]>
Commit | Line | Data |
---|---|---|
758ad80c SH |
1 | /* lxcfs |
2 | * | |
2c51f8dd | 3 | * Copyright © 2014,2015 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 SH |
22 | #include <sched.h> |
23 | #include <linux/sched.h> | |
a05660a6 | 24 | #include <sys/socket.h> |
41bb9357 SH |
25 | #include <sys/mount.h> |
26 | #include <wait.h> | |
758ad80c | 27 | |
758ad80c | 28 | #include "cgmanager.h" |
2e9c0b32 | 29 | #include "config.h" // for VERSION |
758ad80c SH |
30 | |
31 | struct lxcfs_state { | |
32 | /* | |
2c51f8dd | 33 | * a null-terminated list of the mounted subsystems. We |
758ad80c SH |
34 | * detect this at startup. |
35 | */ | |
36 | char **subsystems; | |
37 | }; | |
38 | #define LXCFS_DATA ((struct lxcfs_state *) fuse_get_context()->private_data) | |
39 | ||
443d13f5 SH |
40 | enum { |
41 | LXC_TYPE_CGDIR, | |
42 | LXC_TYPE_CGFILE, | |
43 | LXC_TYPE_PROC_MEMINFO, | |
44 | LXC_TYPE_PROC_CPUINFO, | |
45 | LXC_TYPE_PROC_UPTIME, | |
46 | LXC_TYPE_PROC_STAT, | |
47 | LXC_TYPE_PROC_DISKSTATS, | |
48 | }; | |
49 | ||
c688e1b3 SH |
50 | struct file_info { |
51 | char *controller; | |
52 | char *cgroup; | |
8f6e8f5e | 53 | char *file; |
443d13f5 | 54 | int type; |
c688e1b3 SH |
55 | char *buf; // unused as of yet |
56 | int buflen; | |
97f1f27b | 57 | int size; //actual data size |
c688e1b3 SH |
58 | }; |
59 | ||
97f1f27b YY |
60 | /* reserve buffer size, for cpuall in /proc/stat */ |
61 | #define BUF_RESERVE_SIZE 256 | |
62 | ||
2c51f8dd SH |
63 | /* |
64 | * append pid to *src. | |
65 | * src: a pointer to a char* in which ot append the pid. | |
66 | * sz: the number of characters printed so far, minus trailing \0. | |
67 | * asz: the allocated size so far | |
68 | * pid: the pid to append | |
69 | */ | |
70 | static void must_strcat_pid(char **src, size_t *sz, size_t *asz, pid_t pid) | |
71 | { | |
72 | char *d = *src; | |
73 | char tmp[30]; | |
74 | ||
75 | sprintf(tmp, "%d\n", (int)pid); | |
76 | ||
77 | if (!d) { | |
78 | do { | |
79 | d = malloc(BUF_RESERVE_SIZE); | |
80 | } while (!d); | |
81 | *src = d; | |
82 | *asz = BUF_RESERVE_SIZE; | |
83 | } else if (strlen(tmp) + sz + 1 >= asz) { | |
84 | do { | |
85 | d = realloc(d, *asz + BUF_RESERVE_SIZE); | |
86 | } while (!d); | |
87 | *src = d; | |
88 | *asz += BUF_RESERVE_SIZE; | |
89 | } | |
90 | memcpy(d+*sz, tmp, strlen(tmp)); | |
91 | *sz += strlen(tmp); | |
92 | d[*sz] = '\0'; | |
93 | } | |
94 | ||
bae07053 | 95 | static char *must_copy_string(void *parent, const char *str) |
c688e1b3 | 96 | { |
2c51f8dd | 97 | char *dup = NULL; |
c688e1b3 SH |
98 | if (!str) |
99 | return NULL; | |
2c51f8dd SH |
100 | do { |
101 | dup = strdup(str); | |
102 | } while (!dup); | |
103 | ||
104 | return dup; | |
c688e1b3 SH |
105 | } |
106 | ||
a05660a6 SH |
107 | static int wait_for_pid(pid_t pid) |
108 | { | |
109 | int status, ret; | |
110 | ||
111 | again: | |
112 | ret = waitpid(pid, &status, 0); | |
113 | if (ret == -1) { | |
114 | if (errno == EINTR) | |
115 | goto again; | |
116 | return -1; | |
117 | } | |
118 | if (ret != pid) | |
119 | goto again; | |
120 | if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) | |
121 | return -1; | |
122 | return 0; | |
123 | } | |
124 | ||
053a659d SH |
125 | /* |
126 | * Given a open file * to /proc/pid/{u,g}id_map, and an id | |
127 | * valid in the caller's namespace, return the id mapped into | |
128 | * pid's namespace. | |
129 | * Returns the mapped id, or -1 on error. | |
130 | */ | |
131 | unsigned int | |
132 | convert_id_to_ns(FILE *idfile, unsigned int in_id) | |
133 | { | |
134 | unsigned int nsuid, // base id for a range in the idfile's namespace | |
135 | hostuid, // base id for a range in the caller's namespace | |
136 | count; // number of ids in this range | |
137 | char line[400]; | |
138 | int ret; | |
139 | ||
140 | fseek(idfile, 0L, SEEK_SET); | |
141 | while (fgets(line, 400, idfile)) { | |
142 | ret = sscanf(line, "%u %u %u\n", &nsuid, &hostuid, &count); | |
143 | if (ret != 3) | |
144 | continue; | |
145 | if (hostuid + count < hostuid || nsuid + count < nsuid) { | |
146 | /* | |
147 | * uids wrapped around - unexpected as this is a procfile, | |
148 | * so just bail. | |
149 | */ | |
647c89e5 | 150 | fprintf(stderr, "pid wrapparound at entry %u %u %u in %s\n", |
053a659d SH |
151 | nsuid, hostuid, count, line); |
152 | return -1; | |
153 | } | |
154 | if (hostuid <= in_id && hostuid+count > in_id) { | |
155 | /* | |
156 | * now since hostuid <= in_id < hostuid+count, and | |
157 | * hostuid+count and nsuid+count do not wrap around, | |
158 | * we know that nsuid+(in_id-hostuid) which must be | |
159 | * less that nsuid+(count) must not wrap around | |
160 | */ | |
161 | return (in_id - hostuid) + nsuid; | |
162 | } | |
163 | } | |
164 | ||
165 | // no answer found | |
166 | return -1; | |
167 | } | |
168 | ||
341b21ad SH |
169 | /* |
170 | * for is_privileged_over, | |
171 | * specify whether we require the calling uid to be root in his | |
172 | * namespace | |
173 | */ | |
174 | #define NS_ROOT_REQD true | |
175 | #define NS_ROOT_OPT false | |
176 | ||
2c51f8dd SH |
177 | #define PROCLEN 100 |
178 | ||
341b21ad | 179 | static bool is_privileged_over(pid_t pid, uid_t uid, uid_t victim, bool req_ns_root) |
758ad80c | 180 | { |
2c51f8dd SH |
181 | char fpath[PROCLEN]; |
182 | int ret; | |
053a659d SH |
183 | bool answer = false; |
184 | uid_t nsuid; | |
185 | ||
341b21ad SH |
186 | if (victim == -1 || uid == -1) |
187 | return false; | |
188 | ||
189 | /* | |
190 | * If the request is one not requiring root in the namespace, | |
191 | * then having the same uid suffices. (i.e. uid 1000 has write | |
192 | * access to files owned by uid 1000 | |
193 | */ | |
194 | if (!req_ns_root && uid == victim) | |
758ad80c SH |
195 | return true; |
196 | ||
2c51f8dd SH |
197 | ret = snprintf(fpath, PROCLEN, "/proc/%d/uid_map", pid); |
198 | if (ret < 0 || ret >= PROCLEN) | |
199 | return false; | |
053a659d SH |
200 | FILE *f = fopen(fpath, "r"); |
201 | if (!f) | |
202 | return false; | |
203 | ||
341b21ad | 204 | /* if caller's not root in his namespace, reject */ |
053a659d SH |
205 | nsuid = convert_id_to_ns(f, uid); |
206 | if (nsuid) | |
207 | goto out; | |
208 | ||
341b21ad SH |
209 | /* |
210 | * If victim is not mapped into caller's ns, reject. | |
211 | * XXX I'm not sure this check is needed given that fuse | |
212 | * will be sending requests where the vfs has converted | |
213 | */ | |
053a659d SH |
214 | nsuid = convert_id_to_ns(f, victim); |
215 | if (nsuid == -1) | |
216 | goto out; | |
217 | ||
218 | answer = true; | |
219 | ||
220 | out: | |
221 | fclose(f); | |
222 | return answer; | |
758ad80c SH |
223 | } |
224 | ||
225 | static bool perms_include(int fmode, mode_t req_mode) | |
226 | { | |
2ad6d2bd SH |
227 | mode_t r; |
228 | ||
229 | switch (req_mode & O_ACCMODE) { | |
230 | case O_RDONLY: | |
231 | r = S_IROTH; | |
232 | break; | |
233 | case O_WRONLY: | |
234 | r = S_IWOTH; | |
235 | break; | |
236 | case O_RDWR: | |
237 | r = S_IROTH | S_IWOTH; | |
238 | break; | |
239 | default: | |
240 | return false; | |
241 | } | |
242 | return ((fmode & r) == r); | |
758ad80c SH |
243 | } |
244 | ||
3db25a35 SH |
245 | static char *get_next_cgroup_dir(const char *taskcg, const char *querycg) |
246 | { | |
247 | char *start, *end; | |
248 | ||
249 | if (strlen(taskcg) <= strlen(querycg)) { | |
250 | fprintf(stderr, "%s: I was fed bad input\n", __func__); | |
251 | return NULL; | |
252 | } | |
253 | ||
