]>
Commit | Line | Data |
---|---|---|
a7334b09 LP |
1 | /*** |
2 | This file is part of systemd. | |
3 | ||
4 | Copyright 2010 Lennart Poettering | |
5 | ||
6 | systemd is free software; you can redistribute it and/or modify it | |
5430f7f2 LP |
7 | under the terms of the GNU Lesser General Public License as published by |
8 | the Free Software Foundation; either version 2.1 of the License, or | |
a7334b09 LP |
9 | (at your option) any later version. |
10 | ||
11 | systemd is distributed in the hope that it will be useful, but | |
12 | WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
5430f7f2 | 14 | Lesser General Public License for more details. |
a7334b09 | 15 | |
5430f7f2 | 16 | You should have received a copy of the GNU Lesser General Public License |
a7334b09 LP |
17 | along with systemd; If not, see <http://www.gnu.org/licenses/>. |
18 | ***/ | |
19 | ||
034c6ed7 LP |
20 | #include <errno.h> |
21 | #include <fcntl.h> | |
8dd4c05b LP |
22 | #include <glob.h> |
23 | #include <grp.h> | |
24 | #include <poll.h> | |
309bff19 | 25 | #include <signal.h> |
8dd4c05b | 26 | #include <string.h> |
19c0b0b9 | 27 | #include <sys/capability.h> |
d251207d | 28 | #include <sys/eventfd.h> |
f3e43635 | 29 | #include <sys/mman.h> |
8dd4c05b | 30 | #include <sys/personality.h> |
94f04347 | 31 | #include <sys/prctl.h> |
d2ffa389 | 32 | #include <sys/shm.h> |
8dd4c05b | 33 | #include <sys/socket.h> |
451a074f | 34 | #include <sys/stat.h> |
d2ffa389 | 35 | #include <sys/types.h> |
8dd4c05b LP |
36 | #include <sys/un.h> |
37 | #include <unistd.h> | |
023a4f67 | 38 | #include <utmpx.h> |
5cb5a6ff | 39 | |
5b6319dc LP |
40 | #ifdef HAVE_PAM |
41 | #include <security/pam_appl.h> | |
42 | #endif | |
43 | ||
7b52a628 MS |
44 | #ifdef HAVE_SELINUX |
45 | #include <selinux/selinux.h> | |
46 | #endif | |
47 | ||
17df7223 LP |
48 | #ifdef HAVE_SECCOMP |
49 | #include <seccomp.h> | |
50 | #endif | |
51 | ||
eef65bf3 MS |
52 | #ifdef HAVE_APPARMOR |
53 | #include <sys/apparmor.h> | |
54 | #endif | |
55 | ||
24882e06 | 56 | #include "sd-messages.h" |
8dd4c05b LP |
57 | |
58 | #include "af-list.h" | |
b5efdb8a | 59 | #include "alloc-util.h" |
3ffd4af2 LP |
60 | #ifdef HAVE_APPARMOR |
61 | #include "apparmor-util.h" | |
62 | #endif | |
8dd4c05b LP |
63 | #include "async.h" |
64 | #include "barrier.h" | |
8dd4c05b | 65 | #include "cap-list.h" |
430f0182 | 66 | #include "capability-util.h" |
f6a6225e | 67 | #include "def.h" |
4d1a6904 | 68 | #include "env-util.h" |
17df7223 | 69 | #include "errno-list.h" |
3ffd4af2 | 70 | #include "execute.h" |
8dd4c05b | 71 | #include "exit-status.h" |
3ffd4af2 | 72 | #include "fd-util.h" |
8dd4c05b | 73 | #include "fileio.h" |
f97b34a6 | 74 | #include "format-util.h" |
f4f15635 | 75 | #include "fs-util.h" |
7d50b32a | 76 | #include "glob-util.h" |
c004493c | 77 | #include "io-util.h" |
8dd4c05b LP |
78 | #include "ioprio.h" |
79 | #include "log.h" | |
80 | #include "macro.h" | |
81 | #include "missing.h" | |
82 | #include "mkdir.h" | |
83 | #include "namespace.h" | |
6bedfcbb | 84 | #include "parse-util.h" |
8dd4c05b | 85 | #include "path-util.h" |
0b452006 | 86 | #include "process-util.h" |
78f22b97 | 87 | #include "rlimit-util.h" |
8dd4c05b | 88 | #include "rm-rf.h" |
3ffd4af2 LP |
89 | #ifdef HAVE_SECCOMP |
90 | #include "seccomp-util.h" | |
91 | #endif | |
8dd4c05b LP |
92 | #include "securebits.h" |
93 | #include "selinux-util.h" | |
24882e06 | 94 | #include "signal-util.h" |
8dd4c05b | 95 | #include "smack-util.h" |
fd63e712 | 96 | #include "special.h" |
8b43440b | 97 | #include "string-table.h" |
07630cea | 98 | #include "string-util.h" |
8dd4c05b | 99 | #include "strv.h" |
7ccbd1ae | 100 | #include "syslog-util.h" |
8dd4c05b LP |
101 | #include "terminal-util.h" |
102 | #include "unit.h" | |
b1d4f8e1 | 103 | #include "user-util.h" |
8dd4c05b LP |
104 | #include "util.h" |
105 | #include "utmp-wtmp.h" | |
5cb5a6ff | 106 | |
e056b01d | 107 | #define IDLE_TIMEOUT_USEC (5*USEC_PER_SEC) |
31a7eb86 | 108 | #define IDLE_TIMEOUT2_USEC (1*USEC_PER_SEC) |
e6a26745 | 109 | |
02a51aba LP |
110 | /* This assumes there is a 'tty' group */ |
111 | #define TTY_MODE 0620 | |
112 | ||
531dca78 LP |
113 | #define SNDBUF_SIZE (8*1024*1024) |
114 | ||
034c6ed7 LP |
115 | static int shift_fds(int fds[], unsigned n_fds) { |
116 | int start, restart_from; | |
117 | ||
118 | if (n_fds <= 0) | |
119 | return 0; | |
120 | ||
a0d40ac5 LP |
121 | /* Modifies the fds array! (sorts it) */ |
122 | ||
034c6ed7 LP |
123 | assert(fds); |
124 | ||
125 | start = 0; | |
126 | for (;;) { | |
127 | int i; | |
128 | ||
129 | restart_from = -1; | |
130 | ||
131 | for (i = start; i < (int) n_fds; i++) { | |
132 | int nfd; | |
133 | ||
134 | /* Already at right index? */ | |
135 | if (fds[i] == i+3) | |
136 | continue; | |
137 | ||
3cc2aff1 LP |
138 | nfd = fcntl(fds[i], F_DUPFD, i + 3); |
139 | if (nfd < 0) | |
034c6ed7 LP |
140 | return -errno; |
141 | ||
03e334a1 | 142 | safe_close(fds[i]); |
034c6ed7 LP |
143 | fds[i] = nfd; |
144 | ||
145 | /* Hmm, the fd we wanted isn't free? Then | |
ee33e53a | 146 | * let's remember that and try again from here */ |
034c6ed7 LP |
147 | if (nfd != i+3 && restart_from < 0) |
148 | restart_from = i; | |
149 | } | |
150 | ||
151 | if (restart_from < 0) | |
152 | break; | |
153 | ||
154 | start = restart_from; | |
155 | } | |
156 | ||
157 | return 0; | |
158 | } | |
159 | ||
4c47affc FB |
160 | static int flags_fds(const int fds[], unsigned n_storage_fds, unsigned n_socket_fds, bool nonblock) { |
161 | unsigned i, n_fds; | |
e2c76839 | 162 | int r; |
47a71eed | 163 | |
4c47affc | 164 | n_fds = n_storage_fds + n_socket_fds; |
47a71eed LP |
165 | if (n_fds <= 0) |
166 | return 0; | |
167 | ||
168 | assert(fds); | |
169 | ||
9b141911 FB |
170 | /* Drops/Sets O_NONBLOCK and FD_CLOEXEC from the file flags. |
171 | * O_NONBLOCK only applies to socket activation though. */ | |
47a71eed LP |
172 | |
173 | for (i = 0; i < n_fds; i++) { | |
47a71eed | 174 | |
9b141911 FB |
175 | if (i < n_socket_fds) { |
176 | r = fd_nonblock(fds[i], nonblock); | |
177 | if (r < 0) | |
178 | return r; | |
179 | } | |
47a71eed | 180 | |
451a074f LP |
181 | /* We unconditionally drop FD_CLOEXEC from the fds, |
182 | * since after all we want to pass these fds to our | |
183 | * children */ | |
47a71eed | 184 | |
3cc2aff1 LP |
185 | r = fd_cloexec(fds[i], false); |
186 | if (r < 0) | |
e2c76839 | 187 | return r; |
47a71eed LP |
188 | } |
189 | ||
190 | return 0; | |
191 | } | |
192 | ||
1e22b5cd | 193 | static const char *exec_context_tty_path(const ExecContext *context) { |
80876c20 LP |
194 | assert(context); |
195 | ||
1e22b5cd LP |
196 | if (context->stdio_as_fds) |
197 | return NULL; | |
198 | ||
80876c20 LP |
199 | if (context->tty_path) |
200 | return context->tty_path; | |
201 | ||
202 | return "/dev/console"; | |
203 | } | |
204 | ||
1e22b5cd LP |
205 | static void exec_context_tty_reset(const ExecContext *context, const ExecParameters *p) { |
206 | const char *path; | |
207 | ||
6ea832a2 LP |
208 | assert(context); |
209 | ||
1e22b5cd | 210 | path = exec_context_tty_path(context); |
6ea832a2 | 211 | |
1e22b5cd LP |
212 | if (context->tty_vhangup) { |
213 | if (p && p->stdin_fd >= 0) | |
214 | (void) terminal_vhangup_fd(p->stdin_fd); | |
215 | else if (path) | |
216 | (void) terminal_vhangup(path); | |
217 | } | |
6ea832a2 | 218 | |
1e22b5cd LP |
219 | if (context->tty_reset) { |
220 | if (p && p->stdin_fd >= 0) | |
221 | (void) reset_terminal_fd(p->stdin_fd, true); | |
222 | else if (path) | |
223 | (void) reset_terminal(path); | |
224 | } | |
225 | ||
226 | if (context->tty_vt_disallocate && path) | |
227 | (void) vt_disallocate(path); | |
6ea832a2 LP |
228 | } |
229 | ||
6af760f3 LP |
230 | static bool is_terminal_input(ExecInput i) { |
231 | return IN_SET(i, | |
232 | EXEC_INPUT_TTY, | |
233 | EXEC_INPUT_TTY_FORCE, | |
234 | EXEC_INPUT_TTY_FAIL); | |
235 | } | |
236 | ||
3a1286b6 | 237 | static bool is_terminal_output(ExecOutput o) { |
6af760f3 LP |
238 | return IN_SET(o, |
239 | EXEC_OUTPUT_TTY, | |
240 | EXEC_OUTPUT_SYSLOG_AND_CONSOLE, | |
241 | EXEC_OUTPUT_KMSG_AND_CONSOLE, | |
242 | EXEC_OUTPUT_JOURNAL_AND_CONSOLE); | |
243 | } | |
244 | ||
245 | static bool exec_context_needs_term(const ExecContext *c) { | |
246 | assert(c); | |
247 | ||
248 | /* Return true if the execution context suggests we should set $TERM to something useful. */ | |
249 | ||
250 | if (is_terminal_input(c->std_input)) | |
251 | return true; | |
252 | ||
253 | if (is_terminal_output(c->std_output)) | |
254 | return true; | |
255 | ||
256 | if (is_terminal_output(c->std_error)) | |
257 | return true; | |
258 | ||
259 | return !!c->tty_path; | |
3a1286b6 MS |
260 | } |
261 | ||
80876c20 LP |
262 | static int open_null_as(int flags, int nfd) { |
263 | int fd, r; | |
071830ff | 264 | |
80876c20 | 265 | assert(nfd >= 0); |
071830ff | 266 | |
613b411c LP |
267 | fd = open("/dev/null", flags|O_NOCTTY); |
268 | if (fd < 0) | |
071830ff LP |
269 | return -errno; |
270 | ||
80876c20 LP |
271 | if (fd != nfd) { |
272 | r = dup2(fd, nfd) < 0 ? -errno : nfd; | |
03e334a1 | 273 | safe_close(fd); |
80876c20 LP |
274 | } else |
275 | r = nfd; | |
071830ff | 276 | |
80876c20 | 277 | return r; |
071830ff LP |
278 | } |
279 | ||
524daa8c | 280 | static int connect_journal_socket(int fd, uid_t uid, gid_t gid) { |
b92bea5d ZJS |
281 | union sockaddr_union sa = { |
282 | .un.sun_family = AF_UNIX, | |
283 | .un.sun_path = "/run/systemd/journal/stdout", | |
284 | }; | |
524daa8c ZJS |
285 | uid_t olduid = UID_INVALID; |
286 | gid_t oldgid = GID_INVALID; | |
287 | int r; | |
288 | ||
289 | if (gid != GID_INVALID) { | |
290 | oldgid = getgid(); | |
291 | ||
292 | r = setegid(gid); | |
293 | if (r < 0) | |
294 | return -errno; | |
295 | } | |
296 | ||
297 | if (uid != UID_INVALID) { | |
298 | olduid = getuid(); | |
299 | ||
300 | r = seteuid(uid); | |
301 | if (r < 0) { | |
302 | r = -errno; | |
303 | goto restore_gid; | |
304 | } | |
305 | } | |
306 | ||
fc2fffe7 | 307 | r = connect(fd, &sa.sa, SOCKADDR_UN_LEN(sa.un)); |
524daa8c ZJS |
308 | if (r < 0) |
309 | r = -errno; | |
310 | ||
311 | /* If we fail to restore the uid or gid, things will likely | |
312 | fail later on. This should only happen if an LSM interferes. */ | |
313 | ||
314 | if (uid != UID_INVALID) | |
315 | (void) seteuid(olduid); | |
316 | ||
317 | restore_gid: | |
318 | if (gid != GID_INVALID) | |
319 | (void) setegid(oldgid); | |
320 | ||
321 | return r; | |
322 | } | |
323 | ||
fd1f9c89 | 324 | static int connect_logger_as( |
7a1ab780 | 325 | Unit *unit, |
fd1f9c89 LP |
326 | const ExecContext *context, |
327 | ExecOutput output, | |
328 | const char *ident, | |
fd1f9c89 LP |
329 | int nfd, |
330 | uid_t uid, | |
331 | gid_t gid) { | |
332 | ||
524daa8c | 333 | int fd, r; |
071830ff LP |
334 | |
335 | assert(context); | |
80876c20 LP |
336 | assert(output < _EXEC_OUTPUT_MAX); |
337 | assert(ident); | |
338 | assert(nfd >= 0); | |
071830ff | 339 | |
54fe0cdb LP |
340 | fd = socket(AF_UNIX, SOCK_STREAM, 0); |
341 | if (fd < 0) | |
80876c20 | 342 | return -errno; |
071830ff | 343 | |
524daa8c ZJS |
344 | r = connect_journal_socket(fd, uid, gid); |
345 | if (r < 0) | |
346 | return r; | |
071830ff | 347 | |
80876c20 | 348 | if (shutdown(fd, SHUT_RD) < 0) { |
03e334a1 | 349 | safe_close(fd); |
80876c20 LP |
350 | return -errno; |
351 | } | |
071830ff | 352 | |
fd1f9c89 | 353 | (void) fd_inc_sndbuf(fd, SNDBUF_SIZE); |
531dca78 | 354 | |
80876c20 | 355 | dprintf(fd, |
62bca2c6 | 356 | "%s\n" |
80876c20 LP |
357 | "%s\n" |
358 | "%i\n" | |
54fe0cdb LP |
359 | "%i\n" |
360 | "%i\n" | |
361 | "%i\n" | |
4f4a1dbf | 362 | "%i\n", |
4f4a1dbf | 363 | context->syslog_identifier ? context->syslog_identifier : ident, |
7a1ab780 | 364 | unit->id, |
54fe0cdb LP |
365 | context->syslog_priority, |
366 | !!context->syslog_level_prefix, | |
367 | output == EXEC_OUTPUT_SYSLOG || output == EXEC_OUTPUT_SYSLOG_AND_CONSOLE, | |
368 | output == EXEC_OUTPUT_KMSG || output == EXEC_OUTPUT_KMSG_AND_CONSOLE, | |
3a1286b6 | 369 | is_terminal_output(output)); |
80876c20 | 370 | |
fd1f9c89 LP |
371 | if (fd == nfd) |
372 | return nfd; | |
373 | ||
374 | r = dup2(fd, nfd) < 0 ? -errno : nfd; | |
375 | safe_close(fd); | |
071830ff | 376 | |
80876c20 LP |
377 | return r; |
378 | } | |
379 | static int open_terminal_as(const char *path, mode_t mode, int nfd) { | |
380 | int fd, r; | |
071830ff | 381 | |
80876c20 LP |
382 | assert(path); |
383 | assert(nfd >= 0); | |
071830ff | 384 | |
3cc2aff1 LP |
385 | fd = open_terminal(path, mode | O_NOCTTY); |
386 | if (fd < 0) | |
80876c20 | 387 | return fd; |
071830ff | 388 | |
80876c20 LP |
389 | if (fd != nfd) { |
390 | r = dup2(fd, nfd) < 0 ? -errno : nfd; | |
03e334a1 | 391 | safe_close(fd); |
80876c20 LP |
392 | } else |
393 | r = nfd; | |
071830ff | 394 | |
80876c20 LP |
395 | return r; |
396 | } | |
071830ff | 397 | |
1e3ad081 LP |
398 | static int fixup_input(ExecInput std_input, int socket_fd, bool apply_tty_stdin) { |
399 | ||
400 | if (is_terminal_input(std_input) && !apply_tty_stdin) | |
401 | return EXEC_INPUT_NULL; | |
071830ff | 402 | |
03fd9c49 | 403 | if (std_input == EXEC_INPUT_SOCKET && socket_fd < 0) |
4f2d528d LP |
404 | return EXEC_INPUT_NULL; |
405 | ||
03fd9c49 | 406 | return std_input; |
4f2d528d LP |
407 | } |
408 | ||
03fd9c49 | 409 | static int fixup_output(ExecOutput std_output, int socket_fd) { |
4f2d528d | 410 | |
03fd9c49 | 411 | if (std_output == EXEC_OUTPUT_SOCKET && socket_fd < 0) |
4f2d528d LP |
412 | return EXEC_OUTPUT_INHERIT; |
413 | ||
03fd9c49 | 414 | return std_output; |
4f2d528d LP |
415 | } |
416 | ||
a34ceba6 LP |
417 | static int setup_input( |
418 | const ExecContext *context, | |
419 | const ExecParameters *params, | |
52c239d7 LB |
420 | int socket_fd, |
421 | int named_iofds[3]) { | |
a34ceba6 | 422 | |
4f2d528d LP |
423 | ExecInput i; |
424 | ||
425 | assert(context); | |
a34ceba6 LP |
426 | assert(params); |
427 | ||
428 | if (params->stdin_fd >= 0) { | |
429 | if (dup2(params->stdin_fd, STDIN_FILENO) < 0) | |
430 | return -errno; | |
431 | ||
432 | /* Try to make this the controlling tty, if it is a tty, and reset it */ | |
433 | (void) ioctl(STDIN_FILENO, TIOCSCTTY, context->std_input == EXEC_INPUT_TTY_FORCE); | |
434 | (void) reset_terminal_fd(STDIN_FILENO, true); | |
435 | ||
436 | return STDIN_FILENO; | |
437 | } | |
4f2d528d | 438 | |
c39f1ce2 | 439 | i = fixup_input(context->std_input, socket_fd, params->flags & EXEC_APPLY_TTY_STDIN); |
4f2d528d LP |
440 | |
441 | switch (i) { | |
071830ff | 442 | |
80876c20 LP |
443 | case EXEC_INPUT_NULL: |
444 | return open_null_as(O_RDONLY, STDIN_FILENO); | |
445 | ||
446 | case EXEC_INPUT_TTY: | |
447 | case EXEC_INPUT_TTY_FORCE: | |
448 | case EXEC_INPUT_TTY_FAIL: { | |
449 | int fd, r; | |
071830ff | 450 | |
1e22b5cd | 451 | fd = acquire_terminal(exec_context_tty_path(context), |
970edce6 ZJS |
452 | i == EXEC_INPUT_TTY_FAIL, |
453 | i == EXEC_INPUT_TTY_FORCE, | |
454 | false, | |
3a43da28 | 455 | USEC_INFINITY); |
970edce6 | 456 | if (fd < 0) |
80876c20 LP |
457 | return fd; |
458 | ||
459 | if (fd != STDIN_FILENO) { | |
460 | r = dup2(fd, STDIN_FILENO) < 0 ? -errno : STDIN_FILENO; | |
03e334a1 | 461 | safe_close(fd); |
80876c20 LP |
462 | } else |
463 | r = STDIN_FILENO; | |
464 | ||
465 | return r; | |
466 | } | |
467 | ||
4f2d528d LP |
468 | case EXEC_INPUT_SOCKET: |
469 | return dup2(socket_fd, STDIN_FILENO) < 0 ? -errno : STDIN_FILENO; | |
470 | ||
52c239d7 LB |
471 | case EXEC_INPUT_NAMED_FD: |
472 | (void) fd_nonblock(named_iofds[STDIN_FILENO], false); | |
473 | return dup2(named_iofds[STDIN_FILENO], STDIN_FILENO) < 0 ? -errno : STDIN_FILENO; | |
474 | ||
80876c20 LP |
475 | default: |
476 | assert_not_reached("Unknown input type"); | |
477 | } | |
478 | } | |
479 | ||
a34ceba6 LP |
480 | static int setup_output( |
481 | Unit *unit, | |
482 | const ExecContext *context, | |
483 | const ExecParameters *params, | |
484 | int fileno, | |
485 | int socket_fd, | |
52c239d7 | 486 | int named_iofds[3], |
a34ceba6 | 487 | const char *ident, |
7bce046b LP |
488 | uid_t uid, |
489 | gid_t gid, | |
490 | dev_t *journal_stream_dev, | |
491 | ino_t *journal_stream_ino) { | |
a34ceba6 | 492 | |
4f2d528d LP |
493 | ExecOutput o; |
494 | ExecInput i; | |
47c1d80d | 495 | int r; |
4f2d528d | 496 | |
f2341e0a | 497 | assert(unit); |
80876c20 | 498 | assert(context); |
a34ceba6 | 499 | assert(params); |
80876c20 | 500 | assert(ident); |
7bce046b LP |
501 | assert(journal_stream_dev); |
502 | assert(journal_stream_ino); | |
80876c20 | 503 | |
a34ceba6 LP |
504 | if (fileno == STDOUT_FILENO && params->stdout_fd >= 0) { |
505 | ||
506 | if (dup2(params->stdout_fd, STDOUT_FILENO) < 0) | |
507 | return -errno; | |
508 | ||
509 | return STDOUT_FILENO; | |
510 | } | |
511 | ||
512 | if (fileno == STDERR_FILENO && params->stderr_fd >= 0) { | |
513 | if (dup2(params->stderr_fd, STDERR_FILENO) < 0) | |
514 | return -errno; | |
515 | ||
516 | return STDERR_FILENO; | |
517 | } | |
518 | ||
c39f1ce2 | 519 | i = fixup_input(context->std_input, socket_fd, params->flags & EXEC_APPLY_TTY_STDIN); |
03fd9c49 | 520 | o = fixup_output(context->std_output, socket_fd); |
4f2d528d | 521 | |
eb17e935 MS |
522 | if (fileno == STDERR_FILENO) { |
523 | ExecOutput e; | |
524 | e = fixup_output(context->std_error, socket_fd); | |
80876c20 | 525 | |
eb17e935 MS |
526 | /* This expects the input and output are already set up */ |
527 | ||
528 | /* Don't change the stderr file descriptor if we inherit all | |
529 | * the way and are not on a tty */ | |
530 | if (e == EXEC_OUTPUT_INHERIT && | |
531 | o == EXEC_OUTPUT_INHERIT && | |
532 | i == EXEC_INPUT_NULL && | |
533 | !is_terminal_input(context->std_input) && | |
534 | getppid () != 1) | |
535 | return fileno; | |
536 | ||
537 | /* Duplicate from stdout if possible */ | |
52c239d7 | 538 | if ((e == o && e != EXEC_OUTPUT_NAMED_FD) || e == EXEC_OUTPUT_INHERIT) |
eb17e935 | 539 | return dup2(STDOUT_FILENO, fileno) < 0 ? -errno : fileno; |
071830ff | 540 | |
eb17e935 | 541 | o = e; |
80876c20 | 542 | |
eb17e935 | 543 | } else if (o == EXEC_OUTPUT_INHERIT) { |
21d21ea4 LP |
544 | /* If input got downgraded, inherit the original value */ |
545 | if (i == EXEC_INPUT_NULL && is_terminal_input(context->std_input)) | |
1e22b5cd | 546 | return open_terminal_as(exec_context_tty_path(context), O_WRONLY, fileno); |
21d21ea4 | 547 | |
acb591e4 | 548 | /* If the input is connected to anything that's not a /dev/null, inherit that... */ |
ff876e28 | 549 | if (i != EXEC_INPUT_NULL) |
eb17e935 | 550 | return dup2(STDIN_FILENO, fileno) < 0 ? -errno : fileno; |
071830ff | 551 | |
acb591e4 LP |
552 | /* If we are not started from PID 1 we just inherit STDOUT from our parent process. */ |
553 | if (getppid() != 1) | |
eb17e935 | 554 | return fileno; |
94f04347 | 555 | |
eb17e935 MS |
556 | /* We need to open /dev/null here anew, to get the right access mode. */ |
557 | return open_null_as(O_WRONLY, fileno); | |
071830ff | 558 | } |
94f04347 | 559 | |
eb17e935 | 560 | switch (o) { |
80876c20 LP |
561 | |
562 | case EXEC_OUTPUT_NULL: | |
eb17e935 | 563 | return open_null_as(O_WRONLY, fileno); |
80876c20 LP |
564 | |
