]>
Commit | Line | Data |
---|---|---|
758ad80c SH |
1 | /* lxcfs |
2 | * | |
b11c6ec0 | 3 | * Copyright © 2014-2016 Canonical, Inc |
758ad80c SH |
4 | * Author: Serge Hallyn <serge.hallyn@ubuntu.com> |
5 | * | |
f2799430 | 6 | * See COPYING file for details. |
758ad80c SH |
7 | */ |
8 | ||
758ad80c SH |
9 | #define FUSE_USE_VERSION 26 |
10 | ||
2183082c | 11 | #include <stdio.h> |
758ad80c SH |
12 | #include <dirent.h> |
13 | #include <fcntl.h> | |
14 | #include <fuse.h> | |
15 | #include <unistd.h> | |
16 | #include <errno.h> | |
17 | #include <stdbool.h> | |
18 | #include <time.h> | |
19 | #include <string.h> | |
20 | #include <stdlib.h> | |
21 | #include <libgen.h> | |
41bb9357 | 22 | #include <sched.h> |
b11c6ec0 | 23 | #include <pthread.h> |
41bb9357 | 24 | #include <linux/sched.h> |
a05660a6 | 25 | #include <sys/socket.h> |
41bb9357 | 26 | #include <sys/mount.h> |
5b2dfd85 | 27 | #include <sys/epoll.h> |
41bb9357 | 28 | #include <wait.h> |
758ad80c | 29 | |
977ac879 | 30 | #ifdef FORTRAVIS |
df062bcb SH |
31 | #define GLIB_DISABLE_DEPRECATION_WARNINGS |
32 | #include <glib-object.h> | |
977ac879 | 33 | #endif |
df062bcb | 34 | |
35482f91 | 35 | #include "cgfs.h" |
2e9c0b32 | 36 | #include "config.h" // for VERSION |
758ad80c | 37 | |
443d13f5 SH |
38 | enum { |
39 | LXC_TYPE_CGDIR, | |
40 | LXC_TYPE_CGFILE, | |
41 | LXC_TYPE_PROC_MEMINFO, | |
42 | LXC_TYPE_PROC_CPUINFO, | |
43 | LXC_TYPE_PROC_UPTIME, | |
44 | LXC_TYPE_PROC_STAT, | |
45 | LXC_TYPE_PROC_DISKSTATS, | |
46 | }; | |
47 | ||
c688e1b3 SH |
48 | struct file_info { |
49 | char *controller; | |
50 | char *cgroup; | |
8f6e8f5e | 51 | char *file; |
443d13f5 | 52 | int type; |
c688e1b3 SH |
53 | char *buf; // unused as of yet |
54 | int buflen; | |
97f1f27b | 55 | int size; //actual data size |
b5ad2d21 | 56 | int cached; |
c688e1b3 SH |
57 | }; |
58 | ||
97f1f27b YY |
59 | /* reserve buffer size, for cpuall in /proc/stat */ |
60 | #define BUF_RESERVE_SIZE 256 | |
61 | ||
2c51f8dd | 62 | /* |
b11c6ec0 SH |
63 | * A table caching which pid is init for a pid namespace. |
64 | * When looking up which pid is init for $qpid, we first | |
65 | * 1. Stat /proc/$qpid/ns/pid. | |
66 | * 2. Check whether the ino_t is in our store. | |
67 | * a. if not, fork a child in qpid's ns to send us | |
68 | * ucred.pid = 1, and read the initpid. Cache | |
69 | * initpid and creation time for /proc/initpid | |
70 | * in a new store entry. | |
71 | * b. if so, verify that /proc/initpid still matches | |
72 | * what we have saved. If not, clear the store | |
73 | * entry and go back to a. If so, return the | |
74 | * cached initpid. | |
2c51f8dd | 75 | */ |
b11c6ec0 SH |
76 | struct pidns_init_store { |
77 | ino_t ino; // inode number for /proc/$pid/ns/pid | |
78 | pid_t initpid; // the pid of nit in that ns | |
79 | long int ctime; // the time at which /proc/$initpid was created | |
80 | struct pidns_init_store *next; | |
8e547050 | 81 | long int lastcheck; |
b11c6ec0 SH |
82 | }; |
83 | ||
6d1308cb SH |
84 | /* lol - look at how they are allocated in the kernel */ |
85 | #define PIDNS_HASH_SIZE 4096 | |
a6a559b6 | 86 | #define HASH(x) ((x) % PIDNS_HASH_SIZE) |
6d1308cb SH |
87 | |
88 | struct pidns_init_store *pidns_hash_table[PIDNS_HASH_SIZE]; | |
b11c6ec0 SH |
89 | static pthread_mutex_t pidns_store_mutex = PTHREAD_MUTEX_INITIALIZER; |
90 | static void lock_mutex(pthread_mutex_t *l) | |
2c51f8dd | 91 | { |
b11c6ec0 | 92 | int ret; |
2c51f8dd | 93 | |
b11c6ec0 SH |
94 | if ((ret = pthread_mutex_lock(l)) != 0) { |
95 | fprintf(stderr, "pthread_mutex_lock returned:%d %s\n", ret, strerror(ret)); | |
96 | exit(1); | |
97 | } | |
98 | } | |
2c51f8dd | 99 | |
b11c6ec0 SH |
100 | static void unlock_mutex(pthread_mutex_t *l) |
101 | { | |
102 | int ret; | |
103 | ||
104 | if ((ret = pthread_mutex_unlock(l)) != 0) { | |
105 | fprintf(stderr, "pthread_mutex_unlock returned:%d %s\n", ret, strerror(ret)); | |
106 | exit(1); | |
2c51f8dd | 107 | } |
2c51f8dd SH |
108 | } |
109 | ||
b11c6ec0 SH |
110 | static void store_lock(void) |
111 | { | |
112 | lock_mutex(&pidns_store_mutex); | |
113 | } | |
114 | ||
115 | static void store_unlock(void) | |
116 | { | |
117 | unlock_mutex(&pidns_store_mutex); | |
118 | } | |
119 | ||
120 | /* Must be called under store_lock */ | |
121 | static bool initpid_still_valid(struct pidns_init_store *e, struct stat *nsfdsb) | |
122 | { | |
123 | struct stat initsb; | |
124 | char fnam[100]; | |
125 | ||
126 | snprintf(fnam, 100, "/proc/%d", e->initpid); | |
127 | if (stat(fnam, &initsb) < 0) | |
128 | return false; | |
129 | #if DEBUG | |
130 | fprintf(stderr, "comparing ctime %ld %ld for pid %d\n", | |
131 | e->ctime, initsb.st_ctime, e->initpid); | |
132 | #endif | |
133 | if (e->ctime != initsb.st_ctime) | |
134 | return false; | |
135 | return true; | |
136 | } | |
137 | ||
138 | /* Must be called under store_lock */ | |
139 | static void remove_initpid(struct pidns_init_store *e) | |
140 | { | |
141 | struct pidns_init_store *tmp; | |
6d1308cb | 142 | int h; |
b11c6ec0 SH |
143 | |
144 | #if DEBUG | |
145 | fprintf(stderr, "remove_initpid: removing entry for %d\n", e->initpid); | |
146 | #endif | |
6d1308cb SH |
147 | h = HASH(e->ino); |
148 | if (pidns_hash_table[h] == e) { | |
149 | pidns_hash_table[h] = e->next; | |
b11c6ec0 SH |
150 | free(e); |
151 | return; | |
152 | } | |
153 | ||
6d1308cb | 154 | tmp = pidns_hash_table[h]; |
b11c6ec0 SH |
155 | while (tmp) { |
156 | if (tmp->next == e) { | |
157 | tmp->next = e->next; | |
158 | free(e); | |
159 | return; | |
160 | } | |
161 | tmp = tmp->next; | |
162 | } | |
163 | } | |
164 | ||
8e547050 SH |
165 | #define PURGE_SECS 5 |
166 | /* Must be called under store_lock */ | |
167 | static void prune_initpid_store(void) | |
168 | { | |
169 | static long int last_prune = 0; | |
170 | struct pidns_init_store *e, *prev, *delme; | |
171 | long int now, threshold; | |
6d1308cb | 172 | int i; |
8e547050 SH |
173 | |
174 | if (!last_prune) { | |
175 | last_prune = time(NULL); | |
176 | return; | |
177 | } | |
178 | now = time(NULL); | |
179 | if (now < last_prune + PURGE_SECS) | |
180 | return; | |
181 | #if DEBUG | |
182 | fprintf(stderr, "pruning\n"); | |
183 | #endif | |
184 | last_prune = now; | |
185 | threshold = now - 2 * PURGE_SECS; | |
186 | ||
6d1308cb SH |
187 | for (i = 0; i < PIDNS_HASH_SIZE; i++) { |
188 | for (prev = NULL, e = pidns_hash_table[i]; e; ) { | |
189 | if (e->lastcheck < threshold) { | |
8e547050 | 190 | #if DEBUG |
6d1308cb | 191 | fprintf(stderr, "Removing cached entry for %d\n", e->initpid); |
8e547050 | 192 | #endif |
6d1308cb SH |
193 | delme = e; |
194 | if (prev) | |
195 | prev->next = e->next; | |
196 | else | |
197 | pidns_hash_table[i] = e->next; | |
198 | e = e->next; | |
199 | free(delme); | |
200 | } else { | |
201 | prev = e; | |
202 | e = e->next; | |
203 | } | |
8e547050 SH |
204 | } |
205 | } | |
206 | } | |
207 | ||
b11c6ec0 SH |
208 | /* Must be called under store_lock */ |
209 | static void save_initpid(struct stat *sb, pid_t pid) | |
210 | { | |
211 | struct pidns_init_store *e; | |
212 | char fpath[100]; | |
213 | struct stat procsb; | |
6d1308cb | 214 | int h; |
b11c6ec0 SH |
215 | |
216 | #if DEBUG | |
217 | fprintf(stderr, "save_initpid: adding entry for %d\n", pid); | |
218 | #endif | |
219 | snprintf(fpath, 100, "/proc/%d", pid); | |
220 | if (stat(fpath, &procsb) < 0) | |
221 | return; | |
222 | do { | |
223 | e = malloc(sizeof(*e)); | |
224 | } while (!e); | |
225 | e->ino = sb->st_ino; | |
226 | e->initpid = pid; | |
227 | e->ctime = procsb.st_ctime; | |
6d1308cb SH |
228 | h = HASH(e->ino); |
229 | e->next = pidns_hash_table[h]; | |
8e547050 | 230 | e->lastcheck = time(NULL); |
6d1308cb | 231 | pidns_hash_table[h] = e; |
b11c6ec0 SH |
232 | } |
233 | ||
234 | /* | |
235 | * Given the stat(2) info for a nsfd pid inode, lookup the init_pid_store | |
236 | * entry for the inode number and creation time. Verify that the init pid | |
237 | * is still valid. If not, remove it. Return the entry if valid, NULL | |
238 | * otherwise. | |
239 | * Must be called under store_lock | |
240 | */ | |
241 | static struct pidns_init_store *lookup_verify_initpid(struct stat *sb) | |
242 | { | |
6d1308cb SH |
243 | int h = HASH(sb->st_ino); |
244 | struct pidns_init_store *e = pidns_hash_table[h]; | |
245 | ||
b11c6ec0 SH |
246 | while (e) { |
247 | if (e->ino == sb->st_ino) { | |
8e547050 SH |
248 | if (initpid_still_valid(e, sb)) { |
249 | e->lastcheck = time(NULL); | |
b11c6ec0 | 250 | return e; |
8e547050 | 251 | } |
b11c6ec0 SH |
252 | remove_initpid(e); |
253 | return NULL; | |
254 | } | |
255 | e = e->next; | |
256 | } | |
257 | ||
258 | return NULL; | |
259 | } | |
260 | ||
261 | #define SEND_CREDS_OK 0 | |
262 | #define SEND_CREDS_NOTSK 1 | |
263 | #define SEND_CREDS_FAIL 2 | |
264 | static bool recv_creds(int sock, struct ucred *cred, char *v); | |
265 | static int wait_for_pid(pid_t pid); | |
266 | static int send_creds(int sock, struct ucred *cred, char v, bool pingfirst); | |
267 | ||
268 | /* | |
269 | * fork a task which switches to @task's namespace and writes '1'. | |
270 | * over a unix sock so we can read the task's reaper's pid in our | |
271 | * namespace | |
272 | */ | |
273 | static void write_task_init_pid_exit(int sock, pid_t target) | |
274 | { | |
275 | struct ucred cred; | |
276 | char fnam[100]; | |
277 | pid_t pid; | |
278 | char v; | |
279 | int fd, ret; | |
280 | ||
281 | ret = snprintf(fnam, sizeof(fnam), "/proc/%d/ns/pid", (int)target); | |
282 | if (ret < 0 || ret >= sizeof(fnam)) | |
283 | _exit(1); | |
284 | ||
285 | fd = open(fnam, O_RDONLY); | |
286 | if (fd < 0) { | |
287 | perror("write_task_init_pid_exit open of ns/pid"); | |
288 | _exit(1); | |
289 | } | |
290 | if (setns(fd, 0)) { | |
291 | perror("write_task_init_pid_exit setns 1"); | |
292 | close(fd); | |
293 | _exit(1); | |
294 | } | |
295 | pid = fork(); | |
296 | if (pid < 0) | |
297 | _exit(1); | |
298 | if (pid != 0) { | |
299 | if (!wait_for_pid(pid)) | |
300 | _exit(1); | |
301 | _exit(0); | |
302 | } | |
303 | ||
304 | /* we are the child */ | |
305 | cred.uid = 0; | |
306 | cred.gid = 0; | |
307 | cred.pid = 1; | |
308 | v = '1'; | |
309 | if (send_creds(sock, &cred, v, true) != SEND_CREDS_OK) | |
310 | _exit(1); | |
311 | _exit(0); | |
312 | } | |
313 | ||
314 | static pid_t get_init_pid_for_task(pid_t task) | |
315 | { | |
316 | int sock[2]; | |
317 | pid_t pid; | |
318 | pid_t ret = -1; | |
319 | char v = '0'; | |
320 | struct ucred cred; | |
321 | ||
322 | if (socketpair(AF_UNIX, SOCK_DGRAM, 0, sock) < 0) { | |
323 | perror("socketpair"); | |
324 | return -1; | |
325 | } | |
326 | ||
327 | pid = fork(); | |
328 | if (pid < 0) | |
329 | goto out; | |
330 | if (!pid) { | |
331 | close(sock[1]); | |
332 | write_task_init_pid_exit(sock[0], task); | |
333 | _exit(0); | |
334 | } | |
335 | ||
336 | if (!recv_creds(sock[1], &cred, &v)) | |
337 | goto out; | |
338 | ret = cred.pid; | |
339 | ||
340 | out: | |
341 | close(sock[0]); | |
342 | close(sock[1]); | |
343 | if (pid > 0) | |
344 | wait_for_pid(pid); | |
345 | return ret; | |
346 | } | |
347 | ||
348 | static pid_t lookup_initpid_in_store(pid_t qpid) | |
349 | { | |
350 | pid_t answer = 0; | |
351 | struct stat sb; | |
352 | struct pidns_init_store *e; | |
353 | char fnam[100]; | |
354 | ||
355 | snprintf(fnam, 100, "/proc/%d/ns/pid", qpid); | |
356 | store_lock(); | |
357 | if (stat(fnam, &sb) < 0) | |
358 | goto out; | |
359 | e = lookup_verify_initpid(&sb); | |
360 | if (e) { | |
361 | answer = e->initpid; | |
362 | goto out; | |
363 | } | |
364 | answer = get_init_pid_for_task(qpid); | |
365 | if (answer > 0) | |
366 | save_initpid(&sb, answer); | |
367 | ||
368 | out: | |
8e547050 SH |
369 | /* we prune at end in case we are returning |
370 | * the value we were about to return */ | |
371 | prune_initpid_store(); | |
b11c6ec0 SH |
372 | store_unlock(); |
373 | return answer; | |
374 | } | |
0afd85bd | 375 | |
a05660a6 SH |
376 | static int wait_for_pid(pid_t pid) |
377 | { | |
378 | int status, ret; | |
379 | ||
87dce5f6 SH |
380 | if (pid <= 0) |
381 | return -1; | |
382 | ||
a05660a6 SH |
383 | again: |
384 | ret = waitpid(pid, &status, 0); | |
385 | if (ret == -1) { | |
386 | if (errno == EINTR) | |
387 | goto again; | |
388 | return -1; | |
389 | } | |
390 | if (ret != pid) | |
391 | goto again; | |
392 | if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) | |
393 | return -1; | |
394 | return 0; | |
395 | } | |
396 | ||
b11c6ec0 SH |
397 | |
398 | /* | |
399 | * append pid to *src. | |
400 | * src: a pointer to a char* in which ot append the pid. | |
401 | * sz: the number of characters printed so far, minus trailing \0. | |
402 | * asz: the allocated size so far | |
403 | * pid: the pid to append | |
404 | */ | |
405 | static void must_strcat_pid(char **src, size_t *sz, size_t *asz, pid_t pid) | |
406 | { | |
407 | char tmp[30]; | |
408 | ||
409 | int tmplen = sprintf(tmp, "%d\n", (int)pid); | |
410 | ||
411 | if (!*src || tmplen + *sz + 1 >= *asz) { | |
412 | char *tmp; | |
413 | do { | |
414 | tmp = realloc(*src, *asz + BUF_RESERVE_SIZE); | |
415 | } while (!tmp); | |
416 | *src = tmp; | |
417 | *asz += BUF_RESERVE_SIZE; | |
418 | } | |
419 | memcpy((*src) +*sz , tmp, tmplen); | |
420 | *sz += tmplen; | |
421 | (*src)[*sz] = '\0'; | |
422 | } | |
423 | ||
053a659d SH |
424 | /* |
425 | * Given a open file * to /proc/pid/{u,g}id_map, and an id | |
426 | * valid in the caller's namespace, return the id mapped into | |
427 | * pid's namespace. | |
428 | * Returns the mapped id, or -1 on error. | |
429 | */ | |
430 | unsigned int | |
431 | convert_id_to_ns(FILE *idfile, unsigned int in_id) | |
432 | { | |
433 | unsigned int nsuid, // base id for a range in the idfile's namespace | |
434 | hostuid, // base id for a range in the caller's namespace | |
435 | count; // number of ids in this range | |
436 | char line[400]; | |
437 | int ret; | |
438 | ||
439 | fseek(idfile, 0L, SEEK_SET); | |
440 | while (fgets(line, 400, idfile)) { | |
441 | ret = sscanf(line, "%u %u %u\n", &nsuid, &hostuid, &count); | |
442 | if (ret != 3) | |
443 | continue; | |
444 | if (hostuid + count < hostuid || nsuid + count < nsuid) { | |
445 | /* | |
446 | * uids wrapped around - unexpected as this is a procfile, | |
447 | * so just bail. | |
448 | */ | |
