2407 | int err; | |
2408 | ||
2409 | lxc_list_for_each(iterator, network) { | |
2410 | ||
2411 | netdev = iterator->elem; | |
2412 | ||
2413 | /* empty network namespace, nothing to move */ | |
2414 | if (!netdev->ifindex) | |
2415 | continue; | |
2416 | ||
2417 | err = lxc_netdev_move_by_index(netdev->ifindex, pid); | |
2418 | if (err) { | |
2419 | ERROR("failed to move '%s' to the container : %s", | |
2420 | netdev->link, strerror(-err)); | |
2421 | return -1; | |
2422 | } | |
2423 | ||
2424 | DEBUG("move '%s' to '%d'", netdev->name, pid); | |
2425 | } | |
2426 | ||
2427 | return 0; | |
2428 | } | |
2429 | ||
2430 | int lxc_find_gateway_addresses(struct lxc_handler *handler) | |
2431 | { | |
2432 | struct lxc_list *network = &handler->conf->network; | |
2433 | struct lxc_list *iterator; | |
2434 | struct lxc_netdev *netdev; | |
2435 | int link_index; | |
2436 | ||
2437 | lxc_list_for_each(iterator, network) { | |
2438 | netdev = iterator->elem; | |
2439 | ||
2440 | if (!netdev->ipv4_gateway_auto && !netdev->ipv6_gateway_auto) | |
2441 | continue; | |
2442 | ||
2443 | if (netdev->type != LXC_NET_VETH && netdev->type != LXC_NET_MACVLAN) { | |
2444 | ERROR("gateway = auto only supported for " | |
2445 | "veth and macvlan"); | |
2446 | return -1; | |
2447 | } | |
2448 | ||
2449 | if (!netdev->link) { | |
2450 | ERROR("gateway = auto needs a link interface"); | |
2451 | return -1; | |
2452 | } | |
2453 | ||
2454 | link_index = if_nametoindex(netdev->link); | |
2455 | if (!link_index) | |
2456 | return -EINVAL; | |
2457 | ||
2458 | if (netdev->ipv4_gateway_auto) { | |
2459 | if (lxc_ipv4_addr_get(link_index, &netdev->ipv4_gateway)) { | |
2460 | ERROR("failed to automatically find ipv4 gateway " | |
2461 | "address from link interface '%s'", netdev->link); | |
2462 | return -1; | |
2463 | } | |
2464 | } | |
2465 | ||
2466 | if (netdev->ipv6_gateway_auto) { | |
2467 | if (lxc_ipv6_addr_get(link_index, &netdev->ipv6_gateway)) { | |
2468 | ERROR("failed to automatically find ipv6 gateway " | |
2469 | "address from link interface '%s'", netdev->link); | |
2470 | return -1; | |
2471 | } | |
2472 | } | |
2473 | } | |
2474 | ||
2475 | return 0; | |
2476 | } | |
2477 | ||
2478 | int lxc_create_tty(const char *name, struct lxc_conf *conf) | |
2479 | { | |
2480 | struct lxc_tty_info *tty_info = &conf->tty_info; | |
2481 | int i; | |
2482 | ||
2483 | /* no tty in the configuration */ | |
2484 | if (!conf->tty) | |
2485 | return 0; | |
2486 | ||
2487 | tty_info->pty_info = | |
2488 | malloc(sizeof(*tty_info->pty_info)*conf->tty); | |
2489 | if (!tty_info->pty_info) { | |
2490 | SYSERROR("failed to allocate pty_info"); | |
2491 | return -1; | |
2492 | } | |
2493 | ||
2494 | for (i = 0; i < conf->tty; i++) { | |
2495 | ||
2496 | struct lxc_pty_info *pty_info = &tty_info->pty_info[i]; | |
2497 | ||
2498 | if (openpty(&pty_info->master, &pty_info->slave, | |
2499 | pty_info->name, NULL, NULL)) { | |
2500 | SYSERROR("failed to create pty #%d", i); | |
2501 | tty_info->nbtty = i; | |
2502 | lxc_delete_tty(tty_info); | |
2503 | return -1; | |
2504 | } | |
2505 | ||
2506 | DEBUG("allocated pty '%s' (%d/%d)", | |
2507 | pty_info->name, pty_info->master, pty_info->slave); | |
2508 | ||