565 | case EXEC_OUTPUT_TTY: | |
4f2d528d | 566 | if (is_terminal_input(i)) |
eb17e935 | 567 | return dup2(STDIN_FILENO, fileno) < 0 ? -errno : fileno; |
80876c20 LP |
568 | |
569 | /* We don't reset the terminal if this is just about output */ | |
1e22b5cd | 570 | return open_terminal_as(exec_context_tty_path(context), O_WRONLY, fileno); |
80876c20 LP |
571 | |
572 | case EXEC_OUTPUT_SYSLOG: | |
28dbc1e8 | 573 | case EXEC_OUTPUT_SYSLOG_AND_CONSOLE: |
9a6bca7a | 574 | case EXEC_OUTPUT_KMSG: |
28dbc1e8 | 575 | case EXEC_OUTPUT_KMSG_AND_CONSOLE: |
706343f4 LP |
576 | case EXEC_OUTPUT_JOURNAL: |
577 | case EXEC_OUTPUT_JOURNAL_AND_CONSOLE: | |
7a1ab780 | 578 | r = connect_logger_as(unit, context, o, ident, fileno, uid, gid); |
47c1d80d | 579 | if (r < 0) { |
f2341e0a | 580 | log_unit_error_errno(unit, r, "Failed to connect %s to the journal socket, ignoring: %m", fileno == STDOUT_FILENO ? "stdout" : "stderr"); |
eb17e935 | 581 | r = open_null_as(O_WRONLY, fileno); |
7bce046b LP |
582 | } else { |
583 | struct stat st; | |
584 | ||
585 | /* If we connected this fd to the journal via a stream, patch the device/inode into the passed | |
586 | * parameters, but only then. This is useful so that we can set $JOURNAL_STREAM that permits | |
587 | * services to detect whether they are connected to the journal or not. */ | |
588 | ||
589 | if (fstat(fileno, &st) >= 0) { | |
590 | *journal_stream_dev = st.st_dev; | |
591 | *journal_stream_ino = st.st_ino; | |
592 | } | |
47c1d80d MS |
593 | } |
594 | return r; | |
4f2d528d LP |
595 | |
596 | case EXEC_OUTPUT_SOCKET: | |
597 | assert(socket_fd >= 0); | |
eb17e935 | 598 | return dup2(socket_fd, fileno) < 0 ? -errno : fileno; |
94f04347 | 599 | |
52c239d7 LB |
600 | case EXEC_OUTPUT_NAMED_FD: |
601 | (void) fd_nonblock(named_iofds[fileno], false); | |
602 | return dup2(named_iofds[fileno], fileno) < 0 ? -errno : fileno; | |
603 | ||
94f04347 | 604 | default: |
80876c20 | 605 | assert_not_reached("Unknown error type"); |
94f04347 | 606 | } |
071830ff LP |
607 | } |
608 | ||
02a51aba LP |
609 | static int chown_terminal(int fd, uid_t uid) { |
610 | struct stat st; | |
611 | ||
612 | assert(fd >= 0); | |
02a51aba | 613 | |
1ff74fb6 LP |
614 | /* Before we chown/chmod the TTY, let's ensure this is actually a tty */ |
615 | if (isatty(fd) < 1) | |
616 | return 0; | |
617 | ||
02a51aba | 618 | /* This might fail. What matters are the results. */ |
bab45044 LP |
619 | (void) fchown(fd, uid, -1); |
620 | (void) fchmod(fd, TTY_MODE); | |
02a51aba LP |
621 | |
622 | if (fstat(fd, &st) < 0) | |
623 | return -errno; | |
624 | ||
d8b4e2e9 | 625 | if (st.st_uid != uid || (st.st_mode & 0777) != TTY_MODE) |
02a51aba LP |
626 | return -EPERM; |
627 | ||
628 | return 0; | |
629 | } | |
630 | ||
7d5ceb64 | 631 | static int setup_confirm_stdio(const char *vc, int *_saved_stdin, int *_saved_stdout) { |
3d18b167 LP |
632 | _cleanup_close_ int fd = -1, saved_stdin = -1, saved_stdout = -1; |
633 | int r; | |
80876c20 | 634 | |
80876c20 LP |
635 | assert(_saved_stdin); |
636 | assert(_saved_stdout); | |
637 | ||
af6da548 LP |
638 | saved_stdin = fcntl(STDIN_FILENO, F_DUPFD, 3); |
639 | if (saved_stdin < 0) | |
640 | return -errno; | |
80876c20 | 641 | |
af6da548 | 642 | saved_stdout = fcntl(STDOUT_FILENO, F_DUPFD, 3); |
3d18b167 LP |
643 | if (saved_stdout < 0) |
644 | return -errno; | |
80876c20 | 645 | |
7d5ceb64 | 646 | fd = acquire_terminal(vc, false, false, false, DEFAULT_CONFIRM_USEC); |
3d18b167 LP |
647 | if (fd < 0) |
648 | return fd; | |
80876c20 | 649 | |
af6da548 LP |
650 | r = chown_terminal(fd, getuid()); |
651 | if (r < 0) | |
3d18b167 | 652 | return r; |
02a51aba | 653 | |
3d18b167 LP |
654 | r = reset_terminal_fd(fd, true); |
655 | if (r < 0) | |
656 | return r; | |
80876c20 | 657 | |
3d18b167 LP |
658 | if (dup2(fd, STDIN_FILENO) < 0) |
659 | return -errno; | |
660 | ||
661 | if (dup2(fd, STDOUT_FILENO) < 0) | |
662 | return -errno; | |
80876c20 LP |
663 | |
664 | if (fd >= 2) | |
03e334a1 | 665 | safe_close(fd); |
3d18b167 | 666 | fd = -1; |
80876c20 LP |
667 | |
668 | *_saved_stdin = saved_stdin; | |
669 | *_saved_stdout = saved_stdout; | |
670 | ||
3d18b167 | 671 | saved_stdin = saved_stdout = -1; |
80876c20 | 672 | |
3d18b167 | 673 | return 0; |
80876c20 LP |
674 | } |
675 | ||
63d77c92 | 676 | static void write_confirm_error_fd(int err, int fd, const Unit *u) { |
3b20f877 FB |
677 | assert(err < 0); |
678 | ||
679 | if (err == -ETIMEDOUT) | |
63d77c92 | 680 | dprintf(fd, "Confirmation question timed out for %s, assuming positive response.\n", u->id); |
3b20f877 FB |
681 | else { |
682 | errno = -err; | |
63d77c92 | 683 | dprintf(fd, "Couldn't ask confirmation for %s: %m, assuming positive response.\n", u->id); |
3b20f877 FB |
684 | } |
685 | } | |
686 | ||
63d77c92 | 687 | static void write_confirm_error(int err, const char *vc, const Unit *u) { |
03e334a1 | 688 | _cleanup_close_ int fd = -1; |
80876c20 | 689 | |
3b20f877 | 690 | assert(vc); |
80876c20 | 691 | |
7d5ceb64 | 692 | fd = open_terminal(vc, O_WRONLY|O_NOCTTY|O_CLOEXEC); |
af6da548 | 693 | if (fd < 0) |
3b20f877 | 694 | return; |
80876c20 | 695 | |
63d77c92 | 696 | write_confirm_error_fd(err, fd, u); |
af6da548 | 697 | } |
80876c20 | 698 | |
3d18b167 | 699 | static int restore_confirm_stdio(int *saved_stdin, int *saved_stdout) { |
af6da548 | 700 | int r = 0; |
80876c20 | 701 | |
af6da548 LP |
702 | assert(saved_stdin); |
703 | assert(saved_stdout); | |
704 | ||
705 | release_terminal(); | |
706 | ||
707 | if (*saved_stdin >= 0) | |
80876c20 | 708 | if (dup2(*saved_stdin, STDIN_FILENO) < 0) |
af6da548 | 709 | r = -errno; |
80876c20 | 710 | |
af6da548 | 711 | if (*saved_stdout >= 0) |
80876c20 | 712 | if (dup2(*saved_stdout, STDOUT_FILENO) < 0) |
af6da548 | 713 | r = -errno; |
80876c20 | 714 | |
3d18b167 LP |
715 | *saved_stdin = safe_close(*saved_stdin); |
716 | *saved_stdout = safe_close(*saved_stdout); | |
af6da548 LP |
717 | |
718 | return r; | |
719 | } | |
720 | ||
3b20f877 FB |
721 | enum { |
722 | CONFIRM_PRETEND_FAILURE = -1, | |
723 | CONFIRM_PRETEND_SUCCESS = 0, | |
724 | CONFIRM_EXECUTE = 1, | |
725 | }; | |
726 | ||
eedf223a | 727 | static int ask_for_confirmation(const char *vc, Unit *u, const char *cmdline) { |
af6da548 | 728 | int saved_stdout = -1, saved_stdin = -1, r; |
2bcd3c26 | 729 | _cleanup_free_ char *e = NULL; |
3b20f877 | 730 | char c; |
af6da548 | 731 | |
3b20f877 | 732 | /* For any internal errors, assume a positive response. */ |
7d5ceb64 | 733 | r = setup_confirm_stdio(vc, &saved_stdin, &saved_stdout); |
3b20f877 | 734 | if (r < 0) { |
63d77c92 | 735 | write_confirm_error(r, vc, u); |
3b20f877 FB |
736 | return CONFIRM_EXECUTE; |
737 | } | |
af6da548 | 738 | |
b0eb2944 FB |
739 | /* confirm_spawn might have been disabled while we were sleeping. */ |
740 | if (manager_is_confirm_spawn_disabled(u->manager)) { | |
741 | r = 1; | |
742 | goto restore_stdio; | |
743 | } | |
af6da548 | 744 | |
2bcd3c26 FB |
745 | e = ellipsize(cmdline, 60, 100); |
746 | if (!e) { | |
747 | log_oom(); | |
748 | r = CONFIRM_EXECUTE; | |
749 | goto restore_stdio; | |
750 | } | |
af6da548 | 751 | |
d172b175 | 752 | for (;;) { |
539622bd | 753 | r = ask_char(&c, "yfshiDjcn", "Execute %s? [y, f, s – h for help] ", e); |
d172b175 | 754 | if (r < 0) { |
63d77c92 | 755 | write_confirm_error_fd(r, STDOUT_FILENO, u); |
d172b175 FB |
756 | r = CONFIRM_EXECUTE; |
757 | goto restore_stdio; | |
758 | } | |
af6da548 | 759 | |
d172b175 | 760 | switch (c) { |
b0eb2944 FB |
761 | case 'c': |
762 | printf("Resuming normal execution.\n"); | |
763 | manager_disable_confirm_spawn(); | |
764 | r = 1; | |
765 | break; | |
dd6f9ac0 FB |
766 | case 'D': |
767 | unit_dump(u, stdout, " "); | |
768 | continue; /* ask again */ | |
d172b175 FB |
769 | case 'f': |
770 | printf("Failing execution.\n"); | |
771 | r = CONFIRM_PRETEND_FAILURE; | |
772 | break; | |
773 | case 'h': | |
b0eb2944 FB |
774 | printf(" c - continue, proceed without asking anymore\n" |
775 | " D - dump, show the state of the unit\n" | |
dd6f9ac0 | 776 | " f - fail, don't execute the command and pretend it failed\n" |
d172b175 | 777 | " h - help\n" |
eedf223a | 778 | " i - info, show a short summary of the unit\n" |
56fde33a | 779 | " j - jobs, show jobs that are in progress\n" |
d172b175 FB |
780 | " s - skip, don't execute the command and pretend it succeeded\n" |
781 | " y - yes, execute the command\n"); | |
dd6f9ac0 | 782 | continue; /* ask again */ |
eedf223a FB |
783 | case 'i': |
784 | printf(" Description: %s\n" | |
785 | " Unit: %s\n" | |
786 | " Command: %s\n", | |
787 | u->id, u->description, cmdline); | |
788 | continue; /* ask again */ | |
56fde33a FB |
789 | case 'j': |
790 | manager_dump_jobs(u->manager, stdout, " "); | |
791 | continue; /* ask again */ | |
539622bd FB |
792 | case 'n': |
793 | /* 'n' was removed in favor of 'f'. */ | |
794 | printf("Didn't understand 'n', did you mean 'f'?\n"); | |
795 | continue; /* ask again */ | |
d172b175 FB |
796 | case 's': |
797 | printf("Skipping execution.\n"); | |
798 | r = CONFIRM_PRETEND_SUCCESS; | |
799 | break; | |
800 | case 'y': | |
801 | r = CONFIRM_EXECUTE; | |
802 | break; | |
803 | default: | |
804 | assert_not_reached("Unhandled choice"); | |
805 | } | |
3b20f877 | 806 | break; |
3b20f877 | 807 | } |
af6da548 | 808 | |
3b20f877 | 809 | restore_stdio: |
af6da548 | 810 | restore_confirm_stdio(&saved_stdin, &saved_stdout); |
af6da548 | 811 | return r; |
80876c20 LP |
812 | } |
813 | ||
4d885bd3 DH |
814 | static int get_fixed_user(const ExecContext *c, const char **user, |
815 | uid_t *uid, gid_t *gid, | |
816 | const char **home, const char **shell) { | |
81a2b7ce | 817 | int r; |
4d885bd3 | 818 | const char *name; |
81a2b7ce | 819 | |
4d885bd3 | 820 | assert(c); |
81a2b7ce | 821 | |
23deef88 LP |
822 | if (!c->user) |
823 | return 0; | |
824 | ||
4d885bd3 DH |
825 | /* Note that we don't set $HOME or $SHELL if they are not particularly enlightening anyway |
826 | * (i.e. are "/" or "/bin/nologin"). */ | |
81a2b7ce | 827 | |
23deef88 | 828 | name = c->user; |
4d885bd3 DH |
829 | r = get_user_creds_clean(&name, uid, gid, home, shell); |
830 | if (r < 0) | |
831 | return r; | |
81a2b7ce | 832 | |
4d885bd3 DH |
833 | *user = name; |
834 | return 0; | |
835 | } | |
836 | ||
837 | static int get_fixed_group(const ExecContext *c, const char **group, gid_t *gid) { | |
838 | int r; | |
839 | const char *name; | |
840 | ||
841 | assert(c); | |
842 | ||
843 | if (!c->group) | |
844 | return 0; | |
845 | ||
846 | name = c->group; | |
847 | r = get_group_creds(&name, gid); | |
848 | if (r < 0) | |
849 | return r; | |
850 | ||
851 | *group = name; | |
852 | return 0; | |
853 | } | |
854 | ||
cdc5d5c5 DH |
855 | static int get_supplementary_groups(const ExecContext *c, const char *user, |
856 | const char *group, gid_t gid, | |
857 | gid_t **supplementary_gids, int *ngids) { | |
4d885bd3 DH |
858 | char **i; |
859 | int r, k = 0; | |
860 | int ngroups_max; | |
861 | bool keep_groups = false; | |
862 | gid_t *groups = NULL; | |
863 | _cleanup_free_ gid_t *l_gids = NULL; | |
864 | ||
865 | assert(c); | |
866 | ||
bbeea271 DH |
867 | /* |
868 | * If user is given, then lookup GID and supplementary groups list. | |
869 | * We avoid NSS lookups for gid=0. Also we have to initialize groups | |
cdc5d5c5 DH |
870 | * here and as early as possible so we keep the list of supplementary |
871 | * groups of the caller. | |
bbeea271 DH |
872 | */ |
873 | if (user && gid_is_valid(gid) && gid != 0) { | |
874 | /* First step, initialize groups from /etc/groups */ | |
875 | if (initgroups(user, gid) < 0) | |
876 | return -errno; | |
877 | ||
878 | keep_groups = true; | |
879 | } | |
880 | ||
4d885bd3 DH |
881 | if (!c->supplementary_groups) |
882 | return 0; | |
883 | ||
366ddd25 DH |
884 | /* |
885 | * If SupplementaryGroups= was passed then NGROUPS_MAX has to | |
886 | * be positive, otherwise fail. | |
887 | */ | |
888 | errno = 0; | |
889 | ngroups_max = (int) sysconf(_SC_NGROUPS_MAX); | |
890 | if (ngroups_max <= 0) { | |
891 | if (errno > 0) | |
892 | return -errno; | |
893 | else | |
894 | return -EOPNOTSUPP; /* For all other values */ | |
895 | } | |
896 | ||
4d885bd3 DH |
897 | l_gids = new(gid_t, ngroups_max); |
898 | if (!l_gids) | |
899 | return -ENOMEM; | |
81a2b7ce | 900 | |
4d885bd3 DH |
901 | if (keep_groups) { |
902 | /* | |
903 | * Lookup the list of groups that the user belongs to, we | |
904 | * avoid NSS lookups here too for gid=0. | |
905 | */ | |
906 | k = ngroups_max; | |
907 | if (getgrouplist(user, gid, l_gids, &k) < 0) | |
908 | return -EINVAL; | |
909 | } else | |
910 | k = 0; | |
81a2b7ce | 911 | |
4d885bd3 DH |
912 | STRV_FOREACH(i, c->supplementary_groups) { |
913 | const char *g; | |
81a2b7ce | 914 | |
4d885bd3 DH |
915 | if (k >= ngroups_max) |
916 | return -E2BIG; | |
81a2b7ce | 917 | |
4d885bd3 DH |
918 | g = *i; |
919 | r = get_group_creds(&g, l_gids+k); | |
920 | if (r < 0) | |
921 | return r; | |
81a2b7ce | 922 | |
4d885bd3 DH |
923 | k++; |
924 | } | |
81a2b7ce | 925 | |
4d885bd3 DH |
926 | /* |
927 | * Sets ngids to zero to drop all supplementary groups, happens | |
928 | * when we are under root and SupplementaryGroups= is empty. | |
929 | */ | |
930 | if (k == 0) { | |
931 | *ngids = 0; | |
932 | return 0; | |
933 | } | |
81a2b7ce | 934 | |
4d885bd3 DH |
935 | /* Otherwise get the final list of supplementary groups */ |
936 | groups = memdup(l_gids, sizeof(gid_t) * k); | |
937 | if (!groups) | |
938 | return -ENOMEM; | |
939 | ||
940 | *supplementary_gids = groups; | |
941 | *ngids = k; | |
942 | ||
943 | groups = NULL; | |
944 | ||
945 | return 0; | |
946 | } | |
947 | ||
948 | static int enforce_groups(const ExecContext *context, gid_t gid, | |
949 | gid_t *supplementary_gids, int ngids) { | |
950 | int r; | |
951 | ||
952 | assert(context); | |
953 | ||
954 | /* Handle SupplementaryGroups= even if it is empty */ | |
955 | if (context->supplementary_groups) { | |
956 | r = maybe_setgroups(ngids, supplementary_gids); | |
957 | if (r < 0) | |
97f0e76f | 958 | return r; |
4d885bd3 | 959 | } |
81a2b7ce | 960 | |
4d885bd3 DH |
961 | if (gid_is_valid(gid)) { |
962 | /* Then set our gids */ | |
963 | if (setresgid(gid, gid, gid) < 0) | |
964 | return -errno; | |
81a2b7ce LP |
965 | } |
966 | ||
967 | return 0; | |
968 | } | |
969 | ||
970 | static int enforce_user(const ExecContext *context, uid_t uid) { | |
81a2b7ce LP |
971 | assert(context); |
972 | ||
4d885bd3 DH |
973 | if (!uid_is_valid(uid)) |
974 | return 0; | |
975 | ||
479050b3 | 976 | /* Sets (but doesn't look up) the uid and make sure we keep the |
81a2b7ce LP |
977 | * capabilities while doing so. */ |
978 | ||
479050b3 | 979 | if (context->capability_ambient_set != 0) { |
81a2b7ce LP |
980 | |
981 | /* First step: If we need to keep capabilities but | |
982 | * drop privileges we need to make sure we keep our | |
cbb21cca | 983 | * caps, while we drop privileges. */ |
693ced48 | 984 | if (uid != 0) { |
cbb21cca | 985 | int sb = context->secure_bits | 1<<SECURE_KEEP_CAPS; |
693ced48 LP |
986 | |
987 | if (prctl(PR_GET_SECUREBITS) != sb) | |
988 | if (prctl(PR_SET_SECUREBITS, sb) < 0) | |
989 | return -errno; | |
990 | } | |
81a2b7ce LP |
991 | } |
992 | ||
479050b3 | 993 | /* Second step: actually set the uids */ |
81a2b7ce LP |
994 | if (setresuid(uid, uid, uid) < 0) |
995 | return -errno; | |
996 | ||
997 | /* At this point we should have all necessary capabilities but | |
998 | are otherwise a normal user. However, the caps might got | |
999 | corrupted due to the setresuid() so we need clean them up | |
1000 | later. This is done outside of this call. */ | |
1001 | ||
1002 | return 0; | |
1003 | } | |
1004 | ||
5b6319dc LP |
1005 | #ifdef HAVE_PAM |
1006 | ||
1007 | static int null_conv( | |
1008 | int num_msg, | |
1009 | const struct pam_message **msg, | |
1010 | struct pam_response **resp, | |
1011 | void *appdata_ptr) { | |
1012 | ||
1013 | /* We don't support conversations */ | |
1014 | ||
1015 | return PAM_CONV_ERR; | |
1016 | } | |
1017 | ||
cefc33ae LP |
1018 | #endif |
1019 | ||
5b6319dc LP |
1020 | static int setup_pam( |
1021 | const char *name, | |
1022 | const char *user, | |
940c5210 | 1023 | uid_t uid, |
2d6fce8d | 1024 | gid_t gid, |
5b6319dc | 1025 | const char *tty, |
2065ca69 | 1026 | char ***env, |
5b6319dc LP |
1027 | int fds[], unsigned n_fds) { |
1028 | ||
cefc33ae LP |
1029 | #ifdef HAVE_PAM |
1030 | ||
5b6319dc LP |
1031 | static const struct pam_conv conv = { |
1032 | .conv = null_conv, | |
1033 | .appdata_ptr = NULL | |
1034 | }; | |
1035 | ||
2d7c6aa2 | 1036 | _cleanup_(barrier_destroy) Barrier barrier = BARRIER_NULL; |
5b6319dc | 1037 | pam_handle_t *handle = NULL; |
d6e5f3ad | 1038 | sigset_t old_ss; |
7bb70b6e | 1039 | int pam_code = PAM_SUCCESS, r; |
84eada2f | 1040 | char **nv, **e = NULL; |
5b6319dc LP |
1041 | bool close_session = false; |
1042 | pid_t pam_pid = 0, parent_pid; | |
970edce6 | 1043 | int flags = 0; |
5b6319dc LP |
1044 | |
1045 | assert(name); | |
1046 | assert(user); | |
2065ca69 | 1047 | assert(env); |
5b6319dc LP |
1048 | |
1049 | /* We set up PAM in the parent process, then fork. The child | |
35b8ca3a | 1050 | * will then stay around until killed via PR_GET_PDEATHSIG or |
5b6319dc LP |
1051 | * systemd via the cgroup logic. It will then remove the PAM |
1052 | * session again. The parent process will exec() the actual | |
1053 | * daemon. We do things this way to ensure that the main PID | |
1054 | * of the daemon is the one we initially fork()ed. */ | |
1055 | ||
7bb70b6e LP |
1056 | r = barrier_create(&barrier); |
1057 | if (r < 0) | |
2d7c6aa2 DH |
1058 | goto fail; |
1059 | ||
553d2243 | 1060 | if (log_get_max_level() < LOG_DEBUG) |
970edce6 ZJS |
1061 | flags |= PAM_SILENT; |
1062 | ||
f546241b ZJS |
1063 | pam_code = pam_start(name, user, &conv, &handle); |
1064 | if (pam_code != PAM_SUCCESS) { | |
5b6319dc LP |
1065 | handle = NULL; |
1066 | goto fail; | |
1067 | } | |
1068 | ||
f546241b ZJS |
1069 | if (tty) { |
1070 | pam_code = pam_set_item(handle, PAM_TTY, tty); | |
1071 | if (pam_code != PAM_SUCCESS) | |
5b6319dc | 1072 | goto fail; |
f546241b | 1073 | } |
5b6319dc | 1074 | |
84eada2f JW |
1075 | STRV_FOREACH(nv, *env) { |
1076 | pam_code = pam_putenv(handle, *nv); | |
2065ca69 JW |
1077 | if (pam_code != PAM_SUCCESS) |
1078 | goto fail; | |
1079 | } | |
1080 | ||
970edce6 | 1081 | pam_code = pam_acct_mgmt(handle, flags); |
f546241b | 1082 | if (pam_code != PAM_SUCCESS) |
5b6319dc LP |
1083 | goto fail; |
1084 | ||
970edce6 | 1085 | pam_code = pam_open_session(handle, flags); |
f546241b | 1086 | if (pam_code != PAM_SUCCESS) |
5b6319dc LP |
1087 | goto fail; |
1088 | ||
1089 | close_session = true; | |
1090 | ||
f546241b ZJS |
1091 | e = pam_getenvlist(handle); |
1092 | if (!e) { | |
5b6319dc LP |
1093 | pam_code = PAM_BUF_ERR; |
1094 | goto fail; | |
1095 | } | |
1096 | ||
1097 | /* Block SIGTERM, so that we know that it won't get lost in | |
1098 | * the child */ | |
ce30c8dc | 1099 | |
72c0a2c2 | 1100 | assert_se(sigprocmask_many(SIG_BLOCK, &old_ss, SIGTERM, -1) >= 0); |
5b6319dc LP |
1101 | |
1102 | parent_pid = getpid(); | |
1103 | ||
f546241b | 1104 | pam_pid = fork(); |