647c89e5 | 449 | fprintf(stderr, "pid wrapparound at entry %u %u %u in %s\n", |
053a659d SH |
450 | nsuid, hostuid, count, line); |
451 | return -1; | |
452 | } | |
453 | if (hostuid <= in_id && hostuid+count > in_id) { | |
454 | /* | |
455 | * now since hostuid <= in_id < hostuid+count, and | |
456 | * hostuid+count and nsuid+count do not wrap around, | |
457 | * we know that nsuid+(in_id-hostuid) which must be | |
458 | * less that nsuid+(count) must not wrap around | |
459 | */ | |
460 | return (in_id - hostuid) + nsuid; | |
461 | } | |
462 | } | |
463 | ||
464 | // no answer found | |
465 | return -1; | |
466 | } | |
467 | ||
341b21ad SH |
468 | /* |
469 | * for is_privileged_over, | |
470 | * specify whether we require the calling uid to be root in his | |
471 | * namespace | |
472 | */ | |
473 | #define NS_ROOT_REQD true | |
474 | #define NS_ROOT_OPT false | |
475 | ||
2c51f8dd SH |
476 | #define PROCLEN 100 |
477 | ||
341b21ad | 478 | static bool is_privileged_over(pid_t pid, uid_t uid, uid_t victim, bool req_ns_root) |
758ad80c | 479 | { |
2c51f8dd SH |
480 | char fpath[PROCLEN]; |
481 | int ret; | |
053a659d SH |
482 | bool answer = false; |
483 | uid_t nsuid; | |
484 | ||
341b21ad SH |
485 | if (victim == -1 || uid == -1) |
486 | return false; | |
487 | ||
488 | /* | |
489 | * If the request is one not requiring root in the namespace, | |
490 | * then having the same uid suffices. (i.e. uid 1000 has write | |
491 | * access to files owned by uid 1000 | |
492 | */ | |
493 | if (!req_ns_root && uid == victim) | |
758ad80c SH |
494 | return true; |
495 | ||
2c51f8dd SH |
496 | ret = snprintf(fpath, PROCLEN, "/proc/%d/uid_map", pid); |
497 | if (ret < 0 || ret >= PROCLEN) | |
498 | return false; | |
053a659d SH |
499 | FILE *f = fopen(fpath, "r"); |
500 | if (!f) | |
501 | return false; | |
502 | ||
341b21ad | 503 | /* if caller's not root in his namespace, reject */ |
053a659d SH |
504 | nsuid = convert_id_to_ns(f, uid); |
505 | if (nsuid) | |
506 | goto out; | |
507 | ||
341b21ad SH |
508 | /* |
509 | * If victim is not mapped into caller's ns, reject. | |
510 | * XXX I'm not sure this check is needed given that fuse | |
511 | * will be sending requests where the vfs has converted | |
512 | */ | |
053a659d SH |
513 | nsuid = convert_id_to_ns(f, victim); |
514 | if (nsuid == -1) | |
515 | goto out; | |
516 | ||
517 | answer = true; | |
518 | ||
519 | out: | |
520 | fclose(f); | |
521 | return answer; | |
758ad80c SH |
522 | } |
523 | ||
524 | static bool perms_include(int fmode, mode_t req_mode) | |
525 | { | |
2ad6d2bd SH |
526 | mode_t r; |
527 | ||
528 | switch (req_mode & O_ACCMODE) { | |
529 | case O_RDONLY: | |
530 | r = S_IROTH; | |
531 | break; | |
532 | case O_WRONLY: | |
533 | r = S_IWOTH; | |
534 | break; | |
535 | case O_RDWR: | |
536 | r = S_IROTH | S_IWOTH; | |
537 | break; | |
538 | default: | |
539 | return false; | |
540 | } | |
541 | return ((fmode & r) == r); | |
758ad80c SH |
542 | } |
543 | ||
a8b6c3e0 SH |
544 | |
545 | /* | |
546 | * taskcg is a/b/c | |
547 | * querycg is /a/b/c/d/e | |
548 | * we return 'd' | |
549 | */ | |
3db25a35 SH |
550 | static char *get_next_cgroup_dir(const char *taskcg, const char *querycg) |
551 | { | |
552 | char *start, *end; | |
553 | ||
554 | if (strlen(taskcg) <= strlen(querycg)) { | |
555 | fprintf(stderr, "%s: I was fed bad input\n", __func__); | |
556 | return NULL; | |
557 | } | |
558 | ||
559 | if (strcmp(querycg, "/") == 0) | |
2c51f8dd | 560 | start = strdup(taskcg + 1); |
3db25a35 | 561 | else |
2c51f8dd SH |
562 | start = strdup(taskcg + strlen(querycg) + 1); |
563 | if (!start) | |
564 | return NULL; | |
3db25a35 SH |
565 | end = strchr(start, '/'); |
566 | if (end) | |
567 | *end = '\0'; | |
568 | return start; | |
569 | } | |
570 | ||
2c51f8dd SH |
571 | static void stripnewline(char *x) |
572 | { | |
573 | size_t l = strlen(x); | |
574 | if (l && x[l-1] == '\n') | |
575 | x[l-1] = '\0'; | |
576 | } | |
577 | ||
578 | static char *get_pid_cgroup(pid_t pid, const char *contrl) | |
579 | { | |
580 | char fnam[PROCLEN]; | |
581 | FILE *f; | |
582 | char *answer = NULL; | |
583 | char *line = NULL; | |
584 | size_t len = 0; | |
585 | int ret; | |
777dd831 SH |
586 | const char *h = find_mounted_controller(contrl); |
587 | if (!h) | |
588 | return NULL; | |
2c51f8dd SH |
589 | |
590 | ret = snprintf(fnam, PROCLEN, "/proc/%d/cgroup", pid); | |
591 | if (ret < 0 || ret >= PROCLEN) | |
592 | return NULL; | |
593 | if (!(f = fopen(fnam, "r"))) | |
594 | return NULL; | |
595 | ||
596 | while (getline(&line, &len, f) != -1) { | |
597 | char *c1, *c2; | |
598 | if (!line[0]) | |
599 | continue; | |
600 | c1 = strchr(line, ':'); | |
601 | if (!c1) | |
602 | goto out; | |
603 | c1++; | |
604 | c2 = strchr(c1, ':'); | |
605 | if (!c2) | |
606 | goto out; | |
607 | *c2 = '\0'; | |
777dd831 | 608 | if (strcmp(c1, h) != 0) |
2c51f8dd SH |
609 | continue; |
610 | c2++; | |
611 | stripnewline(c2); | |
612 | do { | |
613 | answer = strdup(c2); | |
614 | } while (!answer); | |
615 | break; | |
616 | } | |
617 | ||
618 | out: | |
619 | fclose(f); | |
620 | free(line); | |
621 | return answer; | |
622 | } | |
623 | ||
758ad80c SH |
624 | /* |
625 | * check whether a fuse context may access a cgroup dir or file | |
626 | * | |
627 | * If file is not null, it is a cgroup file to check under cg. | |
628 | * If file is null, then we are checking perms on cg itself. | |
629 | * | |
630 | * For files we can check the mode of the list_keys result. | |
631 | * For cgroups, we must make assumptions based on the files under the | |
632 | * cgroup, because cgmanager doesn't tell us ownership/perms of cgroups | |
633 | * yet. | |
634 | */ | |
635 | static bool fc_may_access(struct fuse_context *fc, const char *contrl, const char *cg, const char *file, mode_t mode) | |
636 | { | |
35482f91 | 637 | struct cgfs_files *k = NULL; |
2c51f8dd | 638 | bool ret = false; |
758ad80c | 639 | |
35482f91 SH |
640 | k = cgfs_get_key(contrl, cg, file); |
641 | if (!k) | |
758ad80c | 642 | return false; |
35482f91 SH |
643 | |
644 | if (is_privileged_over(fc->pid, fc->uid, k->uid, NS_ROOT_OPT)) { | |
645 | if (perms_include(k->mode >> 6, mode)) { | |
646 | ret = true; | |
2c51f8dd | 647 | goto out; |
758ad80c SH |
648 | } |
649 | } | |
35482f91 SH |
650 | if (fc->gid == k->gid) { |
651 | if (perms_include(k->mode >> 3, mode)) { | |
652 | ret = true; | |
653 | goto out; | |
654 | } | |
655 | } | |
656 | ret = perms_include(k->mode, mode); | |
758ad80c | 657 | |
2c51f8dd | 658 | out: |
35482f91 | 659 | free_key(k); |
2c51f8dd | 660 | return ret; |
3db25a35 SH |
661 | } |
662 | ||
04b5cbdc SH |
663 | #define INITSCOPE "/init.scope" |
664 | static void prune_init_slice(char *cg) | |
665 | { | |
666 | char *point; | |
826297d7 SF |
667 | size_t cg_len = strlen(cg), initscope_len = strlen(INITSCOPE); |
668 | ||
669 | if (cg_len < initscope_len) | |
670 | return; | |
671 | ||
672 | point = cg + cg_len - initscope_len; | |
04b5cbdc SH |
673 | if (strcmp(point, INITSCOPE) == 0) { |
674 | if (point == cg) | |
675 | *(point+1) = '\0'; | |
676 | else | |
677 | *point = '\0'; | |
678 | } | |
679 | } | |
680 | ||
3db25a35 | 681 | /* |
0dcc31ea SH |
682 | * If pid is in /a/b/c/d, he may only act on things under cg=/a/b/c/d. |
683 | * If pid is in /a, he may act on /a/b, but not on /b. | |
3db25a35 | 684 | * if the answer is false and nextcg is not NULL, then *nextcg will point |
2c51f8dd SH |
685 | * to a string containing the next cgroup directory under cg, which must be |
686 | * freed by the caller. | |
3db25a35 SH |
687 | */ |
688 | static bool caller_is_in_ancestor(pid_t pid, const char *contrl, const char *cg, char **nextcg) | |
689 | { | |
3db25a35 | 690 | bool answer = false; |
a8b6c3e0 SH |
691 | char *c2 = get_pid_cgroup(pid, contrl); |
692 | char *linecmp; | |
3db25a35 | 693 | |
a8b6c3e0 | 694 | if (!c2) |
3db25a35 | 695 | return false; |
a8b6c3e0 | 696 | prune_init_slice(c2); |
3db25a35 | 697 | |
a8b6c3e0 SH |
698 | /* |
699 | * callers pass in '/' for root cgroup, otherwise they pass | |
700 | * in a cgroup without leading '/' | |
701 | */ | |
702 | linecmp = *cg == '/' ? c2 : c2+1; | |
703 | if (strncmp(linecmp, cg, strlen(linecmp)) != 0) { | |
704 | if (nextcg) { | |
705 | *nextcg = get_next_cgroup_dir(linecmp, cg); | |
3db25a35 | 706 | } |
a8b6c3e0 SH |
707 | goto out; |
708 | } | |
709 | answer = true; | |
710 | ||
711 | out: | |
712 | free(c2); | |
713 | return answer; | |
714 | } | |
715 | ||
716 | /* | |
0dcc31ea | 717 | * If pid is in /a/b/c, he may see that /a exists, but not /b or /a/c. |
a8b6c3e0 SH |
718 | */ |
719 | static bool caller_may_see_dir(pid_t pid, const char *contrl, const char *cg) | |
720 | { | |
721 | bool answer = false; | |
722 | char *c2, *task_cg; | |
723 | size_t target_len, task_len; | |
724 | ||
725 | if (strcmp(cg, "/") == 0) | |
726 | return true; | |
727 | ||
728 | c2 = get_pid_cgroup(pid, contrl); | |
a8b6c3e0 SH |
729 | if (!c2) |
730 | return false; | |
ec3b236f | 731 | prune_init_slice(c2); |
a8b6c3e0 SH |
732 | |
733 | task_cg = c2 + 1; | |
734 | target_len = strlen(cg); | |
735 | task_len = strlen(task_cg); | |
a57cba3c TA |
736 | if (task_len == 0) { |
737 | /* Task is in the root cg, it can see everything. This case is | |
738 | * not handled by the strmcps below, since they test for the | |
739 | * last /, but that is the first / that we've chopped off | |
740 | * above. | |
741 | */ | |
742 | answer = true; | |
743 | goto out; | |
744 | } | |
a8b6c3e0 | 745 | if (strcmp(cg, task_cg) == 0) { |
3db25a35 SH |
746 | answer = true; |
747 | goto out; | |
748 | } | |
a8b6c3e0 SH |
749 | if (target_len < task_len) { |
750 | /* looking up a parent dir */ | |
751 | if (strncmp(task_cg, cg, target_len) == 0 && task_cg[target_len] == '/') | |
752 | answer = true; | |
753 | goto out; | |
754 | } | |
755 | if (target_len > task_len) { | |
756 | /* looking up a child dir */ | |
757 | if (strncmp(task_cg, cg, task_len) == 0 && cg[task_len] == '/') | |
758 | answer = true; | |
759 | goto out; | |
760 | } | |
3db25a35 SH |
761 | |
762 | out: | |
a8b6c3e0 | 763 | free(c2); |
3db25a35 SH |
764 | return answer; |
765 | } | |
766 | ||
758ad80c | 767 | /* |
2c51f8dd SH |
768 | * given /cgroup/freezer/a/b, return "freezer". |
769 | * the returned char* should NOT be freed. | |
758ad80c SH |
770 | */ |
771 | static char *pick_controller_from_path(struct fuse_context *fc, const char *path) | |
772 | { | |
773 | const char *p1; | |
2c51f8dd | 774 | char *contr, *slash; |
758ad80c SH |
775 | |
776 | if (strlen(path) < 9) | |
777 | return NULL; | |
ac5d9d48 SH |
778 | if (*(path+7) != '/') |
779 | return NULL; | |
758ad80c | 780 | p1 = path+8; |
2c51f8dd SH |
781 | contr = strdupa(p1); |
782 | if (!contr) | |
783 | return NULL; | |
784 | slash = strstr(contr, "/"); | |
758ad80c SH |
785 | if (slash) |
786 | *slash = '\0'; | |
787 | ||
758ad80c | 788 | int i; |
35482f91 SH |
789 | for (i = 0; i < num_hierarchies; i++) { |
790 | if (hierarchies[i] && strcmp(hierarchies[i], contr) == 0) | |
791 | return hierarchies[i]; | |
758ad80c | 792 | } |
758ad80c SH |
793 | return NULL; |
794 | } | |
795 | ||
796 | /* | |
797 | * Find the start of cgroup in /cgroup/controller/the/cgroup/path | |
798 | * Note that the returned value may include files (keynames) etc | |
799 | */ | |
800 | static const char *find_cgroup_in_path(const char *path) | |
801 | { | |
802 | const char *p1; | |
803 | ||
804 | if (strlen(path) < 9) | |
805 | return NULL; | |
806 | p1 = strstr(path+8, "/"); | |
807 | if (!p1) | |
808 | return NULL; | |
809 | return p1+1; | |
810 | } | |
811 | ||
2c51f8dd | 812 | /* |
febf2b87 SH |
813 | * split the last path element from the path in @cg. |
814 | * @dir is newly allocated and should be freed, @last not | |
815 | */ | |
816 | static void get_cgdir_and_path(const char *cg, char **dir, char **last) | |
758ad80c | 817 | { |
758ad80c SH |
818 | char *p; |
819 | ||
2c51f8dd SH |
820 | do { |
821 | *dir = strdup(cg); | |
822 | } while (!*dir); | |
febf2b87 SH |
823 | *last = strrchr(cg, '/'); |
824 | if (!*last) { | |
825 | *last = NULL; | |
758ad80c SH |
826 | return; |
827 | } | |
828 | p = strrchr(*dir, '/'); | |
829 | *p = '\0'; | |
830 | } | |
831 | ||
832 | /* | |
2ad6d2bd | 833 | * FUSE ops for /cgroup |
758ad80c | 834 | */ |
2ad6d2bd | 835 | |
758ad80c SH |
836 | static int cg_getattr(const char *path, struct stat *sb) |
837 | { | |
838 | struct timespec now; | |
839 | struct fuse_context *fc = fuse_get_context(); | |
2c51f8dd | 840 | char * cgdir = NULL; |
febf2b87 | 841 | char *last = NULL, *path1, *path2; |
35482f91 | 842 | struct cgfs_files *k = NULL; |
758ad80c | 843 | const char *cgroup; |
2c51f8dd SH |
844 | const char *controller = NULL; |
845 | int ret = -ENOENT; | |
758ad80c SH |
846 | |
847 | ||
848 | if (!fc) | |
849 | return -EIO; | |
850 | ||
851 | memset(sb, 0, sizeof(struct stat)); | |
852 | ||
853 | if (clock_gettime(CLOCK_REALTIME, &now) < 0) | |
854 | return -EINVAL; | |
855 | ||
856 | sb->st_uid = sb->st_gid = 0; | |
857 | sb->st_atim = sb->st_mtim = sb->st_ctim = now; | |
858 | sb->st_size = 0; | |
859 | ||
860 | if (strcmp(path, "/cgroup") == 0) { | |
861 | sb->st_mode = S_IFDIR | 00755; | |
862 | sb->st_nlink = 2; | |
863 | return 0; | |
864 | } | |
865 | ||
866 | controller = pick_controller_from_path(fc, path); | |
867 | if (!controller) | |
868 | return -EIO; | |
758ad80c SH |
869 | cgroup = find_cgroup_in_path(path); |
870 | if (!cgroup) { | |
871 | /* this is just /cgroup/controller, return it as a dir */ | |
872 | sb->st_mode = S_IFDIR | 00755; | |
873 | sb->st_nlink = 2; | |
874 | return 0; | |
875 | } | |
341b21ad | 876 | |
febf2b87 | 877 | get_cgdir_and_path(cgroup, &cgdir, &last); |
758ad80c | 878 | |
febf2b87 | 879 | if (!last) { |
758ad80c SH |
880 | path1 = "/"; |
881 | path2 = cgdir; | |
882 | } else { | |
883 | path1 = cgdir; | |
febf2b87 | 884 | path2 = last; |
758ad80c SH |
885 | } |
886 | ||
b11c6ec0 | 887 | pid_t initpid = lookup_initpid_in_store(fc->pid); |
87dce5f6 SH |
888 | if (initpid <= 0) |
889 | initpid = fc->pid; | |
758ad80c | 890 | /* check that cgcopy is either a child cgroup of cgdir, or listed in its keys. |
febf2b87 SH |
891 | * Then check that caller's cgroup is under path if last is a child |
892 | * cgroup, or cgdir if last is a file */ | |
758ad80c SH |
893 | |
894 | if (is_child_cgroup(controller, path1, path2)) { | |