254 | if (strcmp(querycg, "/") == 0) | |
2c51f8dd | 255 | start = strdup(taskcg + 1); |
3db25a35 | 256 | else |
2c51f8dd SH |
257 | start = strdup(taskcg + strlen(querycg) + 1); |
258 | if (!start) | |
259 | return NULL; | |
3db25a35 SH |
260 | end = strchr(start, '/'); |
261 | if (end) | |
262 | *end = '\0'; | |
263 | return start; | |
264 | } | |
265 | ||
2c51f8dd SH |
266 | static void stripnewline(char *x) |
267 | { | |
268 | size_t l = strlen(x); | |
269 | if (l && x[l-1] == '\n') | |
270 | x[l-1] = '\0'; | |
271 | } | |
272 | ||
273 | static char *get_pid_cgroup(pid_t pid, const char *contrl) | |
274 | { | |
275 | char fnam[PROCLEN]; | |
276 | FILE *f; | |
277 | char *answer = NULL; | |
278 | char *line = NULL; | |
279 | size_t len = 0; | |
280 | int ret; | |
281 | ||
282 | ret = snprintf(fnam, PROCLEN, "/proc/%d/cgroup", pid); | |
283 | if (ret < 0 || ret >= PROCLEN) | |
284 | return NULL; | |
285 | if (!(f = fopen(fnam, "r"))) | |
286 | return NULL; | |
287 | ||
288 | while (getline(&line, &len, f) != -1) { | |
289 | char *c1, *c2; | |
290 | if (!line[0]) | |
291 | continue; | |
292 | c1 = strchr(line, ':'); | |
293 | if (!c1) | |
294 | goto out; | |
295 | c1++; | |
296 | c2 = strchr(c1, ':'); | |
297 | if (!c2) | |
298 | goto out; | |
299 | *c2 = '\0'; | |
300 | if (strcmp(c1, contrl) != 0) | |
301 | continue; | |
302 | c2++; | |
303 | stripnewline(c2); | |
304 | do { | |
305 | answer = strdup(c2); | |
306 | } while (!answer); | |
307 | break; | |
308 | } | |
309 | ||
310 | out: | |
311 | fclose(f); | |
312 | free(line); | |
313 | return answer; | |
314 | } | |
315 | ||
758ad80c SH |
316 | /* |
317 | * check whether a fuse context may access a cgroup dir or file | |
318 | * | |
319 | * If file is not null, it is a cgroup file to check under cg. | |
320 | * If file is null, then we are checking perms on cg itself. | |
321 | * | |
322 | * For files we can check the mode of the list_keys result. | |
323 | * For cgroups, we must make assumptions based on the files under the | |
324 | * cgroup, because cgmanager doesn't tell us ownership/perms of cgroups | |
325 | * yet. | |
326 | */ | |
327 | static bool fc_may_access(struct fuse_context *fc, const char *contrl, const char *cg, const char *file, mode_t mode) | |
328 | { | |
2c51f8dd SH |
329 | struct cgm_keys **list = NULL; |
330 | bool ret = false; | |
758ad80c SH |
331 | int i; |
332 | ||
333 | if (!file) | |
334 | file = "tasks"; | |
335 | ||
336 | if (*file == '/') | |
337 | file++; | |
338 | ||
339 | if (!cgm_list_keys(contrl, cg, &list)) | |
340 | return false; | |
341 | for (i = 0; list[i]; i++) { | |
342 | if (strcmp(list[i]->name, file) == 0) { | |
343 | struct cgm_keys *k = list[i]; | |
341b21ad | 344 | if (is_privileged_over(fc->pid, fc->uid, k->uid, NS_ROOT_OPT)) { |
2c51f8dd SH |
345 | if (perms_include(k->mode >> 6, mode)) { |
346 | ret = true; | |
347 | goto out; | |
348 | } | |
758ad80c SH |
349 | } |
350 | if (fc->gid == k->gid) { | |
2c51f8dd SH |
351 | if (perms_include(k->mode >> 3, mode)) { |
352 | ret = true; | |
353 | goto out; | |
354 | } | |
758ad80c | 355 | } |
2c51f8dd SH |
356 | ret = perms_include(k->mode, mode); |
357 | goto out; | |
758ad80c SH |
358 | } |
359 | } | |
360 | ||
2c51f8dd SH |
361 | out: |
362 | free_keys(list); | |
363 | return ret; | |
3db25a35 SH |
364 | } |
365 | ||
04b5cbdc SH |
366 | #define INITSCOPE "/init.scope" |
367 | static void prune_init_slice(char *cg) | |
368 | { | |
369 | char *point; | |
370 | point = cg + strlen(cg) - strlen(INITSCOPE); | |
371 | if (point < cg) | |
372 | return; | |
373 | if (strcmp(point, INITSCOPE) == 0) { | |
374 | if (point == cg) | |
375 | *(point+1) = '\0'; | |
376 | else | |
377 | *point = '\0'; | |
378 | } | |
379 | } | |
380 | ||
3db25a35 SH |
381 | /* |
382 | * If caller is in /a/b/c/d, he may only act on things under cg=/a/b/c/d. | |
383 | * If caller is in /a, he may act on /a/b, but not on /b. | |
384 | * if the answer is false and nextcg is not NULL, then *nextcg will point | |
2c51f8dd SH |
385 | * to a string containing the next cgroup directory under cg, which must be |
386 | * freed by the caller. | |
3db25a35 SH |
387 | */ |
388 | static bool caller_is_in_ancestor(pid_t pid, const char *contrl, const char *cg, char **nextcg) | |
389 | { | |
2c51f8dd | 390 | char fnam[PROCLEN]; |
3db25a35 SH |
391 | FILE *f; |
392 | bool answer = false; | |
393 | char *line = NULL; | |
394 | size_t len = 0; | |
2c51f8dd | 395 | int ret; |
3db25a35 | 396 | |
2c51f8dd SH |
397 | ret = snprintf(fnam, PROCLEN, "/proc/%d/cgroup", pid); |
398 | if (ret < 0 || ret >= PROCLEN) | |
399 | return false; | |
3db25a35 SH |
400 | if (!(f = fopen(fnam, "r"))) |
401 | return false; | |
402 | ||
403 | while (getline(&line, &len, f) != -1) { | |
404 | char *c1, *c2, *linecmp; | |
405 | if (!line[0]) | |
406 | continue; | |
407 | c1 = strchr(line, ':'); | |
408 | if (!c1) | |
409 | goto out; | |
410 | c1++; | |
411 | c2 = strchr(c1, ':'); | |
412 | if (!c2) | |
413 | goto out; | |
414 | *c2 = '\0'; | |
415 | if (strcmp(c1, contrl) != 0) | |
416 | continue; | |
417 | c2++; | |
418 | stripnewline(c2); | |
04b5cbdc | 419 | prune_init_slice(c2); |
3db25a35 SH |
420 | /* |
421 | * callers pass in '/' for root cgroup, otherwise they pass | |
422 | * in a cgroup without leading '/' | |
423 | */ | |
424 | linecmp = *cg == '/' ? c2 : c2+1; | |
425 | if (strncmp(linecmp, cg, strlen(linecmp)) != 0) { | |
426 | if (nextcg) | |
427 | *nextcg = get_next_cgroup_dir(linecmp, cg); | |
428 | goto out; | |
429 | } | |
430 | answer = true; | |
431 | goto out; | |
432 | } | |
433 | ||
434 | out: | |
435 | fclose(f); | |
436 | free(line); | |
437 | return answer; | |
438 | } | |
439 | ||
758ad80c | 440 | /* |
2c51f8dd SH |
441 | * given /cgroup/freezer/a/b, return "freezer". |
442 | * the returned char* should NOT be freed. | |
758ad80c SH |
443 | */ |
444 | static char *pick_controller_from_path(struct fuse_context *fc, const char *path) | |
445 | { | |
446 | const char *p1; | |
2c51f8dd | 447 | char *contr, *slash; |
758ad80c SH |
448 | |
449 | if (strlen(path) < 9) | |
450 | return NULL; | |
ac5d9d48 SH |
451 | if (*(path+7) != '/') |
452 | return NULL; | |
758ad80c | 453 | p1 = path+8; |
2c51f8dd SH |
454 | contr = strdupa(p1); |
455 | if (!contr) | |
456 | return NULL; | |
457 | slash = strstr(contr, "/"); | |
758ad80c SH |
458 | if (slash) |
459 | *slash = '\0'; | |
460 | ||
461 | /* verify that it is a subsystem */ | |
462 | char **list = LXCFS_DATA ? LXCFS_DATA->subsystems : NULL; | |
463 | int i; | |
2c51f8dd | 464 | if (!list) |
758ad80c | 465 | return NULL; |
758ad80c | 466 | for (i = 0; list[i]; i++) { |
2c51f8dd SH |
467 | if (strcmp(list[i], contr) == 0) |
468 | return list[i]; | |
758ad80c | 469 | } |
758ad80c SH |
470 | return NULL; |
471 | } | |
472 | ||
473 | /* | |
474 | * Find the start of cgroup in /cgroup/controller/the/cgroup/path | |
475 | * Note that the returned value may include files (keynames) etc | |
476 | */ | |
477 | static const char *find_cgroup_in_path(const char *path) | |
478 | { | |
479 | const char *p1; | |
480 | ||
481 | if (strlen(path) < 9) | |
482 | return NULL; | |
483 | p1 = strstr(path+8, "/"); | |
484 | if (!p1) | |
485 | return NULL; | |
486 | return p1+1; | |
487 | } | |
488 | ||
489 | static bool is_child_cgroup(const char *contr, const char *dir, const char *f) | |
490 | { | |
2c51f8dd SH |
491 | char **list; |
492 | bool ret = false; | |
758ad80c SH |
493 | int i; |
494 | ||
495 | if (!f) | |