7bb70b6e LP |
1105 | if (pam_pid < 0) { |
1106 | r = -errno; | |
5b6319dc | 1107 | goto fail; |
7bb70b6e | 1108 | } |
5b6319dc LP |
1109 | |
1110 | if (pam_pid == 0) { | |
7bb70b6e | 1111 | int sig, ret = EXIT_PAM; |
5b6319dc LP |
1112 | |
1113 | /* The child's job is to reset the PAM session on | |
1114 | * termination */ | |
2d7c6aa2 | 1115 | barrier_set_role(&barrier, BARRIER_CHILD); |
5b6319dc LP |
1116 | |
1117 | /* This string must fit in 10 chars (i.e. the length | |
5d6b1584 LP |
1118 | * of "/sbin/init"), to look pretty in /bin/ps */ |
1119 | rename_process("(sd-pam)"); | |
5b6319dc LP |
1120 | |
1121 | /* Make sure we don't keep open the passed fds in this | |
1122 | child. We assume that otherwise only those fds are | |
1123 | open here that have been opened by PAM. */ | |
1124 | close_many(fds, n_fds); | |
1125 | ||
940c5210 AK |
1126 | /* Drop privileges - we don't need any to pam_close_session |
1127 | * and this will make PR_SET_PDEATHSIG work in most cases. | |
1128 | * If this fails, ignore the error - but expect sd-pam threads | |
1129 | * to fail to exit normally */ | |
2d6fce8d | 1130 | |
97f0e76f LP |
1131 | r = maybe_setgroups(0, NULL); |
1132 | if (r < 0) | |
1133 | log_warning_errno(r, "Failed to setgroups() in sd-pam: %m"); | |
2d6fce8d LP |
1134 | if (setresgid(gid, gid, gid) < 0) |
1135 | log_warning_errno(errno, "Failed to setresgid() in sd-pam: %m"); | |
940c5210 | 1136 | if (setresuid(uid, uid, uid) < 0) |
2d6fce8d | 1137 | log_warning_errno(errno, "Failed to setresuid() in sd-pam: %m"); |
940c5210 | 1138 | |
ce30c8dc LP |
1139 | (void) ignore_signals(SIGPIPE, -1); |
1140 | ||
940c5210 AK |
1141 | /* Wait until our parent died. This will only work if |
1142 | * the above setresuid() succeeds, otherwise the kernel | |
1143 | * will not allow unprivileged parents kill their privileged | |
1144 | * children this way. We rely on the control groups kill logic | |
5b6319dc LP |
1145 | * to do the rest for us. */ |
1146 | if (prctl(PR_SET_PDEATHSIG, SIGTERM) < 0) | |
1147 | goto child_finish; | |
1148 | ||
2d7c6aa2 DH |
1149 | /* Tell the parent that our setup is done. This is especially |
1150 | * important regarding dropping privileges. Otherwise, unit | |
643f4706 ZJS |
1151 | * setup might race against our setresuid(2) call. |
1152 | * | |
1153 | * If the parent aborted, we'll detect this below, hence ignore | |
1154 | * return failure here. */ | |
1155 | (void) barrier_place(&barrier); | |
2d7c6aa2 | 1156 | |
643f4706 | 1157 | /* Check if our parent process might already have died? */ |
5b6319dc | 1158 | if (getppid() == parent_pid) { |
d6e5f3ad DM |
1159 | sigset_t ss; |
1160 | ||
1161 | assert_se(sigemptyset(&ss) >= 0); | |
1162 | assert_se(sigaddset(&ss, SIGTERM) >= 0); | |
1163 | ||
3dead8d9 LP |
1164 | for (;;) { |
1165 | if (sigwait(&ss, &sig) < 0) { | |
1166 | if (errno == EINTR) | |
1167 | continue; | |
1168 | ||
1169 | goto child_finish; | |
1170 | } | |
5b6319dc | 1171 | |
3dead8d9 LP |
1172 | assert(sig == SIGTERM); |
1173 | break; | |
1174 | } | |
5b6319dc LP |
1175 | } |
1176 | ||
3dead8d9 | 1177 | /* If our parent died we'll end the session */ |
f546241b | 1178 | if (getppid() != parent_pid) { |
970edce6 | 1179 | pam_code = pam_close_session(handle, flags); |
f546241b | 1180 | if (pam_code != PAM_SUCCESS) |
5b6319dc | 1181 | goto child_finish; |
f546241b | 1182 | } |
5b6319dc | 1183 | |
7bb70b6e | 1184 | ret = 0; |
5b6319dc LP |
1185 | |
1186 | child_finish: | |
970edce6 | 1187 | pam_end(handle, pam_code | flags); |
7bb70b6e | 1188 | _exit(ret); |
5b6319dc LP |
1189 | } |
1190 | ||
2d7c6aa2 DH |
1191 | barrier_set_role(&barrier, BARRIER_PARENT); |
1192 | ||
5b6319dc LP |
1193 | /* If the child was forked off successfully it will do all the |
1194 | * cleanups, so forget about the handle here. */ | |
1195 | handle = NULL; | |
1196 | ||
3b8bddde | 1197 | /* Unblock SIGTERM again in the parent */ |
72c0a2c2 | 1198 | assert_se(sigprocmask(SIG_SETMASK, &old_ss, NULL) >= 0); |
5b6319dc LP |
1199 | |
1200 | /* We close the log explicitly here, since the PAM modules | |
1201 | * might have opened it, but we don't want this fd around. */ | |
1202 | closelog(); | |
1203 | ||
2d7c6aa2 DH |
1204 | /* Synchronously wait for the child to initialize. We don't care for |
1205 | * errors as we cannot recover. However, warn loudly if it happens. */ | |
1206 | if (!barrier_place_and_sync(&barrier)) | |
1207 | log_error("PAM initialization failed"); | |
1208 | ||
2065ca69 JW |
1209 | strv_free(*env); |
1210 | *env = e; | |
aa87e624 | 1211 | |
5b6319dc LP |
1212 | return 0; |
1213 | ||
1214 | fail: | |
970edce6 ZJS |
1215 | if (pam_code != PAM_SUCCESS) { |
1216 | log_error("PAM failed: %s", pam_strerror(handle, pam_code)); | |
7bb70b6e LP |
1217 | r = -EPERM; /* PAM errors do not map to errno */ |
1218 | } else | |
1219 | log_error_errno(r, "PAM failed: %m"); | |
9ba35398 | 1220 | |
5b6319dc LP |
1221 | if (handle) { |
1222 | if (close_session) | |
970edce6 | 1223 | pam_code = pam_close_session(handle, flags); |
5b6319dc | 1224 | |
970edce6 | 1225 | pam_end(handle, pam_code | flags); |
5b6319dc LP |
1226 | } |
1227 | ||
1228 | strv_free(e); | |
5b6319dc LP |
1229 | closelog(); |
1230 | ||
7bb70b6e | 1231 | return r; |
cefc33ae LP |
1232 | #else |
1233 | return 0; | |
5b6319dc | 1234 | #endif |
cefc33ae | 1235 | } |
5b6319dc | 1236 | |
5d6b1584 LP |
1237 | static void rename_process_from_path(const char *path) { |
1238 | char process_name[11]; | |
1239 | const char *p; | |
1240 | size_t l; | |
1241 | ||
1242 | /* This resulting string must fit in 10 chars (i.e. the length | |
1243 | * of "/sbin/init") to look pretty in /bin/ps */ | |
1244 | ||
2b6bf07d | 1245 | p = basename(path); |
5d6b1584 LP |
1246 | if (isempty(p)) { |
1247 | rename_process("(...)"); | |
1248 | return; | |
1249 | } | |
1250 | ||
1251 | l = strlen(p); | |
1252 | if (l > 8) { | |
1253 | /* The end of the process name is usually more | |
1254 | * interesting, since the first bit might just be | |
1255 | * "systemd-" */ | |
1256 | p = p + l - 8; | |
1257 | l = 8; | |
1258 | } | |
1259 | ||
1260 | process_name[0] = '('; | |
1261 | memcpy(process_name+1, p, l); | |
1262 | process_name[1+l] = ')'; | |
1263 | process_name[1+l+1] = 0; | |
1264 | ||
1265 | rename_process(process_name); | |
1266 | } | |
1267 | ||
469830d1 LP |
1268 | static bool context_has_address_families(const ExecContext *c) { |
1269 | assert(c); | |
1270 | ||
1271 | return c->address_families_whitelist || | |
1272 | !set_isempty(c->address_families); | |
1273 | } | |
1274 | ||
1275 | static bool context_has_syscall_filters(const ExecContext *c) { | |
1276 | assert(c); | |
1277 | ||
1278 | return c->syscall_whitelist || | |
1279 | !set_isempty(c->syscall_filter); | |
1280 | } | |
1281 | ||
1282 | static bool context_has_no_new_privileges(const ExecContext *c) { | |
1283 | assert(c); | |
1284 | ||
1285 | if (c->no_new_privileges) | |
1286 | return true; | |
1287 | ||
1288 | if (have_effective_cap(CAP_SYS_ADMIN)) /* if we are privileged, we don't need NNP */ | |
1289 | return false; | |
1290 | ||
1291 | /* We need NNP if we have any form of seccomp and are unprivileged */ | |
1292 | return context_has_address_families(c) || | |
1293 | c->memory_deny_write_execute || | |
1294 | c->restrict_realtime || | |
1295 | exec_context_restrict_namespaces_set(c) || | |
1296 | c->protect_kernel_tunables || | |
1297 | c->protect_kernel_modules || | |
1298 | c->private_devices || | |
1299 | context_has_syscall_filters(c) || | |
1300 | !set_isempty(c->syscall_archs); | |
1301 | } | |
1302 | ||
c0467cf3 | 1303 | #ifdef HAVE_SECCOMP |
17df7223 | 1304 | |
83f12b27 | 1305 | static bool skip_seccomp_unavailable(const Unit* u, const char* msg) { |
f673b62d LP |
1306 | |
1307 | if (is_seccomp_available()) | |
1308 | return false; | |
1309 | ||
1310 | log_open(); | |
1311 | log_unit_debug(u, "SECCOMP features not detected in the kernel, skipping %s", msg); | |
1312 | log_close(); | |
1313 | return true; | |
83f12b27 FS |
1314 | } |
1315 | ||
469830d1 LP |
1316 | static int apply_syscall_filter(const Unit* u, const ExecContext *c) { |
1317 | uint32_t negative_action, default_action, action; | |
8351ceae | 1318 | |
469830d1 | 1319 | assert(u); |
c0467cf3 | 1320 | assert(c); |
8351ceae | 1321 | |
469830d1 | 1322 | if (!context_has_syscall_filters(c)) |
83f12b27 FS |
1323 | return 0; |
1324 | ||
469830d1 LP |
1325 | if (skip_seccomp_unavailable(u, "SystemCallFilter=")) |
1326 | return 0; | |
e9642be2 | 1327 | |
469830d1 | 1328 | negative_action = c->syscall_errno == 0 ? SCMP_ACT_KILL : SCMP_ACT_ERRNO(c->syscall_errno); |
e9642be2 | 1329 | |
469830d1 LP |
1330 | if (c->syscall_whitelist) { |
1331 | default_action = negative_action; | |
1332 | action = SCMP_ACT_ALLOW; | |
7c66bae2 | 1333 | } else { |
469830d1 LP |
1334 | default_action = SCMP_ACT_ALLOW; |
1335 | action = negative_action; | |
57183d11 | 1336 | } |
8351ceae | 1337 | |
469830d1 | 1338 | return seccomp_load_syscall_filter_set_raw(default_action, c->syscall_filter, action); |
4298d0b5 LP |
1339 | } |
1340 | ||
469830d1 LP |
1341 | static int apply_syscall_archs(const Unit *u, const ExecContext *c) { |
1342 | assert(u); | |
4298d0b5 LP |
1343 | assert(c); |
1344 | ||
469830d1 | 1345 | if (set_isempty(c->syscall_archs)) |
83f12b27 FS |
1346 | return 0; |
1347 | ||
469830d1 LP |
1348 | if (skip_seccomp_unavailable(u, "SystemCallArchitectures=")) |
1349 | return 0; | |
4298d0b5 | 1350 | |
469830d1 LP |
1351 | return seccomp_restrict_archs(c->syscall_archs); |
1352 | } | |
4298d0b5 | 1353 | |
469830d1 LP |
1354 | static int apply_address_families(const Unit* u, const ExecContext *c) { |
1355 | assert(u); | |
1356 | assert(c); | |
4298d0b5 | 1357 | |
469830d1 LP |
1358 | if (!context_has_address_families(c)) |
1359 | return 0; | |
4298d0b5 | 1360 | |
469830d1 LP |
1361 | if (skip_seccomp_unavailable(u, "RestrictAddressFamilies=")) |
1362 | return 0; | |
4298d0b5 | 1363 | |
469830d1 | 1364 | return seccomp_restrict_address_families(c->address_families, c->address_families_whitelist); |
8351ceae | 1365 | } |
4298d0b5 | 1366 | |
83f12b27 | 1367 | static int apply_memory_deny_write_execute(const Unit* u, const ExecContext *c) { |
469830d1 | 1368 | assert(u); |
f3e43635 TM |
1369 | assert(c); |
1370 | ||
469830d1 | 1371 | if (!c->memory_deny_write_execute) |
83f12b27 FS |
1372 | return 0; |
1373 | ||
469830d1 LP |
1374 | if (skip_seccomp_unavailable(u, "MemoryDenyWriteExecute=")) |
1375 | return 0; | |
f3e43635 | 1376 | |
469830d1 | 1377 | return seccomp_memory_deny_write_execute(); |
f3e43635 TM |
1378 | } |
1379 | ||
83f12b27 | 1380 | static int apply_restrict_realtime(const Unit* u, const ExecContext *c) { |
469830d1 | 1381 | assert(u); |
f4170c67 LP |
1382 | assert(c); |
1383 | ||
469830d1 | 1384 | if (!c->restrict_realtime) |
83f12b27 FS |
1385 | return 0; |
1386 | ||
469830d1 LP |
1387 | if (skip_seccomp_unavailable(u, "RestrictRealtime=")) |
1388 | return 0; | |
f4170c67 | 1389 | |
469830d1 | 1390 | return seccomp_restrict_realtime(); |
f4170c67 LP |
1391 | } |
1392 | ||
59e856c7 | 1393 | static int apply_protect_sysctl(const Unit *u, const ExecContext *c) { |
469830d1 | 1394 | assert(u); |
59eeb84b LP |
1395 | assert(c); |
1396 | ||
1397 | /* Turn off the legacy sysctl() system call. Many distributions turn this off while building the kernel, but | |
1398 | * let's protect even those systems where this is left on in the kernel. */ | |
1399 | ||
469830d1 | 1400 | if (!c->protect_kernel_tunables) |
59eeb84b LP |
1401 | return 0; |
1402 | ||
469830d1 LP |
1403 | if (skip_seccomp_unavailable(u, "ProtectKernelTunables=")) |
1404 | return 0; | |
59eeb84b | 1405 | |
469830d1 | 1406 | return seccomp_protect_sysctl(); |
59eeb84b LP |
1407 | } |
1408 | ||
59e856c7 | 1409 | static int apply_protect_kernel_modules(const Unit *u, const ExecContext *c) { |
469830d1 | 1410 | assert(u); |
502d704e DH |
1411 | assert(c); |
1412 | ||
25a8d8a0 | 1413 | /* Turn off module syscalls on ProtectKernelModules=yes */ |
502d704e | 1414 | |
469830d1 LP |
1415 | if (!c->protect_kernel_modules) |
1416 | return 0; | |
1417 | ||
502d704e DH |
1418 | if (skip_seccomp_unavailable(u, "ProtectKernelModules=")) |
1419 | return 0; | |
1420 | ||
469830d1 | 1421 | return seccomp_load_syscall_filter_set(SCMP_ACT_ALLOW, syscall_filter_sets + SYSCALL_FILTER_SET_MODULE, SCMP_ACT_ERRNO(EPERM)); |
502d704e DH |
1422 | } |
1423 | ||
59e856c7 | 1424 | static int apply_private_devices(const Unit *u, const ExecContext *c) { |
469830d1 | 1425 | assert(u); |
ba128bb8 LP |
1426 | assert(c); |
1427 | ||
8f81a5f6 | 1428 | /* If PrivateDevices= is set, also turn off iopl and all @raw-io syscalls. */ |
ba128bb8 | 1429 | |
469830d1 LP |
1430 | if (!c->private_devices) |
1431 | return 0; | |
1432 | ||
ba128bb8 LP |
1433 | if (skip_seccomp_unavailable(u, "PrivateDevices=")) |
1434 | return 0; | |
1435 | ||
469830d1 | 1436 | return seccomp_load_syscall_filter_set(SCMP_ACT_ALLOW, syscall_filter_sets + SYSCALL_FILTER_SET_RAW_IO, SCMP_ACT_ERRNO(EPERM)); |
ba128bb8 LP |
1437 | } |
1438 | ||
add00535 | 1439 | static int apply_restrict_namespaces(Unit *u, const ExecContext *c) { |
469830d1 | 1440 | assert(u); |
add00535 LP |
1441 | assert(c); |
1442 | ||
1443 | if (!exec_context_restrict_namespaces_set(c)) | |
1444 | return 0; | |
1445 | ||
1446 | if (skip_seccomp_unavailable(u, "RestrictNamespaces=")) | |
1447 | return 0; | |
1448 | ||
1449 | return seccomp_restrict_namespaces(c->restrict_namespaces); | |
1450 | } | |
1451 | ||
c0467cf3 | 1452 | #endif |
8351ceae | 1453 | |
31a7eb86 ZJS |
1454 | static void do_idle_pipe_dance(int idle_pipe[4]) { |
1455 | assert(idle_pipe); | |
1456 | ||
54eb2300 LP |
1457 | idle_pipe[1] = safe_close(idle_pipe[1]); |
1458 | idle_pipe[2] = safe_close(idle_pipe[2]); | |
31a7eb86 ZJS |
1459 | |
1460 | if (idle_pipe[0] >= 0) { | |
1461 | int r; | |
1462 | ||
1463 | r = fd_wait_for_event(idle_pipe[0], POLLHUP, IDLE_TIMEOUT_USEC); | |
1464 | ||
1465 | if (idle_pipe[3] >= 0 && r == 0 /* timeout */) { | |
c7cc737f LP |
1466 | ssize_t n; |
1467 | ||
31a7eb86 | 1468 | /* Signal systemd that we are bored and want to continue. */ |
c7cc737f LP |
1469 | n = write(idle_pipe[3], "x", 1); |
1470 | if (n > 0) | |
cd972d69 ZJS |
1471 | /* Wait for systemd to react to the signal above. */ |
1472 | fd_wait_for_event(idle_pipe[0], POLLHUP, IDLE_TIMEOUT2_USEC); | |
31a7eb86 ZJS |
1473 | } |
1474 | ||
54eb2300 | 1475 | idle_pipe[0] = safe_close(idle_pipe[0]); |
31a7eb86 ZJS |
1476 | |
1477 | } | |
1478 | ||
54eb2300 | 1479 | idle_pipe[3] = safe_close(idle_pipe[3]); |
31a7eb86 ZJS |
1480 | } |
1481 | ||
7cae38c4 | 1482 | static int build_environment( |
fd63e712 | 1483 | Unit *u, |
9fa95f85 | 1484 | const ExecContext *c, |
1e22b5cd | 1485 | const ExecParameters *p, |
7cae38c4 LP |
1486 | unsigned n_fds, |
1487 | const char *home, | |
1488 | const char *username, | |
1489 | const char *shell, | |
7bce046b LP |
1490 | dev_t journal_stream_dev, |
1491 | ino_t journal_stream_ino, | |
7cae38c4 LP |
1492 | char ***ret) { |
1493 | ||
1494 | _cleanup_strv_free_ char **our_env = NULL; | |
1495 | unsigned n_env = 0; | |
1496 | char *x; | |
1497 | ||
4b58153d | 1498 | assert(u); |
7cae38c4 LP |
1499 | assert(c); |
1500 | assert(ret); | |
1501 | ||
4b58153d | 1502 | our_env = new0(char*, 14); |
7cae38c4 LP |
1503 | if (!our_env) |
1504 | return -ENOMEM; | |
1505 | ||
1506 | if (n_fds > 0) { | |
8dd4c05b LP |
1507 | _cleanup_free_ char *joined = NULL; |
1508 | ||
ccd06097 | 1509 | if (asprintf(&x, "LISTEN_PID="PID_FMT, getpid()) < 0) |
7cae38c4 LP |
1510 | return -ENOMEM; |
1511 | our_env[n_env++] = x; | |
1512 | ||
1513 | if (asprintf(&x, "LISTEN_FDS=%u", n_fds) < 0) | |
1514 | return -ENOMEM; | |
1515 | our_env[n_env++] = x; | |
8dd4c05b | 1516 | |
1e22b5cd | 1517 | joined = strv_join(p->fd_names, ":"); |
8dd4c05b LP |
1518 | if (!joined) |
1519 | return -ENOMEM; | |
1520 | ||
605405c6 | 1521 | x = strjoin("LISTEN_FDNAMES=", joined); |
8dd4c05b LP |
1522 | if (!x) |
1523 | return -ENOMEM; | |
1524 | our_env[n_env++] = x; | |
7cae38c4 LP |
1525 | } |
1526 | ||
b08af3b1 | 1527 | if ((p->flags & EXEC_SET_WATCHDOG) && p->watchdog_usec > 0) { |
ccd06097 | 1528 | if (asprintf(&x, "WATCHDOG_PID="PID_FMT, getpid()) < 0) |
09812eb7 LP |
1529 | return -ENOMEM; |
1530 | our_env[n_env++] = x; | |
1531 | ||
1e22b5cd | 1532 | if (asprintf(&x, "WATCHDOG_USEC="USEC_FMT, p->watchdog_usec) < 0) |
09812eb7 LP |
1533 | return -ENOMEM; |
1534 | our_env[n_env++] = x; | |
1535 | } | |
1536 | ||
fd63e712 LP |
1537 | /* If this is D-Bus, tell the nss-systemd module, since it relies on being able to use D-Bus look up dynamic |
1538 | * users via PID 1, possibly dead-locking the dbus daemon. This way it will not use D-Bus to resolve names, but | |
1539 | * check the database directly. */ | |
1540 | if (unit_has_name(u, SPECIAL_DBUS_SERVICE)) { | |
1541 | x = strdup("SYSTEMD_NSS_BYPASS_BUS=1"); | |
1542 | if (!x) | |
1543 | return -ENOMEM; | |
1544 | our_env[n_env++] = x; | |
1545 | } | |
1546 | ||
7cae38c4 LP |
1547 | if (home) { |
1548 | x = strappend("HOME=", home); | |
1549 | if (!x) | |
1550 | return -ENOMEM; | |
1551 | our_env[n_env++] = x; | |
1552 | } | |
1553 | ||
1554 | if (username) { | |
1555 | x = strappend("LOGNAME=", username); | |
1556 | if (!x) | |
1557 | return -ENOMEM; | |
1558 | our_env[n_env++] = x; | |
1559 | ||
1560 | x = strappend("USER=", username); | |
1561 | if (!x) | |
1562 | return -ENOMEM; | |
1563 | our_env[n_env++] = x; | |
1564 | } | |
1565 | ||
1566 | if (shell) { | |
1567 | x = strappend("SHELL=", shell); | |
1568 | if (!x) | |
1569 | return -ENOMEM; | |
1570 | our_env[n_env++] = x; | |
1571 | } | |
1572 | ||
4b58153d LP |
1573 | if (!sd_id128_is_null(u->invocation_id)) { |
1574 | if (asprintf(&x, "INVOCATION_ID=" SD_ID128_FORMAT_STR, SD_ID128_FORMAT_VAL(u->invocation_id)) < 0) | |
1575 | return -ENOMEM; | |
1576 | ||
1577 | our_env[n_env++] = x; | |
1578 | } | |
1579 | ||
6af760f3 LP |
1580 | if (exec_context_needs_term(c)) { |
1581 | const char *tty_path, *term = NULL; | |
1582 | ||
1583 | tty_path = exec_context_tty_path(c); | |
1584 | ||
1585 | /* If we are forked off PID 1 and we are supposed to operate on /dev/console, then let's try to inherit | |
1586 | * the $TERM set for PID 1. This is useful for containers so that the $TERM the container manager | |
1587 | * passes to PID 1 ends up all the way in the console login shown. */ | |
1588 | ||
1589 | if (path_equal(tty_path, "/dev/console") && getppid() == 1) | |
1590 | term = getenv("TERM"); | |
1591 | if (!term) | |
1592 | term = default_term_for_tty(tty_path); | |
7cae38c4 | 1593 | |
6af760f3 | 1594 | x = strappend("TERM=", term); |
7cae38c4 LP |
1595 | if (!x) |
1596 | return -ENOMEM; | |
1597 | our_env[n_env++] = x; | |
1598 | } | |
1599 | ||
7bce046b LP |
1600 | if (journal_stream_dev != 0 && journal_stream_ino != 0) { |
1601 | if (asprintf(&x, "JOURNAL_STREAM=" DEV_FMT ":" INO_FMT, journal_stream_dev, journal_stream_ino) < 0) | |
1602 | return -ENOMEM; | |
1603 | ||
1604 | our_env[n_env++] = x; | |
1605 | } | |
1606 | ||
7cae38c4 | 1607 | our_env[n_env++] = NULL; |
7bce046b | 1608 | assert(n_env <= 12); |
7cae38c4 LP |
1609 | |