0dcc31ea | 895 | if (!caller_may_see_dir(initpid, controller, cgroup)) { |
a8b6c3e0 SH |
896 | ret = -ENOENT; |
897 | goto out; | |
898 | } | |
0dcc31ea | 899 | if (!caller_is_in_ancestor(initpid, controller, cgroup, NULL)) { |
f9a05025 SH |
900 | /* this is just /cgroup/controller, return it as a dir */ |
901 | sb->st_mode = S_IFDIR | 00555; | |
902 | sb->st_nlink = 2; | |
2c51f8dd SH |
903 | ret = 0; |
904 | goto out; | |
905 | } | |
906 | if (!fc_may_access(fc, controller, cgroup, NULL, O_RDONLY)) { | |
907 | ret = -EACCES; | |
908 | goto out; | |
f9a05025 | 909 | } |
758ad80c | 910 | |
053a659d SH |
911 | // get uid, gid, from '/tasks' file and make up a mode |
912 | // That is a hack, until cgmanager gains a GetCgroupPerms fn. | |
913 | sb->st_mode = S_IFDIR | 00755; | |
febf2b87 | 914 | k = cgfs_get_key(controller, cgroup, NULL); |
053a659d | 915 | if (!k) { |
053a659d SH |
916 | sb->st_uid = sb->st_gid = 0; |
917 | } else { | |
053a659d SH |
918 | sb->st_uid = k->uid; |
919 | sb->st_gid = k->gid; | |
920 | } | |
2c51f8dd | 921 | free_key(k); |
758ad80c | 922 | sb->st_nlink = 2; |
2c51f8dd SH |
923 | ret = 0; |
924 | goto out; | |
758ad80c SH |
925 | } |
926 | ||
35482f91 | 927 | if ((k = cgfs_get_key(controller, path1, path2)) != NULL) { |
758ad80c | 928 | sb->st_mode = S_IFREG | k->mode; |
053a659d | 929 | sb->st_nlink = 1; |
758ad80c SH |
930 | sb->st_uid = k->uid; |
931 | sb->st_gid = k->gid; | |
7253e0a4 | 932 | sb->st_size = 0; |
2c51f8dd | 933 | free_key(k); |
0dcc31ea | 934 | if (!caller_is_in_ancestor(initpid, controller, path1, NULL)) { |
adc3867b SH |
935 | ret = -ENOENT; |
936 | goto out; | |
937 | } | |
938 | if (!fc_may_access(fc, controller, path1, path2, O_RDONLY)) { | |
939 | ret = -EACCES; | |
940 | goto out; | |
941 | } | |
2c51f8dd SH |
942 | |
943 | ret = 0; | |
758ad80c SH |
944 | } |
945 | ||
2c51f8dd SH |
946 | out: |
947 | free(cgdir); | |
948 | return ret; | |
758ad80c | 949 | } |
2183082c | 950 | |
758ad80c | 951 | static int cg_opendir(const char *path, struct fuse_file_info *fi) |
2183082c | 952 | { |
7f163b71 | 953 | struct fuse_context *fc = fuse_get_context(); |
7f163b71 | 954 | const char *cgroup; |
c688e1b3 | 955 | struct file_info *dir_info; |
2c51f8dd | 956 | char *controller = NULL; |
7f163b71 SH |
957 | |
958 | if (!fc) | |
959 | return -EIO; | |
960 | ||
c688e1b3 SH |
961 | if (strcmp(path, "/cgroup") == 0) { |
962 | cgroup = NULL; | |
963 | controller = NULL; | |
964 | } else { | |
965 | // return list of keys for the controller, and list of child cgroups | |
966 | controller = pick_controller_from_path(fc, path); | |
967 | if (!controller) | |
968 | return -EIO; | |
7f163b71 | 969 | |
c688e1b3 SH |
970 | cgroup = find_cgroup_in_path(path); |
971 | if (!cgroup) { | |
972 | /* this is just /cgroup/controller, return its contents */ | |
973 | cgroup = "/"; | |
974 | } | |
7f163b71 SH |
975 | } |
976 | ||
b11c6ec0 | 977 | pid_t initpid = lookup_initpid_in_store(fc->pid); |
87dce5f6 SH |
978 | if (initpid <= 0) |
979 | initpid = fc->pid; | |
a8b6c3e0 | 980 | if (cgroup) { |
0dcc31ea | 981 | if (!caller_may_see_dir(initpid, controller, cgroup)) |
a8b6c3e0 SH |
982 | return -ENOENT; |
983 | if (!fc_may_access(fc, controller, cgroup, NULL, O_RDONLY)) | |
984 | return -EACCES; | |
2c51f8dd | 985 | } |
c688e1b3 SH |
986 | |
987 | /* we'll free this at cg_releasedir */ | |
2c51f8dd SH |
988 | dir_info = malloc(sizeof(*dir_info)); |
989 | if (!dir_info) | |
990 | return -ENOMEM; | |
35482f91 SH |
991 | dir_info->controller = must_copy_string(controller); |
992 | dir_info->cgroup = must_copy_string(cgroup); | |
443d13f5 | 993 | dir_info->type = LXC_TYPE_CGDIR; |
c688e1b3 | 994 | dir_info->buf = NULL; |
8f6e8f5e | 995 | dir_info->file = NULL; |
c688e1b3 SH |
996 | dir_info->buflen = 0; |
997 | ||
998 | fi->fh = (unsigned long)dir_info; | |
758ad80c SH |
999 | return 0; |
1000 | } | |
1001 | ||
758ad80c SH |
1002 | static int cg_readdir(const char *path, void *buf, fuse_fill_dir_t filler, off_t offset, |
1003 | struct fuse_file_info *fi) | |
1004 | { | |
c688e1b3 | 1005 | struct file_info *d = (struct file_info *)fi->fh; |
35482f91 | 1006 | struct cgfs_files **list = NULL; |
2c51f8dd SH |
1007 | int i, ret; |
1008 | char *nextcg = NULL; | |
758ad80c | 1009 | struct fuse_context *fc = fuse_get_context(); |
2c51f8dd | 1010 | char **clist = NULL; |
758ad80c | 1011 | |
443d13f5 | 1012 | if (d->type != LXC_TYPE_CGDIR) { |
b845ad01 SH |
1013 | fprintf(stderr, "Internal error: file cache info used in readdir\n"); |
1014 | return -EIO; | |
1015 | } | |
c688e1b3 SH |
1016 | if (!d->cgroup && !d->controller) { |
1017 | // ls /var/lib/lxcfs/cgroup - just show list of controllers | |
758ad80c SH |
1018 | int i; |
1019 | ||
35482f91 SH |
1020 | for (i = 0; i < num_hierarchies; i++) { |
1021 | if (hierarchies[i] && filler(buf, hierarchies[i], NULL, 0) != 0) { | |
758ad80c SH |
1022 | return -EIO; |
1023 | } | |
1024 | } | |
1025 | return 0; | |
1026 | } | |
1027 | ||
35482f91 | 1028 | if (!cgfs_list_keys(d->controller, d->cgroup, &list)) { |
3db25a35 | 1029 | // not a valid cgroup |
2c51f8dd SH |
1030 | ret = -EINVAL; |
1031 | goto out; | |
1032 | } | |
3db25a35 | 1033 | |
b11c6ec0 | 1034 | pid_t initpid = lookup_initpid_in_store(fc->pid); |
87dce5f6 SH |
1035 | if (initpid <= 0) |
1036 | initpid = fc->pid; | |
0dcc31ea | 1037 | if (!caller_is_in_ancestor(initpid, d->controller, d->cgroup, &nextcg)) { |
3db25a35 SH |
1038 | if (nextcg) { |
1039 | int ret; | |
1040 | ret = filler(buf, nextcg, NULL, 0); | |
2c51f8dd SH |
1041 | free(nextcg); |
1042 | if (ret != 0) { | |
1043 | ret = -EIO; | |
1044 | goto out; | |
1045 | } | |
3db25a35 | 1046 | } |
2c51f8dd SH |
1047 | ret = 0; |
1048 | goto out; | |
3db25a35 SH |
1049 | } |
1050 | ||
758ad80c | 1051 | for (i = 0; list[i]; i++) { |
758ad80c | 1052 | if (filler(buf, list[i]->name, NULL, 0) != 0) { |
2c51f8dd SH |
1053 | ret = -EIO; |
1054 | goto out; | |
758ad80c SH |
1055 | } |
1056 | } | |
1057 | ||
1058 | // now get the list of child cgroups | |
758ad80c | 1059 | |
35482f91 | 1060 | if (!cgfs_list_children(d->controller, d->cgroup, &clist)) { |
2c51f8dd SH |
1061 | ret = 0; |
1062 | goto out; | |
1063 | } | |
758ad80c | 1064 | for (i = 0; clist[i]; i++) { |
758ad80c | 1065 | if (filler(buf, clist[i], NULL, 0) != 0) { |
2c51f8dd SH |
1066 | ret = -EIO; |
1067 | goto out; | |
758ad80c SH |
1068 | } |
1069 | } | |
2c51f8dd SH |
1070 | ret = 0; |
1071 | ||
1072 | out: | |
1073 | free_keys(list); | |
1074 | if (clist) { | |
1075 | for (i = 0; clist[i]; i++) | |
1076 | free(clist[i]); | |
1077 | free(clist); | |
1078 | } | |
1079 | return ret; | |
758ad80c SH |
1080 | } |
1081 | ||
8f6e8f5e SH |
1082 | static void do_release_file_info(struct file_info *f) |
1083 | { | |
2c51f8dd SH |
1084 | if (!f) |
1085 | return; | |
1086 | free(f->controller); | |
1087 | free(f->cgroup); | |
1088 | free(f->file); | |
1089 | free(f->buf); | |
1090 | free(f); | |
8f6e8f5e SH |
1091 | } |
1092 | ||
758ad80c SH |
1093 | static int cg_releasedir(const char *path, struct fuse_file_info *fi) |
1094 | { | |
c688e1b3 SH |
1095 | struct file_info *d = (struct file_info *)fi->fh; |
1096 | ||
8f6e8f5e | 1097 | do_release_file_info(d); |
758ad80c SH |
1098 | return 0; |
1099 | } | |
1100 | ||
99978832 SH |
1101 | static int cg_open(const char *path, struct fuse_file_info *fi) |
1102 | { | |
99978832 | 1103 | const char *cgroup; |
febf2b87 | 1104 | char *last = NULL, *path1, *path2, * cgdir = NULL, *controller; |
35482f91 | 1105 | struct cgfs_files *k = NULL; |
8f6e8f5e | 1106 | struct file_info *file_info; |
99978832 | 1107 | struct fuse_context *fc = fuse_get_context(); |
2c51f8dd | 1108 | int ret; |
99978832 SH |
1109 | |
1110 | if (!fc) | |
1111 | return -EIO; | |
1112 | ||
1113 | controller = pick_controller_from_path(fc, path); | |
1114 | if (!controller) | |
1115 | return -EIO; | |
1116 | cgroup = find_cgroup_in_path(path); | |
1117 | if (!cgroup) | |
1118 | return -EINVAL; | |
1119 | ||
febf2b87 SH |
1120 | get_cgdir_and_path(cgroup, &cgdir, &last); |
1121 | if (!last) { | |
99978832 SH |
1122 | path1 = "/"; |
1123 | path2 = cgdir; | |
1124 | } else { | |
1125 | path1 = cgdir; | |
febf2b87 | 1126 | path2 = last; |
99978832 SH |
1127 | } |
1128 | ||
35482f91 | 1129 | k = cgfs_get_key(controller, path1, path2); |
2c51f8dd SH |
1130 | if (!k) { |
1131 | ret = -EINVAL; | |
1132 | goto out; | |
1133 | } | |
1134 | free_key(k); | |
99978832 | 1135 | |
b11c6ec0 | 1136 | pid_t initpid = lookup_initpid_in_store(fc->pid); |
87dce5f6 SH |
1137 | if (initpid <= 0) |
1138 | initpid = fc->pid; | |
0dcc31ea | 1139 | if (!caller_may_see_dir(initpid, controller, path1)) { |
a8b6c3e0 SH |
1140 | ret = -ENOENT; |
1141 | goto out; | |
1142 | } | |
2c51f8dd | 1143 | if (!fc_may_access(fc, controller, path1, path2, fi->flags)) { |
8f6e8f5e | 1144 | // should never get here |
2c51f8dd SH |
1145 | ret = -EACCES; |
1146 | goto out; | |
1147 | } | |
99978832 | 1148 | |
8f6e8f5e | 1149 | /* we'll free this at cg_release */ |
2c51f8dd SH |
1150 | file_info = malloc(sizeof(*file_info)); |
1151 | if (!file_info) { | |
1152 | ret = -ENOMEM; | |
1153 | goto out; | |
1154 | } | |
35482f91 SH |
1155 | file_info->controller = must_copy_string(controller); |
1156 | file_info->cgroup = must_copy_string(path1); | |
1157 | file_info->file = must_copy_string(path2); | |
443d13f5 | 1158 | file_info->type = LXC_TYPE_CGFILE; |
8f6e8f5e SH |
1159 | file_info->buf = NULL; |
1160 | file_info->buflen = 0; | |
1161 | ||
1162 | fi->fh = (unsigned long)file_info; | |
2c51f8dd SH |
1163 | ret = 0; |
1164 | ||
1165 | out: | |
1166 | free(cgdir); | |
1167 | return ret; | |
8f6e8f5e SH |
1168 | } |
1169 | ||
1170 | static int cg_release(const char *path, struct fuse_file_info *fi) | |
1171 | { | |
1172 | struct file_info *f = (struct file_info *)fi->fh; | |
1173 | ||
1174 | do_release_file_info(f); | |
1175 | return 0; | |
99978832 SH |
1176 | } |
1177 | ||
5b2dfd85 SH |
1178 | #define POLLIN_SET ( EPOLLIN | EPOLLHUP | EPOLLRDHUP ) |
1179 | ||
1180 | static bool wait_for_sock(int sock, int timeout) | |
a05660a6 | 1181 | { |
5b2dfd85 | 1182 | struct epoll_event ev; |
c26e12cb | 1183 | int epfd, ret, now, starttime, deltatime, saved_errno; |
5b2dfd85 | 1184 | |
c26e12cb SH |
1185 | if ((starttime = time(NULL)) < 0) |
1186 | return false; | |
1187 | ||
1188 | if ((epfd = epoll_create(1)) < 0) { | |
5b2dfd85 SH |
1189 | fprintf(stderr, "Failed to create epoll socket: %m\n"); |
1190 | return false; | |
1191 | } | |
1192 | ||
1193 | ev.events = POLLIN_SET; | |
1194 | ev.data.fd = sock; | |
1195 | if (epoll_ctl(epfd, EPOLL_CTL_ADD, sock, &ev) < 0) { | |
1196 | fprintf(stderr, "Failed adding socket to epoll: %m\n"); | |
1197 | close(epfd); | |
1198 | return false; | |
1199 | } | |
1200 | ||
c26e12cb SH |
1201 | again: |
1202 | if ((now = time(NULL)) < 0) { | |
1203 | close(epfd); | |
1204 | return false; | |
1205 | } | |
a05660a6 | 1206 | |
c26e12cb SH |
1207 | deltatime = (starttime + timeout) - now; |
1208 | if (deltatime < 0) { // timeout | |
1209 | errno = 0; | |
b11c6ec0 | 1210 | close(epfd); |
5b2dfd85 | 1211 | return false; |
c26e12cb SH |
1212 | } |
1213 | ret = epoll_wait(epfd, &ev, 1, 1000*deltatime + 1); | |
1214 | if (ret < 0 && errno == EINTR) | |
1215 | goto again; | |
1216 | saved_errno = errno; | |
1217 | close(epfd); | |
1218 | ||
1219 | if (ret <= 0) { | |
1220 | errno = saved_errno; | |
5b2dfd85 SH |
1221 | return false; |
1222 | } | |
1223 | return true; | |
1224 | } | |
a05660a6 | 1225 | |
5b2dfd85 SH |
1226 | static int msgrecv(int sockfd, void *buf, size_t len) |
1227 | { | |
1228 | if (!wait_for_sock(sockfd, 2)) | |
a05660a6 SH |
1229 | return -1; |
1230 | return recv(sockfd, buf, len, MSG_DONTWAIT); | |
1231 | } | |
1232 | ||
01e71852 | 1233 | static int send_creds(int sock, struct ucred *cred, char v, bool pingfirst) |
a05660a6 SH |
1234 | { |
1235 | struct msghdr msg = { 0 }; | |
1236 | struct iovec iov; | |
1237 | struct cmsghdr *cmsg; | |
1238 | char cmsgbuf[CMSG_SPACE(sizeof(*cred))]; | |
1239 | char buf[1]; | |
1240 | buf[0] = 'p'; | |
1241 | ||
01e71852 SH |
1242 | if (pingfirst) { |
1243 | if (msgrecv(sock, buf, 1) != 1) { | |
1420baf8 | 1244 | fprintf(stderr, "%s: Error getting reply from server over socketpair\n", |
01e71852 SH |
1245 | __func__); |
1246 | return SEND_CREDS_FAIL; | |
1247 | } | |
a05660a6 SH |
1248 | } |
1249 | ||
1250 | msg.msg_control = cmsgbuf; | |
1251 | msg.msg_controllen = sizeof(cmsgbuf); | |
1252 | ||
1253 | cmsg = CMSG_FIRSTHDR(&msg); | |
1254 | cmsg->cmsg_len = CMSG_LEN(sizeof(struct ucred)); | |
1255 | cmsg->cmsg_level = SOL_SOCKET; | |
1256 | cmsg->cmsg_type = SCM_CREDENTIALS; | |
1257 | memcpy(CMSG_DATA(cmsg), cred, sizeof(*cred)); | |
1258 | ||
1259 | msg.msg_name = NULL; | |
1260 | msg.msg_namelen = 0; | |
1261 | ||
1262 | buf[0] = v; | |
1263 | iov.iov_base = buf; | |
1264 | iov.iov_len = sizeof(buf); | |
1265 | msg.msg_iov = &iov; | |
1266 | msg.msg_iovlen = 1; | |
1267 | ||
1268 | if (sendmsg(sock, &msg, 0) < 0) { | |
1420baf8 | 1269 | fprintf(stderr, "%s: failed at sendmsg: %s\n", __func__, |
a05660a6 SH |
1270 | strerror(errno)); |
1271 | if (errno == 3) | |
01e71852 SH |
1272 | return SEND_CREDS_NOTSK; |
1273 | return SEND_CREDS_FAIL; | |
a05660a6 SH |
1274 | } |
1275 | ||
01e71852 | 1276 | return SEND_CREDS_OK; |
a05660a6 SH |
1277 | } |
1278 | ||
1279 | static bool recv_creds(int sock, struct ucred *cred, char *v) | |
1280 | { | |
1281 | struct msghdr msg = { 0 }; | |
1282 | struct iovec iov; | |
1283 | struct cmsghdr *cmsg; | |
1284 | char cmsgbuf[CMSG_SPACE(sizeof(*cred))]; | |
1285 | char buf[1]; | |
1286 | int ret; | |
1287 | int optval = 1; | |
1288 | ||
1289 | *v = '1'; | |
1290 | ||
1291 | cred->pid = -1; | |
1292 | cred->uid = -1; | |
1293 | cred->gid = -1; | |
1294 | ||
1295 | if (setsockopt(sock, SOL_SOCKET, SO_PASSCRED, &optval, sizeof(optval)) == -1) { | |
1420baf8 | 1296 | fprintf(stderr, "Failed to set passcred: %s\n", strerror(errno)); |
a05660a6 SH |
1297 | return false; |
1298 | } | |
1299 | buf[0] = '1'; | |
1300 | if (write(sock, buf, 1) != 1) { | |
1420baf8 | 1301 | fprintf(stderr, "Failed to start write on scm fd: %s\n", strerror(errno)); |
a05660a6 SH |
1302 | return false; |
1303 | } | |
1304 | ||
1305 | msg.msg_name = NULL; | |
1306 | msg.msg_namelen = 0; | |
1307 | msg.msg_control = cmsgbuf; | |