496 | return false; | |
497 | if (*f == '/') | |
498 | f++; | |
499 | ||
500 | if (!cgm_list_children(contr, dir, &list)) | |
501 | return false; | |
502 | for (i = 0; list[i]; i++) { | |
2c51f8dd SH |
503 | if (strcmp(list[i], f) == 0) { |
504 | ret = true; | |
505 | goto out; | |
506 | } | |
758ad80c SH |
507 | } |
508 | ||
2c51f8dd SH |
509 | out: |
510 | for (i = 0; list[i]; i++) | |
511 | free(list[i]); | |
512 | free(list); | |
513 | return ret; | |
758ad80c SH |
514 | } |
515 | ||
516 | static struct cgm_keys *get_cgroup_key(const char *contr, const char *dir, const char *f) | |
517 | { | |
2c51f8dd SH |
518 | struct cgm_keys **list = NULL; |
519 | struct cgm_keys *k = NULL; | |
758ad80c SH |
520 | int i; |
521 | ||
522 | if (!f) | |
523 | return NULL; | |
524 | if (*f == '/') | |
525 | f++; | |
526 | if (!cgm_list_keys(contr, dir, &list)) | |
527 | return NULL; | |
528 | for (i = 0; list[i]; i++) { | |
529 | if (strcmp(list[i]->name, f) == 0) { | |
2c51f8dd SH |
530 | int j; |
531 | // free all the keys we are not returning | |
532 | k = list[i]; | |
533 | for (j = 0; list[j]; j++) { | |
534 | if (i != j) | |
103f104c | 535 | free_key(list[j]); |
2c51f8dd SH |
536 | } |
537 | free(list); | |
758ad80c SH |
538 | return k; |
539 | } | |
540 | } | |
541 | ||
2c51f8dd | 542 | free_keys(list); |
758ad80c SH |
543 | return NULL; |
544 | } | |
545 | ||
2c51f8dd SH |
546 | /* |
547 | * dir should be freed, file not | |
548 | */ | |
758ad80c SH |
549 | static void get_cgdir_and_path(const char *cg, char **dir, char **file) |
550 | { | |
758ad80c SH |
551 | char *p; |
552 | ||
2c51f8dd SH |
553 | do { |
554 | *dir = strdup(cg); | |
555 | } while (!*dir); | |
758ad80c SH |
556 | *file = strrchr(cg, '/'); |
557 | if (!*file) { | |
558 | *file = NULL; | |
559 | return; | |
560 | } | |
561 | p = strrchr(*dir, '/'); | |
562 | *p = '\0'; | |
563 | } | |
564 | ||
565 | /* | |
2ad6d2bd | 566 | * FUSE ops for /cgroup |
758ad80c | 567 | */ |
2ad6d2bd | 568 | |
758ad80c SH |
569 | static int cg_getattr(const char *path, struct stat *sb) |
570 | { | |
571 | struct timespec now; | |
572 | struct fuse_context *fc = fuse_get_context(); | |
2c51f8dd | 573 | char * cgdir = NULL; |
758ad80c | 574 | char *fpath = NULL, *path1, *path2; |
2c51f8dd | 575 | struct cgm_keys *k = NULL; |
758ad80c | 576 | const char *cgroup; |
2c51f8dd SH |
577 | const char *controller = NULL; |
578 | int ret = -ENOENT; | |
758ad80c SH |
579 | |
580 | ||
581 | if (!fc) | |
582 | return -EIO; | |
583 | ||
584 | memset(sb, 0, sizeof(struct stat)); | |
585 | ||
586 | if (clock_gettime(CLOCK_REALTIME, &now) < 0) | |
587 | return -EINVAL; | |
588 | ||
589 | sb->st_uid = sb->st_gid = 0; | |
590 | sb->st_atim = sb->st_mtim = sb->st_ctim = now; | |
591 | sb->st_size = 0; | |
592 | ||
593 | if (strcmp(path, "/cgroup") == 0) { | |
594 | sb->st_mode = S_IFDIR | 00755; | |
595 | sb->st_nlink = 2; | |
596 | return 0; | |
597 | } | |
598 | ||
599 | controller = pick_controller_from_path(fc, path); | |
600 | if (!controller) | |
601 | return -EIO; | |
758ad80c SH |
602 | cgroup = find_cgroup_in_path(path); |
603 | if (!cgroup) { | |
604 | /* this is just /cgroup/controller, return it as a dir */ | |
605 | sb->st_mode = S_IFDIR | 00755; | |
606 | sb->st_nlink = 2; | |
607 | return 0; | |
608 | } | |
341b21ad | 609 | |
758ad80c SH |
610 | get_cgdir_and_path(cgroup, &cgdir, &fpath); |
611 | ||
612 | if (!fpath) { | |
613 | path1 = "/"; | |
614 | path2 = cgdir; | |
615 | } else { | |
616 | path1 = cgdir; | |
617 | path2 = fpath; | |
618 | } | |
619 | ||
758ad80c SH |
620 | /* check that cgcopy is either a child cgroup of cgdir, or listed in its keys. |
621 | * Then check that caller's cgroup is under path if fpath is a child | |
622 | * cgroup, or cgdir if fpath is a file */ | |
623 | ||
624 | if (is_child_cgroup(controller, path1, path2)) { | |
f9a05025 SH |
625 | if (!caller_is_in_ancestor(fc->pid, controller, cgroup, NULL)) { |
626 | /* this is just /cgroup/controller, return it as a dir */ | |
627 | sb->st_mode = S_IFDIR | 00555; | |
628 | sb->st_nlink = 2; | |
2c51f8dd SH |
629 | ret = 0; |
630 | goto out; | |
631 | } | |
632 | if (!fc_may_access(fc, controller, cgroup, NULL, O_RDONLY)) { | |
633 | ret = -EACCES; | |
634 | goto out; | |
f9a05025 | 635 | } |
758ad80c | 636 | |
053a659d SH |
637 | // get uid, gid, from '/tasks' file and make up a mode |
638 | // That is a hack, until cgmanager gains a GetCgroupPerms fn. | |
639 | sb->st_mode = S_IFDIR | 00755; | |
640 | k = get_cgroup_key(controller, cgroup, "tasks"); | |
641 | if (!k) { | |
053a659d SH |
642 | sb->st_uid = sb->st_gid = 0; |
643 | } else { | |
053a659d SH |
644 | sb->st_uid = k->uid; |
645 | sb->st_gid = k->gid; | |
646 | } | |
2c51f8dd | 647 | free_key(k); |
758ad80c | 648 | sb->st_nlink = 2; |
2c51f8dd SH |
649 | ret = 0; |
650 | goto out; | |
758ad80c SH |
651 | } |
652 | ||
653 | if ((k = get_cgroup_key(controller, path1, path2)) != NULL) { | |
758ad80c | 654 | sb->st_mode = S_IFREG | k->mode; |
053a659d | 655 | sb->st_nlink = 1; |
758ad80c SH |
656 | sb->st_uid = k->uid; |
657 | sb->st_gid = k->gid; | |
7253e0a4 | 658 | sb->st_size = 0; |
2c51f8dd SH |
659 | free_key(k); |
660 | if (!caller_is_in_ancestor(fc->pid, controller, path1, NULL)) | |
661 | return -ENOENT; | |
662 | if (!fc_may_access(fc, controller, path1, path2, O_RDONLY)) | |
663 | return -EACCES; | |
664 | ||
665 | ret = 0; | |
758ad80c SH |
666 | } |
667 | ||
2c51f8dd SH |
668 | out: |
669 | free(cgdir); | |
670 | return ret; | |
758ad80c | 671 | } |
2183082c | 672 | |
758ad80c | 673 | static int cg_opendir(const char *path, struct fuse_file_info *fi) |
2183082c | 674 | { |
7f163b71 | 675 | struct fuse_context *fc = fuse_get_context(); |
7f163b71 | 676 | const char *cgroup; |
c688e1b3 | 677 | struct file_info *dir_info; |
2c51f8dd | 678 | char *controller = NULL; |
7f163b71 SH |
679 | |
680 | if (!fc) | |
681 | return -EIO; | |
682 | ||
c688e1b3 SH |
683 | if (strcmp(path, "/cgroup") == 0) { |
684 | cgroup = NULL; | |
685 | controller = NULL; | |
686 | } else { | |
687 | // return list of keys for the controller, and list of child cgroups | |
688 | controller = pick_controller_from_path(fc, path); | |
689 | if (!controller) | |
690 | return -EIO; | |
7f163b71 | 691 | |
c688e1b3 SH |
692 | cgroup = find_cgroup_in_path(path); |
693 | if (!cgroup) { | |
694 | /* this is just /cgroup/controller, return its contents */ | |
695 | cgroup = "/"; | |
696 | } | |
7f163b71 SH |
697 | } |
698 | ||
2c51f8dd | 699 | if (cgroup && !fc_may_access(fc, controller, cgroup, NULL, O_RDONLY)) { |
7f163b71 | 700 | return -EACCES; |
2c51f8dd | 701 | } |
c688e1b3 SH |
702 | |
703 | /* we'll free this at cg_releasedir */ | |
2c51f8dd SH |
704 | dir_info = malloc(sizeof(*dir_info)); |
705 | if (!dir_info) | |
706 | return -ENOMEM; | |
bae07053 SH |
707 | dir_info->controller = must_copy_string(dir_info, controller); |
708 | dir_info->cgroup = must_copy_string(dir_info, cgroup); | |
443d13f5 | 709 | dir_info->type = LXC_TYPE_CGDIR; |
c688e1b3 | 710 | dir_info->buf = NULL; |
8f6e8f5e | 711 | dir_info->file = NULL; |
c688e1b3 SH |
712 | dir_info->buflen = 0; |
713 | ||
714 | fi->fh = (unsigned long)dir_info; | |
758ad80c SH |
715 | return 0; |
716 | } | |
717 | ||
758ad80c SH |
718 | static int cg_readdir(const char *path, void *buf, fuse_fill_dir_t filler, off_t offset, |
719 | struct fuse_file_info *fi) | |
720 | { | |
c688e1b3 | 721 | struct file_info *d = (struct file_info *)fi->fh; |
2c51f8dd SH |