1610 | *ret = our_env; | |
1611 | our_env = NULL; | |
1612 | ||
1613 | return 0; | |
1614 | } | |
1615 | ||
b4c14404 FB |
1616 | static int build_pass_environment(const ExecContext *c, char ***ret) { |
1617 | _cleanup_strv_free_ char **pass_env = NULL; | |
1618 | size_t n_env = 0, n_bufsize = 0; | |
1619 | char **i; | |
1620 | ||
1621 | STRV_FOREACH(i, c->pass_environment) { | |
1622 | _cleanup_free_ char *x = NULL; | |
1623 | char *v; | |
1624 | ||
1625 | v = getenv(*i); | |
1626 | if (!v) | |
1627 | continue; | |
605405c6 | 1628 | x = strjoin(*i, "=", v); |
b4c14404 FB |
1629 | if (!x) |
1630 | return -ENOMEM; | |
1631 | if (!GREEDY_REALLOC(pass_env, n_bufsize, n_env + 2)) | |
1632 | return -ENOMEM; | |
1633 | pass_env[n_env++] = x; | |
1634 | pass_env[n_env] = NULL; | |
1635 | x = NULL; | |
1636 | } | |
1637 | ||
1638 | *ret = pass_env; | |
1639 | pass_env = NULL; | |
1640 | ||
1641 | return 0; | |
1642 | } | |
1643 | ||
8b44a3d2 LP |
1644 | static bool exec_needs_mount_namespace( |
1645 | const ExecContext *context, | |
1646 | const ExecParameters *params, | |
1647 | ExecRuntime *runtime) { | |
1648 | ||
1649 | assert(context); | |
1650 | assert(params); | |
1651 | ||
915e6d16 LP |
1652 | if (context->root_image) |
1653 | return true; | |
1654 | ||
2a624c36 AP |
1655 | if (!strv_isempty(context->read_write_paths) || |
1656 | !strv_isempty(context->read_only_paths) || | |
1657 | !strv_isempty(context->inaccessible_paths)) | |
8b44a3d2 LP |
1658 | return true; |
1659 | ||
d2d6c096 LP |
1660 | if (context->n_bind_mounts > 0) |
1661 | return true; | |
1662 | ||
8b44a3d2 LP |
1663 | if (context->mount_flags != 0) |
1664 | return true; | |
1665 | ||
1666 | if (context->private_tmp && runtime && (runtime->tmp_dir || runtime->var_tmp_dir)) | |
1667 | return true; | |
1668 | ||
8b44a3d2 LP |
1669 | if (context->private_devices || |
1670 | context->protect_system != PROTECT_SYSTEM_NO || | |
59eeb84b LP |
1671 | context->protect_home != PROTECT_HOME_NO || |
1672 | context->protect_kernel_tunables || | |
c575770b | 1673 | context->protect_kernel_modules || |
59eeb84b | 1674 | context->protect_control_groups) |
8b44a3d2 LP |
1675 | return true; |
1676 | ||
9c988f93 | 1677 | if (context->mount_apivfs && (context->root_image || context->root_directory)) |
5d997827 LP |
1678 | return true; |
1679 | ||
8b44a3d2 LP |
1680 | return false; |
1681 | } | |
1682 | ||
d251207d LP |
1683 | static int setup_private_users(uid_t uid, gid_t gid) { |
1684 | _cleanup_free_ char *uid_map = NULL, *gid_map = NULL; | |
1685 | _cleanup_close_pair_ int errno_pipe[2] = { -1, -1 }; | |
1686 | _cleanup_close_ int unshare_ready_fd = -1; | |
1687 | _cleanup_(sigkill_waitp) pid_t pid = 0; | |
1688 | uint64_t c = 1; | |
1689 | siginfo_t si; | |
1690 | ssize_t n; | |
1691 | int r; | |
1692 | ||
1693 | /* Set up a user namespace and map root to root, the selected UID/GID to itself, and everything else to | |
1694 | * nobody. In order to be able to write this mapping we need CAP_SETUID in the original user namespace, which | |
1695 | * we however lack after opening the user namespace. To work around this we fork() a temporary child process, | |
1696 | * which waits for the parent to create the new user namespace while staying in the original namespace. The | |
1697 | * child then writes the UID mapping, under full privileges. The parent waits for the child to finish and | |
1698 | * continues execution normally. */ | |
1699 | ||
587ab01b ZJS |
1700 | if (uid != 0 && uid_is_valid(uid)) { |
1701 | r = asprintf(&uid_map, | |
1702 | "0 0 1\n" /* Map root → root */ | |
1703 | UID_FMT " " UID_FMT " 1\n", /* Map $UID → $UID */ | |
1704 | uid, uid); | |
1705 | if (r < 0) | |
1706 | return -ENOMEM; | |
1707 | } else { | |
e0f3720e | 1708 | uid_map = strdup("0 0 1\n"); /* The case where the above is the same */ |
587ab01b ZJS |
1709 | if (!uid_map) |
1710 | return -ENOMEM; | |
1711 | } | |
d251207d | 1712 | |
587ab01b ZJS |
1713 | if (gid != 0 && gid_is_valid(gid)) { |
1714 | r = asprintf(&gid_map, | |
1715 | "0 0 1\n" /* Map root → root */ | |
1716 | GID_FMT " " GID_FMT " 1\n", /* Map $GID → $GID */ | |
1717 | gid, gid); | |
1718 | if (r < 0) | |
1719 | return -ENOMEM; | |
1720 | } else { | |
d251207d | 1721 | gid_map = strdup("0 0 1\n"); /* The case where the above is the same */ |
587ab01b ZJS |
1722 | if (!gid_map) |
1723 | return -ENOMEM; | |
1724 | } | |
d251207d LP |
1725 | |
1726 | /* Create a communication channel so that the parent can tell the child when it finished creating the user | |
1727 | * namespace. */ | |
1728 | unshare_ready_fd = eventfd(0, EFD_CLOEXEC); | |
1729 | if (unshare_ready_fd < 0) | |
1730 | return -errno; | |
1731 | ||
1732 | /* Create a communication channel so that the child can tell the parent a proper error code in case it | |
1733 | * failed. */ | |
1734 | if (pipe2(errno_pipe, O_CLOEXEC) < 0) | |
1735 | return -errno; | |
1736 | ||
1737 | pid = fork(); | |
1738 | if (pid < 0) | |
1739 | return -errno; | |
1740 | ||
1741 | if (pid == 0) { | |
1742 | _cleanup_close_ int fd = -1; | |
1743 | const char *a; | |
1744 | pid_t ppid; | |
1745 | ||
1746 | /* Child process, running in the original user namespace. Let's update the parent's UID/GID map from | |
1747 | * here, after the parent opened its own user namespace. */ | |
1748 | ||
1749 | ppid = getppid(); | |
1750 | errno_pipe[0] = safe_close(errno_pipe[0]); | |
1751 | ||
1752 | /* Wait until the parent unshared the user namespace */ | |
1753 | if (read(unshare_ready_fd, &c, sizeof(c)) < 0) { | |
1754 | r = -errno; | |
1755 | goto child_fail; | |
1756 | } | |
1757 | ||
1758 | /* Disable the setgroups() system call in the child user namespace, for good. */ | |
1759 | a = procfs_file_alloca(ppid, "setgroups"); | |
1760 | fd = open(a, O_WRONLY|O_CLOEXEC); | |
1761 | if (fd < 0) { | |
1762 | if (errno != ENOENT) { | |
1763 | r = -errno; | |
1764 | goto child_fail; | |
1765 | } | |
1766 | ||
1767 | /* If the file is missing the kernel is too old, let's continue anyway. */ | |
1768 | } else { | |
1769 | if (write(fd, "deny\n", 5) < 0) { | |
1770 | r = -errno; | |
1771 | goto child_fail; | |
1772 | } | |
1773 | ||
1774 | fd = safe_close(fd); | |
1775 | } | |
1776 | ||
1777 | /* First write the GID map */ | |
1778 | a = procfs_file_alloca(ppid, "gid_map"); | |
1779 | fd = open(a, O_WRONLY|O_CLOEXEC); | |
1780 | if (fd < 0) { | |
1781 | r = -errno; | |
1782 | goto child_fail; | |
1783 | } | |
1784 | if (write(fd, gid_map, strlen(gid_map)) < 0) { | |
1785 | r = -errno; | |
1786 | goto child_fail; | |
1787 | } | |
1788 | fd = safe_close(fd); | |
1789 | ||
1790 | /* The write the UID map */ | |
1791 | a = procfs_file_alloca(ppid, "uid_map"); | |
1792 | fd = open(a, O_WRONLY|O_CLOEXEC); | |
1793 | if (fd < 0) { | |
1794 | r = -errno; | |
1795 | goto child_fail; | |
1796 | } | |
1797 | if (write(fd, uid_map, strlen(uid_map)) < 0) { | |
1798 | r = -errno; | |
1799 | goto child_fail; | |
1800 | } | |
1801 | ||
1802 | _exit(EXIT_SUCCESS); | |
1803 | ||
1804 | child_fail: | |
1805 | (void) write(errno_pipe[1], &r, sizeof(r)); | |
1806 | _exit(EXIT_FAILURE); | |
1807 | } | |
1808 | ||
1809 | errno_pipe[1] = safe_close(errno_pipe[1]); | |
1810 | ||
1811 | if (unshare(CLONE_NEWUSER) < 0) | |
1812 | return -errno; | |
1813 | ||
1814 | /* Let the child know that the namespace is ready now */ | |
1815 | if (write(unshare_ready_fd, &c, sizeof(c)) < 0) | |
1816 | return -errno; | |
1817 | ||
1818 | /* Try to read an error code from the child */ | |
1819 | n = read(errno_pipe[0], &r, sizeof(r)); | |
1820 | if (n < 0) | |
1821 | return -errno; | |
1822 | if (n == sizeof(r)) { /* an error code was sent to us */ | |
1823 | if (r < 0) | |
1824 | return r; | |
1825 | return -EIO; | |
1826 | } | |
1827 | if (n != 0) /* on success we should have read 0 bytes */ | |
1828 | return -EIO; | |
1829 | ||
1830 | r = wait_for_terminate(pid, &si); | |
1831 | if (r < 0) | |
1832 | return r; | |
1833 | pid = 0; | |
1834 | ||
1835 | /* If something strange happened with the child, let's consider this fatal, too */ | |
1836 | if (si.si_code != CLD_EXITED || si.si_status != 0) | |
1837 | return -EIO; | |
1838 | ||
1839 | return 0; | |
1840 | } | |
1841 | ||
07689d5d LP |
1842 | static int setup_runtime_directory( |
1843 | const ExecContext *context, | |
1844 | const ExecParameters *params, | |
1845 | uid_t uid, | |
1846 | gid_t gid) { | |
1847 | ||
1848 | char **rt; | |
1849 | int r; | |
1850 | ||
1851 | assert(context); | |
1852 | assert(params); | |
1853 | ||
1854 | STRV_FOREACH(rt, context->runtime_directory) { | |
1855 | _cleanup_free_ char *p; | |
1856 | ||
605405c6 | 1857 | p = strjoin(params->runtime_prefix, "/", *rt); |
07689d5d LP |
1858 | if (!p) |
1859 | return -ENOMEM; | |
1860 | ||
1861 | r = mkdir_p_label(p, context->runtime_directory_mode); | |
1862 | if (r < 0) | |
1863 | return r; | |
1864 | ||
1865 | r = chmod_and_chown(p, context->runtime_directory_mode, uid, gid); | |
1866 | if (r < 0) | |
1867 | return r; | |
1868 | } | |
1869 | ||
1870 | return 0; | |
1871 | } | |
1872 | ||
cefc33ae LP |
1873 | static int setup_smack( |
1874 | const ExecContext *context, | |
1875 | const ExecCommand *command) { | |
1876 | ||
1877 | #ifdef HAVE_SMACK | |
1878 | int r; | |
1879 | ||
1880 | assert(context); | |
1881 | assert(command); | |
1882 | ||
1883 | if (!mac_smack_use()) | |
1884 | return 0; | |
1885 | ||
1886 | if (context->smack_process_label) { | |
1887 | r = mac_smack_apply_pid(0, context->smack_process_label); | |
1888 | if (r < 0) | |
1889 | return r; | |
1890 | } | |
1891 | #ifdef SMACK_DEFAULT_PROCESS_LABEL | |
1892 | else { | |
1893 | _cleanup_free_ char *exec_label = NULL; | |
1894 | ||
1895 | r = mac_smack_read(command->path, SMACK_ATTR_EXEC, &exec_label); | |
1896 | if (r < 0 && r != -ENODATA && r != -EOPNOTSUPP) | |
1897 | return r; | |
1898 | ||
1899 | r = mac_smack_apply_pid(0, exec_label ? : SMACK_DEFAULT_PROCESS_LABEL); | |
1900 | if (r < 0) | |
1901 | return r; | |
1902 | } | |
1903 | #endif | |
1904 | #endif | |
1905 | ||
1906 | return 0; | |
1907 | } | |
1908 | ||
3fbe8dbe LP |
1909 | static int compile_read_write_paths( |
1910 | const ExecContext *context, | |
1911 | const ExecParameters *params, | |
1912 | char ***ret) { | |
1913 | ||
1914 | _cleanup_strv_free_ char **l = NULL; | |
1915 | char **rt; | |
1916 | ||
06ec51d8 ZJS |
1917 | /* Compile the list of writable paths. This is the combination of |
1918 | * the explicitly configured paths, plus all runtime directories. */ | |
3fbe8dbe LP |
1919 | |
1920 | if (strv_isempty(context->read_write_paths) && | |
1921 | strv_isempty(context->runtime_directory)) { | |
1922 | *ret = NULL; /* NOP if neither is set */ | |
1923 | return 0; | |
1924 | } | |
1925 | ||
1926 | l = strv_copy(context->read_write_paths); | |
1927 | if (!l) | |
1928 | return -ENOMEM; | |
1929 | ||
1930 | STRV_FOREACH(rt, context->runtime_directory) { | |
1931 | char *s; | |
1932 | ||
605405c6 | 1933 | s = strjoin(params->runtime_prefix, "/", *rt); |
3fbe8dbe LP |
1934 | if (!s) |
1935 | return -ENOMEM; | |
1936 | ||
1937 | if (strv_consume(&l, s) < 0) | |
1938 | return -ENOMEM; | |
1939 | } | |
1940 | ||
1941 | *ret = l; | |
1942 | l = NULL; | |
1943 | ||
1944 | return 0; | |
1945 | } | |
1946 | ||
6818c54c LP |
1947 | static int apply_mount_namespace( |
1948 | Unit *u, | |
1949 | ExecCommand *command, | |
1950 | const ExecContext *context, | |
1951 | const ExecParameters *params, | |
1952 | ExecRuntime *runtime) { | |
1953 | ||
06ec51d8 | 1954 | _cleanup_strv_free_ char **rw = NULL; |
93c6bb51 | 1955 | char *tmp = NULL, *var = NULL; |
915e6d16 | 1956 | const char *root_dir = NULL, *root_image = NULL; |
93c6bb51 | 1957 | NameSpaceInfo ns_info = { |
af964954 | 1958 | .ignore_protect_paths = false, |
93c6bb51 DH |
1959 | .private_dev = context->private_devices, |
1960 | .protect_control_groups = context->protect_control_groups, | |
1961 | .protect_kernel_tunables = context->protect_kernel_tunables, | |
1962 | .protect_kernel_modules = context->protect_kernel_modules, | |
5d997827 | 1963 | .mount_apivfs = context->mount_apivfs, |
93c6bb51 | 1964 | }; |
6818c54c LP |
1965 | bool apply_restrictions; |
1966 | int r; | |
93c6bb51 | 1967 | |
2b3c1b9e DH |
1968 | assert(context); |
1969 | ||
93c6bb51 DH |
1970 | /* The runtime struct only contains the parent of the private /tmp, |
1971 | * which is non-accessible to world users. Inside of it there's a /tmp | |
1972 | * that is sticky, and that's the one we want to use here. */ | |
1973 | ||
1974 | if (context->private_tmp && runtime) { | |
1975 | if (runtime->tmp_dir) | |
1976 | tmp = strjoina(runtime->tmp_dir, "/tmp"); | |
1977 | if (runtime->var_tmp_dir) | |
1978 | var = strjoina(runtime->var_tmp_dir, "/tmp"); | |
1979 | } | |
1980 | ||
1981 | r = compile_read_write_paths(context, params, &rw); | |
1982 | if (r < 0) | |
1983 | return r; | |
1984 | ||
915e6d16 LP |
1985 | if (params->flags & EXEC_APPLY_CHROOT) { |
1986 | root_image = context->root_image; | |
1987 | ||
1988 | if (!root_image) | |
1989 | root_dir = context->root_directory; | |
1990 | } | |
93c6bb51 | 1991 | |
af964954 DH |
1992 | /* |
1993 | * If DynamicUser=no and RootDirectory= is set then lets pass a relaxed | |
1994 | * sandbox info, otherwise enforce it, don't ignore protected paths and | |
1995 | * fail if we are enable to apply the sandbox inside the mount namespace. | |
1996 | */ | |
1997 | if (!context->dynamic_user && root_dir) | |
1998 | ns_info.ignore_protect_paths = true; | |
1999 | ||
6818c54c LP |
2000 | apply_restrictions = (params->flags & EXEC_APPLY_PERMISSIONS) && !command->privileged; |
2001 | ||
915e6d16 LP |
2002 | r = setup_namespace(root_dir, root_image, |
2003 | &ns_info, rw, | |
6818c54c LP |
2004 | apply_restrictions ? context->read_only_paths : NULL, |
2005 | apply_restrictions ? context->inaccessible_paths : NULL, | |
d2d6c096 LP |
2006 | context->bind_mounts, |
2007 | context->n_bind_mounts, | |
93c6bb51 DH |
2008 | tmp, |
2009 | var, | |
6818c54c LP |
2010 | apply_restrictions ? context->protect_home : PROTECT_HOME_NO, |
2011 | apply_restrictions ? context->protect_system : PROTECT_SYSTEM_NO, | |
915e6d16 LP |
2012 | context->mount_flags, |
2013 | DISSECT_IMAGE_DISCARD_ON_LOOP); | |
93c6bb51 DH |
2014 | |
2015 | /* If we couldn't set up the namespace this is probably due to a | |
2016 | * missing capability. In this case, silently proceeed. */ | |
2017 | if (IN_SET(r, -EPERM, -EACCES)) { | |
2018 | log_open(); | |
2019 | log_unit_debug_errno(u, r, "Failed to set up namespace, assuming containerized execution, ignoring: %m"); | |
2020 | log_close(); | |
2021 | r = 0; | |
2022 | } | |
2023 | ||
2024 | return r; | |
2025 | } | |
2026 | ||
915e6d16 LP |
2027 | static int apply_working_directory( |
2028 | const ExecContext *context, | |
2029 | const ExecParameters *params, | |
2030 | const char *home, | |
376fecf6 LP |
2031 | const bool needs_mount_ns, |
2032 | int *exit_status) { | |
915e6d16 | 2033 | |
6732edab | 2034 | const char *d, *wd; |
2b3c1b9e DH |
2035 | |
2036 | assert(context); | |
376fecf6 | 2037 | assert(exit_status); |
2b3c1b9e | 2038 | |
6732edab LP |
2039 | if (context->working_directory_home) { |
2040 | ||
376fecf6 LP |
2041 | if (!home) { |
2042 | *exit_status = EXIT_CHDIR; | |
6732edab | 2043 | return -ENXIO; |
376fecf6 | 2044 | } |
6732edab | 2045 | |
2b3c1b9e | 2046 | wd = home; |
6732edab LP |
2047 | |
2048 | } else if (context->working_directory) | |
2b3c1b9e DH |
2049 | wd = context->working_directory; |
2050 | else | |
2051 | wd = "/"; | |
e7f1e7c6 DH |
2052 | |
2053 | if (params->flags & EXEC_APPLY_CHROOT) { | |
2054 | if (!needs_mount_ns && context->root_directory) | |
376fecf6 LP |
2055 | if (chroot(context->root_directory) < 0) { |
2056 | *exit_status = EXIT_CHROOT; | |
e7f1e7c6 | 2057 | return -errno; |
376fecf6 | 2058 | } |
e7f1e7c6 | 2059 | |
2b3c1b9e DH |
2060 | d = wd; |
2061 | } else | |
3b0e5bb5 | 2062 | d = prefix_roota(context->root_directory, wd); |
e7f1e7c6 | 2063 | |
376fecf6 LP |
2064 | if (chdir(d) < 0 && !context->working_directory_missing_ok) { |
2065 | *exit_status = EXIT_CHDIR; | |
2b3c1b9e | 2066 | return -errno; |
376fecf6 | 2067 | } |
e7f1e7c6 DH |
2068 | |
2069 | return 0; | |
2070 | } | |
2071 | ||
74dd6b51 LP |
2072 | static int setup_keyring(Unit *u, const ExecParameters *p, uid_t uid, gid_t gid) { |
2073 | key_serial_t keyring; | |
2074 | ||
2075 | assert(u); | |
2076 | assert(p); | |
2077 | ||
2078 | /* Let's set up a new per-service "session" kernel keyring for each system service. This has the benefit that | |
2079 | * each service runs with its own keyring shared among all processes of the service, but with no hook-up beyond | |
2080 | * that scope, and in particular no link to the per-UID keyring. If we don't do this the keyring will be | |
2081 | * automatically created on-demand and then linked to the per-UID keyring, by the kernel. The kernel's built-in | |
2082 | * on-demand behaviour is very appropriate for login users, but probably not so much for system services, where | |
2083 | * UIDs are not necessarily specific to a service but reused (at least in the case of UID 0). */ | |
2084 | ||
2085 | if (!(p->flags & EXEC_NEW_KEYRING)) | |
2086 | return 0; | |
2087 | ||
2088 | keyring = keyctl(KEYCTL_JOIN_SESSION_KEYRING, 0, 0, 0, 0); | |
2089 | if (keyring == -1) { | |
2090 | if (errno == ENOSYS) | |
2091 | log_debug_errno(errno, "Kernel keyring not supported, ignoring."); | |
2092 | else if (IN_SET(errno, EACCES, EPERM)) | |
2093 | log_debug_errno(errno, "Kernel keyring access prohibited, ignoring."); | |
2094 | else if (errno == EDQUOT) | |
2095 | log_debug_errno(errno, "Out of kernel keyrings to allocate, ignoring."); | |
2096 | else | |
2097 | return log_error_errno(errno, "Setting up kernel keyring failed: %m"); | |
2098 | ||
2099 | return 0; | |
2100 | } | |
2101 | ||
b3415f5d LP |
2102 | /* Populate they keyring with the invocation ID by default. */ |
2103 | if (!sd_id128_is_null(u->invocation_id)) { | |
2104 | key_serial_t key; | |
2105 | ||
2106 | key = add_key("user", "invocation_id", &u->invocation_id, sizeof(u->invocation_id), KEY_SPEC_SESSION_KEYRING); | |
2107 | if (key == -1) | |
2108 | log_debug_errno(errno, "Failed to add invocation ID to keyring, ignoring: %m"); | |
2109 | else { | |
2110 | if (keyctl(KEYCTL_SETPERM, key, | |
2111 | KEY_POS_VIEW|KEY_POS_READ|KEY_POS_SEARCH| | |
2112 | KEY_USR_VIEW|KEY_USR_READ|KEY_USR_SEARCH, 0, 0) < 0) | |
2113 | return log_error_errno(errno, "Failed to restrict invocation ID permission: %m"); | |
2114 | } | |
2115 | } | |
2116 | ||
74dd6b51 LP |
2117 | /* And now, make the keyring owned by the service's user */ |
2118 | if (uid_is_valid(uid) || gid_is_valid(gid)) | |
2119 | if (keyctl(KEYCTL_CHOWN, keyring, uid, gid, 0) < 0) | |
2120 | return log_error_errno(errno, "Failed to change ownership of session keyring: %m"); | |
2121 | ||
2122 | return 0; | |
2123 | } | |
2124 | ||
29206d46 LP |
2125 | static void append_socket_pair(int *array, unsigned *n, int pair[2]) { |
2126 | assert(array); | |
2127 | assert(n); | |
2128 | ||
2129 | if (!pair) | |
2130 | return; | |
2131 | ||
2132 | if (pair[0] >= 0) | |
2133 | array[(*n)++] = pair[0]; | |