1308 | msg.msg_controllen = sizeof(cmsgbuf); | |
1309 | ||
1310 | iov.iov_base = buf; | |
1311 | iov.iov_len = sizeof(buf); | |
1312 | msg.msg_iov = &iov; | |
1313 | msg.msg_iovlen = 1; | |
1314 | ||
5b2dfd85 SH |
1315 | if (!wait_for_sock(sock, 2)) { |
1316 | fprintf(stderr, "Timed out waiting for scm_cred: %s\n", | |
6ee867dc SH |
1317 | strerror(errno)); |
1318 | return false; | |
1319 | } | |
1320 | ret = recvmsg(sock, &msg, MSG_DONTWAIT); | |
a05660a6 | 1321 | if (ret < 0) { |
1420baf8 | 1322 | fprintf(stderr, "Failed to receive scm_cred: %s\n", |
a05660a6 SH |
1323 | strerror(errno)); |
1324 | return false; | |
1325 | } | |
1326 | ||
1327 | cmsg = CMSG_FIRSTHDR(&msg); | |
1328 | ||
1329 | if (cmsg && cmsg->cmsg_len == CMSG_LEN(sizeof(struct ucred)) && | |
1330 | cmsg->cmsg_level == SOL_SOCKET && | |
1331 | cmsg->cmsg_type == SCM_CREDENTIALS) { | |
1332 | memcpy(cred, CMSG_DATA(cmsg), sizeof(*cred)); | |
1333 | } | |
1334 | *v = buf[0]; | |
1335 | ||
1336 | return true; | |
1337 | } | |
1338 | ||
1339 | ||
1340 | /* | |
4775fba1 SH |
1341 | * pid_to_ns - reads pids from a ucred over a socket, then writes the |
1342 | * int value back over the socket. This shifts the pid from the | |
1343 | * sender's pidns into tpid's pidns. | |
a05660a6 | 1344 | */ |
4775fba1 | 1345 | static void pid_to_ns(int sock, pid_t tpid) |
a05660a6 SH |
1346 | { |
1347 | char v = '0'; | |
1348 | struct ucred cred; | |
1349 | ||
1350 | while (recv_creds(sock, &cred, &v)) { | |
1351 | if (v == '1') | |
67bd113f | 1352 | _exit(0); |
a05660a6 | 1353 | if (write(sock, &cred.pid, sizeof(pid_t)) != sizeof(pid_t)) |
67bd113f | 1354 | _exit(1); |
a05660a6 | 1355 | } |
67bd113f | 1356 | _exit(0); |
a05660a6 SH |
1357 | } |
1358 | ||
1359 | /* | |
4775fba1 | 1360 | * pid_to_ns_wrapper: when you setns into a pidns, you yourself remain |
a05660a6 | 1361 | * in your old pidns. Only children which you fork will be in the target |
4775fba1 | 1362 | * pidns. So the pid_to_ns_wrapper does the setns, then forks a child to |
a05660a6 SH |
1363 | * actually convert pids |
1364 | */ | |
4775fba1 | 1365 | static void pid_to_ns_wrapper(int sock, pid_t tpid) |
a05660a6 | 1366 | { |
ea56f722 | 1367 | int newnsfd = -1, ret, cpipe[2]; |
a05660a6 SH |
1368 | char fnam[100]; |
1369 | pid_t cpid; | |
ea56f722 | 1370 | char v; |
a05660a6 | 1371 | |
c0adec85 SH |
1372 | ret = snprintf(fnam, sizeof(fnam), "/proc/%d/ns/pid", tpid); |
1373 | if (ret < 0 || ret >= sizeof(fnam)) | |
67bd113f | 1374 | _exit(1); |
a05660a6 SH |
1375 | newnsfd = open(fnam, O_RDONLY); |
1376 | if (newnsfd < 0) | |
67bd113f | 1377 | _exit(1); |
a05660a6 | 1378 | if (setns(newnsfd, 0) < 0) |
67bd113f | 1379 | _exit(1); |
a05660a6 SH |
1380 | close(newnsfd); |
1381 | ||
ea56f722 | 1382 | if (pipe(cpipe) < 0) |
67bd113f | 1383 | _exit(1); |
a05660a6 | 1384 | |
ea56f722 | 1385 | cpid = fork(); |
a05660a6 | 1386 | if (cpid < 0) |
67bd113f | 1387 | _exit(1); |
ea56f722 SH |
1388 | |
1389 | if (!cpid) { | |
1390 | char b = '1'; | |
1391 | close(cpipe[0]); | |
1392 | if (write(cpipe[1], &b, sizeof(char)) < 0) { | |
1393 | fprintf(stderr, "%s (child): erorr on write: %s\n", | |
1394 | __func__, strerror(errno)); | |
1395 | } | |
1396 | close(cpipe[1]); | |
4775fba1 | 1397 | pid_to_ns(sock, tpid); |
ff96a5f9 | 1398 | _exit(1); // not reached |
ea56f722 SH |
1399 | } |
1400 | // give the child 1 second to be done forking and | |
ff96a5f9 | 1401 | // write its ack |
5b2dfd85 | 1402 | if (!wait_for_sock(cpipe[0], 1)) |
ff96a5f9 | 1403 | _exit(1); |
ea56f722 | 1404 | ret = read(cpipe[0], &v, 1); |
ff96a5f9 SH |
1405 | if (ret != sizeof(char) || v != '1') |
1406 | _exit(1); | |
ea56f722 | 1407 | |
a05660a6 | 1408 | if (!wait_for_pid(cpid)) |
67bd113f SH |
1409 | _exit(1); |
1410 | _exit(0); | |
a05660a6 SH |
1411 | } |
1412 | ||
1413 | /* | |
1414 | * To read cgroup files with a particular pid, we will setns into the child | |
1415 | * pidns, open a pipe, fork a child - which will be the first to really be in | |
35482f91 | 1416 | * the child ns - which does the cgfs_get_value and writes the data to the pipe. |
a05660a6 SH |
1417 | */ |
1418 | static bool do_read_pids(pid_t tpid, const char *contrl, const char *cg, const char *file, char **d) | |
1419 | { | |
1420 | int sock[2] = {-1, -1}; | |
2c51f8dd | 1421 | char *tmpdata = NULL; |
a05660a6 SH |
1422 | int ret; |
1423 | pid_t qpid, cpid = -1; | |
1424 | bool answer = false; | |
1425 | char v = '0'; | |
1426 | struct ucred cred; | |
2c51f8dd | 1427 | size_t sz = 0, asz = 0; |
a05660a6 | 1428 | |
35482f91 | 1429 | if (!cgfs_get_value(contrl, cg, file, &tmpdata)) |
a05660a6 SH |
1430 | return false; |
1431 | ||
1432 | /* | |
1433 | * Now we read the pids from returned data one by one, pass | |
1434 | * them into a child in the target namespace, read back the | |
1435 | * translated pids, and put them into our to-return data | |
1436 | */ | |
1437 | ||
1438 | if (socketpair(AF_UNIX, SOCK_DGRAM, 0, sock) < 0) { | |
1439 | perror("socketpair"); | |
2c51f8dd SH |
1440 | free(tmpdata); |
1441 | return false; | |
a05660a6 SH |
1442 | } |
1443 | ||
1444 | cpid = fork(); | |
1445 | if (cpid == -1) | |
1446 | goto out; | |
1447 | ||
ff96a5f9 | 1448 | if (!cpid) // child - exits when done |
4775fba1 | 1449 | pid_to_ns_wrapper(sock[1], tpid); |
a05660a6 SH |
1450 | |
1451 | char *ptr = tmpdata; | |
1452 | cred.uid = 0; | |
1453 | cred.gid = 0; | |
1454 | while (sscanf(ptr, "%d\n", &qpid) == 1) { | |
1455 | cred.pid = qpid; | |
01e71852 SH |
1456 | ret = send_creds(sock[0], &cred, v, true); |
1457 | ||
1458 | if (ret == SEND_CREDS_NOTSK) | |
1459 | goto next; | |
1460 | if (ret == SEND_CREDS_FAIL) | |
a05660a6 SH |
1461 | goto out; |
1462 | ||
1463 | // read converted results | |
5b2dfd85 SH |
1464 | if (!wait_for_sock(sock[0], 2)) { |
1465 | fprintf(stderr, "%s: timed out waiting for pid from child: %s\n", | |
6ee867dc | 1466 | __func__, strerror(errno)); |
a05660a6 SH |
1467 | goto out; |
1468 | } | |
1469 | if (read(sock[0], &qpid, sizeof(qpid)) != sizeof(qpid)) { | |
6ee867dc SH |
1470 | fprintf(stderr, "%s: error reading pid from child: %s\n", |
1471 | __func__, strerror(errno)); | |
a05660a6 SH |
1472 | goto out; |
1473 | } | |
2c51f8dd | 1474 | must_strcat_pid(d, &sz, &asz, qpid); |
01e71852 | 1475 | next: |
a05660a6 SH |
1476 | ptr = strchr(ptr, '\n'); |
1477 | if (!ptr) | |
1478 | break; | |
1479 | ptr++; | |
1480 | } | |
1481 | ||
1482 | cred.pid = getpid(); | |
1483 | v = '1'; | |
01e71852 | 1484 | if (send_creds(sock[0], &cred, v, true) != SEND_CREDS_OK) { |
a05660a6 | 1485 | // failed to ask child to exit |
6ee867dc SH |
1486 | fprintf(stderr, "%s: failed to ask child to exit: %s\n", |
1487 | __func__, strerror(errno)); | |
a05660a6 SH |
1488 | goto out; |
1489 | } | |
1490 | ||
1491 | answer = true; | |
1492 | ||
1493 | out: | |
2c51f8dd | 1494 | free(tmpdata); |
a05660a6 SH |
1495 | if (cpid != -1) |
1496 | wait_for_pid(cpid); | |
1497 | if (sock[0] != -1) { | |
1498 | close(sock[0]); | |
1499 | close(sock[1]); | |
1500 | } | |
1501 | return answer; | |
1502 | } | |
1503 | ||
99978832 SH |
1504 | static int cg_read(const char *path, char *buf, size_t size, off_t offset, |
1505 | struct fuse_file_info *fi) | |
1506 | { | |
99978832 | 1507 | struct fuse_context *fc = fuse_get_context(); |
8f6e8f5e | 1508 | struct file_info *f = (struct file_info *)fi->fh; |
35482f91 | 1509 | struct cgfs_files *k = NULL; |
2c51f8dd SH |
1510 | char *data = NULL; |
1511 | int ret, s; | |
1512 | bool r; | |
99978832 | 1513 | |
443d13f5 | 1514 | if (f->type != LXC_TYPE_CGFILE) { |
b845ad01 SH |
1515 | fprintf(stderr, "Internal error: directory cache info used in cg_read\n"); |
1516 | return -EIO; | |
1517 | } | |
1518 | ||
99978832 | 1519 | if (offset) |
7253e0a4 | 1520 | return 0; |
99978832 SH |
1521 | |
1522 | if (!fc) | |
1523 | return -EIO; | |
1524 | ||
8f6e8f5e | 1525 | if (!f->controller) |
99978832 SH |
1526 | return -EINVAL; |
1527 | ||
35482f91 | 1528 | if ((k = cgfs_get_key(f->controller, f->cgroup, f->file)) == NULL) { |
2c51f8dd SH |
1529 | return -EINVAL; |
1530 | } | |
1531 | free_key(k); | |
99978832 | 1532 | |
99978832 | 1533 | |
2c51f8dd SH |
1534 | if (!fc_may_access(fc, f->controller, f->cgroup, f->file, O_RDONLY)) { // should never get here |
1535 | ret = -EACCES; | |
1536 | goto out; | |
1537 | } | |
a05660a6 | 1538 | |
2c51f8dd SH |
1539 | if (strcmp(f->file, "tasks") == 0 || |
1540 | strcmp(f->file, "/tasks") == 0 || | |
1541 | strcmp(f->file, "/cgroup.procs") == 0 || | |
1542 | strcmp(f->file, "cgroup.procs") == 0) | |
1543 | // special case - we have to translate the pids | |
1544 | r = do_read_pids(fc->pid, f->controller, f->cgroup, f->file, &data); | |
1545 | else | |
35482f91 | 1546 | r = cgfs_get_value(f->controller, f->cgroup, f->file, &data); |
99978832 | 1547 | |
2c51f8dd SH |
1548 | if (!r) { |
1549 | ret = -EINVAL; | |
1550 | goto out; | |
1551 | } | |
99978832 | 1552 | |
2c51f8dd SH |
1553 | if (!data) { |
1554 | ret = 0; | |
1555 | goto out; | |
99978832 | 1556 | } |
2c51f8dd SH |
1557 | s = strlen(data); |
1558 | if (s > size) | |
1559 | s = size; | |
1560 | memcpy(buf, data, s); | |
1561 | if (s > 0 && s < size && data[s-1] != '\n') | |
1562 | buf[s++] = '\n'; | |
99978832 | 1563 | |
2c51f8dd SH |
1564 | ret = s; |
1565 | ||
1566 | out: | |
1567 | free(data); | |
1568 | return ret; | |
99978832 SH |
1569 | } |
1570 | ||
4775fba1 SH |
1571 | static void pid_from_ns(int sock, pid_t tpid) |
1572 | { | |
1573 | pid_t vpid; | |
1574 | struct ucred cred; | |
1575 | char v; | |
6ee867dc | 1576 | int ret; |
4775fba1 SH |
1577 | |
1578 | cred.uid = 0; | |
1579 | cred.gid = 0; | |
6ee867dc | 1580 | while (1) { |
5b2dfd85 SH |
1581 | if (!wait_for_sock(sock, 2)) { |
1582 | fprintf(stderr, "%s: timeout reading from parent\n", __func__); | |
67bd113f | 1583 | _exit(1); |
6ee867dc SH |
1584 | } |
1585 | if ((ret = read(sock, &vpid, sizeof(pid_t))) != sizeof(pid_t)) { | |
1586 | fprintf(stderr, "%s: bad read from parent: %s\n", | |
1587 | __func__, strerror(errno)); | |
67bd113f | 1588 | _exit(1); |
6ee867dc | 1589 | } |
4775fba1 | 1590 | if (vpid == -1) // done |
01e71852 | 1591 | break; |
4775fba1 SH |
1592 | v = '0'; |
1593 | cred.pid = vpid; | |
01e71852 | 1594 | if (send_creds(sock, &cred, v, true) != SEND_CREDS_OK) { |
4775fba1 SH |
1595 | v = '1'; |
1596 | cred.pid = getpid(); | |
01e71852 | 1597 | if (send_creds(sock, &cred, v, false) != SEND_CREDS_OK) |
67bd113f | 1598 | _exit(1); |
4775fba1 SH |
1599 | } |
1600 | } | |
67bd113f | 1601 | _exit(0); |
4775fba1 SH |
1602 | } |
1603 | ||
1604 | static void pid_from_ns_wrapper(int sock, pid_t tpid) | |
1605 | { | |
ea56f722 | 1606 | int newnsfd = -1, ret, cpipe[2]; |
4775fba1 SH |
1607 | char fnam[100]; |
1608 | pid_t cpid; | |
ea56f722 | 1609 | char v; |
4775fba1 | 1610 | |
c0adec85 SH |
1611 | ret = snprintf(fnam, sizeof(fnam), "/proc/%d/ns/pid", tpid); |
1612 | if (ret < 0 || ret >= sizeof(fnam)) | |
67bd113f | 1613 | _exit(1); |
4775fba1 SH |
1614 | newnsfd = open(fnam, O_RDONLY); |
1615 | if (newnsfd < 0) | |
67bd113f | 1616 | _exit(1); |
4775fba1 | 1617 | if (setns(newnsfd, 0) < 0) |
67bd113f | 1618 | _exit(1); |
4775fba1 SH |
1619 | close(newnsfd); |
1620 | ||
ea56f722 | 1621 | if (pipe(cpipe) < 0) |
67bd113f | 1622 | _exit(1); |
ea56f722 SH |
1623 | |
1624 | loop: | |
4775fba1 SH |
1625 | cpid = fork(); |
1626 | ||
1627 | if (cpid < 0) | |
67bd113f | 1628 | _exit(1); |
ea56f722 SH |
1629 | |
1630 | if (!cpid) { | |
1631 | char b = '1'; | |
1632 | close(cpipe[0]); | |
1633 | if (write(cpipe[1], &b, sizeof(char)) < 0) { | |
1634 | fprintf(stderr, "%s (child): erorr on write: %s\n", | |
1635 | __func__, strerror(errno)); | |
1636 | } | |
1637 | close(cpipe[1]); | |
4775fba1 | 1638 | pid_from_ns(sock, tpid); |
ea56f722 SH |
1639 | } |
1640 | ||
1641 | // give the child 1 second to be done forking and | |
c26e12cb | 1642 | // write its ack |
5b2dfd85 | 1643 | if (!wait_for_sock(cpipe[0], 1)) |
ea56f722 SH |
1644 | goto again; |
1645 | ret = read(cpipe[0], &v, 1); | |
1646 | if (ret != sizeof(char) || v != '1') { | |
1647 | goto again; | |
1648 | } | |
1649 | ||
4775fba1 | 1650 | if (!wait_for_pid(cpid)) |
67bd113f SH |
1651 | _exit(1); |
1652 | _exit(0); | |
ea56f722 SH |
1653 | |
1654 | again: | |
1655 | kill(cpid, SIGKILL); | |
1656 | wait_for_pid(cpid); | |
1657 | goto loop; | |
4775fba1 SH |
1658 | } |
1659 | ||
8ee2a503 SH |
1660 | /* |
1661 | * Given host @uid, return the uid to which it maps in | |
1662 | * @pid's user namespace, or -1 if none. | |
1663 | */ | |
1664 | bool hostuid_to_ns(uid_t uid, pid_t pid, uid_t *answer) | |
1665 | { | |
1666 | FILE *f; | |
1667 | char line[400]; | |
1668 | ||
1669 | sprintf(line, "/proc/%d/uid_map", pid); | |
1670 | if ((f = fopen(line, "r")) == NULL) { | |
1671 | return false; | |
1672 | } | |
1673 | ||
1674 | *answer = convert_id_to_ns(f, uid); | |
1675 | fclose(f); | |
1676 | ||
1677 | if (*answer == -1) | |
1678 | return false; | |
1679 | return true; | |
1680 | } | |
1681 | ||
1682 | /* | |
1683 | * get_pid_creds: get the real uid and gid of @pid from | |
1684 | * /proc/$$/status | |
1685 | * (XXX should we use euid here?) | |
1686 | */ | |
1687 | void get_pid_creds(pid_t pid, uid_t *uid, gid_t *gid) | |
1688 | { | |
1689 | char line[400]; | |
1690 | uid_t u; | |
1691 | gid_t g; | |
1692 | FILE *f; | |
1693 | ||
1694 | *uid = -1; | |
1695 | *gid = -1; | |
1696 | sprintf(line, "/proc/%d/status", pid); | |
1697 | if ((f = fopen(line, "r")) == NULL) { | |
1698 | fprintf(stderr, "Error opening %s: %s\n", line, strerror(errno)); | |
1699 | return; | |
1700 | } | |
1701 | while (fgets(line, 400, f)) { | |
1702 | if (strncmp(line, "Uid:", 4) == 0) { | |
1703 | if (sscanf(line+4, "%u", &u) != 1) { | |
1704 | fprintf(stderr, "bad uid line for pid %u\n", pid); | |
1705 | fclose(f); | |
1706 | return; | |
1707 | } | |
1708 | *uid = u; | |
1709 | } else if (strncmp(line, "Gid:", 4) == 0) { | |
1710 | if (sscanf(line+4, "%u", &g) != 1) { | |
1711 | fprintf(stderr, "bad gid line for pid %u\n", pid); | |
1712 | fclose(f); | |
1713 | return; | |
1714 | } | |
1715 | *gid = g; | |
1716 | } | |
1717 | } | |
1718 | fclose(f); | |
1719 | } | |
1720 | ||
1721 | /* | |
1722 | * May the requestor @r move victim @v to a new cgroup? | |
1723 | * This is allowed if | |
1724 | * . they are the same task | |