2509 | /* Prevent leaking the file descriptors to the container */ | |
2510 | fcntl(pty_info->master, F_SETFD, FD_CLOEXEC); | |
2511 | fcntl(pty_info->slave, F_SETFD, FD_CLOEXEC); | |
2512 | ||
2513 | pty_info->busy = 0; | |
2514 | } | |
2515 | ||
2516 | tty_info->nbtty = conf->tty; | |
2517 | ||
2518 | INFO("tty's configured"); | |
2519 | ||
2520 | return 0; | |
2521 | } | |
2522 | ||
2523 | void lxc_delete_tty(struct lxc_tty_info *tty_info) | |
2524 | { | |
2525 | int i; | |
2526 | ||
2527 | for (i = 0; i < tty_info->nbtty; i++) { | |
2528 | struct lxc_pty_info *pty_info = &tty_info->pty_info[i]; | |
2529 | ||
2530 | close(pty_info->master); | |
2531 | close(pty_info->slave); | |
2532 | } | |
2533 | ||
2534 | free(tty_info->pty_info); | |
2535 | tty_info->nbtty = 0; | |
2536 | } | |
2537 | ||
2538 | int lxc_setup(const char *name, struct lxc_conf *lxc_conf) | |
2539 | { | |
2540 | #if HAVE_APPARMOR /* || HAVE_SMACK || HAVE_SELINUX */ | |
2541 | int mounted; | |
2542 | #endif | |
2543 | ||
2544 | if (setup_utsname(lxc_conf->utsname)) { | |
2545 | ERROR("failed to setup the utsname for '%s'", name); | |
2546 | return -1; | |
2547 | } | |
2548 | ||
2549 | if (setup_network(&lxc_conf->network)) { | |
2550 | ERROR("failed to setup the network for '%s'", name); | |
2551 | return -1; | |
2552 | } | |
2553 | ||
2554 | if (run_lxc_hooks(name, "pre-mount", lxc_conf)) { | |
2555 | ERROR("failed to run pre-mount hooks for container '%s'.", name); | |
2556 | return -1; | |
2557 | } | |
2558 | ||
2559 | if (setup_rootfs(lxc_conf)) { | |
2560 | ERROR("failed to setup rootfs for '%s'", name); | |
2561 | return -1; | |
2562 | } | |
2563 | ||
2564 | if (lxc_conf->autodev) { | |
2565 | if (mount_autodev(lxc_conf->rootfs.mount)) { | |
2566 | ERROR("failed to mount /dev in the container"); | |
2567 | return -1; | |
2568 | } | |
2569 | } | |
2570 | ||
2571 | if (setup_mount(&lxc_conf->rootfs, lxc_conf->fstab, name)) { | |
2572 | ERROR("failed to setup the mounts for '%s'", name); | |
2573 | return -1; | |
2574 | } | |
2575 | ||
2576 | if (!lxc_list_empty(&lxc_conf->mount_list) && setup_mount_entries(&lxc_conf->rootfs, &lxc_conf->mount_list, name)) { | |
2577 | ERROR("failed to setup the mount entries for '%s'", name); | |
2578 | return -1; | |
2579 | } | |
2580 | ||
2581 | if (run_lxc_hooks(name, "mount", lxc_conf)) { | |
2582 | ERROR("failed to run mount hooks for container '%s'.", name); | |
2583 | return -1; | |
2584 | } | |
2585 | ||
2586 | if (lxc_conf->autodev) { | |
2587 | if (run_lxc_hooks(name, "autodev", lxc_conf)) { | |
2588 | ERROR("failed to run autodev hooks for container '%s'.", name); | |
2589 | return -1; | |
2590 | } | |
2591 | if (setup_autodev(lxc_conf->rootfs.mount)) { | |
2592 | ERROR("failed to populate /dev in the container"); | |
2593 | return -1; | |
2594 | } | |
2595 | } | |
2596 | ||
2597 | if (setup_console(&lxc_conf->rootfs, &lxc_conf->console, lxc_conf->ttydir)) { | |
2598 | ERROR("failed to setup the console for '%s'", name); | |
2599 | return -1; | |
2600 | } | |
2601 | ||
2602 | if (setup_kmsg(&lxc_conf->rootfs, &lxc_conf->console)) // don't fail | |
2603 | ERROR("failed to setup kmsg for '%s'", name); | |
2604 | ||
2605 | if (setup_tty(&lxc_conf->rootfs, &lxc_conf->tty_info, lxc_conf->ttydir)) { | |