2134 | if (pair[1] >= 0) | |
2135 | array[(*n)++] = pair[1]; | |
2136 | } | |
2137 | ||
a34ceba6 LP |
2138 | static int close_remaining_fds( |
2139 | const ExecParameters *params, | |
2140 | ExecRuntime *runtime, | |
29206d46 | 2141 | DynamicCreds *dcreds, |
00d9ef85 | 2142 | int user_lookup_fd, |
a34ceba6 LP |
2143 | int socket_fd, |
2144 | int *fds, unsigned n_fds) { | |
2145 | ||
2146 | unsigned n_dont_close = 0; | |
00d9ef85 | 2147 | int dont_close[n_fds + 12]; |
a34ceba6 LP |
2148 | |
2149 | assert(params); | |
2150 | ||
2151 | if (params->stdin_fd >= 0) | |
2152 | dont_close[n_dont_close++] = params->stdin_fd; | |
2153 | if (params->stdout_fd >= 0) | |
2154 | dont_close[n_dont_close++] = params->stdout_fd; | |
2155 | if (params->stderr_fd >= 0) | |
2156 | dont_close[n_dont_close++] = params->stderr_fd; | |
2157 | ||
2158 | if (socket_fd >= 0) | |
2159 | dont_close[n_dont_close++] = socket_fd; | |
2160 | if (n_fds > 0) { | |
2161 | memcpy(dont_close + n_dont_close, fds, sizeof(int) * n_fds); | |
2162 | n_dont_close += n_fds; | |
2163 | } | |
2164 | ||
29206d46 LP |
2165 | if (runtime) |
2166 | append_socket_pair(dont_close, &n_dont_close, runtime->netns_storage_socket); | |
2167 | ||
2168 | if (dcreds) { | |
2169 | if (dcreds->user) | |
2170 | append_socket_pair(dont_close, &n_dont_close, dcreds->user->storage_socket); | |
2171 | if (dcreds->group) | |
2172 | append_socket_pair(dont_close, &n_dont_close, dcreds->group->storage_socket); | |
a34ceba6 LP |
2173 | } |
2174 | ||
00d9ef85 LP |
2175 | if (user_lookup_fd >= 0) |
2176 | dont_close[n_dont_close++] = user_lookup_fd; | |
2177 | ||
a34ceba6 LP |
2178 | return close_all_fds(dont_close, n_dont_close); |
2179 | } | |
2180 | ||
00d9ef85 LP |
2181 | static int send_user_lookup( |
2182 | Unit *unit, | |
2183 | int user_lookup_fd, | |
2184 | uid_t uid, | |
2185 | gid_t gid) { | |
2186 | ||
2187 | assert(unit); | |
2188 | ||
2189 | /* Send the resolved UID/GID to PID 1 after we learnt it. We send a single datagram, containing the UID/GID | |
2190 | * data as well as the unit name. Note that we suppress sending this if no user/group to resolve was | |
2191 | * specified. */ | |
2192 | ||
2193 | if (user_lookup_fd < 0) | |
2194 | return 0; | |
2195 | ||
2196 | if (!uid_is_valid(uid) && !gid_is_valid(gid)) | |
2197 | return 0; | |
2198 | ||
2199 | if (writev(user_lookup_fd, | |
2200 | (struct iovec[]) { | |
2201 | { .iov_base = &uid, .iov_len = sizeof(uid) }, | |
2202 | { .iov_base = &gid, .iov_len = sizeof(gid) }, | |
2203 | { .iov_base = unit->id, .iov_len = strlen(unit->id) }}, 3) < 0) | |
2204 | return -errno; | |
2205 | ||
2206 | return 0; | |
2207 | } | |
2208 | ||
6732edab LP |
2209 | static int acquire_home(const ExecContext *c, uid_t uid, const char** home, char **buf) { |
2210 | int r; | |
2211 | ||
2212 | assert(c); | |
2213 | assert(home); | |
2214 | assert(buf); | |
2215 | ||
2216 | /* If WorkingDirectory=~ is set, try to acquire a usable home directory. */ | |
2217 | ||
2218 | if (*home) | |
2219 | return 0; | |
2220 | ||
2221 | if (!c->working_directory_home) | |
2222 | return 0; | |
2223 | ||
2224 | if (uid == 0) { | |
2225 | /* Hardcode /root as home directory for UID 0 */ | |
2226 | *home = "/root"; | |
2227 | return 1; | |
2228 | } | |
2229 | ||
2230 | r = get_home_dir(buf); | |
2231 | if (r < 0) | |
2232 | return r; | |
2233 | ||
2234 | *home = *buf; | |
2235 | return 1; | |
2236 | } | |
2237 | ||
ff0af2a1 | 2238 | static int exec_child( |
f2341e0a | 2239 | Unit *unit, |
ff0af2a1 LP |
2240 | ExecCommand *command, |
2241 | const ExecContext *context, | |
2242 | const ExecParameters *params, | |
2243 | ExecRuntime *runtime, | |
29206d46 | 2244 | DynamicCreds *dcreds, |
ff0af2a1 LP |
2245 | char **argv, |
2246 | int socket_fd, | |
52c239d7 | 2247 | int named_iofds[3], |
4c47affc FB |
2248 | int *fds, |
2249 | unsigned n_storage_fds, | |
9b141911 | 2250 | unsigned n_socket_fds, |
ff0af2a1 | 2251 | char **files_env, |
00d9ef85 | 2252 | int user_lookup_fd, |
70dd455c ZJS |
2253 | int *exit_status, |
2254 | char **error_message) { | |
d35fbf6b | 2255 | |
2065ca69 | 2256 | _cleanup_strv_free_ char **our_env = NULL, **pass_env = NULL, **accum_env = NULL, **final_argv = NULL; |
6732edab | 2257 | _cleanup_free_ char *mac_selinux_context_net = NULL, *home_buffer = NULL; |
4d885bd3 DH |
2258 | _cleanup_free_ gid_t *supplementary_gids = NULL; |
2259 | const char *username = NULL, *groupname = NULL; | |
2b3c1b9e | 2260 | const char *home = NULL, *shell = NULL; |
7bce046b LP |
2261 | dev_t journal_stream_dev = 0; |
2262 | ino_t journal_stream_ino = 0; | |
2263 | bool needs_mount_namespace; | |
fed1e721 LP |
2264 | uid_t uid = UID_INVALID; |
2265 | gid_t gid = GID_INVALID; | |
4d885bd3 | 2266 | int i, r, ngids = 0; |
4c47affc | 2267 | unsigned n_fds; |
034c6ed7 | 2268 | |
f2341e0a | 2269 | assert(unit); |
5cb5a6ff LP |
2270 | assert(command); |
2271 | assert(context); | |
d35fbf6b | 2272 | assert(params); |
ff0af2a1 | 2273 | assert(exit_status); |
70dd455c ZJS |
2274 | assert(error_message); |
2275 | /* We don't always set error_message, hence it must be initialized */ | |
2276 | assert(*error_message == NULL); | |
d35fbf6b DM |
2277 | |
2278 | rename_process_from_path(command->path); | |
2279 | ||
2280 | /* We reset exactly these signals, since they are the | |
2281 | * only ones we set to SIG_IGN in the main daemon. All | |
2282 | * others we leave untouched because we set them to | |
2283 | * SIG_DFL or a valid handler initially, both of which | |
2284 | * will be demoted to SIG_DFL. */ | |
ce30c8dc LP |
2285 | (void) default_signals(SIGNALS_CRASH_HANDLER, |
2286 | SIGNALS_IGNORE, -1); | |
d35fbf6b DM |
2287 | |
2288 | if (context->ignore_sigpipe) | |
ce30c8dc | 2289 | (void) ignore_signals(SIGPIPE, -1); |
d35fbf6b | 2290 | |
ff0af2a1 LP |
2291 | r = reset_signal_mask(); |
2292 | if (r < 0) { | |
2293 | *exit_status = EXIT_SIGNAL_MASK; | |
70dd455c ZJS |
2294 | *error_message = strdup("Failed to reset signal mask"); |
2295 | /* If strdup fails, here and below, we will just print the generic error message. */ | |
ff0af2a1 | 2296 | return r; |
d35fbf6b | 2297 | } |
034c6ed7 | 2298 | |
d35fbf6b DM |
2299 | if (params->idle_pipe) |
2300 | do_idle_pipe_dance(params->idle_pipe); | |
4f2d528d | 2301 | |
d35fbf6b DM |
2302 | /* Close sockets very early to make sure we don't |
2303 | * block init reexecution because it cannot bind its | |
2304 | * sockets */ | |
ff0af2a1 | 2305 | |
d35fbf6b | 2306 | log_forget_fds(); |
4f2d528d | 2307 | |
4c47affc | 2308 | n_fds = n_storage_fds + n_socket_fds; |
00d9ef85 | 2309 | r = close_remaining_fds(params, runtime, dcreds, user_lookup_fd, socket_fd, fds, n_fds); |
ff0af2a1 LP |
2310 | if (r < 0) { |
2311 | *exit_status = EXIT_FDS; | |
70dd455c | 2312 | *error_message = strdup("Failed to close remaining fds"); |
ff0af2a1 | 2313 | return r; |
8c7be95e LP |
2314 | } |
2315 | ||
d35fbf6b DM |
2316 | if (!context->same_pgrp) |
2317 | if (setsid() < 0) { | |
ff0af2a1 | 2318 | *exit_status = EXIT_SETSID; |
d35fbf6b DM |
2319 | return -errno; |
2320 | } | |
9e2f7c11 | 2321 | |
1e22b5cd | 2322 | exec_context_tty_reset(context, params); |
d35fbf6b | 2323 | |
c891efaf | 2324 | if (unit_shall_confirm_spawn(unit)) { |
7d5ceb64 | 2325 | const char *vc = params->confirm_spawn; |
3b20f877 FB |
2326 | _cleanup_free_ char *cmdline = NULL; |
2327 | ||
2328 | cmdline = exec_command_line(argv); | |
2329 | if (!cmdline) { | |
2330 | *exit_status = EXIT_CONFIRM; | |
2331 | return -ENOMEM; | |
2332 | } | |
d35fbf6b | 2333 | |
eedf223a | 2334 | r = ask_for_confirmation(vc, unit, cmdline); |
3b20f877 FB |
2335 | if (r != CONFIRM_EXECUTE) { |
2336 | if (r == CONFIRM_PRETEND_SUCCESS) { | |
2337 | *exit_status = EXIT_SUCCESS; | |
2338 | return 0; | |
2339 | } | |
ff0af2a1 | 2340 | *exit_status = EXIT_CONFIRM; |
70dd455c | 2341 | *error_message = strdup("Execution cancelled"); |
d35fbf6b | 2342 | return -ECANCELED; |
d35fbf6b DM |
2343 | } |
2344 | } | |
1a63a750 | 2345 | |
29206d46 LP |
2346 | if (context->dynamic_user && dcreds) { |
2347 | ||
409093fe LP |
2348 | /* Make sure we bypass our own NSS module for any NSS checks */ |
2349 | if (putenv((char*) "SYSTEMD_NSS_DYNAMIC_BYPASS=1") != 0) { | |
2350 | *exit_status = EXIT_USER; | |
70dd455c | 2351 | *error_message = strdup("Failed to update environment"); |
409093fe LP |
2352 | return -errno; |
2353 | } | |
2354 | ||
29206d46 | 2355 | r = dynamic_creds_realize(dcreds, &uid, &gid); |
ff0af2a1 LP |
2356 | if (r < 0) { |
2357 | *exit_status = EXIT_USER; | |
70dd455c | 2358 | *error_message = strdup("Failed to update dynamic user credentials"); |
ff0af2a1 | 2359 | return r; |
524daa8c | 2360 | } |
524daa8c | 2361 | |
70dd455c | 2362 | if (!uid_is_valid(uid)) { |
29206d46 | 2363 | *exit_status = EXIT_USER; |
70dd455c ZJS |
2364 | (void) asprintf(error_message, "UID validation failed for \""UID_FMT"\"", uid); |
2365 | /* If asprintf fails, here and below, we will just print the generic error message. */ | |
2366 | return -ESRCH; | |
2367 | } | |
2368 | ||
2369 | if (!gid_is_valid(gid)) { | |
2370 | *exit_status = EXIT_USER; | |
2371 | (void) asprintf(error_message, "GID validation failed for \""GID_FMT"\"", gid); | |
29206d46 LP |
2372 | return -ESRCH; |
2373 | } | |
5bc7452b | 2374 | |
29206d46 LP |
2375 | if (dcreds->user) |
2376 | username = dcreds->user->name; | |
2377 | ||
2378 | } else { | |
4d885bd3 DH |
2379 | r = get_fixed_user(context, &username, &uid, &gid, &home, &shell); |
2380 | if (r < 0) { | |
2381 | *exit_status = EXIT_USER; | |
70dd455c | 2382 | *error_message = strdup("Failed to determine user credentials"); |
4d885bd3 | 2383 | return r; |
5bc7452b | 2384 | } |
5bc7452b | 2385 | |
4d885bd3 DH |
2386 | r = get_fixed_group(context, &groupname, &gid); |
2387 | if (r < 0) { | |
2388 | *exit_status = EXIT_GROUP; | |
70dd455c | 2389 | *error_message = strdup("Failed to determine group credentials"); |
4d885bd3 DH |
2390 | return r; |
2391 | } | |
cdc5d5c5 | 2392 | } |
29206d46 | 2393 | |
cdc5d5c5 DH |
2394 | /* Initialize user supplementary groups and get SupplementaryGroups= ones */ |
2395 | r = get_supplementary_groups(context, username, groupname, gid, | |
2396 | &supplementary_gids, &ngids); | |
2397 | if (r < 0) { | |
2398 | *exit_status = EXIT_GROUP; | |
70dd455c | 2399 | *error_message = strdup("Failed to determine supplementary groups"); |
cdc5d5c5 | 2400 | return r; |
29206d46 | 2401 | } |
5bc7452b | 2402 | |
00d9ef85 LP |
2403 | r = send_user_lookup(unit, user_lookup_fd, uid, gid); |
2404 | if (r < 0) { | |
2405 | *exit_status = EXIT_USER; | |
70dd455c | 2406 | *error_message = strdup("Failed to send user credentials to PID1"); |
00d9ef85 LP |
2407 | return r; |
2408 | } | |
2409 | ||
2410 | user_lookup_fd = safe_close(user_lookup_fd); | |
2411 | ||
6732edab LP |
2412 | r = acquire_home(context, uid, &home, &home_buffer); |
2413 | if (r < 0) { | |
2414 | *exit_status = EXIT_CHDIR; | |
2415 | *error_message = strdup("Failed to determine $HOME for user"); | |
2416 | return r; | |
2417 | } | |
2418 | ||
d35fbf6b DM |
2419 | /* If a socket is connected to STDIN/STDOUT/STDERR, we |
2420 | * must sure to drop O_NONBLOCK */ | |
2421 | if (socket_fd >= 0) | |
a34ceba6 | 2422 | (void) fd_nonblock(socket_fd, false); |
acbb0225 | 2423 | |
52c239d7 | 2424 | r = setup_input(context, params, socket_fd, named_iofds); |
ff0af2a1 LP |
2425 | if (r < 0) { |
2426 | *exit_status = EXIT_STDIN; | |
70dd455c | 2427 | *error_message = strdup("Failed to set up stdin"); |
ff0af2a1 | 2428 | return r; |
d35fbf6b | 2429 | } |
034c6ed7 | 2430 | |
52c239d7 | 2431 | r = setup_output(unit, context, params, STDOUT_FILENO, socket_fd, named_iofds, basename(command->path), uid, gid, &journal_stream_dev, &journal_stream_ino); |
ff0af2a1 LP |
2432 | if (r < 0) { |
2433 | *exit_status = EXIT_STDOUT; | |
70dd455c | 2434 | *error_message = strdup("Failed to set up stdout"); |
ff0af2a1 | 2435 | return r; |
d35fbf6b DM |
2436 | } |
2437 | ||
52c239d7 | 2438 | r = setup_output(unit, context, params, STDERR_FILENO, socket_fd, named_iofds, basename(command->path), uid, gid, &journal_stream_dev, &journal_stream_ino); |
ff0af2a1 LP |
2439 | if (r < 0) { |
2440 | *exit_status = EXIT_STDERR; | |
70dd455c | 2441 | *error_message = strdup("Failed to set up stderr"); |
ff0af2a1 | 2442 | return r; |
d35fbf6b DM |
2443 | } |
2444 | ||
2445 | if (params->cgroup_path) { | |
ff0af2a1 LP |
2446 | r = cg_attach_everywhere(params->cgroup_supported, params->cgroup_path, 0, NULL, NULL); |
2447 | if (r < 0) { | |
2448 | *exit_status = EXIT_CGROUP; | |
70dd455c | 2449 | (void) asprintf(error_message, "Failed to attach to cgroup %s", params->cgroup_path); |
ff0af2a1 | 2450 | return r; |
309bff19 | 2451 | } |
d35fbf6b | 2452 | } |
309bff19 | 2453 | |
d35fbf6b | 2454 | if (context->oom_score_adjust_set) { |
d5243d62 | 2455 | char t[DECIMAL_STR_MAX(context->oom_score_adjust)]; |
f2b68789 | 2456 | |
d5243d62 LP |
2457 | /* When we can't make this change due to EPERM, then |
2458 | * let's silently skip over it. User namespaces | |
2459 | * prohibit write access to this file, and we | |
2460 | * shouldn't trip up over that. */ | |
613b411c | 2461 | |
d5243d62 | 2462 | sprintf(t, "%i", context->oom_score_adjust); |
ad118bda | 2463 | r = write_string_file("/proc/self/oom_score_adj", t, 0); |
6cb7fa17 | 2464 | if (r == -EPERM || r == -EACCES) { |
ff0af2a1 | 2465 | log_open(); |
f2341e0a | 2466 | log_unit_debug_errno(unit, r, "Failed to adjust OOM setting, assuming containerized execution, ignoring: %m"); |
ff0af2a1 LP |
2467 | log_close(); |
2468 | } else if (r < 0) { | |
2469 | *exit_status = EXIT_OOM_ADJUST; | |
70dd455c | 2470 | *error_message = strdup("Failed to write /proc/self/oom_score_adj"); |
d35fbf6b | 2471 | return -errno; |
613b411c | 2472 | } |
d35fbf6b DM |
2473 | } |
2474 | ||
2475 | if (context->nice_set) | |
2476 | if (setpriority(PRIO_PROCESS, 0, context->nice) < 0) { | |
ff0af2a1 | 2477 | *exit_status = EXIT_NICE; |
d35fbf6b | 2478 | return -errno; |
613b411c LP |
2479 | } |
2480 | ||
d35fbf6b DM |
2481 | if (context->cpu_sched_set) { |
2482 | struct sched_param param = { | |
2483 | .sched_priority = context->cpu_sched_priority, | |
2484 | }; | |
2485 | ||
ff0af2a1 LP |
2486 | r = sched_setscheduler(0, |
2487 | context->cpu_sched_policy | | |
2488 | (context->cpu_sched_reset_on_fork ? | |
2489 | SCHED_RESET_ON_FORK : 0), | |
2490 | ¶m); | |
2491 | if (r < 0) { | |
2492 | *exit_status = EXIT_SETSCHEDULER; | |
d35fbf6b | 2493 | return -errno; |
fc9b2a84 | 2494 | } |
d35fbf6b | 2495 | } |
fc9b2a84 | 2496 | |
d35fbf6b DM |
2497 | if (context->cpuset) |
2498 | if (sched_setaffinity(0, CPU_ALLOC_SIZE(context->cpuset_ncpus), context->cpuset) < 0) { | |
ff0af2a1 | 2499 | *exit_status = EXIT_CPUAFFINITY; |
d35fbf6b | 2500 | return -errno; |
034c6ed7 LP |
2501 | } |
2502 | ||
d35fbf6b DM |
2503 | if (context->ioprio_set) |
2504 | if (ioprio_set(IOPRIO_WHO_PROCESS, 0, context->ioprio) < 0) { | |
ff0af2a1 | 2505 | *exit_status = EXIT_IOPRIO; |
d35fbf6b DM |
2506 | return -errno; |
2507 | } | |
da726a4d | 2508 | |
d35fbf6b DM |
2509 | if (context->timer_slack_nsec != NSEC_INFINITY) |
2510 | if (prctl(PR_SET_TIMERSLACK, context->timer_slack_nsec) < 0) { | |
ff0af2a1 | 2511 | *exit_status = EXIT_TIMERSLACK; |
d35fbf6b | 2512 | return -errno; |
4c2630eb | 2513 | } |
9eba9da4 | 2514 | |
050f7277 | 2515 | if (context->personality != PERSONALITY_INVALID) |
d35fbf6b | 2516 | if (personality(context->personality) < 0) { |
ff0af2a1 | 2517 | *exit_status = EXIT_PERSONALITY; |
d35fbf6b | 2518 | return -errno; |
4c2630eb | 2519 | } |
94f04347 | 2520 | |
d35fbf6b | 2521 | if (context->utmp_id) |
6a93917d ZJS |
2522 | utmp_put_init_process(context->utmp_id, getpid(), getsid(0), |
2523 | context->tty_path, | |
023a4f67 LP |
2524 | context->utmp_mode == EXEC_UTMP_INIT ? INIT_PROCESS : |
2525 | context->utmp_mode == EXEC_UTMP_LOGIN ? LOGIN_PROCESS : | |
2526 | USER_PROCESS, | |
6a93917d | 2527 | username); |
d35fbf6b | 2528 | |
e0d2adfd | 2529 | if (context->user) { |
ff0af2a1 LP |
2530 | r = chown_terminal(STDIN_FILENO, uid); |
2531 | if (r < 0) { | |
2532 | *exit_status = EXIT_STDIN; | |
2533 | return r; | |
071830ff | 2534 | } |
d35fbf6b | 2535 | } |
8e274523 | 2536 | |
a931ad47 LP |
2537 | /* If delegation is enabled we'll pass ownership of the cgroup |
2538 | * (but only in systemd's own controller hierarchy!) to the | |
2539 | * user of the new process. */ | |
2540 | if (params->cgroup_path && context->user && params->cgroup_delegate) { | |
ff0af2a1 LP |
2541 | r = cg_set_task_access(SYSTEMD_CGROUP_CONTROLLER, params->cgroup_path, 0644, uid, gid); |
2542 | if (r < 0) { | |
2543 | *exit_status = EXIT_CGROUP; | |
2544 | return r; | |
d35fbf6b | 2545 | } |
034c6ed7 | 2546 | |
034c6ed7 | 2547 | |
ff0af2a1 LP |
2548 | r = cg_set_group_access(SYSTEMD_CGROUP_CONTROLLER, params->cgroup_path, 0755, uid, gid); |
2549 | if (r < 0) { | |
2550 | *exit_status = EXIT_CGROUP; | |
2551 | return r; | |
034c6ed7 | 2552 | } |
d35fbf6b | 2553 | } |
034c6ed7 | 2554 | |
d35fbf6b | 2555 | if (!strv_isempty(context->runtime_directory) && params->runtime_prefix) { |
07689d5d LP |
2556 | r = setup_runtime_directory(context, params, uid, gid); |
2557 | if (r < 0) { | |
2558 | *exit_status = EXIT_RUNTIME_DIRECTORY; | |
2559 | return r; | |
d35fbf6b DM |
2560 | } |
2561 | } | |
94f04347 | 2562 | |
7bce046b | 2563 | r = build_environment( |
fd63e712 | 2564 | unit, |
7bce046b LP |
2565 | context, |
2566 | params, | |
2567 | n_fds, | |
2568 | home, | |
2569 | username, | |
2570 | shell, | |
2571 | journal_stream_dev, | |
2572 | journal_stream_ino, | |
2573 | &our_env); | |
2065ca69 JW |
2574 | if (r < 0) { |
2575 | *exit_status = EXIT_MEMORY; | |
2576 | return r; | |
2577 | } | |
2578 | ||
2579 | r = build_pass_environment(context, &pass_env); | |
2580 | if (r < 0) { | |
2581 | *exit_status = EXIT_MEMORY; | |
2582 | return r; | |
2583 | } | |
2584 | ||
2585 | accum_env = strv_env_merge(5, | |
2586 | params->environment, | |
2587 | our_env, | |
2588 | pass_env, | |
2589 | context->environment, | |
2590 | files_env, | |
2591 | NULL); | |
2592 | if (!accum_env) { | |
2593 | *exit_status = EXIT_MEMORY; | |
2594 | return -ENOMEM; | |
2595 | } | |
1280503b | 2596 | accum_env = strv_env_clean(accum_env); |
2065ca69 | 2597 | |
096424d1 | 2598 | (void) umask(context->umask); |
b213e1c1 | 2599 | |
74dd6b51 LP |
2600 | r = setup_keyring(unit, params, uid, gid); |
2601 | if (r < 0) { | |
2602 | *exit_status = EXIT_KEYRING; | |
2603 | return r; | |
2604 | } | |
2605 | ||
c39f1ce2 | 2606 | if ((params->flags & EXEC_APPLY_PERMISSIONS) && !command->privileged) { |
b213e1c1 | 2607 | if (context->pam_name && username) { |
2d6fce8d | 2608 | r = setup_pam(context->pam_name, username, uid, gid, context->tty_path, &accum_env, fds, n_fds); |
b213e1c1 SW |
2609 | if (r < 0) { |
2610 | *exit_status = EXIT_PAM; | |
2611 | return r; | |
2612 | } | |
d35fbf6b | 2613 | } |
b213e1c1 | 2614 | } |
ac45f971 | 2615 | |
d35fbf6b | 2616 | if (context->private_network && runtime && runtime->netns_storage_socket[0] >= 0) { |
ff0af2a1 LP |
2617 | r = setup_netns(runtime->netns_storage_socket); |
2618 | if (r < 0) { | |