722 | struct cgm_keys **list = NULL; |
723 | int i, ret; | |
724 | char *nextcg = NULL; | |
758ad80c | 725 | struct fuse_context *fc = fuse_get_context(); |
2c51f8dd | 726 | char **clist = NULL; |
758ad80c | 727 | |
443d13f5 | 728 | if (d->type != LXC_TYPE_CGDIR) { |
b845ad01 SH |
729 | fprintf(stderr, "Internal error: file cache info used in readdir\n"); |
730 | return -EIO; | |
731 | } | |
c688e1b3 SH |
732 | if (!d->cgroup && !d->controller) { |
733 | // ls /var/lib/lxcfs/cgroup - just show list of controllers | |
758ad80c SH |
734 | char **list = LXCFS_DATA ? LXCFS_DATA->subsystems : NULL; |
735 | int i; | |
736 | ||
737 | if (!list) | |
738 | return -EIO; | |
7f163b71 | 739 | |
758ad80c SH |
740 | for (i = 0; list[i]; i++) { |
741 | if (filler(buf, list[i], NULL, 0) != 0) { | |
742 | return -EIO; | |
743 | } | |
744 | } | |
745 | return 0; | |
746 | } | |
747 | ||
2c51f8dd | 748 | if (!cgm_list_keys(d->controller, d->cgroup, &list)) { |
3db25a35 | 749 | // not a valid cgroup |
2c51f8dd SH |
750 | ret = -EINVAL; |
751 | goto out; | |
752 | } | |
3db25a35 | 753 | |
c688e1b3 | 754 | if (!caller_is_in_ancestor(fc->pid, d->controller, d->cgroup, &nextcg)) { |
3db25a35 SH |
755 | if (nextcg) { |
756 | int ret; | |
757 | ret = filler(buf, nextcg, NULL, 0); | |
2c51f8dd SH |
758 | free(nextcg); |
759 | if (ret != 0) { | |
760 | ret = -EIO; | |
761 | goto out; | |
762 | } | |
3db25a35 | 763 | } |
2c51f8dd SH |
764 | ret = 0; |
765 | goto out; | |
3db25a35 SH |
766 | } |
767 | ||
758ad80c | 768 | for (i = 0; list[i]; i++) { |
758ad80c | 769 | if (filler(buf, list[i]->name, NULL, 0) != 0) { |
2c51f8dd SH |
770 | ret = -EIO; |
771 | goto out; | |
758ad80c SH |
772 | } |
773 | } | |
774 | ||
775 | // now get the list of child cgroups | |
758ad80c | 776 | |
2c51f8dd SH |
777 | if (!cgm_list_children(d->controller, d->cgroup, &clist)) { |
778 | ret = 0; | |
779 | goto out; | |
780 | } | |
758ad80c | 781 | for (i = 0; clist[i]; i++) { |
758ad80c | 782 | if (filler(buf, clist[i], NULL, 0) != 0) { |
2c51f8dd SH |
783 | ret = -EIO; |
784 | goto out; | |
758ad80c SH |
785 | } |
786 | } | |
2c51f8dd SH |
787 | ret = 0; |
788 | ||
789 | out: | |
790 | free_keys(list); | |
791 | if (clist) { | |
792 | for (i = 0; clist[i]; i++) | |
793 | free(clist[i]); | |
794 | free(clist); | |
795 | } | |
796 | return ret; | |
758ad80c SH |
797 | } |
798 | ||
8f6e8f5e SH |
799 | static void do_release_file_info(struct file_info *f) |
800 | { | |
2c51f8dd SH |
801 | if (!f) |
802 | return; | |
803 | free(f->controller); | |
804 | free(f->cgroup); | |
805 | free(f->file); | |
806 | free(f->buf); | |
807 | free(f); | |
8f6e8f5e SH |
808 | } |
809 | ||
758ad80c SH |
810 | static int cg_releasedir(const char *path, struct fuse_file_info *fi) |
811 | { | |
c688e1b3 SH |
812 | struct file_info *d = (struct file_info *)fi->fh; |
813 | ||
8f6e8f5e | 814 | do_release_file_info(d); |
758ad80c SH |
815 | return 0; |
816 | } | |
817 | ||
99978832 SH |
818 | static int cg_open(const char *path, struct fuse_file_info *fi) |
819 | { | |
99978832 | 820 | const char *cgroup; |
2c51f8dd SH |
821 | char *fpath = NULL, *path1, *path2, * cgdir = NULL, *controller; |
822 | struct cgm_keys *k = NULL; | |
8f6e8f5e | 823 | struct file_info *file_info; |
99978832 | 824 | struct fuse_context *fc = fuse_get_context(); |
2c51f8dd | 825 | int ret; |
99978832 SH |
826 | |
827 | if (!fc) | |
828 | return -EIO; | |
829 | ||
830 | controller = pick_controller_from_path(fc, path); | |
831 | if (!controller) | |
832 | return -EIO; | |
833 | cgroup = find_cgroup_in_path(path); | |
834 | if (!cgroup) | |
835 | return -EINVAL; | |
836 | ||
837 | get_cgdir_and_path(cgroup, &cgdir, &fpath); | |
838 | if (!fpath) { | |
839 | path1 = "/"; | |
840 | path2 = cgdir; | |
841 | } else { | |
842 | path1 = cgdir; | |
843 | path2 = fpath; | |
844 | } | |
845 | ||
8f6e8f5e | 846 | k = get_cgroup_key(controller, path1, path2); |
2c51f8dd SH |
847 | if (!k) { |
848 | ret = -EINVAL; | |
849 | goto out; | |
850 | } | |
851 | free_key(k); | |
99978832 | 852 | |
2c51f8dd | 853 | if (!fc_may_access(fc, controller, path1, path2, fi->flags)) { |
8f6e8f5e | 854 | // should never get here |
2c51f8dd SH |
855 | ret = -EACCES; |
856 | goto out; | |
857 | } | |
99978832 | 858 | |
8f6e8f5e | 859 | /* we'll free this at cg_release */ |
2c51f8dd SH |
860 | file_info = malloc(sizeof(*file_info)); |
861 | if (!file_info) { | |
862 | ret = -ENOMEM; | |
863 | goto out; | |
864 | } | |
bae07053 SH |
865 | file_info->controller = must_copy_string(file_info, controller); |
866 | file_info->cgroup = must_copy_string(file_info, path1); | |
867 | file_info->file = must_copy_string(file_info, path2); | |
443d13f5 | 868 | file_info->type = LXC_TYPE_CGFILE; |
8f6e8f5e SH |
869 | file_info->buf = NULL; |
870 | file_info->buflen = 0; | |
871 | ||
872 | fi->fh = (unsigned long)file_info; | |
2c51f8dd SH |
873 | ret = 0; |
874 | ||
875 | out: | |
876 | free(cgdir); | |
877 | return ret; | |
8f6e8f5e SH |
878 | } |
879 | ||
880 | static int cg_release(const char *path, struct fuse_file_info *fi) | |
881 | { | |
882 | struct file_info *f = (struct file_info *)fi->fh; | |
883 | ||
884 | do_release_file_info(f); | |
885 | return 0; | |
99978832 SH |
886 | } |
887 | ||
a05660a6 SH |
888 | static int msgrecv(int sockfd, void *buf, size_t len) |
889 | { | |
890 | struct timeval tv; | |
891 | fd_set rfds; | |
892 | ||
893 | FD_ZERO(&rfds); | |
894 | FD_SET(sockfd, &rfds); | |
895 | tv.tv_sec = 2; | |
896 | tv.tv_usec = 0; | |
897 | ||
ea56f722 | 898 | if (select(sockfd+1, &rfds, NULL, NULL, &tv) <= 0) |
a05660a6 SH |
899 | return -1; |
900 | return recv(sockfd, buf, len, MSG_DONTWAIT); | |
901 | } | |
902 | ||
01e71852 SH |
903 | #define SEND_CREDS_OK 0 |
904 | #define SEND_CREDS_NOTSK 1 | |
905 | #define SEND_CREDS_FAIL 2 | |
906 | static int send_creds(int sock, struct ucred *cred, char v, bool pingfirst) | |
a05660a6 SH |
907 | { |
908 | struct msghdr msg = { 0 }; | |
909 | struct iovec iov; | |
910 | struct cmsghdr *cmsg; | |
911 | char cmsgbuf[CMSG_SPACE(sizeof(*cred))]; | |
912 | char buf[1]; | |
913 | buf[0] = 'p'; | |
914 | ||
01e71852 SH |
915 | if (pingfirst) { |
916 | if (msgrecv(sock, buf, 1) != 1) { | |
1420baf8 | 917 | fprintf(stderr, "%s: Error getting reply from server over socketpair\n", |
01e71852 SH |
918 | __func__); |
919 | return SEND_CREDS_FAIL; | |
920 | } | |
a05660a6 SH |
921 | } |
922 | ||
923 | msg.msg_control = cmsgbuf; | |
924 | msg.msg_controllen = sizeof(cmsgbuf); | |
925 | ||
926 | cmsg = CMSG_FIRSTHDR(&msg); | |
927 | cmsg->cmsg_len = CMSG_LEN(sizeof(struct ucred)); | |
928 | cmsg->cmsg_level = SOL_SOCKET; | |
929 | cmsg->cmsg_type = SCM_CREDENTIALS; | |
930 | memcpy(CMSG_DATA(cmsg), cred, sizeof(*cred)); | |
931 | ||
932 | msg.msg_name = NULL; | |
933 | msg.msg_namelen = 0; | |
934 | ||
935 | buf[0] = v; | |
936 | iov.iov_base = buf; | |
937 | iov.iov_len = sizeof(buf); | |
938 | msg.msg_iov = &iov; | |
939 | msg.msg_iovlen = 1; | |
940 | ||
941 | if (sendmsg(sock, &msg, 0) < 0) { | |
1420baf8 | 942 | fprintf(stderr, "%s: failed at sendmsg: %s\n", __func__, |
a05660a6 SH |
943 | strerror(errno)); |
944 | if (errno == 3) | |
01e71852 SH |
945 | return SEND_CREDS_NOTSK; |
946 | return SEND_CREDS_FAIL; | |
a05660a6 SH |
947 | } |
948 | ||
01e71852 | 949 | return SEND_CREDS_OK; |
a05660a6 SH |
950 | } |
951 | ||
952 | static bool recv_creds(int sock, struct ucred *cred, char *v) | |
953 | { | |
954 | struct msghdr msg = { 0 }; | |