1725 | * . they are ownedy by the same uid | |
1726 | * . @r is root on the host, or | |
1727 | * . @v's uid is mapped into @r's where @r is root. | |
1728 | */ | |
1729 | bool may_move_pid(pid_t r, uid_t r_uid, pid_t v) | |
1730 | { | |
1731 | uid_t v_uid, tmpuid; | |
1732 | gid_t v_gid; | |
1733 | ||
1734 | if (r == v) | |
1735 | return true; | |
1736 | if (r_uid == 0) | |
1737 | return true; | |
1738 | get_pid_creds(v, &v_uid, &v_gid); | |
1739 | if (r_uid == v_uid) | |
1740 | return true; | |
1741 | if (hostuid_to_ns(r_uid, r, &tmpuid) && tmpuid == 0 | |
1742 | && hostuid_to_ns(v_uid, r, &tmpuid)) | |
1743 | return true; | |
1744 | return false; | |
1745 | } | |
1746 | ||
1747 | static bool do_write_pids(pid_t tpid, uid_t tuid, const char *contrl, const char *cg, | |
1748 | const char *file, const char *buf) | |
4775fba1 SH |
1749 | { |
1750 | int sock[2] = {-1, -1}; | |
1751 | pid_t qpid, cpid = -1; | |
35482f91 | 1752 | FILE *pids_file = NULL; |
4775fba1 SH |
1753 | bool answer = false, fail = false; |
1754 | ||
35482f91 SH |
1755 | pids_file = open_pids_file(contrl, cg); |
1756 | if (!pids_file) | |
1757 | return false; | |
1758 | ||
4775fba1 SH |
1759 | /* |
1760 | * write the pids to a socket, have helper in writer's pidns | |
1761 | * call movepid for us | |
1762 | */ | |
1763 | if (socketpair(AF_UNIX, SOCK_DGRAM, 0, sock) < 0) { | |
1764 | perror("socketpair"); | |
35482f91 | 1765 | goto out; |
4775fba1 SH |
1766 | } |
1767 | ||
1768 | cpid = fork(); | |
1769 | if (cpid == -1) | |
1770 | goto out; | |
1771 | ||
35482f91 SH |
1772 | if (!cpid) { // child |
1773 | fclose(pids_file); | |
4775fba1 | 1774 | pid_from_ns_wrapper(sock[1], tpid); |
35482f91 | 1775 | } |
4775fba1 SH |
1776 | |
1777 | const char *ptr = buf; | |
1778 | while (sscanf(ptr, "%d", &qpid) == 1) { | |
1779 | struct ucred cred; | |
1780 | char v; | |
1781 | ||
1782 | if (write(sock[0], &qpid, sizeof(qpid)) != sizeof(qpid)) { | |
6ee867dc SH |
1783 | fprintf(stderr, "%s: error writing pid to child: %s\n", |
1784 | __func__, strerror(errno)); | |
4775fba1 SH |
1785 | goto out; |
1786 | } | |
1787 | ||
01e71852 SH |
1788 | if (recv_creds(sock[0], &cred, &v)) { |
1789 | if (v == '0') { | |
8ee2a503 SH |
1790 | if (!may_move_pid(tpid, tuid, cred.pid)) { |
1791 | fail = true; | |
1792 | break; | |
1793 | } | |
35482f91 | 1794 | if (fprintf(pids_file, "%d", (int) cred.pid) < 0) |
01e71852 SH |
1795 | fail = true; |
1796 | } | |
4775fba1 SH |
1797 | } |
1798 | ||
1799 | ptr = strchr(ptr, '\n'); | |
1800 | if (!ptr) | |
1801 | break; | |
1802 | ptr++; | |
1803 | } | |
1804 | ||
1805 | /* All good, write the value */ | |
1806 | qpid = -1; | |
1807 | if (write(sock[0], &qpid ,sizeof(qpid)) != sizeof(qpid)) | |
1420baf8 | 1808 | fprintf(stderr, "Warning: failed to ask child to exit\n"); |
4775fba1 SH |
1809 | |
1810 | if (!fail) | |
1811 | answer = true; | |
1812 | ||
1813 | out: | |
1814 | if (cpid != -1) | |
1815 | wait_for_pid(cpid); | |
1816 | if (sock[0] != -1) { | |
1817 | close(sock[0]); | |
1818 | close(sock[1]); | |
1819 | } | |
35482f91 SH |
1820 | if (pids_file) { |
1821 | if (fclose(pids_file) != 0) | |
1822 | answer = false; | |
1823 | } | |
4775fba1 SH |
1824 | return answer; |
1825 | } | |
1826 | ||
2ad6d2bd SH |
1827 | int cg_write(const char *path, const char *buf, size_t size, off_t offset, |
1828 | struct fuse_file_info *fi) | |
1829 | { | |
2ad6d2bd | 1830 | struct fuse_context *fc = fuse_get_context(); |
2c51f8dd | 1831 | char *localbuf = NULL; |
35482f91 | 1832 | struct cgfs_files *k = NULL; |
8f6e8f5e | 1833 | struct file_info *f = (struct file_info *)fi->fh; |
2c51f8dd | 1834 | bool r; |
2ad6d2bd | 1835 | |
443d13f5 | 1836 | if (f->type != LXC_TYPE_CGFILE) { |
b845ad01 SH |
1837 | fprintf(stderr, "Internal error: directory cache info used in cg_write\n"); |
1838 | return -EIO; | |
1839 | } | |
1840 | ||
2ad6d2bd | 1841 | if (offset) |
7253e0a4 | 1842 | return 0; |
2ad6d2bd SH |
1843 | |
1844 | if (!fc) | |
1845 | return -EIO; | |
1846 | ||
2c51f8dd | 1847 | localbuf = alloca(size+1); |
47cbf0e5 SH |
1848 | localbuf[size] = '\0'; |
1849 | memcpy(localbuf, buf, size); | |
2ad6d2bd | 1850 | |
35482f91 | 1851 | if ((k = cgfs_get_key(f->controller, f->cgroup, f->file)) == NULL) { |
2c51f8dd SH |
1852 | size = -EINVAL; |
1853 | goto out; | |
1854 | } | |
2ad6d2bd | 1855 | |
2c51f8dd SH |
1856 | if (!fc_may_access(fc, f->controller, f->cgroup, f->file, O_WRONLY)) { |
1857 | size = -EACCES; | |
1858 | goto out; | |
1859 | } | |
4775fba1 | 1860 | |
2c51f8dd SH |
1861 | if (strcmp(f->file, "tasks") == 0 || |
1862 | strcmp(f->file, "/tasks") == 0 || | |
1863 | strcmp(f->file, "/cgroup.procs") == 0 || | |
1864 | strcmp(f->file, "cgroup.procs") == 0) | |
1865 | // special case - we have to translate the pids | |
8ee2a503 | 1866 | r = do_write_pids(fc->pid, fc->uid, f->controller, f->cgroup, f->file, localbuf); |
2c51f8dd | 1867 | else |
35482f91 | 1868 | r = cgfs_set_value(f->controller, f->cgroup, f->file, localbuf); |
2ad6d2bd | 1869 | |
2c51f8dd SH |
1870 | if (!r) |
1871 | size = -EINVAL; | |
2ad6d2bd | 1872 | |
2c51f8dd SH |
1873 | out: |
1874 | free_key(k); | |
1875 | return size; | |
2ad6d2bd SH |
1876 | } |
1877 | ||
341b21ad SH |
1878 | int cg_chown(const char *path, uid_t uid, gid_t gid) |
1879 | { | |
1880 | struct fuse_context *fc = fuse_get_context(); | |
febf2b87 | 1881 | char *cgdir = NULL, *last = NULL, *path1, *path2, *controller; |
35482f91 | 1882 | struct cgfs_files *k = NULL; |
341b21ad | 1883 | const char *cgroup; |
2c51f8dd | 1884 | int ret; |
341b21ad SH |
1885 | |
1886 | if (!fc) | |
1887 | return -EIO; | |
1888 | ||
1889 | if (strcmp(path, "/cgroup") == 0) | |
1890 | return -EINVAL; | |
1891 | ||
1892 | controller = pick_controller_from_path(fc, path); | |
1893 | if (!controller) | |
f9a05025 | 1894 | return -EINVAL; |
341b21ad SH |
1895 | cgroup = find_cgroup_in_path(path); |
1896 | if (!cgroup) | |
1897 | /* this is just /cgroup/controller */ | |
1898 | return -EINVAL; | |
1899 | ||
febf2b87 | 1900 | get_cgdir_and_path(cgroup, &cgdir, &last); |
341b21ad | 1901 | |
febf2b87 | 1902 | if (!last) { |
341b21ad SH |
1903 | path1 = "/"; |
1904 | path2 = cgdir; | |
1905 | } else { | |
1906 | path1 = cgdir; | |
febf2b87 | 1907 | path2 = last; |
341b21ad SH |
1908 | } |
1909 | ||
1910 | if (is_child_cgroup(controller, path1, path2)) { | |
1911 | // get uid, gid, from '/tasks' file and make up a mode | |
1912 | // That is a hack, until cgmanager gains a GetCgroupPerms fn. | |
35482f91 | 1913 | k = cgfs_get_key(controller, cgroup, "tasks"); |
341b21ad SH |
1914 | |
1915 | } else | |
35482f91 | 1916 | k = cgfs_get_key(controller, path1, path2); |
341b21ad | 1917 | |
2c51f8dd SH |
1918 | if (!k) { |
1919 | ret = -EINVAL; | |
1920 | goto out; | |
1921 | } | |
341b21ad SH |
1922 | |
1923 | /* | |
1924 | * This being a fuse request, the uid and gid must be valid | |
1925 | * in the caller's namespace. So we can just check to make | |
1926 | * sure that the caller is root in his uid, and privileged | |
1927 | * over the file's current owner. | |
1928 | */ | |
2c51f8dd SH |
1929 | if (!is_privileged_over(fc->pid, fc->uid, k->uid, NS_ROOT_REQD)) { |
1930 | ret = -EACCES; | |
1931 | goto out; | |
1932 | } | |
341b21ad | 1933 | |
1f69d62e | 1934 | ret = cgfs_chown_file(controller, cgroup, uid, gid); |
2c51f8dd SH |
1935 | |
1936 | out: | |
1937 | free_key(k); | |
1938 | free(cgdir); | |
1939 | ||
1940 | return ret; | |
341b21ad | 1941 | } |
2ad6d2bd | 1942 | |
fd2e4e03 SH |
1943 | int cg_chmod(const char *path, mode_t mode) |
1944 | { | |
0a1bb5ea | 1945 | struct fuse_context *fc = fuse_get_context(); |
febf2b87 | 1946 | char * cgdir = NULL, *last = NULL, *path1, *path2, *controller; |
35482f91 | 1947 | struct cgfs_files *k = NULL; |
0a1bb5ea | 1948 | const char *cgroup; |
2c51f8dd | 1949 | int ret; |
0a1bb5ea SH |
1950 | |
1951 | if (!fc) | |
1952 | return -EIO; | |
1953 | ||
1954 | if (strcmp(path, "/cgroup") == 0) | |
1955 | return -EINVAL; | |
1956 | ||
1957 | controller = pick_controller_from_path(fc, path); | |
1958 | if (!controller) | |
f9a05025 | 1959 | return -EINVAL; |
0a1bb5ea SH |
1960 | cgroup = find_cgroup_in_path(path); |
1961 | if (!cgroup) | |
1962 | /* this is just /cgroup/controller */ | |
1963 | return -EINVAL; | |
1964 | ||
febf2b87 | 1965 | get_cgdir_and_path(cgroup, &cgdir, &last); |
0a1bb5ea | 1966 | |
febf2b87 | 1967 | if (!last) { |
0a1bb5ea SH |
1968 | path1 = "/"; |
1969 | path2 = cgdir; | |
1970 | } else { | |
1971 | path1 = cgdir; | |
febf2b87 | 1972 | path2 = last; |
0a1bb5ea SH |
1973 | } |
1974 | ||
1975 | if (is_child_cgroup(controller, path1, path2)) { | |
1976 | // get uid, gid, from '/tasks' file and make up a mode | |
1977 | // That is a hack, until cgmanager gains a GetCgroupPerms fn. | |
35482f91 | 1978 | k = cgfs_get_key(controller, cgroup, "tasks"); |
0a1bb5ea SH |
1979 | |
1980 | } else | |
35482f91 | 1981 | k = cgfs_get_key(controller, path1, path2); |
0a1bb5ea | 1982 | |
2c51f8dd SH |
1983 | if (!k) { |
1984 | ret = -EINVAL; | |
1985 | goto out; | |
1986 | } | |
0a1bb5ea SH |
1987 | |
1988 | /* | |
1989 | * This being a fuse request, the uid and gid must be valid | |
1990 | * in the caller's namespace. So we can just check to make | |
1991 | * sure that the caller is root in his uid, and privileged | |
1992 | * over the file's current owner. | |
1993 | */ | |
2c51f8dd SH |
1994 | if (!is_privileged_over(fc->pid, fc->uid, k->uid, NS_ROOT_OPT)) { |
1995 | ret = -EPERM; | |
1996 | goto out; | |
1997 | } | |
0a1bb5ea | 1998 | |
35482f91 | 1999 | if (!cgfs_chmod_file(controller, cgroup, mode)) { |
2c51f8dd SH |
2000 | ret = -EINVAL; |
2001 | goto out; | |
2002 | } | |
2003 | ||
2004 | ret = 0; | |
2005 | out: | |
2006 | free_key(k); | |
2007 | free(cgdir); | |
2008 | return ret; | |
fd2e4e03 SH |
2009 | } |
2010 | ||
ab54b798 SH |
2011 | int cg_mkdir(const char *path, mode_t mode) |
2012 | { | |
2013 | struct fuse_context *fc = fuse_get_context(); | |
febf2b87 | 2014 | char *last = NULL, *path1, *cgdir = NULL, *controller, *next = NULL; |
ab54b798 | 2015 | const char *cgroup; |
2c51f8dd | 2016 | int ret; |
ab54b798 | 2017 | |
ab54b798 SH |
2018 | if (!fc) |
2019 | return -EIO; | |
2020 | ||
2021 | ||
2022 | controller = pick_controller_from_path(fc, path); | |
2023 | if (!controller) | |
f9a05025 | 2024 | return -EINVAL; |
ab54b798 SH |
2025 | |
2026 | cgroup = find_cgroup_in_path(path); | |
2027 | if (!cgroup) | |
f9a05025 | 2028 | return -EINVAL; |
ab54b798 | 2029 | |
febf2b87 SH |
2030 | get_cgdir_and_path(cgroup, &cgdir, &last); |
2031 | if (!last) | |
ab54b798 SH |
2032 | path1 = "/"; |
2033 | else | |
2034 | path1 = cgdir; | |
2035 | ||
b11c6ec0 | 2036 | pid_t initpid = lookup_initpid_in_store(fc->pid); |
87dce5f6 SH |
2037 | if (initpid <= 0) |
2038 | initpid = fc->pid; | |
0dcc31ea | 2039 | if (!caller_is_in_ancestor(initpid, controller, path1, &next)) { |
87dce5f6 SH |
2040 | if (!next) |
2041 | ret = -EINVAL; | |
2042 | else if (last && strcmp(next, last) == 0) | |
a8b6c3e0 SH |
2043 | ret = -EEXIST; |
2044 | else | |
2045 | ret = -ENOENT; | |
2046 | goto out; | |
2047 | } | |
2048 | ||
2c51f8dd SH |
2049 | if (!fc_may_access(fc, controller, path1, NULL, O_RDWR)) { |
2050 | ret = -EACCES; | |
2051 | goto out; | |
2052 | } | |
0dcc31ea | 2053 | if (!caller_is_in_ancestor(initpid, controller, path1, NULL)) { |
2c51f8dd SH |
2054 | ret = -EACCES; |
2055 | goto out; | |
2056 | } | |
ab54b798 | 2057 | |
af869b9c | 2058 | ret = cgfs_create(controller, cgroup, fc->uid, fc->gid); |
ab54b798 | 2059 | |
2c51f8dd SH |
2060 | out: |
2061 | free(cgdir); | |
a8b6c3e0 | 2062 | free(next); |
2c51f8dd | 2063 | return ret; |
ab54b798 SH |
2064 | } |
2065 | ||
50d8d5b5 SH |
2066 | static int cg_rmdir(const char *path) |
2067 | { | |
2068 | struct fuse_context *fc = fuse_get_context(); | |
febf2b87 | 2069 | char *last = NULL, *cgdir = NULL, *controller, *next = NULL; |
50d8d5b5 | 2070 | const char *cgroup; |
2c51f8dd | 2071 | int ret; |
50d8d5b5 SH |
2072 | |
2073 | if (!fc) | |
2074 | return -EIO; | |
2075 | ||
50d8d5b5 SH |
2076 | controller = pick_controller_from_path(fc, path); |
2077 | if (!controller) | |
f9a05025 | 2078 | return -EINVAL; |
50d8d5b5 SH |
2079 | |
2080 | cgroup = find_cgroup_in_path(path); | |
2081 | if (!cgroup) | |
f9a05025 | 2082 | return -EINVAL; |
50d8d5b5 | 2083 | |
febf2b87 SH |
2084 | get_cgdir_and_path(cgroup, &cgdir, &last); |
2085 | if (!last) { | |
2c51f8dd SH |
2086 | ret = -EINVAL; |
2087 | goto out; | |
2088 | } | |
50d8d5b5 | 2089 | |
b11c6ec0 | 2090 | pid_t initpid = lookup_initpid_in_store(fc->pid); |
87dce5f6 SH |
2091 | if (initpid <= 0) |
2092 | initpid = fc->pid; | |
0dcc31ea | 2093 | if (!caller_is_in_ancestor(initpid, controller, cgroup, &next)) { |
febf2b87 | 2094 | if (!last || strcmp(next, last) == 0) |
a8b6c3e0 SH |
2095 | ret = -EBUSY; |
2096 | else | |
2097 | ret = -ENOENT; | |
2098 | goto out; | |
2099 | } | |
2100 | ||
2c51f8dd SH |
2101 | if (!fc_may_access(fc, controller, cgdir, NULL, O_WRONLY)) { |
2102 | ret = -EACCES; | |
2103 | goto out; | |
2104 | } | |
0dcc31ea | 2105 | if (!caller_is_in_ancestor(initpid, controller, cgroup, NULL)) { |
2c51f8dd SH |
2106 | ret = -EACCES; |
2107 | goto out; | |
2108 | } | |
50d8d5b5 | 2109 | |
35482f91 | 2110 | if (!cgfs_remove(controller, cgroup)) { |
2c51f8dd SH |
2111 | ret = -EINVAL; |
2112 | goto out; | |
2113 | } | |
50d8d5b5 | 2114 | |
2c51f8dd SH |
2115 | ret = 0; |
2116 | ||
2117 | out: | |
2118 | free(cgdir); | |
a8b6c3e0 | 2119 | free(next); |
2c51f8dd | 2120 | return ret; |
50d8d5b5 SH |
2121 | } |
2122 | ||
2dc17609 SH |
2123 | static bool startswith(const char *line, const char *pref) |
2124 | { | |
2125 | if (strncmp(line, pref, strlen(pref)) == 0) | |
2126 | return true; | |
2127 | return false; | |
2128 | } | |
2129 | ||
2130 | static void get_mem_cached(char *memstat, unsigned long *v) | |
2131 | { | |
2132 | char *eol; | |
2133 | ||
2134 | *v = 0; | |
2135 | while (*memstat) { | |
2136 | if (startswith(memstat, "total_cache")) { | |
2137 | sscanf(memstat + 11, "%lu", v); | |
2138 | *v /= 1024; | |
2139 | return; | |
2140 | } | |
2141 | eol = strchr(memstat, '\n'); | |
2142 | if (!eol) | |
2143 | return; | |
2144 | memstat = eol+1; | |
2145 | } | |
2146 | } | |
2147 | ||
49878439 | 2148 | static void get_blkio_io_value(char *str, unsigned major, unsigned minor, char *iotype, unsigned long *v) |