2619 | *exit_status = EXIT_NETWORK; | |
2620 | return r; | |
d35fbf6b DM |
2621 | } |
2622 | } | |
169c1bda | 2623 | |
ee818b89 | 2624 | needs_mount_namespace = exec_needs_mount_namespace(context, params, runtime); |
ee818b89 | 2625 | if (needs_mount_namespace) { |
6818c54c | 2626 | r = apply_mount_namespace(unit, command, context, params, runtime); |
3fbe8dbe LP |
2627 | if (r < 0) { |
2628 | *exit_status = EXIT_NAMESPACE; | |
2629 | return r; | |
2630 | } | |
d35fbf6b | 2631 | } |
81a2b7ce | 2632 | |
50b3dfb9 | 2633 | /* Apply just after mount namespace setup */ |
376fecf6 LP |
2634 | r = apply_working_directory(context, params, home, needs_mount_namespace, exit_status); |
2635 | if (r < 0) | |
50b3dfb9 | 2636 | return r; |
50b3dfb9 | 2637 | |
bbeea271 | 2638 | /* Drop groups as early as possbile */ |
096424d1 | 2639 | if ((params->flags & EXEC_APPLY_PERMISSIONS) && !command->privileged) { |
4d885bd3 | 2640 | r = enforce_groups(context, gid, supplementary_gids, ngids); |
096424d1 LP |
2641 | if (r < 0) { |
2642 | *exit_status = EXIT_GROUP; | |
2643 | return r; | |
2644 | } | |
2645 | } | |
2646 | ||
9008e1ac | 2647 | #ifdef HAVE_SELINUX |
c39f1ce2 LP |
2648 | if ((params->flags & EXEC_APPLY_PERMISSIONS) && |
2649 | mac_selinux_use() && | |
2650 | params->selinux_context_net && | |
2651 | socket_fd >= 0 && | |
2652 | !command->privileged) { | |
2653 | ||
ff0af2a1 LP |
2654 | r = mac_selinux_get_child_mls_label(socket_fd, command->path, context->selinux_context, &mac_selinux_context_net); |
2655 | if (r < 0) { | |
2656 | *exit_status = EXIT_SELINUX_CONTEXT; | |
2657 | return r; | |
9008e1ac MS |
2658 | } |
2659 | } | |
2660 | #endif | |
2661 | ||
d87a2ef7 | 2662 | if ((params->flags & EXEC_APPLY_PERMISSIONS) && context->private_users) { |
d251207d LP |
2663 | r = setup_private_users(uid, gid); |
2664 | if (r < 0) { | |
2665 | *exit_status = EXIT_USER; | |
2666 | return r; | |
2667 | } | |
2668 | } | |
2669 | ||
d35fbf6b DM |
2670 | /* We repeat the fd closing here, to make sure that |
2671 | * nothing is leaked from the PAM modules. Note that | |
2672 | * we are more aggressive this time since socket_fd | |
e44da745 DM |
2673 | * and the netns fds we don't need anymore. The custom |
2674 | * endpoint fd was needed to upload the policy and can | |
2675 | * now be closed as well. */ | |
ff0af2a1 LP |
2676 | r = close_all_fds(fds, n_fds); |
2677 | if (r >= 0) | |
2678 | r = shift_fds(fds, n_fds); | |
2679 | if (r >= 0) | |
4c47affc | 2680 | r = flags_fds(fds, n_storage_fds, n_socket_fds, context->non_blocking); |
ff0af2a1 LP |
2681 | if (r < 0) { |
2682 | *exit_status = EXIT_FDS; | |
2683 | return r; | |
d35fbf6b | 2684 | } |
e66cf1a3 | 2685 | |
c39f1ce2 | 2686 | if ((params->flags & EXEC_APPLY_PERMISSIONS) && !command->privileged) { |
e66cf1a3 | 2687 | |
755d4b67 IP |
2688 | int secure_bits = context->secure_bits; |
2689 | ||
d35fbf6b | 2690 | for (i = 0; i < _RLIMIT_MAX; i++) { |
03857c43 | 2691 | |
d35fbf6b DM |
2692 | if (!context->rlimit[i]) |
2693 | continue; | |
2694 | ||
03857c43 LP |
2695 | r = setrlimit_closest(i, context->rlimit[i]); |
2696 | if (r < 0) { | |
ff0af2a1 | 2697 | *exit_status = EXIT_LIMITS; |
03857c43 | 2698 | return r; |
e66cf1a3 LP |
2699 | } |
2700 | } | |
2701 | ||
f4170c67 LP |
2702 | /* Set the RTPRIO resource limit to 0, but only if nothing else was explicitly requested. */ |
2703 | if (context->restrict_realtime && !context->rlimit[RLIMIT_RTPRIO]) { | |
2704 | if (setrlimit(RLIMIT_RTPRIO, &RLIMIT_MAKE_CONST(0)) < 0) { | |
2705 | *exit_status = EXIT_LIMITS; | |
2706 | return -errno; | |
2707 | } | |
2708 | } | |
2709 | ||
a103496c IP |
2710 | if (!cap_test_all(context->capability_bounding_set)) { |
2711 | r = capability_bounding_set_drop(context->capability_bounding_set, false); | |
ff0af2a1 LP |
2712 | if (r < 0) { |
2713 | *exit_status = EXIT_CAPABILITIES; | |
70dd455c | 2714 | *error_message = strdup("Failed to drop capabilities"); |
ff0af2a1 | 2715 | return r; |
3b8bddde | 2716 | } |
4c2630eb | 2717 | } |
3b8bddde | 2718 | |
755d4b67 IP |
2719 | /* This is done before enforce_user, but ambient set |
2720 | * does not survive over setresuid() if keep_caps is not set. */ | |
2721 | if (context->capability_ambient_set != 0) { | |
2722 | r = capability_ambient_set_apply(context->capability_ambient_set, true); | |
2723 | if (r < 0) { | |
2724 | *exit_status = EXIT_CAPABILITIES; | |
70dd455c | 2725 | *error_message = strdup("Failed to apply ambient capabilities (before UID change)"); |
755d4b67 IP |
2726 | return r; |
2727 | } | |
755d4b67 IP |
2728 | } |
2729 | ||
d35fbf6b | 2730 | if (context->user) { |
ff0af2a1 LP |
2731 | r = enforce_user(context, uid); |
2732 | if (r < 0) { | |
2733 | *exit_status = EXIT_USER; | |
70dd455c | 2734 | (void) asprintf(error_message, "Failed to change UID to "UID_FMT, uid); |
ff0af2a1 | 2735 | return r; |
5b6319dc | 2736 | } |
755d4b67 IP |
2737 | if (context->capability_ambient_set != 0) { |
2738 | ||
2739 | /* Fix the ambient capabilities after user change. */ | |
2740 | r = capability_ambient_set_apply(context->capability_ambient_set, false); | |
2741 | if (r < 0) { | |
2742 | *exit_status = EXIT_CAPABILITIES; | |
70dd455c | 2743 | *error_message = strdup("Failed to apply ambient capabilities (after UID change)"); |
755d4b67 IP |
2744 | return r; |
2745 | } | |
2746 | ||
2747 | /* If we were asked to change user and ambient capabilities | |
2748 | * were requested, we had to add keep-caps to the securebits | |
2749 | * so that we would maintain the inherited capability set | |
2750 | * through the setresuid(). Make sure that the bit is added | |
2751 | * also to the context secure_bits so that we don't try to | |
2752 | * drop the bit away next. */ | |
2753 | ||
7f508f2c | 2754 | secure_bits |= 1<<SECURE_KEEP_CAPS; |
755d4b67 | 2755 | } |
5b6319dc | 2756 | } |
d35fbf6b | 2757 | |
5cd9cd35 LP |
2758 | /* Apply the MAC contexts late, but before seccomp syscall filtering, as those should really be last to |
2759 | * influence our own codepaths as little as possible. Moreover, applying MAC contexts usually requires | |
2760 | * syscalls that are subject to seccomp filtering, hence should probably be applied before the syscalls | |
2761 | * are restricted. */ | |
2762 | ||
2763 | #ifdef HAVE_SELINUX | |
2764 | if (mac_selinux_use()) { | |
2765 | char *exec_context = mac_selinux_context_net ?: context->selinux_context; | |
2766 | ||
2767 | if (exec_context) { | |
2768 | r = setexeccon(exec_context); | |
2769 | if (r < 0) { | |
2770 | *exit_status = EXIT_SELINUX_CONTEXT; | |
70dd455c | 2771 | (void) asprintf(error_message, "Failed to set SELinux context to %s", exec_context); |
5cd9cd35 LP |
2772 | return r; |
2773 | } | |
2774 | } | |
2775 | } | |
2776 | #endif | |
2777 | ||
2778 | r = setup_smack(context, command); | |
2779 | if (r < 0) { | |
2780 | *exit_status = EXIT_SMACK_PROCESS_LABEL; | |
70dd455c | 2781 | *error_message = strdup("Failed to set SMACK process label"); |
5cd9cd35 LP |
2782 | return r; |
2783 | } | |
2784 | ||
2785 | #ifdef HAVE_APPARMOR | |
2786 | if (context->apparmor_profile && mac_apparmor_use()) { | |
2787 | r = aa_change_onexec(context->apparmor_profile); | |
2788 | if (r < 0 && !context->apparmor_profile_ignore) { | |
2789 | *exit_status = EXIT_APPARMOR_PROFILE; | |
70dd455c ZJS |
2790 | (void) asprintf(error_message, |
2791 | "Failed to prepare AppArmor profile change to %s", | |
2792 | context->apparmor_profile); | |
5cd9cd35 LP |
2793 | return -errno; |
2794 | } | |
2795 | } | |
2796 | #endif | |
2797 | ||
d35fbf6b DM |
2798 | /* PR_GET_SECUREBITS is not privileged, while |
2799 | * PR_SET_SECUREBITS is. So to suppress | |
2800 | * potential EPERMs we'll try not to call | |
2801 | * PR_SET_SECUREBITS unless necessary. */ | |
755d4b67 IP |
2802 | if (prctl(PR_GET_SECUREBITS) != secure_bits) |
2803 | if (prctl(PR_SET_SECUREBITS, secure_bits) < 0) { | |
ff0af2a1 | 2804 | *exit_status = EXIT_SECUREBITS; |
70dd455c | 2805 | *error_message = strdup("Failed to set secure bits"); |
d35fbf6b | 2806 | return -errno; |
ff01d048 | 2807 | } |
5b6319dc | 2808 | |
59eeb84b | 2809 | if (context_has_no_new_privileges(context)) |
d35fbf6b | 2810 | if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) { |
ff0af2a1 | 2811 | *exit_status = EXIT_NO_NEW_PRIVILEGES; |
70dd455c | 2812 | *error_message = strdup("Failed to disable new privileges"); |
d35fbf6b DM |
2813 | return -errno; |
2814 | } | |
2815 | ||
2816 | #ifdef HAVE_SECCOMP | |
469830d1 LP |
2817 | r = apply_address_families(unit, context); |
2818 | if (r < 0) { | |
2819 | *exit_status = EXIT_ADDRESS_FAMILIES; | |
5b3637b4 | 2820 | *error_message = strdup("Failed to restrict address families"); |
469830d1 | 2821 | return r; |
4c2630eb | 2822 | } |
04aa0cb9 | 2823 | |
469830d1 LP |
2824 | r = apply_memory_deny_write_execute(unit, context); |
2825 | if (r < 0) { | |
2826 | *exit_status = EXIT_SECCOMP; | |
5b3637b4 | 2827 | *error_message = strdup("Failed to disable writing to executable memory"); |
469830d1 | 2828 | return r; |
f3e43635 | 2829 | } |
f4170c67 | 2830 | |
469830d1 LP |
2831 | r = apply_restrict_realtime(unit, context); |
2832 | if (r < 0) { | |
2833 | *exit_status = EXIT_SECCOMP; | |
5b3637b4 | 2834 | *error_message = strdup("Failed to apply realtime restrictions"); |
469830d1 | 2835 | return r; |
f4170c67 LP |
2836 | } |
2837 | ||
add00535 LP |
2838 | r = apply_restrict_namespaces(unit, context); |
2839 | if (r < 0) { | |
2840 | *exit_status = EXIT_SECCOMP; | |
70dd455c | 2841 | *error_message = strdup("Failed to apply namespace restrictions"); |
add00535 LP |
2842 | return r; |
2843 | } | |
2844 | ||
469830d1 LP |
2845 | r = apply_protect_sysctl(unit, context); |
2846 | if (r < 0) { | |
2847 | *exit_status = EXIT_SECCOMP; | |
5b3637b4 | 2848 | *error_message = strdup("Failed to apply sysctl restrictions"); |
469830d1 | 2849 | return r; |
502d704e DH |
2850 | } |
2851 | ||
469830d1 LP |
2852 | r = apply_protect_kernel_modules(unit, context); |
2853 | if (r < 0) { | |
2854 | *exit_status = EXIT_SECCOMP; | |
5b3637b4 | 2855 | *error_message = strdup("Failed to apply module loading restrictions"); |
469830d1 | 2856 | return r; |
59eeb84b LP |
2857 | } |
2858 | ||
469830d1 LP |
2859 | r = apply_private_devices(unit, context); |
2860 | if (r < 0) { | |
2861 | *exit_status = EXIT_SECCOMP; | |
5b3637b4 | 2862 | *error_message = strdup("Failed to set up private devices"); |
469830d1 LP |
2863 | return r; |
2864 | } | |
2865 | ||
2866 | r = apply_syscall_archs(unit, context); | |
2867 | if (r < 0) { | |
2868 | *exit_status = EXIT_SECCOMP; | |
5b3637b4 | 2869 | *error_message = strdup("Failed to apply syscall architecture restrictions"); |
469830d1 | 2870 | return r; |
ba128bb8 LP |
2871 | } |
2872 | ||
5cd9cd35 LP |
2873 | /* This really should remain the last step before the execve(), to make sure our own code is unaffected |
2874 | * by the filter as little as possible. */ | |
469830d1 LP |
2875 | r = apply_syscall_filter(unit, context); |
2876 | if (r < 0) { | |
2877 | *exit_status = EXIT_SECCOMP; | |
5b3637b4 | 2878 | *error_message = strdup("Failed to apply syscall filters"); |
469830d1 | 2879 | return r; |
d35fbf6b DM |
2880 | } |
2881 | #endif | |
d35fbf6b | 2882 | } |
034c6ed7 | 2883 | |
2065ca69 | 2884 | final_argv = replace_env_argv(argv, accum_env); |
d35fbf6b | 2885 | if (!final_argv) { |
ff0af2a1 | 2886 | *exit_status = EXIT_MEMORY; |
70dd455c | 2887 | *error_message = strdup("Failed to prepare process arguments"); |
d35fbf6b DM |
2888 | return -ENOMEM; |
2889 | } | |
034c6ed7 | 2890 | |
553d2243 | 2891 | if (_unlikely_(log_get_max_level() >= LOG_DEBUG)) { |
d35fbf6b | 2892 | _cleanup_free_ char *line; |
81a2b7ce | 2893 | |
d35fbf6b DM |
2894 | line = exec_command_line(final_argv); |
2895 | if (line) { | |
2896 | log_open(); | |
f2341e0a | 2897 | log_struct(LOG_DEBUG, |
f2341e0a LP |
2898 | "EXECUTABLE=%s", command->path, |
2899 | LOG_UNIT_MESSAGE(unit, "Executing: %s", line), | |
ba360bb0 | 2900 | LOG_UNIT_ID(unit), |
f2341e0a | 2901 | NULL); |
d35fbf6b DM |
2902 | log_close(); |
2903 | } | |
2904 | } | |
dd305ec9 | 2905 | |
2065ca69 | 2906 | execve(command->path, final_argv, accum_env); |
ff0af2a1 | 2907 | *exit_status = EXIT_EXEC; |
d35fbf6b DM |
2908 | return -errno; |
2909 | } | |
81a2b7ce | 2910 | |
f2341e0a LP |
2911 | int exec_spawn(Unit *unit, |
2912 | ExecCommand *command, | |
d35fbf6b DM |
2913 | const ExecContext *context, |
2914 | const ExecParameters *params, | |
2915 | ExecRuntime *runtime, | |
29206d46 | 2916 | DynamicCreds *dcreds, |
d35fbf6b | 2917 | pid_t *ret) { |
8351ceae | 2918 | |
d35fbf6b | 2919 | _cleanup_strv_free_ char **files_env = NULL; |
9b141911 | 2920 | int *fds = NULL; |
4c47affc | 2921 | unsigned n_storage_fds = 0, n_socket_fds = 0; |
ff0af2a1 LP |
2922 | _cleanup_free_ char *line = NULL; |
2923 | int socket_fd, r; | |
52c239d7 | 2924 | int named_iofds[3] = { -1, -1, -1 }; |
ff0af2a1 | 2925 | char **argv; |
d35fbf6b | 2926 | pid_t pid; |
8351ceae | 2927 | |
f2341e0a | 2928 | assert(unit); |
d35fbf6b DM |
2929 | assert(command); |
2930 | assert(context); | |
2931 | assert(ret); | |
2932 | assert(params); | |
4c47affc | 2933 | assert(params->fds || (params->n_storage_fds + params->n_socket_fds <= 0)); |
4298d0b5 | 2934 | |
d35fbf6b DM |
2935 | if (context->std_input == EXEC_INPUT_SOCKET || |
2936 | context->std_output == EXEC_OUTPUT_SOCKET || | |
2937 | context->std_error == EXEC_OUTPUT_SOCKET) { | |
17df7223 | 2938 | |
4c47affc | 2939 | if (params->n_socket_fds > 1) { |
f2341e0a | 2940 | log_unit_error(unit, "Got more than one socket."); |
d35fbf6b | 2941 | return -EINVAL; |
ff0af2a1 | 2942 | } |
eef65bf3 | 2943 | |
4c47affc | 2944 | if (params->n_socket_fds == 0) { |
488ab41c AA |
2945 | log_unit_error(unit, "Got no socket."); |
2946 | return -EINVAL; | |
2947 | } | |
2948 | ||
d35fbf6b DM |
2949 | socket_fd = params->fds[0]; |
2950 | } else { | |
2951 | socket_fd = -1; | |
2952 | fds = params->fds; | |
4c47affc | 2953 | n_storage_fds = params->n_storage_fds; |
9b141911 | 2954 | n_socket_fds = params->n_socket_fds; |
d35fbf6b | 2955 | } |
94f04347 | 2956 | |
52c239d7 LB |
2957 | r = exec_context_named_iofds(unit, context, params, named_iofds); |
2958 | if (r < 0) | |
2959 | return log_unit_error_errno(unit, r, "Failed to load a named file descriptor: %m"); | |
2960 | ||
f2341e0a | 2961 | r = exec_context_load_environment(unit, context, &files_env); |
ff0af2a1 | 2962 | if (r < 0) |
f2341e0a | 2963 | return log_unit_error_errno(unit, r, "Failed to load environment files: %m"); |
034c6ed7 | 2964 | |
d35fbf6b | 2965 | argv = params->argv ?: command->argv; |
d35fbf6b DM |
2966 | line = exec_command_line(argv); |
2967 | if (!line) | |
2968 | return log_oom(); | |
fab56fc5 | 2969 | |
f2341e0a | 2970 | log_struct(LOG_DEBUG, |
f2341e0a LP |
2971 | LOG_UNIT_MESSAGE(unit, "About to execute: %s", line), |
2972 | "EXECUTABLE=%s", command->path, | |
ba360bb0 | 2973 | LOG_UNIT_ID(unit), |
f2341e0a | 2974 | NULL); |
d35fbf6b DM |
2975 | pid = fork(); |
2976 | if (pid < 0) | |
74129a12 | 2977 | return log_unit_error_errno(unit, errno, "Failed to fork: %m"); |
d35fbf6b DM |
2978 | |
2979 | if (pid == 0) { | |
ff0af2a1 | 2980 | int exit_status; |
70dd455c | 2981 | _cleanup_free_ char *error_message = NULL; |
ff0af2a1 | 2982 | |
f2341e0a LP |
2983 | r = exec_child(unit, |
2984 | command, | |
ff0af2a1 LP |
2985 | context, |
2986 | params, | |
2987 | runtime, | |
29206d46 | 2988 | dcreds, |
ff0af2a1 LP |
2989 | argv, |
2990 | socket_fd, | |
52c239d7 | 2991 | named_iofds, |
4c47affc FB |
2992 | fds, |
2993 | n_storage_fds, | |
9b141911 | 2994 | n_socket_fds, |
ff0af2a1 | 2995 | files_env, |
00d9ef85 | 2996 | unit->manager->user_lookup_fds[1], |
70dd455c ZJS |
2997 | &exit_status, |
2998 | &error_message); | |
ff0af2a1 | 2999 | if (r < 0) { |
4c2630eb | 3000 | log_open(); |
70dd455c ZJS |
3001 | if (error_message) |
3002 | log_struct_errno(LOG_ERR, r, | |
2b044526 | 3003 | "MESSAGE_ID=" SD_MESSAGE_SPAWN_FAILED_STR, |
70dd455c ZJS |
3004 | LOG_UNIT_ID(unit), |
3005 | LOG_UNIT_MESSAGE(unit, "%s: %m", | |
3006 | error_message), | |
3007 | "EXECUTABLE=%s", command->path, | |
3008 | NULL); | |
4d8b0f0f YW |
3009 | else if (r == -ENOENT && command->ignore) |
3010 | log_struct_errno(LOG_INFO, r, | |
3011 | "MESSAGE_ID=" SD_MESSAGE_SPAWN_FAILED_STR, | |
3012 | LOG_UNIT_ID(unit), | |
3013 | LOG_UNIT_MESSAGE(unit, "Skipped spawning %s: %m", | |
3014 | command->path), | |
3015 | "EXECUTABLE=%s", command->path, | |
3016 | NULL); | |
70dd455c ZJS |
3017 | else |
3018 | log_struct_errno(LOG_ERR, r, | |
2b044526 | 3019 | "MESSAGE_ID=" SD_MESSAGE_SPAWN_FAILED_STR, |
70dd455c ZJS |
3020 | LOG_UNIT_ID(unit), |
3021 | LOG_UNIT_MESSAGE(unit, "Failed at step %s spawning %s: %m", | |
3022 | exit_status_to_string(exit_status, EXIT_STATUS_SYSTEMD), | |
3023 | command->path), | |
3024 | "EXECUTABLE=%s", command->path, | |
3025 | NULL); | |
4c2630eb MS |
3026 | } |
3027 | ||
ff0af2a1 | 3028 | _exit(exit_status); |
034c6ed7 LP |
3029 | } |
3030 | ||
f2341e0a | 3031 | log_unit_debug(unit, "Forked %s as "PID_FMT, command->path, pid); |
23635a85 | 3032 | |
80876c20 LP |
3033 | /* We add the new process to the cgroup both in the child (so |
3034 | * that we can be sure that no user code is ever executed | |
3035 | * outside of the cgroup) and in the parent (so that we can be | |
3036 | * sure that when we kill the cgroup the process will be | |
3037 | * killed too). */ | |
d35fbf6b | 3038 | if (params->cgroup_path) |
dd305ec9 | 3039 | (void) cg_attach(SYSTEMD_CGROUP_CONTROLLER, params->cgroup_path, pid); |
2da3263a | 3040 | |
b58b4116 | 3041 | exec_status_start(&command->exec_status, pid); |
9fb86720 | 3042 | |
034c6ed7 | 3043 | *ret = pid; |
5cb5a6ff LP |
3044 | return 0; |
3045 | } | |
3046 | ||
034c6ed7 LP |
3047 | void exec_context_init(ExecContext *c) { |
3048 | assert(c); | |
3049 | ||
4c12626c | 3050 | c->umask = 0022; |
9eba9da4 | 3051 | c->ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 0); |
94f04347 | 3052 | c->cpu_sched_policy = SCHED_OTHER; |
071830ff | 3053 | c->syslog_priority = LOG_DAEMON|LOG_INFO; |
74922904 | 3054 | c->syslog_level_prefix = true; |
353e12c2 | 3055 | c->ignore_sigpipe = true; |
3a43da28 | 3056 | c->timer_slack_nsec = NSEC_INFINITY; |
050f7277 | 3057 | c->personality = PERSONALITY_INVALID; |
e66cf1a3 | 3058 | c->runtime_directory_mode = 0755; |
a103496c | 3059 | c->capability_bounding_set = CAP_ALL; |
add00535 | 3060 | c->restrict_namespaces = NAMESPACE_FLAGS_ALL; |
034c6ed7 LP |
3061 | } |
3062 | ||
613b411c | 3063 | void exec_context_done(ExecContext *c) { |
5cb5a6ff LP |
3064 | unsigned l; |
3065 | ||
3066 | assert(c); | |
3067 | ||
6796073e LP |
3068 | c->environment = strv_free(c->environment); |
3069 | c->environment_files = strv_free(c->environment_files); | |
b4c14404 | 3070 | c->pass_environment = strv_free(c->pass_environment); |
8c7be95e | 3071 | |
1f6b4113 | 3072 | for (l = 0; l < ELEMENTSOF(c->rlimit); l++) |
a1e58e8e | 3073 | c->rlimit[l] = mfree(c->rlimit[l]); |
034c6ed7 | 3074 | |
52c239d7 LB |
3075 | for (l = 0; l < 3; l++) |
3076 | c->stdio_fdname[l] = mfree(c->stdio_fdname[l]); | |
3077 | ||
a1e58e8e LP |
3078 | c->working_directory = mfree(c->working_directory); |