955 | struct iovec iov; | |
956 | struct cmsghdr *cmsg; | |
957 | char cmsgbuf[CMSG_SPACE(sizeof(*cred))]; | |
958 | char buf[1]; | |
959 | int ret; | |
960 | int optval = 1; | |
6ee867dc SH |
961 | struct timeval tv; |
962 | fd_set rfds; | |
a05660a6 SH |
963 | |
964 | *v = '1'; | |
965 | ||
966 | cred->pid = -1; | |
967 | cred->uid = -1; | |
968 | cred->gid = -1; | |
969 | ||
970 | if (setsockopt(sock, SOL_SOCKET, SO_PASSCRED, &optval, sizeof(optval)) == -1) { | |
1420baf8 | 971 | fprintf(stderr, "Failed to set passcred: %s\n", strerror(errno)); |
a05660a6 SH |
972 | return false; |
973 | } | |
974 | buf[0] = '1'; | |
975 | if (write(sock, buf, 1) != 1) { | |
1420baf8 | 976 | fprintf(stderr, "Failed to start write on scm fd: %s\n", strerror(errno)); |
a05660a6 SH |
977 | return false; |
978 | } | |
979 | ||
980 | msg.msg_name = NULL; | |
981 | msg.msg_namelen = 0; | |
982 | msg.msg_control = cmsgbuf; | |
983 | msg.msg_controllen = sizeof(cmsgbuf); | |
984 | ||
985 | iov.iov_base = buf; | |
986 | iov.iov_len = sizeof(buf); | |
987 | msg.msg_iov = &iov; | |
988 | msg.msg_iovlen = 1; | |
989 | ||
6ee867dc SH |
990 | FD_ZERO(&rfds); |
991 | FD_SET(sock, &rfds); | |
992 | tv.tv_sec = 2; | |
993 | tv.tv_usec = 0; | |
ea56f722 | 994 | if (select(sock+1, &rfds, NULL, NULL, &tv) <= 0) { |
6ee867dc SH |
995 | fprintf(stderr, "Failed to select for scm_cred: %s\n", |
996 | strerror(errno)); | |
997 | return false; | |
998 | } | |
999 | ret = recvmsg(sock, &msg, MSG_DONTWAIT); | |
a05660a6 | 1000 | if (ret < 0) { |
1420baf8 | 1001 | fprintf(stderr, "Failed to receive scm_cred: %s\n", |
a05660a6 SH |
1002 | strerror(errno)); |
1003 | return false; | |
1004 | } | |
1005 | ||
1006 | cmsg = CMSG_FIRSTHDR(&msg); | |
1007 | ||
1008 | if (cmsg && cmsg->cmsg_len == CMSG_LEN(sizeof(struct ucred)) && | |
1009 | cmsg->cmsg_level == SOL_SOCKET && | |
1010 | cmsg->cmsg_type == SCM_CREDENTIALS) { | |
1011 | memcpy(cred, CMSG_DATA(cmsg), sizeof(*cred)); | |
1012 | } | |
1013 | *v = buf[0]; | |
1014 | ||
1015 | return true; | |
1016 | } | |
1017 | ||
1018 | ||
1019 | /* | |
4775fba1 SH |
1020 | * pid_to_ns - reads pids from a ucred over a socket, then writes the |
1021 | * int value back over the socket. This shifts the pid from the | |
1022 | * sender's pidns into tpid's pidns. | |
a05660a6 | 1023 | */ |
4775fba1 | 1024 | static void pid_to_ns(int sock, pid_t tpid) |
a05660a6 SH |
1025 | { |
1026 | char v = '0'; | |
1027 | struct ucred cred; | |
1028 | ||
1029 | while (recv_creds(sock, &cred, &v)) { | |
1030 | if (v == '1') | |
67bd113f | 1031 | _exit(0); |
a05660a6 | 1032 | if (write(sock, &cred.pid, sizeof(pid_t)) != sizeof(pid_t)) |
67bd113f | 1033 | _exit(1); |
a05660a6 | 1034 | } |
67bd113f | 1035 | _exit(0); |
a05660a6 SH |
1036 | } |
1037 | ||
1038 | /* | |
4775fba1 | 1039 | * pid_to_ns_wrapper: when you setns into a pidns, you yourself remain |
a05660a6 | 1040 | * in your old pidns. Only children which you fork will be in the target |
4775fba1 | 1041 | * pidns. So the pid_to_ns_wrapper does the setns, then forks a child to |
a05660a6 SH |
1042 | * actually convert pids |
1043 | */ | |
4775fba1 | 1044 | static void pid_to_ns_wrapper(int sock, pid_t tpid) |
a05660a6 | 1045 | { |
ea56f722 | 1046 | int newnsfd = -1, ret, cpipe[2]; |
a05660a6 SH |
1047 | char fnam[100]; |
1048 | pid_t cpid; | |
ea56f722 SH |
1049 | struct timeval tv; |
1050 | fd_set s; | |
1051 | char v; | |
a05660a6 | 1052 | |
c0adec85 SH |
1053 | ret = snprintf(fnam, sizeof(fnam), "/proc/%d/ns/pid", tpid); |
1054 | if (ret < 0 || ret >= sizeof(fnam)) | |
67bd113f | 1055 | _exit(1); |
a05660a6 SH |
1056 | newnsfd = open(fnam, O_RDONLY); |
1057 | if (newnsfd < 0) | |
67bd113f | 1058 | _exit(1); |
a05660a6 | 1059 | if (setns(newnsfd, 0) < 0) |
67bd113f | 1060 | _exit(1); |
a05660a6 SH |
1061 | close(newnsfd); |
1062 | ||
ea56f722 | 1063 | if (pipe(cpipe) < 0) |
67bd113f | 1064 | _exit(1); |
a05660a6 | 1065 | |
ea56f722 SH |
1066 | loop: |
1067 | cpid = fork(); | |
a05660a6 | 1068 | if (cpid < 0) |
67bd113f | 1069 | _exit(1); |
ea56f722 SH |
1070 | |
1071 | if (!cpid) { | |
1072 | char b = '1'; | |
1073 | close(cpipe[0]); | |
1074 | if (write(cpipe[1], &b, sizeof(char)) < 0) { | |
1075 | fprintf(stderr, "%s (child): erorr on write: %s\n", | |
1076 | __func__, strerror(errno)); | |
1077 | } | |
1078 | close(cpipe[1]); | |
4775fba1 | 1079 | pid_to_ns(sock, tpid); |
ea56f722 SH |
1080 | } |
1081 | // give the child 1 second to be done forking and | |
1082 | // write it's ack | |
1083 | FD_ZERO(&s); | |
1084 | FD_SET(cpipe[0], &s); | |
1085 | tv.tv_sec = 1; | |
1086 | tv.tv_usec = 0; | |
1087 | ret = select(cpipe[0]+1, &s, NULL, NULL, &tv); | |
1088 | if (ret <= 0) | |
1089 | goto again; | |
1090 | ret = read(cpipe[0], &v, 1); | |
1091 | if (ret != sizeof(char) || v != '1') { | |
1092 | goto again; | |
1093 | } | |
1094 | ||
a05660a6 | 1095 | if (!wait_for_pid(cpid)) |
67bd113f SH |
1096 | _exit(1); |
1097 | _exit(0); | |
ea56f722 SH |
1098 | |
1099 | again: | |
1100 | kill(cpid, SIGKILL); | |
1101 | wait_for_pid(cpid); | |
1102 | goto loop; | |
a05660a6 SH |
1103 | } |
1104 | ||
1105 | /* | |
1106 | * To read cgroup files with a particular pid, we will setns into the child | |
1107 | * pidns, open a pipe, fork a child - which will be the first to really be in | |
1108 | * the child ns - which does the cgm_get_value and writes the data to the pipe. | |
1109 | */ | |
1110 | static bool do_read_pids(pid_t tpid, const char *contrl, const char *cg, const char *file, char **d) | |
1111 | { | |
1112 | int sock[2] = {-1, -1}; | |
2c51f8dd | 1113 | char *tmpdata = NULL; |
a05660a6 SH |
1114 | int ret; |
1115 | pid_t qpid, cpid = -1; | |
1116 | bool answer = false; | |
1117 | char v = '0'; | |
1118 | struct ucred cred; | |
1119 | struct timeval tv; | |
2c51f8dd | 1120 | size_t sz = 0, asz = 0; |
a05660a6 SH |
1121 | fd_set s; |
1122 | ||
1123 | if (!cgm_get_value(contrl, cg, file, &tmpdata)) | |
1124 | return false; | |
1125 | ||
1126 | /* | |
1127 | * Now we read the pids from returned data one by one, pass | |
1128 | * them into a child in the target namespace, read back the | |
1129 | * translated pids, and put them into our to-return data | |
1130 | */ | |
1131 | ||
1132 | if (socketpair(AF_UNIX, SOCK_DGRAM, 0, sock) < 0) { | |
1133 | perror("socketpair"); | |
2c51f8dd SH |
1134 | free(tmpdata); |
1135 | return false; | |
a05660a6 SH |
1136 | } |
1137 | ||
1138 | cpid = fork(); | |
1139 | if (cpid == -1) | |
1140 | goto out; | |
1141 | ||
1142 | if (!cpid) // child | |
4775fba1 | 1143 | pid_to_ns_wrapper(sock[1], tpid); |
a05660a6 SH |
1144 | |
1145 | char *ptr = tmpdata; | |
1146 | cred.uid = 0; | |
1147 | cred.gid = 0; | |
1148 | while (sscanf(ptr, "%d\n", &qpid) == 1) { | |
1149 | cred.pid = qpid; | |
01e71852 SH |
1150 | ret = send_creds(sock[0], &cred, v, true); |
1151 | ||
1152 | if (ret == SEND_CREDS_NOTSK) | |
1153 | goto next; | |
1154 | if (ret == SEND_CREDS_FAIL) | |
a05660a6 SH |
1155 | goto out; |
1156 | ||
1157 | // read converted results | |
1158 | FD_ZERO(&s); | |
1159 | FD_SET(sock[0], &s); | |
6ee867dc | 1160 | tv.tv_sec = 2; |
a05660a6 SH |
1161 | tv.tv_usec = 0; |
1162 | ret = select(sock[0]+1, &s, NULL, NULL, &tv); | |
1163 | if (ret <= 0) { | |
6ee867dc SH |
1164 | fprintf(stderr, "%s: select error waiting for pid from child: %s\n", |
1165 | __func__, strerror(errno)); | |
a05660a6 SH |
1166 | goto out; |
1167 | } | |
1168 | if (read(sock[0], &qpid, sizeof(qpid)) != sizeof(qpid)) { | |