2f919d9d | 2149 | { |
49878439 YY |
2150 | char *eol; |
2151 | char key[32]; | |
2f919d9d | 2152 | |
49878439 YY |
2153 | memset(key, 0, 32); |
2154 | snprintf(key, 32, "%u:%u %s", major, minor, iotype); | |
2f919d9d | 2155 | |
49878439 YY |
2156 | size_t len = strlen(key); |
2157 | *v = 0; | |
2158 | ||
2159 | while (*str) { | |
2160 | if (startswith(str, key)) { | |
2f919d9d SH |
2161 | sscanf(str + len, "%lu", v); |
2162 | return; | |
2163 | } | |
2164 | eol = strchr(str, '\n'); | |
49878439 | 2165 | if (!eol) |
2f919d9d | 2166 | return; |
49878439 YY |
2167 | str = eol+1; |
2168 | } | |
2169 | } | |
2170 | ||
53b43826 SH |
2171 | static int read_file(const char *path, char *buf, size_t size, |
2172 | struct file_info *d) | |
2173 | { | |
2174 | size_t linelen = 0, total_len = 0, rv = 0; | |
2175 | char *line = NULL; | |
2176 | char *cache = d->buf; | |
2177 | size_t cache_size = d->buflen; | |
2178 | FILE *f = fopen(path, "r"); | |
2179 | if (!f) | |
2180 | return 0; | |
2181 | ||
2182 | while (getline(&line, &linelen, f) != -1) { | |
2183 | size_t l = snprintf(cache, cache_size, "%s", line); | |
2184 | if (l < 0) { | |
2185 | perror("Error writing to cache"); | |
2186 | rv = 0; | |
2187 | goto err; | |
2188 | } | |
2189 | if (l >= cache_size) { | |
2190 | fprintf(stderr, "Internal error: truncated write to cache\n"); | |
2191 | rv = 0; | |
2192 | goto err; | |
2193 | } | |
711b4030 SF |
2194 | cache += l; |
2195 | cache_size -= l; | |
2196 | total_len += l; | |
53b43826 SH |
2197 | } |
2198 | ||
2199 | d->size = total_len; | |
2200 | if (total_len > size ) total_len = size; | |
2201 | ||
2202 | /* read from off 0 */ | |
2203 | memcpy(buf, d->buf, total_len); | |
2204 | rv = total_len; | |
2205 | err: | |
2206 | fclose(f); | |
2207 | free(line); | |
2208 | return rv; | |
2209 | } | |
2210 | ||
758ad80c | 2211 | /* |
2ad6d2bd | 2212 | * FUSE ops for /proc |
758ad80c | 2213 | */ |
758ad80c | 2214 | |
7bc95a75 SH |
2215 | static unsigned long get_memlimit(const char *cgroup) |
2216 | { | |
2217 | char *memlimit_str = NULL; | |
2218 | unsigned long memlimit = -1; | |
2219 | ||
35482f91 | 2220 | if (cgfs_get_value("memory", cgroup, "memory.limit_in_bytes", &memlimit_str)) |
7bc95a75 SH |
2221 | memlimit = strtoul(memlimit_str, NULL, 10); |
2222 | ||
2223 | free(memlimit_str); | |
2224 | ||
2225 | return memlimit; | |
2226 | } | |
2227 | ||
2228 | static unsigned long get_min_memlimit(const char *cgroup) | |
2229 | { | |
2230 | char *copy = strdupa(cgroup); | |
2231 | unsigned long memlimit = 0, retlimit; | |
2232 | ||
2233 | retlimit = get_memlimit(copy); | |
2234 | ||
2235 | while (strcmp(copy, "/") != 0) { | |
2236 | copy = dirname(copy); | |
2237 | memlimit = get_memlimit(copy); | |
2238 | if (memlimit != -1 && memlimit < retlimit) | |
2239 | retlimit = memlimit; | |
2240 | }; | |
2241 | ||
2242 | return retlimit; | |
2243 | } | |
2244 | ||
23ce2127 SH |
2245 | static int proc_meminfo_read(char *buf, size_t size, off_t offset, |
2246 | struct fuse_file_info *fi) | |
2247 | { | |
2dc17609 | 2248 | struct fuse_context *fc = fuse_get_context(); |
97f1f27b | 2249 | struct file_info *d = (struct file_info *)fi->fh; |
2c51f8dd | 2250 | char *cg; |
4622ad78 | 2251 | char *memusage_str = NULL, *memstat_str = NULL, |
b731895e NW |
2252 | *memswlimit_str = NULL, *memswusage_str = NULL, |
2253 | *memswlimit_default_str = NULL, *memswusage_default_str = NULL; | |
4622ad78 TG |
2254 | unsigned long memlimit = 0, memusage = 0, memswlimit = 0, memswusage = 0, |
2255 | cached = 0, hosttotal = 0; | |
2dc17609 | 2256 | char *line = NULL; |
e1068397 | 2257 | size_t linelen = 0, total_len = 0, rv = 0; |
97f1f27b YY |
2258 | char *cache = d->buf; |
2259 | size_t cache_size = d->buflen; | |
2c51f8dd | 2260 | FILE *f = NULL; |
2dc17609 | 2261 | |
97f1f27b YY |
2262 | if (offset){ |
2263 | if (offset > d->size) | |
2264 | return -EINVAL; | |
b5ad2d21 SH |
2265 | if (!d->cached) |
2266 | return 0; | |
97f1f27b YY |
2267 | int left = d->size - offset; |
2268 | total_len = left > size ? size: left; | |
2269 | memcpy(buf, cache + offset, total_len); | |
2270 | return total_len; | |
2271 | } | |
2dc17609 | 2272 | |
b11c6ec0 | 2273 | pid_t initpid = lookup_initpid_in_store(fc->pid); |
87dce5f6 SH |
2274 | if (initpid <= 0) |
2275 | initpid = fc->pid; | |
0dcc31ea | 2276 | cg = get_pid_cgroup(initpid, "memory"); |
2dc17609 | 2277 | if (!cg) |
53b43826 | 2278 | return read_file("/proc/meminfo", buf, size, d); |
2dc17609 | 2279 | |
7bc95a75 | 2280 | memlimit = get_min_memlimit(cg); |
35482f91 | 2281 | if (!cgfs_get_value("memory", cg, "memory.usage_in_bytes", &memusage_str)) |
2c51f8dd | 2282 | goto err; |
35482f91 | 2283 | if (!cgfs_get_value("memory", cg, "memory.stat", &memstat_str)) |
2c51f8dd | 2284 | goto err; |
4622ad78 TG |
2285 | |
2286 | // Following values are allowed to fail, because swapaccount might be turned | |
2287 | // off for current kernel | |
2288 | if(cgfs_get_value("memory", cg, "memory.memsw.limit_in_bytes", &memswlimit_str) && | |
2289 | cgfs_get_value("memory", cg, "memory.memsw.usage_in_bytes", &memswusage_str)) | |
2290 | { | |
b731895e NW |
2291 | /* If swapaccounting is turned on, then default value is assumed to be that of cgroup / */ |
2292 | if (!cgfs_get_value("memory", "/", "memory.memsw.limit_in_bytes", &memswlimit_default_str)) | |
2293 | goto err; | |
2294 | if (!cgfs_get_value("memory", "/", "memory.memsw.usage_in_bytes", &memswusage_default_str)) | |
2295 | goto err; | |
2296 | ||
4622ad78 TG |
2297 | memswlimit = strtoul(memswlimit_str, NULL, 10); |
2298 | memswusage = strtoul(memswusage_str, NULL, 10); | |
b731895e NW |
2299 | |
2300 | if (!strcmp(memswlimit_str, memswlimit_default_str)) | |
a2de34ba | 2301 | memswlimit = 0; |
b731895e | 2302 | if (!strcmp(memswusage_str, memswusage_default_str)) |
a2de34ba SH |
2303 | memswusage = 0; |
2304 | ||
b731895e NW |
2305 | memswlimit = memswlimit / 1024; |
2306 | memswusage = memswusage / 1024; | |
4622ad78 | 2307 | } |
b731895e NW |
2308 | |
2309 | memusage = strtoul(memusage_str, NULL, 10); | |
2310 | memlimit /= 1024; | |
2311 | memusage /= 1024; | |
2312 | ||
2dc17609 SH |
2313 | get_mem_cached(memstat_str, &cached); |
2314 | ||
2315 | f = fopen("/proc/meminfo", "r"); | |
2316 | if (!f) | |
2c51f8dd | 2317 | goto err; |
2dc17609 SH |
2318 | |
2319 | while (getline(&line, &linelen, f) != -1) { | |
2320 | size_t l; | |
2321 | char *printme, lbuf[100]; | |
2322 | ||
2323 | memset(lbuf, 0, 100); | |
2324 | if (startswith(line, "MemTotal:")) { | |
2325 | sscanf(line+14, "%lu", &hosttotal); | |
2326 | if (hosttotal < memlimit) | |
2327 | memlimit = hosttotal; | |
2328 | snprintf(lbuf, 100, "MemTotal: %8lu kB\n", memlimit); | |
2329 | printme = lbuf; | |
2330 | } else if (startswith(line, "MemFree:")) { | |
2331 | snprintf(lbuf, 100, "MemFree: %8lu kB\n", memlimit - memusage); | |
2332 | printme = lbuf; | |
2333 | } else if (startswith(line, "MemAvailable:")) { | |
2334 | snprintf(lbuf, 100, "MemAvailable: %8lu kB\n", memlimit - memusage); | |
2335 | printme = lbuf; | |
4622ad78 TG |
2336 | } else if (startswith(line, "SwapTotal:") && memswlimit > 0) { |
2337 | snprintf(lbuf, 100, "SwapTotal: %8lu kB\n", memswlimit - memlimit); | |
2338 | printme = lbuf; | |
2339 | } else if (startswith(line, "SwapFree:") && memswlimit > 0 && memswusage > 0) { | |
2340 | snprintf(lbuf, 100, "SwapFree: %8lu kB\n", | |
2341 | (memswlimit - memlimit) - (memswusage - memusage)); | |
2342 | printme = lbuf; | |
2dc17609 SH |
2343 | } else if (startswith(line, "Buffers:")) { |
2344 | snprintf(lbuf, 100, "Buffers: %8lu kB\n", 0UL); | |
2345 | printme = lbuf; | |
2346 | } else if (startswith(line, "Cached:")) { | |
2347 | snprintf(lbuf, 100, "Cached: %8lu kB\n", cached); | |
2348 | printme = lbuf; | |
2349 | } else if (startswith(line, "SwapCached:")) { | |
2350 | snprintf(lbuf, 100, "SwapCached: %8lu kB\n", 0UL); | |
2351 | printme = lbuf; | |
2352 | } else | |
2353 | printme = line; | |
97f1f27b YY |
2354 | |
2355 | l = snprintf(cache, cache_size, "%s", printme); | |
e1068397 MM |
2356 | if (l < 0) { |
2357 | perror("Error writing to cache"); | |
2358 | rv = 0; | |
2359 | goto err; | |
2360 | ||
2361 | } | |
2362 | if (l >= cache_size) { | |
2363 | fprintf(stderr, "Internal error: truncated write to cache\n"); | |
2364 | rv = 0; | |
2365 | goto err; | |
2366 | } | |
2367 | ||
97f1f27b YY |
2368 | cache += l; |
2369 | cache_size -= l; | |
2f919d9d | 2370 | total_len += l; |
2dc17609 SH |
2371 | } |
2372 | ||
b5ad2d21 | 2373 | d->cached = 1; |
97f1f27b YY |
2374 | d->size = total_len; |
2375 | if (total_len > size ) total_len = size; | |
2376 | memcpy(buf, d->buf, total_len); | |
2377 | ||
e1068397 | 2378 | rv = total_len; |
2c51f8dd SH |
2379 | err: |
2380 | if (f) | |
2381 | fclose(f); | |
92c84dc4 | 2382 | free(line); |
2c51f8dd | 2383 | free(cg); |
2c51f8dd | 2384 | free(memusage_str); |
4622ad78 TG |
2385 | free(memswlimit_str); |
2386 | free(memswusage_str); | |
2c51f8dd | 2387 | free(memstat_str); |
b731895e NW |
2388 | free(memswlimit_default_str); |
2389 | free(memswusage_default_str); | |
e1068397 | 2390 | return rv; |
23ce2127 SH |
2391 | } |
2392 | ||
2393 | /* | |
2394 | * Read the cpuset.cpus for cg | |
2c51f8dd | 2395 | * Return the answer in a newly allocated string which must be freed |
23ce2127 SH |
2396 | */ |
2397 | static char *get_cpuset(const char *cg) | |
2398 | { | |
2399 | char *answer; | |
2400 | ||
35482f91 | 2401 | if (!cgfs_get_value("cpuset", cg, "cpuset.cpus", &answer)) |
23ce2127 SH |
2402 | return NULL; |
2403 | return answer; | |
2404 | } | |
2405 | ||
fa47bb52 | 2406 | bool cpu_in_cpuset(int cpu, const char *cpuset); |
23ce2127 | 2407 | |
aeb56147 SH |
2408 | static bool cpuline_in_cpuset(const char *line, const char *cpuset) |
2409 | { | |
2410 | int cpu; | |
2411 | ||
2412 | if (sscanf(line, "processor : %d", &cpu) != 1) | |
2413 | return false; | |
2414 | return cpu_in_cpuset(cpu, cpuset); | |
2415 | } | |
2416 | ||
23ce2127 SH |
2417 | /* |
2418 | * check whether this is a '^processor" line in /proc/cpuinfo | |
2419 | */ | |
2420 | static bool is_processor_line(const char *line) | |
2421 | { | |
2422 | int cpu; | |
2423 | ||
2424 | if (sscanf(line, "processor : %d", &cpu) == 1) | |
2425 | return true; | |
2426 | return false; | |
2427 | } | |
2428 | ||
23ce2127 SH |
2429 | static int proc_cpuinfo_read(char *buf, size_t size, off_t offset, |
2430 | struct fuse_file_info *fi) | |
2431 | { | |
2432 | struct fuse_context *fc = fuse_get_context(); | |
97f1f27b | 2433 | struct file_info *d = (struct file_info *)fi->fh; |
2c51f8dd SH |
2434 | char *cg; |
2435 | char *cpuset = NULL; | |
23ce2127 | 2436 | char *line = NULL; |
e1068397 | 2437 | size_t linelen = 0, total_len = 0, rv = 0; |
23ce2127 SH |
2438 | bool am_printing = false; |
2439 | int curcpu = -1; | |
97f1f27b YY |
2440 | char *cache = d->buf; |
2441 | size_t cache_size = d->buflen; | |
2c51f8dd | 2442 | FILE *f = NULL; |
23ce2127 | 2443 | |
97f1f27b YY |
2444 | if (offset){ |
2445 | if (offset > d->size) | |
2446 | return -EINVAL; | |
b5ad2d21 SH |
2447 | if (!d->cached) |
2448 | return 0; | |
97f1f27b YY |
2449 | int left = d->size - offset; |
2450 | total_len = left > size ? size: left; | |
2451 | memcpy(buf, cache + offset, total_len); | |
2f919d9d | 2452 | return total_len; |
97f1f27b | 2453 | } |
23ce2127 | 2454 | |
b11c6ec0 | 2455 | pid_t initpid = lookup_initpid_in_store(fc->pid); |
87dce5f6 SH |
2456 | if (initpid <= 0) |
2457 | initpid = fc->pid; | |
0dcc31ea | 2458 | cg = get_pid_cgroup(initpid, "cpuset"); |
23ce2127 | 2459 | if (!cg) |
53b43826 | 2460 | return read_file("proc/cpuinfo", buf, size, d); |
23ce2127 SH |
2461 | |
2462 | cpuset = get_cpuset(cg); | |
2463 | if (!cpuset) | |
2c51f8dd | 2464 | goto err; |
23ce2127 SH |
2465 | |
2466 | f = fopen("/proc/cpuinfo", "r"); | |
2467 | if (!f) | |
2c51f8dd | 2468 | goto err; |
23ce2127 SH |
2469 | |
2470 | while (getline(&line, &linelen, f) != -1) { | |
2471 | size_t l; | |
2472 | if (is_processor_line(line)) { | |
aeb56147 | 2473 | am_printing = cpuline_in_cpuset(line, cpuset); |
23ce2127 SH |
2474 | if (am_printing) { |
2475 | curcpu ++; | |
97f1f27b | 2476 | l = snprintf(cache, cache_size, "processor : %d\n", curcpu); |
e1068397 MM |
2477 | if (l < 0) { |
2478 | perror("Error writing to cache"); | |
2479 | rv = 0; | |
2480 | goto err; | |
2481 | } | |
2482 | if (l >= cache_size) { | |
2483 | fprintf(stderr, "Internal error: truncated write to cache\n"); | |
2484 | rv = 0; | |
2485 | goto err; | |
2486 | } | |
711b4030 SF |
2487 | cache += l; |
2488 | cache_size -= l; | |
2489 | total_len += l; | |
23ce2127 SH |
2490 | } |
2491 | continue; | |
2492 | } | |
2493 | if (am_printing) { | |
97f1f27b | 2494 | l = snprintf(cache, cache_size, "%s", line); |
e1068397 MM |
2495 | if (l < 0) { |
2496 | perror("Error writing to cache"); | |
2497 | rv = 0; | |
2498 | goto err; | |
2499 | } | |
2500 | if (l >= cache_size) { | |
2501 | fprintf(stderr, "Internal error: truncated write to cache\n"); | |
2502 | rv = 0; | |
2503 | goto err; | |
2504 | } | |
711b4030 SF |
2505 | cache += l; |
2506 | cache_size -= l; | |
2507 | total_len += l; | |
23ce2127 SH |
2508 | } |
2509 | } | |
2510 | ||
b5ad2d21 | 2511 | d->cached = 1; |
97f1f27b YY |
2512 | d->size = total_len; |
2513 | if (total_len > size ) total_len = size; | |
2514 | ||
2515 | /* read from off 0 */ | |
2516 | memcpy(buf, d->buf, total_len); | |
e1068397 | 2517 | rv = total_len; |
2c51f8dd SH |
2518 | err: |
2519 | if (f) | |
2520 | fclose(f); | |
92c84dc4 | 2521 | free(line); |
2c51f8dd SH |
2522 | free(cpuset); |
2523 | free(cg); | |
e1068397 | 2524 | return rv; |
23ce2127 SH |
2525 | } |
2526 | ||
2527 | static int proc_stat_read(char *buf, size_t size, off_t offset, | |
2528 | struct fuse_file_info *fi) | |
2529 | { | |
aeb56147 | 2530 | struct fuse_context *fc = fuse_get_context(); |
97f1f27b | 2531 | struct file_info *d = (struct file_info *)fi->fh; |
2c51f8dd SH |
2532 | char *cg; |
2533 | char *cpuset = NULL; | |
aeb56147 | 2534 | char *line = NULL; |
e1068397 | 2535 | size_t linelen = 0, total_len = 0, rv = 0; |
2a0fde62 | 2536 | int curcpu = -1; /* cpu numbering starts at 0 */ |
97f1f27b YY |
2537 | unsigned long user = 0, nice = 0, system = 0, idle = 0, iowait = 0, irq = 0, softirq = 0, steal = 0, guest = 0; |