2606 | ERROR("failed to setup the ttys for '%s'", name); | |
2607 | return -1; | |
2608 | } | |
2609 | ||
2610 | #if HAVE_APPARMOR /* || HAVE_SMACK || HAVE_SELINUX */ | |
2611 | INFO("rootfs path is .%s., mount is .%s.", lxc_conf->rootfs.path, | |
2612 | lxc_conf->rootfs.mount); | |
2613 | if (lxc_conf->rootfs.path == NULL || strlen(lxc_conf->rootfs.path) == 0) | |
2614 | mounted = 0; | |
2615 | else | |
2616 | mounted = lsm_mount_proc_if_needed(lxc_conf->rootfs.path, lxc_conf->rootfs.mount); | |
2617 | if (mounted == -1) { | |
2618 | SYSERROR("failed to mount /proc in the container."); | |
2619 | return -1; | |
2620 | } else if (mounted == 1) { | |
2621 | lxc_conf->lsm_umount_proc = 1; | |
2622 | } | |
2623 | #endif | |
2624 | ||
2625 | if (setup_pivot_root(&lxc_conf->rootfs)) { | |
2626 | ERROR("failed to set rootfs for '%s'", name); | |
2627 | return -1; | |
2628 | } | |
2629 | ||
2630 | if (setup_pts(lxc_conf->pts)) { | |
2631 | ERROR("failed to setup the new pts instance"); | |
2632 | return -1; | |
2633 | } | |
2634 | ||
2635 | if (setup_personality(lxc_conf->personality)) { | |
2636 | ERROR("failed to setup personality"); | |
2637 | return -1; | |
2638 | } | |
2639 | ||
2640 | if (setup_caps(&lxc_conf->caps)) { | |
2641 | ERROR("failed to drop capabilities"); | |
2642 | return -1; | |
2643 | } | |
2644 | ||
2645 | NOTICE("'%s' is setup.", name); | |
2646 | ||
2647 | return 0; | |
2648 | } | |
2649 | ||
2650 | int run_lxc_hooks(const char *name, char *hook, struct lxc_conf *conf) | |
2651 | { | |
2652 | int which = -1; | |
2653 | struct lxc_list *it; | |
2654 | ||
2655 | if (strcmp(hook, "pre-start") == 0) | |
2656 | which = LXCHOOK_PRESTART; | |
2657 | else if (strcmp(hook, "pre-mount") == 0) | |
2658 | which = LXCHOOK_PREMOUNT; | |
2659 | else if (strcmp(hook, "mount") == 0) | |
2660 | which = LXCHOOK_MOUNT; | |
2661 | else if (strcmp(hook, "autodev") == 0) | |
2662 | which = LXCHOOK_AUTODEV; | |
2663 | else if (strcmp(hook, "start") == 0) | |
2664 | which = LXCHOOK_START; | |
2665 | else if (strcmp(hook, "post-stop") == 0) | |
2666 | which = LXCHOOK_POSTSTOP; | |
2667 | else | |
2668 | return -1; | |
2669 | lxc_list_for_each(it, &conf->hooks[which]) { | |
2670 | int ret; | |
2671 | char *hookname = it->elem; | |
2672 | ret = run_script(name, "lxc", hookname, hook, NULL); | |
2673 | if (ret) | |
2674 | return ret; | |
2675 | } | |
2676 | return 0; | |
2677 | } | |
2678 | ||
2679 | static void lxc_remove_nic(struct lxc_list *it) | |
2680 | { | |
2681 | struct lxc_netdev *netdev = it->elem; | |
2682 | struct lxc_list *it2,*next; | |
2683 | ||
2684 | lxc_list_del(it); | |
2685 | ||
2686 | if (netdev->link) | |
2687 | free(netdev->link); | |
2688 | if (netdev->name) | |
2689 | free(netdev->name); | |
2690 | if (netdev->upscript) | |
2691 | free(netdev->upscript); | |
2692 | if (netdev->hwaddr) | |
2693 | free(netdev->hwaddr); | |
2694 | if (netdev->mtu) | |
2695 | free(netdev->mtu); | |
2696 | if (netdev->ipv4_gateway) | |
2697 | free(netdev->ipv4_gateway); | |
2698 | if (netdev->ipv6_gateway) | |
2699 | free(netdev->ipv6_gateway); | |
2700 | lxc_list_for_each_safe(it2, &netdev->ipv4, next) { | |
2701 | lxc_list_del(it2); | |
2702 | free(it2->elem); | |
2703 | free(it2); | |
2704 | } | |