6ee867dc SH |
1169 | fprintf(stderr, "%s: error reading pid from child: %s\n", |
1170 | __func__, strerror(errno)); | |
a05660a6 SH |
1171 | goto out; |
1172 | } | |
2c51f8dd | 1173 | must_strcat_pid(d, &sz, &asz, qpid); |
01e71852 | 1174 | next: |
a05660a6 SH |
1175 | ptr = strchr(ptr, '\n'); |
1176 | if (!ptr) | |
1177 | break; | |
1178 | ptr++; | |
1179 | } | |
1180 | ||
1181 | cred.pid = getpid(); | |
1182 | v = '1'; | |
01e71852 | 1183 | if (send_creds(sock[0], &cred, v, true) != SEND_CREDS_OK) { |
a05660a6 | 1184 | // failed to ask child to exit |
6ee867dc SH |
1185 | fprintf(stderr, "%s: failed to ask child to exit: %s\n", |
1186 | __func__, strerror(errno)); | |
a05660a6 SH |
1187 | goto out; |
1188 | } | |
1189 | ||
1190 | answer = true; | |
1191 | ||
1192 | out: | |
2c51f8dd | 1193 | free(tmpdata); |
a05660a6 SH |
1194 | if (cpid != -1) |
1195 | wait_for_pid(cpid); | |
1196 | if (sock[0] != -1) { | |
1197 | close(sock[0]); | |
1198 | close(sock[1]); | |
1199 | } | |
1200 | return answer; | |
1201 | } | |
1202 | ||
99978832 SH |
1203 | static int cg_read(const char *path, char *buf, size_t size, off_t offset, |
1204 | struct fuse_file_info *fi) | |
1205 | { | |
99978832 | 1206 | struct fuse_context *fc = fuse_get_context(); |
8f6e8f5e | 1207 | struct file_info *f = (struct file_info *)fi->fh; |
2c51f8dd SH |
1208 | struct cgm_keys *k = NULL; |
1209 | char *data = NULL; | |
1210 | int ret, s; | |
1211 | bool r; | |
99978832 | 1212 | |
443d13f5 | 1213 | if (f->type != LXC_TYPE_CGFILE) { |
b845ad01 SH |
1214 | fprintf(stderr, "Internal error: directory cache info used in cg_read\n"); |
1215 | return -EIO; | |
1216 | } | |
1217 | ||
99978832 | 1218 | if (offset) |
7253e0a4 | 1219 | return 0; |
99978832 SH |
1220 | |
1221 | if (!fc) | |
1222 | return -EIO; | |
1223 | ||
8f6e8f5e | 1224 | if (!f->controller) |
99978832 SH |
1225 | return -EINVAL; |
1226 | ||
2c51f8dd SH |
1227 | if ((k = get_cgroup_key(f->controller, f->cgroup, f->file)) == NULL) { |
1228 | return -EINVAL; | |
1229 | } | |
1230 | free_key(k); | |
99978832 | 1231 | |
99978832 | 1232 | |
2c51f8dd SH |
1233 | if (!fc_may_access(fc, f->controller, f->cgroup, f->file, O_RDONLY)) { // should never get here |
1234 | ret = -EACCES; | |
1235 | goto out; | |
1236 | } | |
a05660a6 | 1237 | |
2c51f8dd SH |
1238 | if (strcmp(f->file, "tasks") == 0 || |
1239 | strcmp(f->file, "/tasks") == 0 || | |
1240 | strcmp(f->file, "/cgroup.procs") == 0 || | |
1241 | strcmp(f->file, "cgroup.procs") == 0) | |
1242 | // special case - we have to translate the pids | |
1243 | r = do_read_pids(fc->pid, f->controller, f->cgroup, f->file, &data); | |
1244 | else | |
1245 | r = cgm_get_value(f->controller, f->cgroup, f->file, &data); | |
99978832 | 1246 | |
2c51f8dd SH |
1247 | if (!r) { |
1248 | ret = -EINVAL; | |
1249 | goto out; | |
1250 | } | |
99978832 | 1251 | |
2c51f8dd SH |
1252 | if (!data) { |
1253 | ret = 0; | |
1254 | goto out; | |
99978832 | 1255 | } |
2c51f8dd SH |
1256 | s = strlen(data); |
1257 | if (s > size) | |
1258 | s = size; | |
1259 | memcpy(buf, data, s); | |
1260 | if (s > 0 && s < size && data[s-1] != '\n') | |
1261 | buf[s++] = '\n'; | |
99978832 | 1262 | |
2c51f8dd SH |
1263 | ret = s; |
1264 | ||
1265 | out: | |
1266 | free(data); | |
1267 | return ret; | |
99978832 SH |
1268 | } |
1269 | ||
4775fba1 SH |
1270 | static void pid_from_ns(int sock, pid_t tpid) |
1271 | { | |
1272 | pid_t vpid; | |
1273 | struct ucred cred; | |
1274 | char v; | |
6ee867dc SH |
1275 | struct timeval tv; |
1276 | fd_set s; | |
1277 | int ret; | |
4775fba1 SH |
1278 | |
1279 | cred.uid = 0; | |
1280 | cred.gid = 0; | |
6ee867dc SH |
1281 | while (1) { |
1282 | FD_ZERO(&s); | |
1283 | FD_SET(sock, &s); | |
1284 | tv.tv_sec = 2; | |
1285 | tv.tv_usec = 0; | |
1286 | ret = select(sock+1, &s, NULL, NULL, &tv); | |
ea56f722 SH |
1287 | if (ret <= 0) { |
1288 | fprintf(stderr, "%s: bad select before read from parent: %s\n", | |
6ee867dc | 1289 | __func__, strerror(errno)); |
67bd113f | 1290 | _exit(1); |
6ee867dc SH |
1291 | } |
1292 | if ((ret = read(sock, &vpid, sizeof(pid_t))) != sizeof(pid_t)) { | |
1293 | fprintf(stderr, "%s: bad read from parent: %s\n", | |
1294 | __func__, strerror(errno)); | |
67bd113f | 1295 | _exit(1); |
6ee867dc | 1296 | } |
4775fba1 | 1297 | if (vpid == -1) // done |
01e71852 | 1298 | break; |
4775fba1 SH |
1299 | v = '0'; |
1300 | cred.pid = vpid; | |
01e71852 | 1301 | if (send_creds(sock, &cred, v, true) != SEND_CREDS_OK) { |
4775fba1 SH |
1302 | v = '1'; |
1303 | cred.pid = getpid(); | |
01e71852 | 1304 | if (send_creds(sock, &cred, v, false) != SEND_CREDS_OK) |
67bd113f | 1305 | _exit(1); |
4775fba1 SH |
1306 | } |
1307 | } | |
67bd113f | 1308 | _exit(0); |
4775fba1 SH |
1309 | } |
1310 | ||
1311 | static void pid_from_ns_wrapper(int sock, pid_t tpid) | |
1312 | { | |
ea56f722 | 1313 | int newnsfd = -1, ret, cpipe[2]; |
4775fba1 SH |
1314 | char fnam[100]; |
1315 | pid_t cpid; | |
ea56f722 SH |
1316 | fd_set s; |
1317 | struct timeval tv; | |
1318 | char v; | |
4775fba1 | 1319 | |
c0adec85 SH |
1320 | ret = snprintf(fnam, sizeof(fnam), "/proc/%d/ns/pid", tpid); |
1321 | if (ret < 0 || ret >= sizeof(fnam)) | |
67bd113f | 1322 | _exit(1); |
4775fba1 SH |
1323 | newnsfd = open(fnam, O_RDONLY); |
1324 | if (newnsfd < 0) | |
67bd113f | 1325 | _exit(1); |
4775fba1 | 1326 | if (setns(newnsfd, 0) < 0) |
67bd113f | 1327 | _exit(1); |
4775fba1 SH |
1328 | close(newnsfd); |
1329 | ||
ea56f722 | 1330 | if (pipe(cpipe) < 0) |
67bd113f | 1331 | _exit(1); |
ea56f722 SH |
1332 | |
1333 | loop: | |
4775fba1 SH |
1334 | cpid = fork(); |
1335 | ||
1336 | if (cpid < 0) | |
67bd113f | 1337 | _exit(1); |
ea56f722 SH |
1338 | |
1339 | if (!cpid) { | |
1340 | char b = '1'; | |
1341 | close(cpipe[0]); | |
1342 | if (write(cpipe[1], &b, sizeof(char)) < 0) { | |
1343 | fprintf(stderr, "%s (child): erorr on write: %s\n", | |
1344 | __func__, strerror(errno)); | |
1345 | } | |
1346 | close(cpipe[1]); | |
4775fba1 | 1347 | pid_from_ns(sock, tpid); |
ea56f722 SH |
1348 | } |
1349 | ||
1350 | // give the child 1 second to be done forking and | |
1351 | // write it's ack | |
1352 | FD_ZERO(&s); | |
1353 | FD_SET(cpipe[0], &s); | |
1354 | tv.tv_sec = 1; | |
1355 | tv.tv_usec = 0; | |
1356 | ret = select(cpipe[0]+1, &s, NULL, NULL, &tv); | |
1357 | if (ret <= 0) | |
1358 | goto again; | |
1359 | ret = read(cpipe[0], &v, 1); | |
1360 | if (ret != sizeof(char) || v != '1') { | |
1361 | goto again; | |
1362 | } | |
1363 | ||
4775fba1 | 1364 | if (!wait_for_pid(cpid)) |
67bd113f SH |
1365 | _exit(1); |
1366 | _exit(0); | |
ea56f722 SH |
1367 | |
1368 | again: | |
1369 | kill(cpid, SIGKILL); | |
1370 | wait_for_pid(cpid); | |
1371 | goto loop; | |
4775fba1 SH |
1372 | } |
1373 | ||
1374 | static bool do_write_pids(pid_t tpid, const char *contrl, const char *cg, const char *file, const char *buf) | |
1375 | { | |
1376 | int sock[2] = {-1, -1}; | |
1377 | pid_t qpid, cpid = -1; | |
1378 | bool answer = false, fail = false; | |
1379 | ||
1380 | /* | |
1381 | * write the pids to a socket, have helper in writer's pidns | |
1382 | * call movepid for us | |
1383 | */ | |
1384 | if (socketpair(AF_UNIX, SOCK_DGRAM, 0, sock) < 0) { | |
1385 | perror("socketpair"); | |
1386 | exit(1); | |
1387 | } | |
1388 | ||
1389 | cpid = fork(); | |
1390 | if (cpid == -1) | |
1391 | goto out; | |
1392 | ||
1393 | if (!cpid) // child | |
1394 | pid_from_ns_wrapper(sock[1], tpid); | |
1395 | ||
1396 | const char *ptr = buf; | |