2538 | unsigned long user_sum = 0, nice_sum = 0, system_sum = 0, idle_sum = 0, iowait_sum = 0, | |
2539 | irq_sum = 0, softirq_sum = 0, steal_sum = 0, guest_sum = 0; | |
2540 | #define CPUALL_MAX_SIZE BUF_RESERVE_SIZE | |
2541 | char cpuall[CPUALL_MAX_SIZE]; | |
2542 | /* reserve for cpu all */ | |
2543 | char *cache = d->buf + CPUALL_MAX_SIZE; | |
2544 | size_t cache_size = d->buflen - CPUALL_MAX_SIZE; | |
2c51f8dd | 2545 | FILE *f = NULL; |
aeb56147 | 2546 | |
97f1f27b YY |
2547 | if (offset){ |
2548 | if (offset > d->size) | |
2549 | return -EINVAL; | |
b5ad2d21 SH |
2550 | if (!d->cached) |
2551 | return 0; | |
97f1f27b YY |
2552 | int left = d->size - offset; |
2553 | total_len = left > size ? size: left; | |
2554 | memcpy(buf, d->buf + offset, total_len); | |
2f919d9d | 2555 | return total_len; |
97f1f27b | 2556 | } |
aeb56147 | 2557 | |
b11c6ec0 | 2558 | pid_t initpid = lookup_initpid_in_store(fc->pid); |
87dce5f6 SH |
2559 | if (initpid <= 0) |
2560 | initpid = fc->pid; | |
0dcc31ea | 2561 | cg = get_pid_cgroup(initpid, "cpuset"); |
aeb56147 | 2562 | if (!cg) |
53b43826 | 2563 | return read_file("/proc/stat", buf, size, d); |
aeb56147 SH |
2564 | |
2565 | cpuset = get_cpuset(cg); | |
2566 | if (!cpuset) | |
2c51f8dd | 2567 | goto err; |
aeb56147 SH |
2568 | |
2569 | f = fopen("/proc/stat", "r"); | |
2570 | if (!f) | |
2c51f8dd | 2571 | goto err; |
aeb56147 | 2572 | |
97f1f27b YY |
2573 | //skip first line |
2574 | if (getline(&line, &linelen, f) < 0) { | |
2575 | fprintf(stderr, "proc_stat_read read first line failed\n"); | |
2c51f8dd | 2576 | goto err; |
97f1f27b YY |
2577 | } |
2578 | ||
aeb56147 SH |
2579 | while (getline(&line, &linelen, f) != -1) { |
2580 | size_t l; | |
2581 | int cpu; | |
2a0fde62 | 2582 | char cpu_char[10]; /* That's a lot of cores */ |
aeb56147 SH |
2583 | char *c; |
2584 | ||
2a0fde62 CB |
2585 | if (sscanf(line, "cpu%9[^ ]", cpu_char) != 1) { |
2586 | /* not a ^cpuN line containing a number N, just print it */ | |
97f1f27b | 2587 | l = snprintf(cache, cache_size, "%s", line); |
e1068397 MM |
2588 | if (l < 0) { |
2589 | perror("Error writing to cache"); | |
2590 | rv = 0; | |
2591 | goto err; | |
2592 | } | |
2593 | if (l >= cache_size) { | |
2594 | fprintf(stderr, "Internal error: truncated write to cache\n"); | |
2595 | rv = 0; | |
2596 | goto err; | |
2597 | } | |
711b4030 SF |
2598 | cache += l; |
2599 | cache_size -= l; | |
2600 | total_len += l; | |
2601 | continue; | |
aeb56147 | 2602 | } |
2a0fde62 CB |
2603 | |
2604 | if (sscanf(cpu_char, "%d", &cpu) != 1) | |
2605 | continue; | |
aeb56147 SH |
2606 | if (!cpu_in_cpuset(cpu, cpuset)) |
2607 | continue; | |
2608 | curcpu ++; | |
2609 | ||
2610 | c = strchr(line, ' '); | |
2611 | if (!c) | |
2612 | continue; | |
25c5e8fb | 2613 | l = snprintf(cache, cache_size, "cpu%d%s", curcpu, c); |
e1068397 MM |
2614 | if (l < 0) { |
2615 | perror("Error writing to cache"); | |
2616 | rv = 0; | |
2617 | goto err; | |
2618 | ||
2619 | } | |
2620 | if (l >= cache_size) { | |
2621 | fprintf(stderr, "Internal error: truncated write to cache\n"); | |
2622 | rv = 0; | |
2623 | goto err; | |
2624 | } | |
2625 | ||
97f1f27b YY |
2626 | cache += l; |
2627 | cache_size -= l; | |
aeb56147 | 2628 | total_len += l; |
2f919d9d | 2629 | |
97f1f27b YY |
2630 | if (sscanf(line, "%*s %lu %lu %lu %lu %lu %lu %lu %lu %lu", &user, &nice, &system, &idle, &iowait, &irq, |
2631 | &softirq, &steal, &guest) != 9) | |
2632 | continue; | |
2633 | user_sum += user; | |
2634 | nice_sum += nice; | |
2635 | system_sum += system; | |
2636 | idle_sum += idle; | |
2637 | iowait_sum += iowait; | |
2638 | irq_sum += irq; | |
2639 | softirq_sum += softirq; | |
2640 | steal_sum += steal; | |
2f919d9d | 2641 | guest_sum += guest; |
97f1f27b YY |
2642 | } |
2643 | ||
2644 | cache = d->buf; | |
2645 | ||
2f919d9d | 2646 | int cpuall_len = snprintf(cpuall, CPUALL_MAX_SIZE, "%s %lu %lu %lu %lu %lu %lu %lu %lu %lu\n", |
97f1f27b YY |
2647 | "cpu ", user_sum, nice_sum, system_sum, idle_sum, iowait_sum, irq_sum, softirq_sum, steal_sum, guest_sum); |
2648 | if (cpuall_len > 0 && cpuall_len < CPUALL_MAX_SIZE){ | |
2649 | memcpy(cache, cpuall, cpuall_len); | |
2f919d9d | 2650 | cache += cpuall_len; |
2c51f8dd | 2651 | } else{ |
97f1f27b YY |
2652 | /* shouldn't happen */ |
2653 | fprintf(stderr, "proc_stat_read copy cpuall failed, cpuall_len=%d\n", cpuall_len); | |
2654 | cpuall_len = 0; | |
2655 | } | |
2656 | ||
2657 | memmove(cache, d->buf + CPUALL_MAX_SIZE, total_len); | |
2658 | total_len += cpuall_len; | |
b5ad2d21 | 2659 | d->cached = 1; |
97f1f27b YY |
2660 | d->size = total_len; |
2661 | if (total_len > size ) total_len = size; | |
2662 | ||
2663 | memcpy(buf, d->buf, total_len); | |
e1068397 | 2664 | rv = total_len; |
2c51f8dd SH |
2665 | |
2666 | err: | |
2667 | if (f) | |
2668 | fclose(f); | |
92c84dc4 | 2669 | free(line); |
2c51f8dd SH |
2670 | free(cpuset); |
2671 | free(cg); | |
e1068397 | 2672 | return rv; |
23ce2127 SH |
2673 | } |
2674 | ||
0afd85bd | 2675 | static long int getreaperage(pid_t pid) |
41bb9357 SH |
2676 | { |
2677 | char fnam[100]; | |
41bb9357 | 2678 | struct stat sb; |
0afd85bd SH |
2679 | int ret; |
2680 | pid_t qpid; | |
5ca64c2a | 2681 | |
b11c6ec0 | 2682 | qpid = lookup_initpid_in_store(pid); |
87dce5f6 | 2683 | if (qpid <= 0) |
c0adec85 SH |
2684 | return 0; |
2685 | ||
0afd85bd SH |
2686 | ret = snprintf(fnam, 100, "/proc/%d", qpid); |
2687 | if (ret < 0 || ret >= 100) | |
41bb9357 | 2688 | return 0; |
ea56f722 | 2689 | |
0afd85bd | 2690 | if (lstat(fnam, &sb) < 0) |
41bb9357 | 2691 | return 0; |
41bb9357 | 2692 | |
0afd85bd | 2693 | return time(NULL) - sb.st_ctime; |
41bb9357 SH |
2694 | } |
2695 | ||
0b6af11b SH |
2696 | static unsigned long get_reaper_busy(pid_t task) |
2697 | { | |
b11c6ec0 | 2698 | pid_t initpid = lookup_initpid_in_store(task); |
0b6af11b SH |
2699 | char *cgroup = NULL, *usage_str = NULL; |
2700 | unsigned long usage = 0; | |
2701 | ||
87dce5f6 | 2702 | if (initpid <= 0) |
41bb9357 | 2703 | return 0; |
0b6af11b | 2704 | |
0dcc31ea | 2705 | cgroup = get_pid_cgroup(initpid, "cpuacct"); |
0b6af11b SH |
2706 | if (!cgroup) |
2707 | goto out; | |
2708 | if (!cgfs_get_value("cpuacct", cgroup, "cpuacct.usage", &usage_str)) | |
2709 | goto out; | |
2710 | usage = strtoul(usage_str, NULL, 10); | |
87e96963 | 2711 | usage /= 1000000000; |
0b6af11b SH |
2712 | |
2713 | out: | |
2714 | free(cgroup); | |
2715 | free(usage_str); | |
2716 | return usage; | |
41bb9357 SH |
2717 | } |
2718 | ||
2719 | /* | |
2720 | * We read /proc/uptime and reuse its second field. | |
2721 | * For the first field, we use the mtime for the reaper for | |
2722 | * the calling pid as returned by getreaperage | |
2723 | */ | |
23ce2127 SH |
2724 | static int proc_uptime_read(char *buf, size_t size, off_t offset, |
2725 | struct fuse_file_info *fi) | |
2726 | { | |
41bb9357 | 2727 | struct fuse_context *fc = fuse_get_context(); |
97f1f27b | 2728 | struct file_info *d = (struct file_info *)fi->fh; |
ff96a5f9 | 2729 | long int reaperage = getreaperage(fc->pid); |
0b6af11b | 2730 | unsigned long int busytime = get_reaper_busy(fc->pid), idletime; |
b5ad2d21 | 2731 | char *cache = d->buf; |
97f1f27b | 2732 | size_t total_len = 0; |
41bb9357 | 2733 | |
97f1f27b YY |
2734 | if (offset){ |
2735 | if (offset > d->size) | |
2736 | return -EINVAL; | |
b5ad2d21 SH |
2737 | if (!d->cached) |
2738 | return 0; | |
2739 | int left = d->size - offset; | |
2740 | total_len = left > size ? size: left; | |
2741 | memcpy(buf, cache + offset, total_len); | |
2742 | return total_len; | |
97f1f27b YY |
2743 | } |
2744 | ||
0b6af11b | 2745 | idletime = reaperage - busytime; |
f6c0b279 SH |
2746 | if (idletime > reaperage) |
2747 | idletime = reaperage; | |
2748 | ||
b5ad2d21 | 2749 | total_len = snprintf(d->buf, d->size, "%ld.0 %lu.0\n", reaperage, idletime); |
e1068397 MM |
2750 | if (total_len < 0){ |
2751 | perror("Error writing to cache"); | |
2752 | return 0; | |
2753 | } | |
cdcdb29b | 2754 | |
b5ad2d21 SH |
2755 | d->size = (int)total_len; |
2756 | d->cached = 1; | |
2757 | ||
2758 | if (total_len > size) total_len = size; | |
2759 | ||
2760 | memcpy(buf, d->buf, total_len); | |
97f1f27b | 2761 | return total_len; |
23ce2127 SH |
2762 | } |
2763 | ||
49878439 YY |
2764 | static int proc_diskstats_read(char *buf, size_t size, off_t offset, |
2765 | struct fuse_file_info *fi) | |
2766 | { | |
2767 | char dev_name[72]; | |
2768 | struct fuse_context *fc = fuse_get_context(); | |
97f1f27b | 2769 | struct file_info *d = (struct file_info *)fi->fh; |
2c51f8dd SH |
2770 | char *cg; |
2771 | char *io_serviced_str = NULL, *io_merged_str = NULL, *io_service_bytes_str = NULL, | |
49878439 YY |
2772 | *io_wait_time_str = NULL, *io_service_time_str = NULL; |
2773 | unsigned long read = 0, write = 0; | |
2774 | unsigned long read_merged = 0, write_merged = 0; | |
2775 | unsigned long read_sectors = 0, write_sectors = 0; | |
2776 | unsigned long read_ticks = 0, write_ticks = 0; | |
2777 | unsigned long ios_pgr = 0, tot_ticks = 0, rq_ticks = 0; | |
2778 | unsigned long rd_svctm = 0, wr_svctm = 0, rd_wait = 0, wr_wait = 0; | |
b5ad2d21 SH |
2779 | char *cache = d->buf; |
2780 | size_t cache_size = d->buflen; | |
49878439 | 2781 | char *line = NULL; |
e1068397 | 2782 | size_t linelen = 0, total_len = 0, rv = 0; |
49878439 YY |
2783 | unsigned int major = 0, minor = 0; |
2784 | int i = 0; | |
2c51f8dd | 2785 | FILE *f = NULL; |
49878439 | 2786 | |
97f1f27b YY |
2787 | if (offset){ |
2788 | if (offset > d->size) | |
2789 | return -EINVAL; | |
b5ad2d21 SH |
2790 | if (!d->cached) |
2791 | return 0; | |
2792 | int left = d->size - offset; | |
2793 | total_len = left > size ? size: left; | |
2794 | memcpy(buf, cache + offset, total_len); | |
2795 | return total_len; | |
97f1f27b | 2796 | } |
49878439 | 2797 | |
b11c6ec0 | 2798 | pid_t initpid = lookup_initpid_in_store(fc->pid); |
87dce5f6 SH |
2799 | if (initpid <= 0) |
2800 | initpid = fc->pid; | |
0dcc31ea | 2801 | cg = get_pid_cgroup(initpid, "blkio"); |
49878439 | 2802 | if (!cg) |
53b43826 | 2803 | return read_file("/proc/diskstats", buf, size, d); |
49878439 | 2804 | |
35482f91 | 2805 | if (!cgfs_get_value("blkio", cg, "blkio.io_serviced", &io_serviced_str)) |
2c51f8dd | 2806 | goto err; |
35482f91 | 2807 | if (!cgfs_get_value("blkio", cg, "blkio.io_merged", &io_merged_str)) |
2c51f8dd | 2808 | goto err; |
35482f91 | 2809 | if (!cgfs_get_value("blkio", cg, "blkio.io_service_bytes", &io_service_bytes_str)) |
2c51f8dd | 2810 | goto err; |
35482f91 | 2811 | if (!cgfs_get_value("blkio", cg, "blkio.io_wait_time", &io_wait_time_str)) |
2c51f8dd | 2812 | goto err; |
35482f91 | 2813 | if (!cgfs_get_value("blkio", cg, "blkio.io_service_time", &io_service_time_str)) |
2c51f8dd | 2814 | goto err; |
49878439 YY |
2815 | |
2816 | ||
2817 | f = fopen("/proc/diskstats", "r"); | |
2818 | if (!f) | |
2c51f8dd | 2819 | goto err; |
49878439 YY |
2820 | |
2821 | while (getline(&line, &linelen, f) != -1) { | |
2822 | size_t l; | |
2823 | char *printme, lbuf[256]; | |
2824 | ||
c0adec85 | 2825 | i = sscanf(line, "%u %u %71s", &major, &minor, dev_name); |
49878439 YY |
2826 | if(i == 3){ |
2827 | get_blkio_io_value(io_serviced_str, major, minor, "Read", &read); | |
2828 | get_blkio_io_value(io_serviced_str, major, minor, "Write", &write); | |
2829 | get_blkio_io_value(io_merged_str, major, minor, "Read", &read_merged); | |
2830 | get_blkio_io_value(io_merged_str, major, minor, "Write", &write_merged); | |
2831 | get_blkio_io_value(io_service_bytes_str, major, minor, "Read", &read_sectors); | |
2832 | read_sectors = read_sectors/512; | |
2833 | get_blkio_io_value(io_service_bytes_str, major, minor, "Write", &write_sectors); | |
2834 | write_sectors = write_sectors/512; | |
2f919d9d | 2835 | |
49878439 YY |
2836 | get_blkio_io_value(io_service_time_str, major, minor, "Read", &rd_svctm); |
2837 | rd_svctm = rd_svctm/1000000; | |
2838 | get_blkio_io_value(io_wait_time_str, major, minor, "Read", &rd_wait); | |
2839 | rd_wait = rd_wait/1000000; | |
2840 | read_ticks = rd_svctm + rd_wait; | |
2841 | ||
2842 | get_blkio_io_value(io_service_time_str, major, minor, "Write", &wr_svctm); | |
2843 | wr_svctm = wr_svctm/1000000; | |
2844 | get_blkio_io_value(io_wait_time_str, major, minor, "Write", &wr_wait); | |
2845 | wr_wait = wr_wait/1000000; | |
2846 | write_ticks = wr_svctm + wr_wait; | |
2847 | ||
2848 | get_blkio_io_value(io_service_time_str, major, minor, "Total", &tot_ticks); | |
2849 | tot_ticks = tot_ticks/1000000; | |
2850 | }else{ | |
2851 | continue; | |
2852 | } | |
2853 | ||
2854 | memset(lbuf, 0, 256); | |
2855 | if (read || write || read_merged || write_merged || read_sectors || write_sectors || read_ticks || write_ticks) { | |
2f919d9d | 2856 | snprintf(lbuf, 256, "%u %u %s %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu\n", |
49878439 YY |
2857 | major, minor, dev_name, read, read_merged, read_sectors, read_ticks, |
2858 | write, write_merged, write_sectors, write_ticks, ios_pgr, tot_ticks, rq_ticks); | |
2859 | printme = lbuf; | |
2860 | } else | |
2861 | continue; | |
2862 | ||
b5ad2d21 | 2863 | l = snprintf(cache, cache_size, "%s", printme); |
e1068397 MM |
2864 | if (l < 0) { |
2865 | perror("Error writing to fuse buf"); | |
2866 | rv = 0; | |
2867 | goto err; | |
2868 | } | |
b5ad2d21 | 2869 | if (l >= cache_size) { |
e1068397 MM |
2870 | fprintf(stderr, "Internal error: truncated write to cache\n"); |
2871 | rv = 0; | |
2872 | goto err; | |
2873 | } | |
b5ad2d21 SH |
2874 | cache += l; |
2875 | cache_size -= l; | |
49878439 YY |
2876 | total_len += l; |
2877 | } | |
2878 | ||
b5ad2d21 | 2879 | d->cached = 1; |
97f1f27b | 2880 | d->size = total_len; |
b5ad2d21 SH |
2881 | if (total_len > size ) total_len = size; |
2882 | memcpy(buf, d->buf, total_len); | |
2883 | ||
e1068397 | 2884 | rv = total_len; |
2c51f8dd SH |
2885 | err: |
2886 | free(cg); | |
2887 | if (f) | |
2888 | fclose(f); | |
49878439 | 2889 | free(line); |
2c51f8dd SH |
2890 | free(io_serviced_str); |
2891 | free(io_merged_str); | |
2892 | free(io_service_bytes_str); | |
2893 | free(io_wait_time_str); | |
2894 | free(io_service_time_str); | |
e1068397 | 2895 | return rv; |
49878439 YY |
2896 | } |
2897 | ||
23ce2127 SH |
2898 | static off_t get_procfile_size(const char *which) |
2899 | { | |
2900 | FILE *f = fopen(which, "r"); | |
2901 | char *line = NULL; | |