2705 | lxc_list_for_each_safe(it2, &netdev->ipv6, next) { | |
2706 | lxc_list_del(it2); | |
2707 | free(it2->elem); | |
2708 | free(it2); | |
2709 | } | |
2710 | free(netdev); | |
2711 | free(it); | |
2712 | } | |
2713 | ||
2714 | /* we get passed in something like '0', '0.ipv4' or '1.ipv6' */ | |
2715 | int lxc_clear_nic(struct lxc_conf *c, const char *key) | |
2716 | { | |
2717 | char *p1; | |
2718 | int ret, idx, i; | |
2719 | struct lxc_list *it; | |
2720 | struct lxc_netdev *netdev; | |
2721 | ||
2722 | p1 = index(key, '.'); | |
2723 | if (!p1 || *(p1+1) == '\0') | |
2724 | p1 = NULL; | |
2725 | ||
2726 | ret = sscanf(key, "%d", &idx); | |
2727 | if (ret != 1) return -1; | |
2728 | if (idx < 0) | |
2729 | return -1; | |
2730 | ||
2731 | i = 0; | |
2732 | lxc_list_for_each(it, &c->network) { | |
2733 | if (i == idx) | |
2734 | break; | |
2735 | i++; | |
2736 | } | |
2737 | if (i < idx) // we don't have that many nics defined | |
2738 | return -1; | |
2739 | ||
2740 | if (!it || !it->elem) | |
2741 | return -1; | |
2742 | ||
2743 | netdev = it->elem; | |
2744 | ||
2745 | if (!p1) { | |
2746 | lxc_remove_nic(it); | |
2747 | } else if (strcmp(p1, "ipv4") == 0) { | |
2748 | struct lxc_list *it2,*next; | |
2749 | lxc_list_for_each_safe(it2, &netdev->ipv4, next) { | |
2750 | lxc_list_del(it2); | |
2751 | free(it2->elem); | |
2752 | free(it2); | |
2753 | } | |
2754 | } else if (strcmp(p1, "ipv6") == 0) { | |
2755 | struct lxc_list *it2,*next; | |
2756 | lxc_list_for_each_safe(it2, &netdev->ipv6, next) { | |
2757 | lxc_list_del(it2); | |
2758 | free(it2->elem); | |
2759 | free(it2); | |
2760 | } | |
2761 | } else if (strcmp(p1, "link") == 0) { | |
2762 | if (netdev->link) { | |
2763 | free(netdev->link); | |
2764 | netdev->link = NULL; | |
2765 | } | |
2766 | } else if (strcmp(p1, "name") == 0) { | |
2767 | if (netdev->name) { | |
2768 | free(netdev->name); | |
2769 | netdev->name = NULL; | |
2770 | } | |
2771 | } else if (strcmp(p1, "script.up") == 0) { | |
2772 | if (netdev->upscript) { | |
2773 | free(netdev->upscript); | |
2774 | netdev->upscript = NULL; | |
2775 | } | |
2776 | } else if (strcmp(p1, "hwaddr") == 0) { | |
2777 | if (netdev->hwaddr) { | |
2778 | free(netdev->hwaddr); | |
2779 | netdev->hwaddr = NULL; | |
2780 | } | |
2781 | } else if (strcmp(p1, "mtu") == 0) { | |
2782 | if (netdev->mtu) { | |
2783 | free(netdev->mtu); | |
2784 | netdev->mtu = NULL; | |
2785 | } | |
2786 | } else if (strcmp(p1, "ipv4_gateway") == 0) { | |
2787 | if (netdev->ipv4_gateway) { | |
2788 | free(netdev->ipv4_gateway); | |
2789 | netdev->ipv4_gateway = NULL; | |
2790 | } | |
2791 | } else if (strcmp(p1, "ipv6_gateway") == 0) { | |
2792 | if (netdev->ipv6_gateway) { | |
2793 | free(netdev->ipv6_gateway); | |
2794 | netdev->ipv6_gateway = NULL; | |
2795 | } | |
2796 | } | |
2797 | else return -1; | |
2798 | ||
2799 | return 0; | |
2800 | } | |
2801 | ||
2802 | int lxc_clear_config_network(struct lxc_conf *c) | |
2803 | { | |
2804 | struct lxc_list *it,*next; | |
2805 | lxc_list_for_each_safe(it, &c->network, next) { | |
2806 | lxc_remove_nic(it); | |
2807 | } | |
2808 | return 0; | |
2809 | } | |
2810 | ||
2811 | int lxc_clear_config_caps(struct lxc_conf *c) | |
2812 | { | |
2813 | struct lxc_list *it,*next; | |