1397 | while (sscanf(ptr, "%d", &qpid) == 1) { | |
1398 | struct ucred cred; | |
1399 | char v; | |
1400 | ||
1401 | if (write(sock[0], &qpid, sizeof(qpid)) != sizeof(qpid)) { | |
6ee867dc SH |
1402 | fprintf(stderr, "%s: error writing pid to child: %s\n", |
1403 | __func__, strerror(errno)); | |
4775fba1 SH |
1404 | goto out; |
1405 | } | |
1406 | ||
01e71852 SH |
1407 | if (recv_creds(sock[0], &cred, &v)) { |
1408 | if (v == '0') { | |
1409 | if (!cgm_move_pid(contrl, cg, cred.pid)) | |
1410 | fail = true; | |
1411 | } | |
4775fba1 SH |
1412 | } |
1413 | ||
1414 | ptr = strchr(ptr, '\n'); | |
1415 | if (!ptr) | |
1416 | break; | |
1417 | ptr++; | |
1418 | } | |
1419 | ||
1420 | /* All good, write the value */ | |
1421 | qpid = -1; | |
1422 | if (write(sock[0], &qpid ,sizeof(qpid)) != sizeof(qpid)) | |
1420baf8 | 1423 | fprintf(stderr, "Warning: failed to ask child to exit\n"); |
4775fba1 SH |
1424 | |
1425 | if (!fail) | |
1426 | answer = true; | |
1427 | ||
1428 | out: | |
1429 | if (cpid != -1) | |
1430 | wait_for_pid(cpid); | |
1431 | if (sock[0] != -1) { | |
1432 | close(sock[0]); | |
1433 | close(sock[1]); | |
1434 | } | |
1435 | return answer; | |
1436 | } | |
1437 | ||
2ad6d2bd SH |
1438 | int cg_write(const char *path, const char *buf, size_t size, off_t offset, |
1439 | struct fuse_file_info *fi) | |
1440 | { | |
2ad6d2bd | 1441 | struct fuse_context *fc = fuse_get_context(); |
2c51f8dd SH |
1442 | char *localbuf = NULL; |
1443 | struct cgm_keys *k = NULL; | |
8f6e8f5e | 1444 | struct file_info *f = (struct file_info *)fi->fh; |
2c51f8dd | 1445 | bool r; |
2ad6d2bd | 1446 | |
443d13f5 | 1447 | if (f->type != LXC_TYPE_CGFILE) { |
b845ad01 SH |
1448 | fprintf(stderr, "Internal error: directory cache info used in cg_write\n"); |
1449 | return -EIO; | |
1450 | } | |
1451 | ||
2ad6d2bd | 1452 | if (offset) |
7253e0a4 | 1453 | return 0; |
2ad6d2bd SH |
1454 | |
1455 | if (!fc) | |
1456 | return -EIO; | |
1457 | ||
2c51f8dd | 1458 | localbuf = alloca(size+1); |
47cbf0e5 SH |
1459 | localbuf[size] = '\0'; |
1460 | memcpy(localbuf, buf, size); | |
2ad6d2bd | 1461 | |
2c51f8dd SH |
1462 | if ((k = get_cgroup_key(f->controller, f->cgroup, f->file)) == NULL) { |
1463 | size = -EINVAL; | |
1464 | goto out; | |
1465 | } | |
2ad6d2bd | 1466 | |
2c51f8dd SH |
1467 | if (!fc_may_access(fc, f->controller, f->cgroup, f->file, O_WRONLY)) { |
1468 | size = -EACCES; | |
1469 | goto out; | |
1470 | } | |
4775fba1 | 1471 | |
2c51f8dd SH |
1472 | if (strcmp(f->file, "tasks") == 0 || |
1473 | strcmp(f->file, "/tasks") == 0 || | |
1474 | strcmp(f->file, "/cgroup.procs") == 0 || | |
1475 | strcmp(f->file, "cgroup.procs") == 0) | |
1476 | // special case - we have to translate the pids | |
1477 | r = do_write_pids(fc->pid, f->controller, f->cgroup, f->file, localbuf); | |
1478 | else | |
1479 | r = cgm_set_value(f->controller, f->cgroup, f->file, localbuf); | |
2ad6d2bd | 1480 | |
2c51f8dd SH |
1481 | if (!r) |
1482 | size = -EINVAL; | |
2ad6d2bd | 1483 | |
2c51f8dd SH |
1484 | out: |
1485 | free_key(k); | |
1486 | return size; | |
2ad6d2bd SH |
1487 | } |
1488 | ||
341b21ad SH |
1489 | int cg_chown(const char *path, uid_t uid, gid_t gid) |
1490 | { | |
1491 | struct fuse_context *fc = fuse_get_context(); | |
2c51f8dd SH |
1492 | char *cgdir = NULL, *fpath = NULL, *path1, *path2, *controller; |
1493 | struct cgm_keys *k = NULL; | |
341b21ad | 1494 | const char *cgroup; |
2c51f8dd | 1495 | int ret; |
341b21ad SH |
1496 | |
1497 | if (!fc) | |
1498 | return -EIO; | |
1499 | ||
1500 | if (strcmp(path, "/cgroup") == 0) | |
1501 | return -EINVAL; | |
1502 | ||
1503 | controller = pick_controller_from_path(fc, path); | |
1504 | if (!controller) | |
f9a05025 | 1505 | return -EINVAL; |
341b21ad SH |
1506 | cgroup = find_cgroup_in_path(path); |
1507 | if (!cgroup) | |
1508 | /* this is just /cgroup/controller */ | |
1509 | return -EINVAL; | |
1510 | ||
1511 | get_cgdir_and_path(cgroup, &cgdir, &fpath); | |
1512 | ||
1513 | if (!fpath) { | |
1514 | path1 = "/"; | |
1515 | path2 = cgdir; | |
1516 | } else { | |
1517 | path1 = cgdir; | |
1518 | path2 = fpath; | |
1519 | } | |
1520 | ||
1521 | if (is_child_cgroup(controller, path1, path2)) { | |
1522 | // get uid, gid, from '/tasks' file and make up a mode | |
1523 | // That is a hack, until cgmanager gains a GetCgroupPerms fn. | |
1524 | k = get_cgroup_key(controller, cgroup, "tasks"); | |
1525 | ||
1526 | } else | |
1527 | k = get_cgroup_key(controller, path1, path2); | |
1528 | ||
2c51f8dd SH |
1529 | if (!k) { |
1530 | ret = -EINVAL; | |
1531 | goto out; | |
1532 | } | |
341b21ad SH |
1533 | |
1534 | /* | |
1535 | * This being a fuse request, the uid and gid must be valid | |
1536 | * in the caller's namespace. So we can just check to make | |
1537 | * sure that the caller is root in his uid, and privileged | |
1538 | * over the file's current owner. | |
1539 | */ | |
2c51f8dd SH |
1540 | if (!is_privileged_over(fc->pid, fc->uid, k->uid, NS_ROOT_REQD)) { |
1541 | ret = -EACCES; | |
1542 | goto out; | |
1543 | } | |
341b21ad | 1544 | |
2c51f8dd SH |
1545 | if (!cgm_chown_file(controller, cgroup, uid, gid)) { |
1546 | ret = -EINVAL; | |
1547 | goto out; | |
1548 | } | |
1549 | ||
1550 | ret = 0; | |
1551 | ||
1552 | out: | |
1553 | free_key(k); | |
1554 | free(cgdir); | |
1555 | ||
1556 | return ret; | |
341b21ad | 1557 | } |
2ad6d2bd | 1558 | |
fd2e4e03 SH |
1559 | int cg_chmod(const char *path, mode_t mode) |
1560 | { | |
0a1bb5ea | 1561 | struct fuse_context *fc = fuse_get_context(); |
2c51f8dd SH |
1562 | char * cgdir = NULL, *fpath = NULL, *path1, *path2, *controller; |
1563 | struct cgm_keys *k = NULL; | |
0a1bb5ea | 1564 | const char *cgroup; |
2c51f8dd | 1565 | int ret; |
0a1bb5ea SH |
1566 | |
1567 | if (!fc) | |
1568 | return -EIO; | |
1569 | ||
1570 | if (strcmp(path, "/cgroup") == 0) | |
1571 | return -EINVAL; | |
1572 | ||
1573 | controller = pick_controller_from_path(fc, path); | |
1574 | if (!controller) | |
f9a05025 | 1575 | return -EINVAL; |
0a1bb5ea SH |
1576 | cgroup = find_cgroup_in_path(path); |
1577 | if (!cgroup) | |
1578 | /* this is just /cgroup/controller */ | |
1579 | return -EINVAL; | |
1580 | ||
1581 | get_cgdir_and_path(cgroup, &cgdir, &fpath); | |
1582 | ||
1583 | if (!fpath) { | |
1584 | path1 = "/"; | |
1585 | path2 = cgdir; | |
1586 | } else { | |
1587 | path1 = cgdir; | |
1588 | path2 = fpath; | |
1589 | } | |
1590 | ||
1591 | if (is_child_cgroup(controller, path1, path2)) { | |
1592 | // get uid, gid, from '/tasks' file and make up a mode | |
1593 | // That is a hack, until cgmanager gains a GetCgroupPerms fn. | |
1594 | k = get_cgroup_key(controller, cgroup, "tasks"); | |
1595 | ||
1596 | } else | |
1597 | k = get_cgroup_key(controller, path1, path2); | |
1598 | ||
2c51f8dd SH |
1599 | if (!k) { |
1600 | ret = -EINVAL; | |
1601 | goto out; | |
1602 | } | |
0a1bb5ea SH |
1603 | |
1604 | /* | |
1605 | * This being a fuse request, the uid and gid must be valid | |
1606 | * in the caller's namespace. So we can just check to make | |
1607 | * sure that the caller is root in his uid, and privileged | |
1608 | * over the file's current owner. | |
1609 | */ | |
2c51f8dd SH |
1610 | if (!is_privileged_over(fc->pid, fc->uid, k->uid, NS_ROOT_OPT)) { |
1611 | ret = -EPERM; | |
1612 | goto out; | |
1613 | } | |
0a1bb5ea | 1614 | |
2c51f8dd SH |
1615 | if (!cgm_chmod_file(controller, cgroup, mode)) { |
1616 | ret = -EINVAL; | |
1617 | goto out; | |
1618 | } | |
1619 | ||
1620 | ret = 0; | |
1621 | out: | |