3079 | c->root_directory = mfree(c->root_directory); | |
915e6d16 | 3080 | c->root_image = mfree(c->root_image); |
a1e58e8e LP |
3081 | c->tty_path = mfree(c->tty_path); |
3082 | c->syslog_identifier = mfree(c->syslog_identifier); | |
3083 | c->user = mfree(c->user); | |
3084 | c->group = mfree(c->group); | |
034c6ed7 | 3085 | |
6796073e | 3086 | c->supplementary_groups = strv_free(c->supplementary_groups); |
94f04347 | 3087 | |
a1e58e8e | 3088 | c->pam_name = mfree(c->pam_name); |
5b6319dc | 3089 | |
2a624c36 AP |
3090 | c->read_only_paths = strv_free(c->read_only_paths); |
3091 | c->read_write_paths = strv_free(c->read_write_paths); | |
3092 | c->inaccessible_paths = strv_free(c->inaccessible_paths); | |
82c121a4 | 3093 | |
d2d6c096 LP |
3094 | bind_mount_free_many(c->bind_mounts, c->n_bind_mounts); |
3095 | ||
82c121a4 LP |
3096 | if (c->cpuset) |
3097 | CPU_FREE(c->cpuset); | |
86a3475b | 3098 | |
a1e58e8e LP |
3099 | c->utmp_id = mfree(c->utmp_id); |
3100 | c->selinux_context = mfree(c->selinux_context); | |
3101 | c->apparmor_profile = mfree(c->apparmor_profile); | |
5b8e1b77 | 3102 | c->smack_process_label = mfree(c->smack_process_label); |
eef65bf3 | 3103 | |
525d3cc7 LP |
3104 | c->syscall_filter = set_free(c->syscall_filter); |
3105 | c->syscall_archs = set_free(c->syscall_archs); | |
3106 | c->address_families = set_free(c->address_families); | |
e66cf1a3 | 3107 | |
6796073e | 3108 | c->runtime_directory = strv_free(c->runtime_directory); |
e66cf1a3 LP |
3109 | } |
3110 | ||
3111 | int exec_context_destroy_runtime_directory(ExecContext *c, const char *runtime_prefix) { | |
3112 | char **i; | |
3113 | ||
3114 | assert(c); | |
3115 | ||
3116 | if (!runtime_prefix) | |
3117 | return 0; | |
3118 | ||
3119 | STRV_FOREACH(i, c->runtime_directory) { | |
3120 | _cleanup_free_ char *p; | |
3121 | ||
605405c6 | 3122 | p = strjoin(runtime_prefix, "/", *i); |
e66cf1a3 LP |
3123 | if (!p) |
3124 | return -ENOMEM; | |
3125 | ||
3126 | /* We execute this synchronously, since we need to be | |
3127 | * sure this is gone when we start the service | |
3128 | * next. */ | |
c6878637 | 3129 | (void) rm_rf(p, REMOVE_ROOT); |
e66cf1a3 LP |
3130 | } |
3131 | ||
3132 | return 0; | |
5cb5a6ff LP |
3133 | } |
3134 | ||
43d0fcbd LP |
3135 | void exec_command_done(ExecCommand *c) { |
3136 | assert(c); | |
3137 | ||
a1e58e8e | 3138 | c->path = mfree(c->path); |
43d0fcbd | 3139 | |
6796073e | 3140 | c->argv = strv_free(c->argv); |
43d0fcbd LP |
3141 | } |
3142 | ||
3143 | void exec_command_done_array(ExecCommand *c, unsigned n) { | |
3144 | unsigned i; | |
3145 | ||
3146 | for (i = 0; i < n; i++) | |
3147 | exec_command_done(c+i); | |
3148 | } | |
3149 | ||
f1acf85a | 3150 | ExecCommand* exec_command_free_list(ExecCommand *c) { |
5cb5a6ff LP |
3151 | ExecCommand *i; |
3152 | ||
3153 | while ((i = c)) { | |
71fda00f | 3154 | LIST_REMOVE(command, c, i); |
43d0fcbd | 3155 | exec_command_done(i); |
5cb5a6ff LP |
3156 | free(i); |
3157 | } | |
f1acf85a ZJS |
3158 | |
3159 | return NULL; | |
5cb5a6ff LP |
3160 | } |
3161 | ||
034c6ed7 LP |
3162 | void exec_command_free_array(ExecCommand **c, unsigned n) { |
3163 | unsigned i; | |
3164 | ||
f1acf85a ZJS |
3165 | for (i = 0; i < n; i++) |
3166 | c[i] = exec_command_free_list(c[i]); | |
034c6ed7 LP |
3167 | } |
3168 | ||
039f0e70 | 3169 | typedef struct InvalidEnvInfo { |
f2341e0a | 3170 | Unit *unit; |
039f0e70 LP |
3171 | const char *path; |
3172 | } InvalidEnvInfo; | |
3173 | ||
3174 | static void invalid_env(const char *p, void *userdata) { | |
3175 | InvalidEnvInfo *info = userdata; | |
3176 | ||
f2341e0a | 3177 | log_unit_error(info->unit, "Ignoring invalid environment assignment '%s': %s", p, info->path); |
039f0e70 LP |
3178 | } |
3179 | ||
52c239d7 LB |
3180 | const char* exec_context_fdname(const ExecContext *c, int fd_index) { |
3181 | assert(c); | |
3182 | ||
3183 | switch (fd_index) { | |
3184 | case STDIN_FILENO: | |
3185 | if (c->std_input != EXEC_INPUT_NAMED_FD) | |
3186 | return NULL; | |
3187 | return c->stdio_fdname[STDIN_FILENO] ?: "stdin"; | |
3188 | case STDOUT_FILENO: | |
3189 | if (c->std_output != EXEC_OUTPUT_NAMED_FD) | |
3190 | return NULL; | |
3191 | return c->stdio_fdname[STDOUT_FILENO] ?: "stdout"; | |
3192 | case STDERR_FILENO: | |
3193 | if (c->std_error != EXEC_OUTPUT_NAMED_FD) | |
3194 | return NULL; | |
3195 | return c->stdio_fdname[STDERR_FILENO] ?: "stderr"; | |
3196 | default: | |
3197 | return NULL; | |
3198 | } | |
3199 | } | |
3200 | ||
3201 | int exec_context_named_iofds(Unit *unit, const ExecContext *c, const ExecParameters *p, int named_iofds[3]) { | |
3202 | unsigned i, targets; | |
56fbd561 | 3203 | const char* stdio_fdname[3]; |
4c47affc | 3204 | unsigned n_fds; |
52c239d7 LB |
3205 | |
3206 | assert(c); | |
3207 | assert(p); | |
3208 | ||
3209 | targets = (c->std_input == EXEC_INPUT_NAMED_FD) + | |
3210 | (c->std_output == EXEC_OUTPUT_NAMED_FD) + | |
3211 | (c->std_error == EXEC_OUTPUT_NAMED_FD); | |
3212 | ||
3213 | for (i = 0; i < 3; i++) | |
3214 | stdio_fdname[i] = exec_context_fdname(c, i); | |
3215 | ||
4c47affc FB |
3216 | n_fds = p->n_storage_fds + p->n_socket_fds; |
3217 | ||
3218 | for (i = 0; i < n_fds && targets > 0; i++) | |
56fbd561 ZJS |
3219 | if (named_iofds[STDIN_FILENO] < 0 && |
3220 | c->std_input == EXEC_INPUT_NAMED_FD && | |
3221 | stdio_fdname[STDIN_FILENO] && | |
3222 | streq(p->fd_names[i], stdio_fdname[STDIN_FILENO])) { | |
3223 | ||
52c239d7 LB |
3224 | named_iofds[STDIN_FILENO] = p->fds[i]; |
3225 | targets--; | |
56fbd561 ZJS |
3226 | |
3227 | } else if (named_iofds[STDOUT_FILENO] < 0 && | |
3228 | c->std_output == EXEC_OUTPUT_NAMED_FD && | |
3229 | stdio_fdname[STDOUT_FILENO] && | |
3230 | streq(p->fd_names[i], stdio_fdname[STDOUT_FILENO])) { | |
3231 | ||
52c239d7 LB |
3232 | named_iofds[STDOUT_FILENO] = p->fds[i]; |
3233 | targets--; | |
56fbd561 ZJS |
3234 | |
3235 | } else if (named_iofds[STDERR_FILENO] < 0 && | |
3236 | c->std_error == EXEC_OUTPUT_NAMED_FD && | |
3237 | stdio_fdname[STDERR_FILENO] && | |
3238 | streq(p->fd_names[i], stdio_fdname[STDERR_FILENO])) { | |
3239 | ||
52c239d7 LB |
3240 | named_iofds[STDERR_FILENO] = p->fds[i]; |
3241 | targets--; | |
3242 | } | |
3243 | ||
56fbd561 | 3244 | return targets == 0 ? 0 : -ENOENT; |
52c239d7 LB |
3245 | } |
3246 | ||
f2341e0a | 3247 | int exec_context_load_environment(Unit *unit, const ExecContext *c, char ***l) { |
8c7be95e LP |
3248 | char **i, **r = NULL; |
3249 | ||
3250 | assert(c); | |
3251 | assert(l); | |
3252 | ||
3253 | STRV_FOREACH(i, c->environment_files) { | |
3254 | char *fn; | |
52511fae ZJS |
3255 | int k; |
3256 | unsigned n; | |
8c7be95e LP |
3257 | bool ignore = false; |
3258 | char **p; | |
7fd1b19b | 3259 | _cleanup_globfree_ glob_t pglob = {}; |
8c7be95e LP |
3260 | |
3261 | fn = *i; | |
3262 | ||
3263 | if (fn[0] == '-') { | |
3264 | ignore = true; | |
313cefa1 | 3265 | fn++; |
8c7be95e LP |
3266 | } |
3267 | ||
3268 | if (!path_is_absolute(fn)) { | |
8c7be95e LP |
3269 | if (ignore) |
3270 | continue; | |
3271 | ||
3272 | strv_free(r); | |
3273 | return -EINVAL; | |
3274 | } | |
3275 | ||
2bef10ab | 3276 | /* Filename supports globbing, take all matching files */ |
d8c92e8b ZJS |
3277 | k = safe_glob(fn, 0, &pglob); |
3278 | if (k < 0) { | |
2bef10ab PL |
3279 | if (ignore) |
3280 | continue; | |
8c7be95e | 3281 | |
2bef10ab | 3282 | strv_free(r); |
d8c92e8b | 3283 | return k; |
2bef10ab | 3284 | } |
8c7be95e | 3285 | |
d8c92e8b ZJS |
3286 | /* When we don't match anything, -ENOENT should be returned */ |
3287 | assert(pglob.gl_pathc > 0); | |
3288 | ||
3289 | for (n = 0; n < pglob.gl_pathc; n++) { | |
717603e3 | 3290 | k = load_env_file(NULL, pglob.gl_pathv[n], NULL, &p); |
2bef10ab PL |
3291 | if (k < 0) { |
3292 | if (ignore) | |
3293 | continue; | |
8c7be95e | 3294 | |
2bef10ab | 3295 | strv_free(r); |
2bef10ab | 3296 | return k; |
e9c1ea9d | 3297 | } |
ebc05a09 | 3298 | /* Log invalid environment variables with filename */ |
039f0e70 LP |
3299 | if (p) { |
3300 | InvalidEnvInfo info = { | |
f2341e0a | 3301 | .unit = unit, |
039f0e70 LP |
3302 | .path = pglob.gl_pathv[n] |
3303 | }; | |
3304 | ||
3305 | p = strv_env_clean_with_callback(p, invalid_env, &info); | |
3306 | } | |
8c7be95e | 3307 | |
2bef10ab PL |
3308 | if (r == NULL) |
3309 | r = p; | |
3310 | else { | |
3311 | char **m; | |
8c7be95e | 3312 | |
2bef10ab PL |
3313 | m = strv_env_merge(2, r, p); |
3314 | strv_free(r); | |
3315 | strv_free(p); | |
c84a9488 | 3316 | if (!m) |
2bef10ab | 3317 | return -ENOMEM; |
2bef10ab PL |
3318 | |
3319 | r = m; | |
3320 | } | |
8c7be95e LP |
3321 | } |
3322 | } | |
3323 | ||
3324 | *l = r; | |
3325 | ||
3326 | return 0; | |
3327 | } | |
3328 | ||
6ac8fdc9 | 3329 | static bool tty_may_match_dev_console(const char *tty) { |
e1d75803 | 3330 | _cleanup_free_ char *active = NULL; |
7d6884b6 | 3331 | char *console; |
6ac8fdc9 | 3332 | |
1e22b5cd LP |
3333 | if (!tty) |
3334 | return true; | |
3335 | ||
6ac8fdc9 MS |
3336 | if (startswith(tty, "/dev/")) |
3337 | tty += 5; | |
3338 | ||
3339 | /* trivial identity? */ | |
3340 | if (streq(tty, "console")) | |
3341 | return true; | |
3342 | ||
3343 | console = resolve_dev_console(&active); | |
3344 | /* if we could not resolve, assume it may */ | |
3345 | if (!console) | |
3346 | return true; | |
3347 | ||
3348 | /* "tty0" means the active VC, so it may be the same sometimes */ | |
e1d75803 | 3349 | return streq(console, tty) || (streq(console, "tty0") && tty_is_vc(tty)); |
6ac8fdc9 MS |
3350 | } |
3351 | ||
3352 | bool exec_context_may_touch_console(ExecContext *ec) { | |
1e22b5cd LP |
3353 | |
3354 | return (ec->tty_reset || | |
3355 | ec->tty_vhangup || | |
3356 | ec->tty_vt_disallocate || | |
6ac8fdc9 MS |
3357 | is_terminal_input(ec->std_input) || |
3358 | is_terminal_output(ec->std_output) || | |
3359 | is_terminal_output(ec->std_error)) && | |
1e22b5cd | 3360 | tty_may_match_dev_console(exec_context_tty_path(ec)); |
6ac8fdc9 MS |
3361 | } |
3362 | ||
15ae422b LP |
3363 | static void strv_fprintf(FILE *f, char **l) { |
3364 | char **g; | |
3365 | ||
3366 | assert(f); | |
3367 | ||
3368 | STRV_FOREACH(g, l) | |
3369 | fprintf(f, " %s", *g); | |
3370 | } | |
3371 | ||
5cb5a6ff | 3372 | void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) { |
c2bbd90b | 3373 | char **e, **d; |
94f04347 | 3374 | unsigned i; |
add00535 | 3375 | int r; |
9eba9da4 | 3376 | |
5cb5a6ff LP |
3377 | assert(c); |
3378 | assert(f); | |
3379 | ||
4ad49000 | 3380 | prefix = strempty(prefix); |
5cb5a6ff LP |
3381 | |
3382 | fprintf(f, | |
94f04347 LP |
3383 | "%sUMask: %04o\n" |
3384 | "%sWorkingDirectory: %s\n" | |
451a074f | 3385 | "%sRootDirectory: %s\n" |
15ae422b | 3386 | "%sNonBlocking: %s\n" |
64747e2d | 3387 | "%sPrivateTmp: %s\n" |
7f112f50 | 3388 | "%sPrivateDevices: %s\n" |
59eeb84b | 3389 | "%sProtectKernelTunables: %s\n" |
e66a2f65 | 3390 | "%sProtectKernelModules: %s\n" |
59eeb84b | 3391 | "%sProtectControlGroups: %s\n" |
d251207d LP |
3392 | "%sPrivateNetwork: %s\n" |
3393 | "%sPrivateUsers: %s\n" | |
1b8689f9 LP |
3394 | "%sProtectHome: %s\n" |
3395 | "%sProtectSystem: %s\n" | |
5d997827 | 3396 | "%sMountAPIVFS: %s\n" |
f3e43635 | 3397 | "%sIgnoreSIGPIPE: %s\n" |
f4170c67 LP |
3398 | "%sMemoryDenyWriteExecute: %s\n" |
3399 | "%sRestrictRealtime: %s\n", | |
5cb5a6ff | 3400 | prefix, c->umask, |
9eba9da4 | 3401 | prefix, c->working_directory ? c->working_directory : "/", |
451a074f | 3402 | prefix, c->root_directory ? c->root_directory : "/", |
15ae422b | 3403 | prefix, yes_no(c->non_blocking), |
64747e2d | 3404 | prefix, yes_no(c->private_tmp), |
7f112f50 | 3405 | prefix, yes_no(c->private_devices), |
59eeb84b | 3406 | prefix, yes_no(c->protect_kernel_tunables), |
e66a2f65 | 3407 | prefix, yes_no(c->protect_kernel_modules), |
59eeb84b | 3408 | prefix, yes_no(c->protect_control_groups), |
d251207d LP |
3409 | prefix, yes_no(c->private_network), |
3410 | prefix, yes_no(c->private_users), | |
1b8689f9 LP |
3411 | prefix, protect_home_to_string(c->protect_home), |
3412 | prefix, protect_system_to_string(c->protect_system), | |
5d997827 | 3413 | prefix, yes_no(c->mount_apivfs), |
f3e43635 | 3414 | prefix, yes_no(c->ignore_sigpipe), |
f4170c67 LP |
3415 | prefix, yes_no(c->memory_deny_write_execute), |
3416 | prefix, yes_no(c->restrict_realtime)); | |
fb33a393 | 3417 | |
915e6d16 LP |
3418 | if (c->root_image) |
3419 | fprintf(f, "%sRootImage: %s\n", prefix, c->root_image); | |
3420 | ||
8c7be95e LP |
3421 | STRV_FOREACH(e, c->environment) |
3422 | fprintf(f, "%sEnvironment: %s\n", prefix, *e); | |
3423 | ||
3424 | STRV_FOREACH(e, c->environment_files) | |
3425 | fprintf(f, "%sEnvironmentFile: %s\n", prefix, *e); | |
94f04347 | 3426 | |
b4c14404 FB |
3427 | STRV_FOREACH(e, c->pass_environment) |
3428 | fprintf(f, "%sPassEnvironment: %s\n", prefix, *e); | |
3429 | ||
c2bbd90b EV |
3430 | fprintf(f, "%sRuntimeDirectoryMode: %04o\n", prefix, c->runtime_directory_mode); |
3431 | ||
3432 | STRV_FOREACH(d, c->runtime_directory) | |
3433 | fprintf(f, "%sRuntimeDirectory: %s\n", prefix, *d); | |
3434 | ||
fb33a393 LP |
3435 | if (c->nice_set) |
3436 | fprintf(f, | |
3437 | "%sNice: %i\n", | |
3438 | prefix, c->nice); | |
3439 | ||
dd6c17b1 | 3440 | if (c->oom_score_adjust_set) |
fb33a393 | 3441 | fprintf(f, |
dd6c17b1 LP |
3442 | "%sOOMScoreAdjust: %i\n", |
3443 | prefix, c->oom_score_adjust); | |
9eba9da4 | 3444 | |
94f04347 | 3445 | for (i = 0; i < RLIM_NLIMITS; i++) |
3c11da9d EV |
3446 | if (c->rlimit[i]) { |
3447 | fprintf(f, "%s%s: " RLIM_FMT "\n", | |
3448 | prefix, rlimit_to_string(i), c->rlimit[i]->rlim_max); | |
3449 | fprintf(f, "%s%sSoft: " RLIM_FMT "\n", | |
3450 | prefix, rlimit_to_string(i), c->rlimit[i]->rlim_cur); | |
3451 | } | |
94f04347 | 3452 | |
f8b69d1d | 3453 | if (c->ioprio_set) { |
1756a011 | 3454 | _cleanup_free_ char *class_str = NULL; |
f8b69d1d | 3455 | |
1756a011 | 3456 | ioprio_class_to_string_alloc(IOPRIO_PRIO_CLASS(c->ioprio), &class_str); |
9eba9da4 LP |
3457 | fprintf(f, |
3458 | "%sIOSchedulingClass: %s\n" | |
3459 | "%sIOPriority: %i\n", | |
f8b69d1d | 3460 | prefix, strna(class_str), |
9eba9da4 | 3461 | prefix, (int) IOPRIO_PRIO_DATA(c->ioprio)); |
f8b69d1d | 3462 | } |
94f04347 | 3463 | |
f8b69d1d | 3464 | if (c->cpu_sched_set) { |
1756a011 | 3465 | _cleanup_free_ char *policy_str = NULL; |
f8b69d1d | 3466 | |
1756a011 | 3467 | sched_policy_to_string_alloc(c->cpu_sched_policy, &policy_str); |
94f04347 LP |
3468 | fprintf(f, |
3469 | "%sCPUSchedulingPolicy: %s\n" | |
38b48754 LP |
3470 | "%sCPUSchedulingPriority: %i\n" |
3471 | "%sCPUSchedulingResetOnFork: %s\n", | |
f8b69d1d | 3472 | prefix, strna(policy_str), |
38b48754 LP |
3473 | prefix, c->cpu_sched_priority, |
3474 | prefix, yes_no(c->cpu_sched_reset_on_fork)); | |
b929bf04 | 3475 | } |
94f04347 | 3476 | |
82c121a4 | 3477 | if (c->cpuset) { |
94f04347 | 3478 | fprintf(f, "%sCPUAffinity:", prefix); |
82c121a4 LP |
3479 | for (i = 0; i < c->cpuset_ncpus; i++) |
3480 | if (CPU_ISSET_S(i, CPU_ALLOC_SIZE(c->cpuset_ncpus), c->cpuset)) | |
43a99a7a | 3481 | fprintf(f, " %u", i); |
94f04347 LP |
3482 | fputs("\n", f); |
3483 | } | |
3484 | ||
3a43da28 | 3485 | if (c->timer_slack_nsec != NSEC_INFINITY) |
ccd06097 | 3486 | fprintf(f, "%sTimerSlackNSec: "NSEC_FMT "\n", prefix, c->timer_slack_nsec); |
94f04347 LP |
3487 | |
3488 | fprintf(f, | |
80876c20 LP |
3489 | "%sStandardInput: %s\n" |
3490 | "%sStandardOutput: %s\n" | |
3491 | "%sStandardError: %s\n", | |
3492 | prefix, exec_input_to_string(c->std_input), | |
3493 | prefix, exec_output_to_string(c->std_output), | |
3494 | prefix, exec_output_to_string(c->std_error)); | |
3495 | ||
3496 | if (c->tty_path) | |
3497 | fprintf(f, | |
6ea832a2 LP |
3498 | "%sTTYPath: %s\n" |
3499 | "%sTTYReset: %s\n" | |
3500 | "%sTTYVHangup: %s\n" | |
3501 | "%sTTYVTDisallocate: %s\n", | |
3502 | prefix, c->tty_path, | |
3503 | prefix, yes_no(c->tty_reset), | |
3504 | prefix, yes_no(c->tty_vhangup), | |
3505 | prefix, yes_no(c->tty_vt_disallocate)); | |
94f04347 | 3506 | |
5ce70e5b ZJS |
3507 | if (c->std_output == EXEC_OUTPUT_SYSLOG || |
3508 | c->std_output == EXEC_OUTPUT_KMSG || | |
3509 | c->std_output == EXEC_OUTPUT_JOURNAL || | |
3510 | c->std_output == EXEC_OUTPUT_SYSLOG_AND_CONSOLE || | |
3511 | c->std_output == EXEC_OUTPUT_KMSG_AND_CONSOLE || | |
3512 | c->std_output == EXEC_OUTPUT_JOURNAL_AND_CONSOLE || | |
3513 | c->std_error == EXEC_OUTPUT_SYSLOG || | |
3514 | c->std_error == EXEC_OUTPUT_KMSG || | |
3515 | c->std_error == EXEC_OUTPUT_JOURNAL || | |
3516 | c->std_error == EXEC_OUTPUT_SYSLOG_AND_CONSOLE || | |
3517 | c->std_error == EXEC_OUTPUT_KMSG_AND_CONSOLE || | |
3518 | c->std_error == EXEC_OUTPUT_JOURNAL_AND_CONSOLE) { | |
f8b69d1d | 3519 | |
5ce70e5b | 3520 | _cleanup_free_ char *fac_str = NULL, *lvl_str = NULL; |
f8b69d1d | 3521 | |
5ce70e5b ZJS |
3522 | log_facility_unshifted_to_string_alloc(c->syslog_priority >> 3, &fac_str); |
3523 | log_level_to_string_alloc(LOG_PRI(c->syslog_priority), &lvl_str); | |
f8b69d1d | 3524 | |
94f04347 LP |
3525 | fprintf(f, |
3526 | "%sSyslogFacility: %s\n" | |
3527 | "%sSyslogLevel: %s\n", | |
f8b69d1d MS |
3528 | prefix, strna(fac_str), |
3529 | prefix, strna(lvl_str)); | |
f8b69d1d | 3530 | } |
94f04347 | 3531 | |
94f04347 LP |
3532 | if (c->secure_bits) |
3533 | fprintf(f, "%sSecure Bits:%s%s%s%s%s%s\n", | |
3534 | prefix, | |
cbb21cca ZJS |
3535 | (c->secure_bits & 1<<SECURE_KEEP_CAPS) ? " keep-caps" : "", |
3536 | (c->secure_bits & 1<<SECURE_KEEP_CAPS_LOCKED) ? " keep-caps-locked" : "", | |
3537 | (c->secure_bits & 1<<SECURE_NO_SETUID_FIXUP) ? " no-setuid-fixup" : "", | |
3538 | (c->secure_bits & 1<<SECURE_NO_SETUID_FIXUP_LOCKED) ? " no-setuid-fixup-locked" : "", | |
3539 | (c->secure_bits & 1<<SECURE_NOROOT) ? " noroot" : "", | |
3540 | (c->secure_bits & 1<<SECURE_NOROOT_LOCKED) ? "noroot-locked" : ""); | |
94f04347 | 3541 | |
a103496c | 3542 | if (c->capability_bounding_set != CAP_ALL) { |
ae556c21 | 3543 | unsigned long l; |
260abb78 | 3544 | fprintf(f, "%sCapabilityBoundingSet:", prefix); |
94f04347 | 3545 | |
64685e0c | 3546 | for (l = 0; l <= cap_last_cap(); l++) |
a103496c | 3547 | if (c->capability_bounding_set & (UINT64_C(1) << l)) |
2822da4f | 3548 | fprintf(f, " %s", strna(capability_to_name(l))); |
94f04347 LP |
3549 | |
3550 | fputs("\n", f); | |
755d4b67 IP |
3551 | } |
3552 | ||
3553 | if (c->capability_ambient_set != 0) { | |
3554 | unsigned long l; | |
3555 | fprintf(f, "%sAmbientCapabilities:", prefix); | |
3556 | ||
3557 | for (l = 0; l <= cap_last_cap(); l++) | |
3558 | if (c->capability_ambient_set & (UINT64_C(1) << l)) | |
3559 | fprintf(f, " %s", strna(capability_to_name(l))); | |
3560 | ||
3561 | fputs("\n", f); | |
94f04347 LP |
3562 | } |
3563 | ||
3564 | if (c->user) | |
f2d3769a | 3565 | fprintf(f, "%sUser: %s\n", prefix, c->user); |
94f04347 | 3566 | if (c->group) |
f2d3769a | 3567 | fprintf(f, "%sGroup: %s\n", prefix, c->group); |
94f04347 | 3568 | |
29206d46 LP |
3569 | fprintf(f, "%sDynamicUser: %s\n", prefix, yes_no(c->dynamic_user)); |
3570 | ||
15ae422b | 3571 | if (strv_length(c->supplementary_groups) > 0) { |
94f04347 | 3572 | fprintf(f, "%sSupplementaryGroups:", prefix); |