2902 | size_t len = 0; | |
2903 | ssize_t sz, answer = 0; | |
2904 | if (!f) | |
2905 | return 0; | |
2906 | ||
2907 | while ((sz = getline(&line, &len, f)) != -1) | |
2908 | answer += sz; | |
2909 | fclose (f); | |
92c84dc4 | 2910 | free(line); |
23ce2127 SH |
2911 | |
2912 | return answer; | |
2913 | } | |
2914 | ||
758ad80c SH |
2915 | static int proc_getattr(const char *path, struct stat *sb) |
2916 | { | |
35629743 SH |
2917 | struct timespec now; |
2918 | ||
2919 | memset(sb, 0, sizeof(struct stat)); | |
2920 | if (clock_gettime(CLOCK_REALTIME, &now) < 0) | |
2921 | return -EINVAL; | |
2922 | sb->st_uid = sb->st_gid = 0; | |
2923 | sb->st_atim = sb->st_mtim = sb->st_ctim = now; | |
2924 | if (strcmp(path, "/proc") == 0) { | |
2925 | sb->st_mode = S_IFDIR | 00555; | |
2926 | sb->st_nlink = 2; | |
2927 | return 0; | |
2928 | } | |
2929 | if (strcmp(path, "/proc/meminfo") == 0 || | |
2930 | strcmp(path, "/proc/cpuinfo") == 0 || | |
2931 | strcmp(path, "/proc/uptime") == 0 || | |
49878439 YY |
2932 | strcmp(path, "/proc/stat") == 0 || |
2933 | strcmp(path, "/proc/diskstats") == 0) { | |
7253e0a4 | 2934 | sb->st_size = 0; |
35629743 SH |
2935 | sb->st_mode = S_IFREG | 00444; |
2936 | sb->st_nlink = 1; | |
2937 | return 0; | |
2938 | } | |
2939 | ||
2940 | return -ENOENT; | |
2941 | } | |
2942 | ||
2943 | static int proc_readdir(const char *path, void *buf, fuse_fill_dir_t filler, off_t offset, | |
2944 | struct fuse_file_info *fi) | |
2945 | { | |
2946 | if (filler(buf, "cpuinfo", NULL, 0) != 0 || | |
2947 | filler(buf, "meminfo", NULL, 0) != 0 || | |
2948 | filler(buf, "stat", NULL, 0) != 0 || | |
49878439 YY |
2949 | filler(buf, "uptime", NULL, 0) != 0 || |
2950 | filler(buf, "diskstats", NULL, 0) != 0) | |
758ad80c | 2951 | return -EINVAL; |
758ad80c SH |
2952 | return 0; |
2953 | } | |
2954 | ||
35629743 SH |
2955 | static int proc_open(const char *path, struct fuse_file_info *fi) |
2956 | { | |
96fc5ee6 SH |
2957 | int type = -1; |
2958 | struct file_info *info; | |
2959 | ||
2960 | if (strcmp(path, "/proc/meminfo") == 0) | |
2961 | type = LXC_TYPE_PROC_MEMINFO; | |
2962 | else if (strcmp(path, "/proc/cpuinfo") == 0) | |
2963 | type = LXC_TYPE_PROC_CPUINFO; | |
2964 | else if (strcmp(path, "/proc/uptime") == 0) | |
2965 | type = LXC_TYPE_PROC_UPTIME; | |
2966 | else if (strcmp(path, "/proc/stat") == 0) | |
2967 | type = LXC_TYPE_PROC_STAT; | |
2968 | else if (strcmp(path, "/proc/diskstats") == 0) | |
2969 | type = LXC_TYPE_PROC_DISKSTATS; | |
2970 | if (type == -1) | |
2971 | return -ENOENT; | |
2972 | ||
2c51f8dd SH |
2973 | info = malloc(sizeof(*info)); |
2974 | if (!info) | |
2975 | return -ENOMEM; | |
2976 | ||
96fc5ee6 SH |
2977 | memset(info, 0, sizeof(*info)); |
2978 | info->type = type; | |
2979 | ||
97f1f27b | 2980 | info->buflen = get_procfile_size(path) + BUF_RESERVE_SIZE; |
2c51f8dd SH |
2981 | do { |
2982 | info->buf = malloc(info->buflen); | |
2983 | } while (!info->buf); | |
97f1f27b YY |
2984 | memset(info->buf, 0, info->buflen); |
2985 | /* set actual size to buffer size */ | |
2f919d9d | 2986 | info->size = info->buflen; |
97f1f27b | 2987 | |
96fc5ee6 SH |
2988 | fi->fh = (unsigned long)info; |
2989 | return 0; | |
2990 | } | |
2991 | ||
2992 | static int proc_release(const char *path, struct fuse_file_info *fi) | |
2993 | { | |
2994 | struct file_info *f = (struct file_info *)fi->fh; | |
2995 | ||
2996 | do_release_file_info(f); | |
2997 | return 0; | |
35629743 SH |
2998 | } |
2999 | ||
35629743 SH |
3000 | static int proc_read(const char *path, char *buf, size_t size, off_t offset, |
3001 | struct fuse_file_info *fi) | |
3002 | { | |
96fc5ee6 SH |
3003 | struct file_info *f = (struct file_info *) fi->fh; |
3004 | ||
3005 | switch (f->type) { | |
2f919d9d | 3006 | case LXC_TYPE_PROC_MEMINFO: |
23ce2127 | 3007 | return proc_meminfo_read(buf, size, offset, fi); |
96fc5ee6 | 3008 | case LXC_TYPE_PROC_CPUINFO: |
23ce2127 | 3009 | return proc_cpuinfo_read(buf, size, offset, fi); |
96fc5ee6 | 3010 | case LXC_TYPE_PROC_UPTIME: |
23ce2127 | 3011 | return proc_uptime_read(buf, size, offset, fi); |
96fc5ee6 | 3012 | case LXC_TYPE_PROC_STAT: |
23ce2127 | 3013 | return proc_stat_read(buf, size, offset, fi); |
96fc5ee6 | 3014 | case LXC_TYPE_PROC_DISKSTATS: |
49878439 | 3015 | return proc_diskstats_read(buf, size, offset, fi); |
96fc5ee6 SH |
3016 | default: |
3017 | return -EINVAL; | |
3018 | } | |
35629743 SH |
3019 | } |
3020 | ||
2ad6d2bd SH |
3021 | /* |
3022 | * FUSE ops for / | |
3023 | * these just delegate to the /proc and /cgroup ops as | |
3024 | * needed | |
3025 | */ | |
758ad80c SH |
3026 | |
3027 | static int lxcfs_getattr(const char *path, struct stat *sb) | |
3028 | { | |
3029 | if (strcmp(path, "/") == 0) { | |
3030 | sb->st_mode = S_IFDIR | 00755; | |
3031 | sb->st_nlink = 2; | |
3032 | return 0; | |
3033 | } | |
3034 | if (strncmp(path, "/cgroup", 7) == 0) { | |
3035 | return cg_getattr(path, sb); | |
3036 | } | |
35629743 | 3037 | if (strncmp(path, "/proc", 5) == 0) { |
758ad80c SH |
3038 | return proc_getattr(path, sb); |
3039 | } | |
3040 | return -EINVAL; | |
3041 | } | |
3042 | ||
3043 | static int lxcfs_opendir(const char *path, struct fuse_file_info *fi) | |
3044 | { | |
3045 | if (strcmp(path, "/") == 0) | |
3046 | return 0; | |
3047 | ||
3048 | if (strncmp(path, "/cgroup", 7) == 0) { | |
3049 | return cg_opendir(path, fi); | |
3050 | } | |
35629743 SH |
3051 | if (strcmp(path, "/proc") == 0) |
3052 | return 0; | |
3053 | return -ENOENT; | |
758ad80c SH |
3054 | } |
3055 | ||
3056 | static int lxcfs_readdir(const char *path, void *buf, fuse_fill_dir_t filler, off_t offset, | |
3057 | struct fuse_file_info *fi) | |
3058 | { | |
3059 | if (strcmp(path, "/") == 0) { | |
3060 | if (filler(buf, "proc", NULL, 0) != 0 || | |
3061 | filler(buf, "cgroup", NULL, 0) != 0) | |
3062 | return -EINVAL; | |
3063 | return 0; | |
3064 | } | |
35629743 | 3065 | if (strncmp(path, "/cgroup", 7) == 0) |
758ad80c | 3066 | return cg_readdir(path, buf, filler, offset, fi); |
35629743 SH |
3067 | if (strcmp(path, "/proc") == 0) |
3068 | return proc_readdir(path, buf, filler, offset, fi); | |
758ad80c SH |
3069 | return -EINVAL; |
3070 | } | |
3071 | ||
3072 | static int lxcfs_releasedir(const char *path, struct fuse_file_info *fi) | |
3073 | { | |
3074 | if (strcmp(path, "/") == 0) | |
3075 | return 0; | |
3076 | if (strncmp(path, "/cgroup", 7) == 0) { | |
3077 | return cg_releasedir(path, fi); | |
3078 | } | |
35629743 SH |
3079 | if (strcmp(path, "/proc") == 0) |
3080 | return 0; | |
758ad80c SH |
3081 | return -EINVAL; |
3082 | } | |
3083 | ||
99978832 SH |
3084 | static int lxcfs_open(const char *path, struct fuse_file_info *fi) |
3085 | { | |
35629743 | 3086 | if (strncmp(path, "/cgroup", 7) == 0) |
99978832 | 3087 | return cg_open(path, fi); |
35629743 SH |
3088 | if (strncmp(path, "/proc", 5) == 0) |
3089 | return proc_open(path, fi); | |
99978832 SH |
3090 | |
3091 | return -EINVAL; | |
3092 | } | |
3093 | ||
3094 | static int lxcfs_read(const char *path, char *buf, size_t size, off_t offset, | |
3095 | struct fuse_file_info *fi) | |
3096 | { | |
35629743 | 3097 | if (strncmp(path, "/cgroup", 7) == 0) |
99978832 | 3098 | return cg_read(path, buf, size, offset, fi); |
35629743 SH |
3099 | if (strncmp(path, "/proc", 5) == 0) |
3100 | return proc_read(path, buf, size, offset, fi); | |
99978832 SH |
3101 | |
3102 | return -EINVAL; | |
3103 | } | |
3104 | ||
2ad6d2bd SH |
3105 | int lxcfs_write(const char *path, const char *buf, size_t size, off_t offset, |
3106 | struct fuse_file_info *fi) | |
3107 | { | |
3108 | if (strncmp(path, "/cgroup", 7) == 0) { | |
3109 | return cg_write(path, buf, size, offset, fi); | |
3110 | } | |
3111 | ||
3112 | return -EINVAL; | |
3113 | } | |
3114 | ||
99978832 SH |
3115 | static int lxcfs_flush(const char *path, struct fuse_file_info *fi) |
3116 | { | |
3117 | return 0; | |
3118 | } | |
3119 | ||
3120 | static int lxcfs_release(const char *path, struct fuse_file_info *fi) | |
758ad80c | 3121 | { |
8f6e8f5e SH |
3122 | if (strncmp(path, "/cgroup", 7) == 0) |
3123 | return cg_release(path, fi); | |
8f6e8f5e | 3124 | if (strncmp(path, "/proc", 5) == 0) |
96fc5ee6 | 3125 | return proc_release(path, fi); |
8f6e8f5e SH |
3126 | |
3127 | return -EINVAL; | |
99978832 SH |
3128 | } |
3129 | ||
3130 | static int lxcfs_fsync(const char *path, int datasync, struct fuse_file_info *fi) | |
3131 | { | |
3132 | return 0; | |
758ad80c SH |
3133 | } |
3134 | ||
ab54b798 SH |
3135 | int lxcfs_mkdir(const char *path, mode_t mode) |
3136 | { | |
3137 | if (strncmp(path, "/cgroup", 7) == 0) | |
3138 | return cg_mkdir(path, mode); | |
3139 | ||
3140 | return -EINVAL; | |
3141 | } | |
3142 | ||
341b21ad SH |
3143 | int lxcfs_chown(const char *path, uid_t uid, gid_t gid) |
3144 | { | |
3145 | if (strncmp(path, "/cgroup", 7) == 0) | |
3146 | return cg_chown(path, uid, gid); | |
3147 | ||
3148 | return -EINVAL; | |
3149 | } | |
3150 | ||
2ad6d2bd SH |
3151 | /* |
3152 | * cat first does a truncate before doing ops->write. This doesn't | |
3153 | * really make sense for cgroups. So just return 0 always but do | |
3154 | * nothing. | |
3155 | */ | |
3156 | int lxcfs_truncate(const char *path, off_t newsize) | |
3157 | { | |
3158 | if (strncmp(path, "/cgroup", 7) == 0) | |
3159 | return 0; | |
3160 | return -EINVAL; | |
3161 | } | |
3162 | ||
50d8d5b5 SH |
3163 | int lxcfs_rmdir(const char *path) |
3164 | { | |
3165 | if (strncmp(path, "/cgroup", 7) == 0) | |
3166 | return cg_rmdir(path); | |
3167 | return -EINVAL; | |
3168 | } | |
3169 | ||
fd2e4e03 SH |
3170 | int lxcfs_chmod(const char *path, mode_t mode) |
3171 | { | |
3172 | if (strncmp(path, "/cgroup", 7) == 0) | |
3173 | return cg_chmod(path, mode); | |
3174 | return -EINVAL; | |
3175 | } | |
3176 | ||
758ad80c SH |
3177 | const struct fuse_operations lxcfs_ops = { |
3178 | .getattr = lxcfs_getattr, | |
3179 | .readlink = NULL, | |
3180 | .getdir = NULL, | |
3181 | .mknod = NULL, | |
ab54b798 | 3182 | .mkdir = lxcfs_mkdir, |
758ad80c | 3183 | .unlink = NULL, |
50d8d5b5 | 3184 | .rmdir = lxcfs_rmdir, |
758ad80c SH |
3185 | .symlink = NULL, |
3186 | .rename = NULL, | |
3187 | .link = NULL, | |
fd2e4e03 | 3188 | .chmod = lxcfs_chmod, |
341b21ad | 3189 | .chown = lxcfs_chown, |
2ad6d2bd | 3190 | .truncate = lxcfs_truncate, |
758ad80c | 3191 | .utime = NULL, |
99978832 SH |
3192 | |
3193 | .open = lxcfs_open, | |
3194 | .read = lxcfs_read, | |
3195 | .release = lxcfs_release, | |
2ad6d2bd | 3196 | .write = lxcfs_write, |
99978832 | 3197 | |
758ad80c | 3198 | .statfs = NULL, |
99978832 SH |
3199 | .flush = lxcfs_flush, |
3200 | .fsync = lxcfs_fsync, | |
758ad80c SH |
3201 | |
3202 | .setxattr = NULL, | |
3203 | .getxattr = NULL, | |
3204 | .listxattr = NULL, | |
3205 | .removexattr = NULL, | |
3206 | ||
3207 | .opendir = lxcfs_opendir, | |
3208 | .readdir = lxcfs_readdir, | |
3209 | .releasedir = lxcfs_releasedir, | |
3210 | ||
3211 | .fsyncdir = NULL, | |
3212 | .init = NULL, | |
3213 | .destroy = NULL, | |
3214 | .access = NULL, | |
3215 | .create = NULL, | |
3216 | .ftruncate = NULL, | |
3217 | .fgetattr = NULL, | |
3218 | }; | |
3219 | ||
99978832 | 3220 | static void usage(const char *me) |
758ad80c SH |
3221 | { |
3222 | fprintf(stderr, "Usage:\n"); | |
3223 | fprintf(stderr, "\n"); | |
0b0f73db SH |
3224 | fprintf(stderr, "%s mountpoint\n", me); |
3225 | fprintf(stderr, "%s -h\n", me); | |
758ad80c SH |
3226 | exit(1); |
3227 | } | |
3228 | ||
99978832 | 3229 | static bool is_help(char *w) |
758ad80c SH |
3230 | { |
3231 | if (strcmp(w, "-h") == 0 || | |
3232 | strcmp(w, "--help") == 0 || | |
3233 | strcmp(w, "-help") == 0 || | |
3234 | strcmp(w, "help") == 0) | |
3235 | return true; | |
3236 | return false; | |
3237 | } | |
3238 | ||
0b0f73db SH |
3239 | void swallow_arg(int *argcp, char *argv[], char *which) |
3240 | { | |
3241 | int i; | |
3242 | ||
3243 | for (i = 1; argv[i]; i++) { | |
3244 | if (strcmp(argv[i], which) != 0) | |
3245 | continue; | |
3246 | for (; argv[i]; i++) { | |
3247 | argv[i] = argv[i+1]; | |
3248 | } | |
3249 | (*argcp)--; | |
3250 | return; | |
3251 | } | |
3252 | } | |
3253 | ||
3254 | void swallow_option(int *argcp, char *argv[], char *opt, char *v) | |
3255 | { | |
3256 | int i; | |
3257 | ||
3258 | for (i = 1; argv[i]; i++) { | |
3259 | if (!argv[i+1]) | |
3260 | continue; | |
3261 | if (strcmp(argv[i], opt) != 0) | |
3262 | continue; | |
3263 | if (strcmp(argv[i+1], v) != 0) { | |
3264 | fprintf(stderr, "Warning: unexpected fuse option %s\n", v); | |
3265 | exit(1); | |
3266 | } | |
3267 | for (; argv[i+1]; i++) { | |
3268 | argv[i] = argv[i+2]; | |
3269 | } | |
3270 | (*argcp) -= 2; | |
3271 | return; | |
3272 | } | |
3273 | } | |
3274 | ||
758ad80c SH |
3275 | int main(int argc, char *argv[]) |
3276 | { | |
c0adec85 | 3277 | int ret = -1; |
0b0f73db SH |
3278 | /* |
3279 | * what we pass to fuse_main is: | |
3280 | * argv[0] -s -f -o allow_other,directio argv[1] NULL | |
3281 | */ | |
2c51f8dd SH |
3282 | int nargs = 5, cnt = 0; |
3283 | char *newargv[6]; | |
758ad80c | 3284 | |
977ac879 | 3285 | #ifdef FORTRAVIS |
df062bcb SH |
3286 | /* for travis which runs on 12.04 */ |
3287 | if (glib_check_version (2, 36, 0) != NULL) | |
3288 | g_type_init (); | |
977ac879 | 3289 | #endif |
df062bcb | 3290 | |
0b0f73db SH |
3291 | /* accomodate older init scripts */ |
3292 | swallow_arg(&argc, argv, "-s"); | |
3293 | swallow_arg(&argc, argv, "-f"); | |
3294 | swallow_option(&argc, argv, "-o", "allow_other"); | |
3295 | ||
2e9c0b32 SH |
3296 | if (argc == 2 && strcmp(argv[1], "--version") == 0) { |
3297 | fprintf(stderr, "%s\n", VERSION); | |
3298 | exit(0); | |
3299 | } | |
0b0f73db | 3300 | if (argc != 2 || is_help(argv[1])) |
758ad80c SH |
3301 | usage(argv[0]); |
3302 | ||
38a76a91 | 3303 | newargv[cnt++] = argv[0]; |
38a76a91 SH |
3304 | newargv[cnt++] = "-f"; |
3305 | newargv[cnt++] = "-o"; | |
f466a31e | 3306 | newargv[cnt++] = "allow_other,direct_io,entry_timeout=0.5,attr_timeout=0.5"; |
38a76a91 SH |
3307 | newargv[cnt++] = argv[1]; |
3308 | newargv[cnt++] = NULL; | |
758ad80c | 3309 | |
35482f91 | 3310 | if (!cgfs_setup_controllers()) |
c0adec85 | 3311 | goto out; |
758ad80c | 3312 | |
35482f91 | 3313 | ret = fuse_main(nargs, newargv, &lxcfs_ops, NULL); |
758ad80c | 3314 | |
c0adec85 | 3315 | out: |
758ad80c | 3316 | return ret; |
2183082c | 3317 | } |