2814 | ||
2815 | lxc_list_for_each_safe(it, &c->caps, next) { | |
2816 | lxc_list_del(it); | |
2817 | free(it->elem); | |
2818 | free(it); | |
2819 | } | |
2820 | return 0; | |
2821 | } | |
2822 | ||
2823 | int lxc_clear_cgroups(struct lxc_conf *c, const char *key) | |
2824 | { | |
2825 | struct lxc_list *it,*next; | |
2826 | bool all = false; | |
2827 | const char *k = key + 11; | |
2828 | ||
2829 | if (strcmp(key, "lxc.cgroup") == 0) | |
2830 | all = true; | |
2831 | ||
2832 | lxc_list_for_each_safe(it, &c->cgroup, next) { | |
2833 | struct lxc_cgroup *cg = it->elem; | |
2834 | if (!all && strcmp(cg->subsystem, k) != 0) | |
2835 | continue; | |
2836 | lxc_list_del(it); | |
2837 | free(cg->subsystem); | |
2838 | free(cg->value); | |
2839 | free(cg); | |
2840 | free(it); | |
2841 | } | |
2842 | return 0; | |
2843 | } | |
2844 | ||
2845 | int lxc_clear_mount_entries(struct lxc_conf *c) | |
2846 | { | |
2847 | struct lxc_list *it,*next; | |
2848 | ||
2849 | lxc_list_for_each_safe(it, &c->mount_list, next) { | |
2850 | lxc_list_del(it); | |
2851 | free(it->elem); | |
2852 | free(it); | |
2853 | } | |
2854 | return 0; | |
2855 | } | |
2856 | ||
2857 | int lxc_clear_hooks(struct lxc_conf *c, const char *key) | |
2858 | { | |
2859 | struct lxc_list *it,*next; | |
2860 | bool all = false, done = false; | |
2861 | const char *k = key + 9; | |
2862 | int i; | |
2863 | ||
2864 | if (strcmp(key, "lxc.hook") == 0) | |
2865 | all = true; | |
2866 | ||
2867 | for (i=0; i<NUM_LXC_HOOKS; i++) { | |
2868 | if (all || strcmp(k, lxchook_names[i]) == 0) { | |
2869 | lxc_list_for_each_safe(it, &c->hooks[i], next) { | |
2870 | lxc_list_del(it); | |
2871 | free(it->elem); | |
2872 | free(it); | |
2873 | } | |
2874 | done = true; | |
2875 | } | |
2876 | } | |
2877 | ||
2878 | if (!done) { | |
2879 | ERROR("Invalid hook key: %s", key); | |
2880 | return -1; | |
2881 | } | |
2882 | return 0; | |
2883 | } | |
2884 | ||
2885 | void lxc_clear_saved_nics(struct lxc_conf *conf) | |
2886 | { | |
2887 | int i; | |
2888 | ||
2889 | if (!conf->num_savednics) | |
2890 | return; | |
2891 | for (i=0; i < conf->num_savednics; i++) | |
2892 | free(conf->saved_nics[i].orig_name); | |
2893 | conf->saved_nics = 0; | |
2894 | free(conf->saved_nics); | |
2895 | } | |
2896 | ||
2897 | void lxc_conf_free(struct lxc_conf *conf) | |
2898 | { | |
2899 | if (!conf) | |
2900 | return; | |
2901 | if (conf->console.path) | |
2902 | free(conf->console.path); | |
2903 | if (conf->rootfs.mount != default_rootfs_mount) | |
2904 | free(conf->rootfs.mount); | |
2905 | if (conf->rootfs.path) | |
2906 | free(conf->rootfs.path); | |
2907 | if (conf->utsname) | |
2908 | free(conf->utsname); | |
2909 | if (conf->ttydir) | |
2910 | free(conf->ttydir); | |
2911 | if (conf->fstab) | |
2912 | free(conf->fstab); | |
2913 | lxc_clear_config_network(conf); | |
2914 | #if HAVE_APPARMOR | |
2915 | if (conf->aa_profile) | |
2916 | free(conf->aa_profile); | |
2917 | #endif | |
2918 | lxc_seccomp_free(conf); | |
2919 | lxc_clear_config_caps(conf); | |
2920 | lxc_clear_cgroups(conf, "lxc.cgroup"); | |
2921 | lxc_clear_hooks(conf, "lxc.hook"); | |
2922 | lxc_clear_mount_entries(conf); | |
2923 | lxc_clear_saved_nics(conf); | |
2924 | free(conf); | |
2925 | } |