15ae422b LP |
3573 | strv_fprintf(f, c->supplementary_groups); |
3574 | fputs("\n", f); | |
3575 | } | |
94f04347 | 3576 | |
5b6319dc | 3577 | if (c->pam_name) |
f2d3769a | 3578 | fprintf(f, "%sPAMName: %s\n", prefix, c->pam_name); |
5b6319dc | 3579 | |
2a624c36 AP |
3580 | if (strv_length(c->read_write_paths) > 0) { |
3581 | fprintf(f, "%sReadWritePaths:", prefix); | |
3582 | strv_fprintf(f, c->read_write_paths); | |
15ae422b LP |
3583 | fputs("\n", f); |
3584 | } | |
3585 | ||
2a624c36 AP |
3586 | if (strv_length(c->read_only_paths) > 0) { |
3587 | fprintf(f, "%sReadOnlyPaths:", prefix); | |
3588 | strv_fprintf(f, c->read_only_paths); | |
15ae422b LP |
3589 | fputs("\n", f); |
3590 | } | |
94f04347 | 3591 | |
2a624c36 AP |
3592 | if (strv_length(c->inaccessible_paths) > 0) { |
3593 | fprintf(f, "%sInaccessiblePaths:", prefix); | |
3594 | strv_fprintf(f, c->inaccessible_paths); | |
94f04347 LP |
3595 | fputs("\n", f); |
3596 | } | |
2e22afe9 | 3597 | |
d2d6c096 LP |
3598 | if (c->n_bind_mounts > 0) |
3599 | for (i = 0; i < c->n_bind_mounts; i++) { | |
3600 | fprintf(f, "%s%s: %s:%s:%s\n", prefix, | |
3601 | c->bind_mounts[i].read_only ? "BindReadOnlyPaths" : "BindPaths", | |
3602 | c->bind_mounts[i].source, | |
3603 | c->bind_mounts[i].destination, | |
3604 | c->bind_mounts[i].recursive ? "rbind" : "norbind"); | |
3605 | } | |
3606 | ||
169c1bda LP |
3607 | if (c->utmp_id) |
3608 | fprintf(f, | |
3609 | "%sUtmpIdentifier: %s\n", | |
3610 | prefix, c->utmp_id); | |
7b52a628 MS |
3611 | |
3612 | if (c->selinux_context) | |
3613 | fprintf(f, | |
5f8640fb LP |
3614 | "%sSELinuxContext: %s%s\n", |
3615 | prefix, c->selinux_context_ignore ? "-" : "", c->selinux_context); | |
17df7223 | 3616 | |
050f7277 | 3617 | if (c->personality != PERSONALITY_INVALID) |
ac45f971 LP |
3618 | fprintf(f, |
3619 | "%sPersonality: %s\n", | |
3620 | prefix, strna(personality_to_string(c->personality))); | |
3621 | ||
17df7223 | 3622 | if (c->syscall_filter) { |
351a19b1 | 3623 | #ifdef HAVE_SECCOMP |
17df7223 LP |
3624 | Iterator j; |
3625 | void *id; | |
3626 | bool first = true; | |
351a19b1 | 3627 | #endif |
17df7223 LP |
3628 | |
3629 | fprintf(f, | |
57183d11 | 3630 | "%sSystemCallFilter: ", |
17df7223 LP |
3631 | prefix); |
3632 | ||
3633 | if (!c->syscall_whitelist) | |
3634 | fputc('~', f); | |
3635 | ||
351a19b1 | 3636 | #ifdef HAVE_SECCOMP |
17df7223 LP |
3637 | SET_FOREACH(id, c->syscall_filter, j) { |
3638 | _cleanup_free_ char *name = NULL; | |
3639 | ||
3640 | if (first) | |
3641 | first = false; | |
3642 | else | |
3643 | fputc(' ', f); | |
3644 | ||
57183d11 | 3645 | name = seccomp_syscall_resolve_num_arch(SCMP_ARCH_NATIVE, PTR_TO_INT(id) - 1); |
17df7223 LP |
3646 | fputs(strna(name), f); |
3647 | } | |
351a19b1 | 3648 | #endif |
17df7223 LP |
3649 | |
3650 | fputc('\n', f); | |
3651 | } | |
3652 | ||
57183d11 LP |
3653 | if (c->syscall_archs) { |
3654 | #ifdef HAVE_SECCOMP | |
3655 | Iterator j; | |
3656 | void *id; | |
3657 | #endif | |
3658 | ||
3659 | fprintf(f, | |
3660 | "%sSystemCallArchitectures:", | |
3661 | prefix); | |
3662 | ||
3663 | #ifdef HAVE_SECCOMP | |
3664 | SET_FOREACH(id, c->syscall_archs, j) | |
3665 | fprintf(f, " %s", strna(seccomp_arch_to_string(PTR_TO_UINT32(id) - 1))); | |
3666 | #endif | |
3667 | fputc('\n', f); | |
3668 | } | |
3669 | ||
add00535 LP |
3670 | if (exec_context_restrict_namespaces_set(c)) { |
3671 | _cleanup_free_ char *s = NULL; | |
3672 | ||
3673 | r = namespace_flag_to_string_many(c->restrict_namespaces, &s); | |
3674 | if (r >= 0) | |
3675 | fprintf(f, "%sRestrictNamespaces: %s\n", | |
3676 | prefix, s); | |
3677 | } | |
3678 | ||
b3267152 | 3679 | if (c->syscall_errno > 0) |
17df7223 LP |
3680 | fprintf(f, |
3681 | "%sSystemCallErrorNumber: %s\n", | |
3682 | prefix, strna(errno_to_name(c->syscall_errno))); | |
eef65bf3 MS |
3683 | |
3684 | if (c->apparmor_profile) | |
3685 | fprintf(f, | |
3686 | "%sAppArmorProfile: %s%s\n", | |
3687 | prefix, c->apparmor_profile_ignore ? "-" : "", c->apparmor_profile); | |
5cb5a6ff LP |
3688 | } |
3689 | ||
a931ad47 LP |
3690 | bool exec_context_maintains_privileges(ExecContext *c) { |
3691 | assert(c); | |
3692 | ||
61233823 | 3693 | /* Returns true if the process forked off would run under |
a931ad47 LP |
3694 | * an unchanged UID or as root. */ |
3695 | ||
3696 | if (!c->user) | |
3697 | return true; | |
3698 | ||
3699 | if (streq(c->user, "root") || streq(c->user, "0")) | |
3700 | return true; | |
3701 | ||
3702 | return false; | |
3703 | } | |
3704 | ||
7f452159 LP |
3705 | int exec_context_get_effective_ioprio(ExecContext *c) { |
3706 | int p; | |
3707 | ||
3708 | assert(c); | |
3709 | ||
3710 | if (c->ioprio_set) | |
3711 | return c->ioprio; | |
3712 | ||
3713 | p = ioprio_get(IOPRIO_WHO_PROCESS, 0); | |
3714 | if (p < 0) | |
3715 | return IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 4); | |
3716 | ||
3717 | return p; | |
3718 | } | |
3719 | ||
b58b4116 | 3720 | void exec_status_start(ExecStatus *s, pid_t pid) { |
034c6ed7 | 3721 | assert(s); |
5cb5a6ff | 3722 | |
b58b4116 LP |
3723 | zero(*s); |
3724 | s->pid = pid; | |
3725 | dual_timestamp_get(&s->start_timestamp); | |
3726 | } | |
3727 | ||
6ea832a2 | 3728 | void exec_status_exit(ExecStatus *s, ExecContext *context, pid_t pid, int code, int status) { |
b58b4116 LP |
3729 | assert(s); |
3730 | ||
0b1f4ae6 | 3731 | if (s->pid && s->pid != pid) |
b58b4116 LP |
3732 | zero(*s); |
3733 | ||
034c6ed7 | 3734 | s->pid = pid; |
63983207 | 3735 | dual_timestamp_get(&s->exit_timestamp); |
9fb86720 | 3736 | |
034c6ed7 LP |
3737 | s->code = code; |
3738 | s->status = status; | |
169c1bda | 3739 | |
6ea832a2 LP |
3740 | if (context) { |
3741 | if (context->utmp_id) | |
3742 | utmp_put_dead_process(context->utmp_id, pid, code, status); | |
3743 | ||
1e22b5cd | 3744 | exec_context_tty_reset(context, NULL); |
6ea832a2 | 3745 | } |
9fb86720 LP |
3746 | } |
3747 | ||
3748 | void exec_status_dump(ExecStatus *s, FILE *f, const char *prefix) { | |
3749 | char buf[FORMAT_TIMESTAMP_MAX]; | |
3750 | ||
3751 | assert(s); | |
3752 | assert(f); | |
3753 | ||
9fb86720 LP |
3754 | if (s->pid <= 0) |
3755 | return; | |
3756 | ||
4c940960 LP |
3757 | prefix = strempty(prefix); |
3758 | ||
9fb86720 | 3759 | fprintf(f, |
ccd06097 ZJS |
3760 | "%sPID: "PID_FMT"\n", |
3761 | prefix, s->pid); | |
9fb86720 | 3762 | |
af9d16e1 | 3763 | if (dual_timestamp_is_set(&s->start_timestamp)) |
9fb86720 LP |
3764 | fprintf(f, |
3765 | "%sStart Timestamp: %s\n", | |
63983207 | 3766 | prefix, format_timestamp(buf, sizeof(buf), s->start_timestamp.realtime)); |
9fb86720 | 3767 | |
af9d16e1 | 3768 | if (dual_timestamp_is_set(&s->exit_timestamp)) |
9fb86720 LP |
3769 | fprintf(f, |
3770 | "%sExit Timestamp: %s\n" | |
3771 | "%sExit Code: %s\n" | |
3772 | "%sExit Status: %i\n", | |
63983207 | 3773 | prefix, format_timestamp(buf, sizeof(buf), s->exit_timestamp.realtime), |
9fb86720 LP |
3774 | prefix, sigchld_code_to_string(s->code), |
3775 | prefix, s->status); | |
5cb5a6ff | 3776 | } |
44d8db9e | 3777 | |
9e2f7c11 | 3778 | char *exec_command_line(char **argv) { |
44d8db9e LP |
3779 | size_t k; |
3780 | char *n, *p, **a; | |
3781 | bool first = true; | |
3782 | ||
9e2f7c11 | 3783 | assert(argv); |
44d8db9e | 3784 | |
9164977d | 3785 | k = 1; |
9e2f7c11 | 3786 | STRV_FOREACH(a, argv) |
44d8db9e LP |
3787 | k += strlen(*a)+3; |
3788 | ||
5cd9cd35 LP |
3789 | n = new(char, k); |
3790 | if (!n) | |
44d8db9e LP |
3791 | return NULL; |
3792 | ||
3793 | p = n; | |
9e2f7c11 | 3794 | STRV_FOREACH(a, argv) { |
44d8db9e LP |
3795 | |
3796 | if (!first) | |
3797 | *(p++) = ' '; | |
3798 | else | |
3799 | first = false; | |
3800 | ||
3801 | if (strpbrk(*a, WHITESPACE)) { | |
3802 | *(p++) = '\''; | |
3803 | p = stpcpy(p, *a); | |
3804 | *(p++) = '\''; | |
3805 | } else | |
3806 | p = stpcpy(p, *a); | |
3807 | ||
3808 | } | |
3809 | ||
9164977d LP |
3810 | *p = 0; |
3811 | ||
44d8db9e LP |
3812 | /* FIXME: this doesn't really handle arguments that have |
3813 | * spaces and ticks in them */ | |
3814 | ||
3815 | return n; | |
3816 | } | |
3817 | ||
3818 | void exec_command_dump(ExecCommand *c, FILE *f, const char *prefix) { | |
e1d75803 | 3819 | _cleanup_free_ char *cmd = NULL; |
4c940960 | 3820 | const char *prefix2; |
44d8db9e LP |
3821 | |
3822 | assert(c); | |
3823 | assert(f); | |
3824 | ||
4c940960 | 3825 | prefix = strempty(prefix); |
63c372cb | 3826 | prefix2 = strjoina(prefix, "\t"); |
44d8db9e | 3827 | |
9e2f7c11 | 3828 | cmd = exec_command_line(c->argv); |
44d8db9e LP |
3829 | fprintf(f, |
3830 | "%sCommand Line: %s\n", | |
3831 | prefix, cmd ? cmd : strerror(ENOMEM)); | |
3832 | ||
9fb86720 | 3833 | exec_status_dump(&c->exec_status, f, prefix2); |
44d8db9e LP |
3834 | } |
3835 | ||
3836 | void exec_command_dump_list(ExecCommand *c, FILE *f, const char *prefix) { | |
3837 | assert(f); | |
3838 | ||
4c940960 | 3839 | prefix = strempty(prefix); |
44d8db9e LP |
3840 | |
3841 | LIST_FOREACH(command, c, c) | |
3842 | exec_command_dump(c, f, prefix); | |
3843 | } | |
94f04347 | 3844 | |
a6a80b4f LP |
3845 | void exec_command_append_list(ExecCommand **l, ExecCommand *e) { |
3846 | ExecCommand *end; | |
3847 | ||
3848 | assert(l); | |
3849 | assert(e); | |
3850 | ||
3851 | if (*l) { | |
35b8ca3a | 3852 | /* It's kind of important, that we keep the order here */ |
71fda00f LP |
3853 | LIST_FIND_TAIL(command, *l, end); |
3854 | LIST_INSERT_AFTER(command, *l, end, e); | |
a6a80b4f LP |
3855 | } else |
3856 | *l = e; | |
3857 | } | |
3858 | ||
26fd040d LP |
3859 | int exec_command_set(ExecCommand *c, const char *path, ...) { |
3860 | va_list ap; | |
3861 | char **l, *p; | |
3862 | ||
3863 | assert(c); | |
3864 | assert(path); | |
3865 | ||
3866 | va_start(ap, path); | |
3867 | l = strv_new_ap(path, ap); | |
3868 | va_end(ap); | |
3869 | ||
3870 | if (!l) | |
3871 | return -ENOMEM; | |
3872 | ||
250a918d LP |
3873 | p = strdup(path); |
3874 | if (!p) { | |
26fd040d LP |
3875 | strv_free(l); |
3876 | return -ENOMEM; | |
3877 | } | |
3878 | ||
3879 | free(c->path); | |
3880 | c->path = p; | |
3881 | ||
3882 | strv_free(c->argv); | |
3883 | c->argv = l; | |
3884 | ||
3885 | return 0; | |
3886 | } | |
3887 | ||
86b23b07 | 3888 | int exec_command_append(ExecCommand *c, const char *path, ...) { |
e63ff941 | 3889 | _cleanup_strv_free_ char **l = NULL; |
86b23b07 | 3890 | va_list ap; |
86b23b07 JS |
3891 | int r; |
3892 | ||
3893 | assert(c); | |
3894 | assert(path); | |
3895 | ||
3896 | va_start(ap, path); | |
3897 | l = strv_new_ap(path, ap); | |
3898 | va_end(ap); | |
3899 | ||
3900 | if (!l) | |
3901 | return -ENOMEM; | |
3902 | ||
e287086b | 3903 | r = strv_extend_strv(&c->argv, l, false); |
e63ff941 | 3904 | if (r < 0) |
86b23b07 | 3905 | return r; |
86b23b07 JS |
3906 | |
3907 | return 0; | |
3908 | } | |
3909 | ||
3910 | ||
613b411c LP |
3911 | static int exec_runtime_allocate(ExecRuntime **rt) { |
3912 | ||
3913 | if (*rt) | |
3914 | return 0; | |
3915 | ||
3916 | *rt = new0(ExecRuntime, 1); | |
f146f5e1 | 3917 | if (!*rt) |
613b411c LP |
3918 | return -ENOMEM; |
3919 | ||
3920 | (*rt)->n_ref = 1; | |
3921 | (*rt)->netns_storage_socket[0] = (*rt)->netns_storage_socket[1] = -1; | |
3922 | ||
3923 | return 0; | |
3924 | } | |
3925 | ||
3926 | int exec_runtime_make(ExecRuntime **rt, ExecContext *c, const char *id) { | |
3927 | int r; | |
3928 | ||
3929 | assert(rt); | |
3930 | assert(c); | |
3931 | assert(id); | |
3932 | ||
3933 | if (*rt) | |
3934 | return 1; | |
3935 | ||
3936 | if (!c->private_network && !c->private_tmp) | |
3937 | return 0; | |
3938 | ||
3939 | r = exec_runtime_allocate(rt); | |
3940 | if (r < 0) | |
3941 | return r; | |
3942 | ||
3943 | if (c->private_network && (*rt)->netns_storage_socket[0] < 0) { | |
33df919d | 3944 | if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, (*rt)->netns_storage_socket) < 0) |
613b411c LP |
3945 | return -errno; |
3946 | } | |
3947 | ||
3948 | if (c->private_tmp && !(*rt)->tmp_dir) { | |
3949 | r = setup_tmp_dirs(id, &(*rt)->tmp_dir, &(*rt)->var_tmp_dir); | |
3950 | if (r < 0) | |
3951 | return r; | |
3952 | } | |
3953 | ||
3954 | return 1; | |
3955 | } | |
3956 | ||
3957 | ExecRuntime *exec_runtime_ref(ExecRuntime *r) { | |
3958 | assert(r); | |
3959 | assert(r->n_ref > 0); | |
3960 | ||
3961 | r->n_ref++; | |
3962 | return r; | |
3963 | } | |
3964 | ||
3965 | ExecRuntime *exec_runtime_unref(ExecRuntime *r) { | |
3966 | ||
3967 | if (!r) | |
3968 | return NULL; | |
3969 | ||
3970 | assert(r->n_ref > 0); | |
3971 | ||
3972 | r->n_ref--; | |
f2341e0a LP |
3973 | if (r->n_ref > 0) |
3974 | return NULL; | |
3975 | ||
3976 | free(r->tmp_dir); | |
3977 | free(r->var_tmp_dir); | |
3978 | safe_close_pair(r->netns_storage_socket); | |
6b430fdb | 3979 | return mfree(r); |
613b411c LP |
3980 | } |
3981 | ||
f2341e0a | 3982 | int exec_runtime_serialize(Unit *u, ExecRuntime *rt, FILE *f, FDSet *fds) { |
613b411c LP |
3983 | assert(u); |
3984 | assert(f); | |
3985 | assert(fds); | |
3986 | ||
3987 | if (!rt) | |
3988 | return 0; | |
3989 | ||
3990 | if (rt->tmp_dir) | |
3991 | unit_serialize_item(u, f, "tmp-dir", rt->tmp_dir); | |
3992 | ||
3993 | if (rt->var_tmp_dir) | |
3994 | unit_serialize_item(u, f, "var-tmp-dir", rt->var_tmp_dir); | |
3995 | ||
3996 | if (rt->netns_storage_socket[0] >= 0) { | |
3997 | int copy; | |
3998 | ||
3999 | copy = fdset_put_dup(fds, rt->netns_storage_socket[0]); | |
4000 | if (copy < 0) | |
4001 | return copy; | |
4002 | ||
4003 | unit_serialize_item_format(u, f, "netns-socket-0", "%i", copy); | |
4004 | } | |
4005 | ||
4006 | if (rt->netns_storage_socket[1] >= 0) { | |
4007 | int copy; | |
4008 | ||
4009 | copy = fdset_put_dup(fds, rt->netns_storage_socket[1]); | |
4010 | if (copy < 0) | |
4011 | return copy; | |
4012 | ||
4013 | unit_serialize_item_format(u, f, "netns-socket-1", "%i", copy); | |
4014 | } | |
4015 | ||
4016 | return 0; | |
4017 | } | |
4018 | ||
f2341e0a | 4019 | int exec_runtime_deserialize_item(Unit *u, ExecRuntime **rt, const char *key, const char *value, FDSet *fds) { |
613b411c LP |
4020 | int r; |
4021 | ||
4022 | assert(rt); | |
4023 | assert(key); | |
4024 | assert(value); | |
4025 | ||
4026 | if (streq(key, "tmp-dir")) { | |
4027 | char *copy; | |
4028 | ||
4029 | r = exec_runtime_allocate(rt); | |
4030 | if (r < 0) | |
f2341e0a | 4031 | return log_oom(); |
613b411c LP |
4032 | |
4033 | copy = strdup(value); | |
4034 | if (!copy) | |
4035 | return log_oom(); | |
4036 | ||
4037 | free((*rt)->tmp_dir); | |
4038 | (*rt)->tmp_dir = copy; | |
4039 | ||
4040 | } else if (streq(key, "var-tmp-dir")) { | |
4041 | char *copy; | |
4042 | ||
4043 | r = exec_runtime_allocate(rt); | |
4044 | if (r < 0) | |
f2341e0a | 4045 | return log_oom(); |
613b411c LP |
4046 | |
4047 | copy = strdup(value); | |
4048 | if (!copy) | |
4049 | return log_oom(); | |
4050 | ||
4051 | free((*rt)->var_tmp_dir); | |
4052 | (*rt)->var_tmp_dir = copy; | |
4053 | ||
4054 | } else if (streq(key, "netns-socket-0")) { | |
4055 | int fd; | |
4056 | ||
4057 | r = exec_runtime_allocate(rt); | |
4058 | if (r < 0) | |
f2341e0a | 4059 | return log_oom(); |
613b411c LP |
4060 | |
4061 | if (safe_atoi(value, &fd) < 0 || !fdset_contains(fds, fd)) | |
f2341e0a | 4062 | log_unit_debug(u, "Failed to parse netns socket value: %s", value); |
613b411c | 4063 | else { |
03e334a1 | 4064 | safe_close((*rt)->netns_storage_socket[0]); |
613b411c LP |
4065 | (*rt)->netns_storage_socket[0] = fdset_remove(fds, fd); |
4066 | } | |
4067 | } else if (streq(key, "netns-socket-1")) { | |
4068 | int fd; | |
4069 | ||
4070 | r = exec_runtime_allocate(rt); | |
4071 | if (r < 0) | |
f2341e0a | 4072 | return log_oom(); |
613b411c LP |
4073 | |
4074 | if (safe_atoi(value, &fd) < 0 || !fdset_contains(fds, fd)) | |
f2341e0a | 4075 | log_unit_debug(u, "Failed to parse netns socket value: %s", value); |
613b411c | 4076 | else { |
03e334a1 | 4077 | safe_close((*rt)->netns_storage_socket[1]); |
613b411c LP |
4078 | (*rt)->netns_storage_socket[1] = fdset_remove(fds, fd); |
4079 | } | |
4080 | } else | |
4081 | return 0; | |
4082 | ||
4083 | return 1; | |
4084 | } | |
4085 | ||
4086 | static void *remove_tmpdir_thread(void *p) { | |
4087 | _cleanup_free_ char *path = p; | |
4088 | ||
c6878637 | 4089 | (void) rm_rf(path, REMOVE_ROOT|REMOVE_PHYSICAL); |
613b411c LP |
4090 | return NULL; |
4091 | } | |
4092 | ||
4093 | void exec_runtime_destroy(ExecRuntime *rt) { | |
98b47d54 LP |
4094 | int r; |
4095 | ||
613b411c LP |
4096 | if (!rt) |
4097 | return; | |
4098 | ||
4099 | /* If there are multiple users of this, let's leave the stuff around */ | |
4100 | if (rt->n_ref > 1) | |
4101 | return; | |
4102 | ||
4103 | if (rt->tmp_dir) { | |
4104 | log_debug("Spawning thread to nuke %s", rt->tmp_dir); | |
98b47d54 LP |
4105 | |
4106 | r = asynchronous_job(remove_tmpdir_thread, rt->tmp_dir); | |
4107 | if (r < 0) { | |
da927ba9 | 4108 | log_warning_errno(r, "Failed to nuke %s: %m", rt->tmp_dir); |
98b47d54 LP |
4109 | free(rt->tmp_dir); |
4110 | } | |
4111 | ||
613b411c LP |
4112 | rt->tmp_dir = NULL; |
4113 | } | |
4114 | ||
4115 | if (rt->var_tmp_dir) { | |
4116 | log_debug("Spawning thread to nuke %s", rt->var_tmp_dir); | |
98b47d54 LP |
4117 | |
4118 | r = asynchronous_job(remove_tmpdir_thread, rt->var_tmp_dir); | |
4119 | if (r < 0) { | |
da927ba9 | 4120 | log_warning_errno(r, "Failed to nuke %s: %m", rt->var_tmp_dir); |
98b47d54 LP |
4121 | free(rt->var_tmp_dir); |
4122 | } | |
4123 | ||
613b411c LP |
4124 | rt->var_tmp_dir = NULL; |
4125 | } | |
4126 | ||
3d94f76c | 4127 | safe_close_pair(rt->netns_storage_socket); |
613b411c LP |
4128 | } |
4129 | ||
80876c20 LP |
4130 | static const char* const exec_input_table[_EXEC_INPUT_MAX] = { |
4131 | [EXEC_INPUT_NULL] = "null", | |
4132 | [EXEC_INPUT_TTY] = "tty", | |
4133 | [EXEC_INPUT_TTY_FORCE] = "tty-force", | |
4f2d528d | 4134 | [EXEC_INPUT_TTY_FAIL] = "tty-fail", |
52c239d7 LB |
4135 | [EXEC_INPUT_SOCKET] = "socket", |
4136 | [EXEC_INPUT_NAMED_FD] = "fd", | |
80876c20 LP |
4137 | }; |
4138 | ||
8a0867d6 LP |
4139 | DEFINE_STRING_TABLE_LOOKUP(exec_input, ExecInput); |
4140 | ||
94f04347 | 4141 | static const char* const exec_output_table[_EXEC_OUTPUT_MAX] = { |
80876c20 | 4142 | [EXEC_OUTPUT_INHERIT] = "inherit", |
94f04347 | 4143 | [EXEC_OUTPUT_NULL] = "null", |
80876c20 | 4144 | [EXEC_OUTPUT_TTY] = "tty", |
94f04347 | 4145 | [EXEC_OUTPUT_SYSLOG] = "syslog", |
28dbc1e8 | 4146 | [EXEC_OUTPUT_SYSLOG_AND_CONSOLE] = "syslog+console", |
9a6bca7a | 4147 | [EXEC_OUTPUT_KMSG] = "kmsg", |
28dbc1e8 | 4148 | [EXEC_OUTPUT_KMSG_AND_CONSOLE] = "kmsg+console", |
706343f4 LP |
4149 | [EXEC_OUTPUT_JOURNAL] = "journal", |
4150 | [EXEC_OUTPUT_JOURNAL_AND_CONSOLE] = "journal+console", | |
52c239d7 LB |
4151 | [EXEC_OUTPUT_SOCKET] = "socket", |
4152 | [EXEC_OUTPUT_NAMED_FD] = "fd", | |
94f04347 LP |
4153 | }; |
4154 | ||
4155 | DEFINE_STRING_TABLE_LOOKUP(exec_output, ExecOutput); | |
023a4f67 LP |
4156 | |
4157 | static const char* const exec_utmp_mode_table[_EXEC_UTMP_MODE_MAX] = { | |
4158 | [EXEC_UTMP_INIT] = "init", | |
4159 | [EXEC_UTMP_LOGIN] = "login", | |
4160 | [EXEC_UTMP_USER] = "user", | |
4161 | }; | |
4162 | ||
4163 | DEFINE_STRING_TABLE_LOOKUP(exec_utmp_mode, ExecUtmpMode); |