1622 | free_key(k); | |
1623 | free(cgdir); | |
1624 | return ret; | |
fd2e4e03 SH |
1625 | } |
1626 | ||
ab54b798 SH |
1627 | int cg_mkdir(const char *path, mode_t mode) |
1628 | { | |
1629 | struct fuse_context *fc = fuse_get_context(); | |
2c51f8dd | 1630 | char *fpath = NULL, *path1, *cgdir = NULL, *controller; |
ab54b798 | 1631 | const char *cgroup; |
2c51f8dd | 1632 | int ret; |
ab54b798 | 1633 | |
ab54b798 SH |
1634 | if (!fc) |
1635 | return -EIO; | |
1636 | ||
1637 | ||
1638 | controller = pick_controller_from_path(fc, path); | |
1639 | if (!controller) | |
f9a05025 | 1640 | return -EINVAL; |
ab54b798 SH |
1641 | |
1642 | cgroup = find_cgroup_in_path(path); | |
1643 | if (!cgroup) | |
f9a05025 | 1644 | return -EINVAL; |
ab54b798 SH |
1645 | |
1646 | get_cgdir_and_path(cgroup, &cgdir, &fpath); | |
1647 | if (!fpath) | |
1648 | path1 = "/"; | |
1649 | else | |
1650 | path1 = cgdir; | |
1651 | ||
2c51f8dd SH |
1652 | if (!fc_may_access(fc, controller, path1, NULL, O_RDWR)) { |
1653 | ret = -EACCES; | |
1654 | goto out; | |
1655 | } | |
1656 | if (!caller_is_in_ancestor(fc->pid, controller, path1, NULL)) { | |
1657 | ret = -EACCES; | |
1658 | goto out; | |
1659 | } | |
ab54b798 | 1660 | |
2c51f8dd SH |
1661 | if (fc->uid == 0 && fc->gid == 0) { |
1662 | if (!cgm_create(controller, cgroup)) { | |
1663 | ret = -EINVAL; | |
1664 | goto out; | |
1665 | } | |
1666 | } else { | |
1667 | /* | |
40b8c791 | 1668 | * exec a helper so as to get a clean dbus connection |
2c51f8dd SH |
1669 | * 17 for lxcfs_mkdir, and spaces and newline and \0. 50 for two ints. |
1670 | * 50 for two ints | |
1671 | */ | |
1672 | size_t len = strlen(cgroup) + strlen(controller) + 17 + 50; | |
1673 | char *cmd = alloca(len); | |
1674 | ret = snprintf(cmd, len, "lxcfs_mkdir %d %d %s %s\n", | |
1675 | fc->uid, fc->gid, controller, cgroup); | |
1676 | if (ret < 0 || ret >= len) { | |
1677 | ret = -EINVAL; | |
1678 | goto out; | |
1679 | } | |
1680 | ret = system(cmd); | |
1681 | if (ret != 0) | |
1682 | goto out; | |
1683 | } | |
ab54b798 | 1684 | |
2c51f8dd | 1685 | ret = 0; |
ab54b798 | 1686 | |
2c51f8dd SH |
1687 | out: |
1688 | free(cgdir); | |
1689 | return ret; | |
ab54b798 SH |
1690 | } |
1691 | ||
50d8d5b5 SH |
1692 | static int cg_rmdir(const char *path) |
1693 | { | |
1694 | struct fuse_context *fc = fuse_get_context(); | |
2c51f8dd | 1695 | char *fpath = NULL, *cgdir = NULL, *controller; |
50d8d5b5 | 1696 | const char *cgroup; |
2c51f8dd | 1697 | int ret; |
50d8d5b5 SH |
1698 | |
1699 | if (!fc) | |
1700 | return -EIO; | |
1701 | ||
50d8d5b5 SH |
1702 | controller = pick_controller_from_path(fc, path); |
1703 | if (!controller) | |
f9a05025 | 1704 | return -EINVAL; |
50d8d5b5 SH |
1705 | |
1706 | cgroup = find_cgroup_in_path(path); | |
1707 | if (!cgroup) | |
f9a05025 | 1708 | return -EINVAL; |
50d8d5b5 SH |
1709 | |
1710 | get_cgdir_and_path(cgroup, &cgdir, &fpath); | |
2c51f8dd SH |
1711 | if (!fpath) { |
1712 | ret = -EINVAL; | |
1713 | goto out; | |
1714 | } | |
50d8d5b5 | 1715 | |
2c51f8dd SH |
1716 | fprintf(stderr, "rmdir: verifying access to %s:%s (req path %s)\n", |
1717 | controller, cgdir, path); | |
1718 | if (!fc_may_access(fc, controller, cgdir, NULL, O_WRONLY)) { | |
1719 | ret = -EACCES; | |
1720 | goto out; | |
1721 | } | |
1722 | if (!caller_is_in_ancestor(fc->pid, controller, cgroup, NULL)) { | |
1723 | ret = -EACCES; | |
1724 | goto out; | |
1725 | } | |
50d8d5b5 | 1726 | |
2c51f8dd SH |
1727 | if (!cgm_remove(controller, cgroup)) { |
1728 | ret = -EINVAL; | |
1729 | goto out; | |
1730 | } | |
50d8d5b5 | 1731 | |
2c51f8dd SH |
1732 | ret = 0; |
1733 | ||
1734 | out: | |
1735 | free(cgdir); | |
1736 | return ret; | |
50d8d5b5 SH |
1737 | } |
1738 | ||
2dc17609 SH |
1739 | static bool startswith(const char *line, const char *pref) |
1740 | { | |
1741 | if (strncmp(line, pref, strlen(pref)) == 0) | |
1742 | return true; | |
1743 | return false; | |
1744 | } | |
1745 | ||
1746 | static void get_mem_cached(char *memstat, unsigned long *v) | |
1747 | { | |
1748 | char *eol; | |
1749 | ||
1750 | *v = 0; | |
1751 | while (*memstat) { | |
1752 | if (startswith(memstat, "total_cache")) { | |
1753 | sscanf(memstat + 11, "%lu", v); | |
1754 | *v /= 1024; | |
1755 | return; | |
1756 | } | |
1757 | eol = strchr(memstat, '\n'); | |
1758 | if (!eol) | |
1759 | return; | |
1760 | memstat = eol+1; | |
1761 | } | |
1762 | } | |
1763 | ||
49878439 | 1764 | static void get_blkio_io_value(char *str, unsigned major, unsigned minor, char *iotype, unsigned long *v) |
2f919d9d | 1765 | { |
49878439 YY |
1766 | char *eol; |
1767 | char key[32]; | |
2f919d9d | 1768 | |
49878439 YY |
1769 | memset(key, 0, 32); |
1770 | snprintf(key, 32, "%u:%u %s", major, minor, iotype); | |
2f919d9d | 1771 | |
49878439 YY |
1772 | size_t len = strlen(key); |
1773 | *v = 0; | |
1774 | ||
1775 | while (*str) { | |
1776 | if (startswith(str, key)) { | |
2f919d9d SH |
1777 | sscanf(str + len, "%lu", v); |
1778 | return; | |
1779 | } | |
1780 | eol = strchr(str, '\n'); | |
49878439 | 1781 | if (!eol) |
2f919d9d | 1782 | return; |
49878439 YY |
1783 | str = eol+1; |
1784 | } | |
1785 | } | |
1786 | ||
53b43826 SH |
1787 | static int read_file(const char *path, char *buf, size_t size, |
1788 | struct file_info *d) | |
1789 | { | |
1790 | size_t linelen = 0, total_len = 0, rv = 0; | |
1791 | char *line = NULL; | |
1792 | char *cache = d->buf; | |
1793 | size_t cache_size = d->buflen; | |
1794 | FILE *f = fopen(path, "r"); | |
1795 | if (!f) | |
1796 | return 0; | |
1797 | ||
1798 | while (getline(&line, &linelen, f) != -1) { | |
1799 | size_t l = snprintf(cache, cache_size, "%s", line); | |
1800 | if (l < 0) { | |
1801 | perror("Error writing to cache"); | |
1802 | rv = 0; | |
1803 | goto err; | |
1804 | } | |
1805 | if (l >= cache_size) { | |
1806 | fprintf(stderr, "Internal error: truncated write to cache\n"); | |
1807 | rv = 0; | |
1808 | goto err; | |
1809 | } | |
1810 | if (l < cache_size) { | |
1811 | cache += l; | |
1812 | cache_size -= l; | |
1813 | total_len += l; | |
1814 | } else { | |
1815 | cache += cache_size; | |
1816 | total_len += cache_size; | |
1817 | cache_size = 0; | |
1818 | break; | |
1819 | } | |
1820 | } | |
1821 | ||
1822 | d->size = total_len; | |
1823 | if (total_len > size ) total_len = size; | |
1824 | ||
1825 | /* read from off 0 */ | |
1826 | memcpy(buf, d->buf, total_len); | |
1827 | rv = total_len; | |
1828 | err: | |
1829 | fclose(f); | |
1830 | free(line); | |
1831 | return rv; | |
1832 | } | |
1833 | ||
758ad80c | 1834 | /* |
2ad6d2bd | 1835 | * FUSE ops for /proc |
758ad80c | 1836 | */ |
758ad80c | 1837 | |
23ce2127 SH |
1838 | static int proc_meminfo_read(char *buf, size_t size, off_t offset, |
1839 | struct fuse_file_info *fi) | |
1840 | { | |
2dc17609 | 1841 | struct fuse_context *fc = fuse_get_context(); |
97f1f27b | 1842 | struct file_info *d = (struct file_info *)fi->fh; |
2c51f8dd SH |
1843 | char *cg; |
1844 | char *memlimit_str = NULL, *memusage_str = NULL, *memstat_str = NULL; | |
2dc17609 SH |
1845 | unsigned long memlimit = 0, memusage = 0, cached = 0, hosttotal = 0; |
1846 | char *line = NULL; | |
e1068397 | 1847 | size_t linelen = 0, total_len = 0, rv = 0; |
97f1f27b YY |
1848 | char *cache = d->buf; |
1849 | size_t cache_size = d->buflen; | |
2c51f8dd | 1850 | FILE *f = NULL; |
2dc17609 | 1851 | |
97f1f27b YY |
1852 | if (offset){ |
1853 | if (offset > d->size) | |
1854 | return -EINVAL; | |
1855 | int left = d->size - offset; | |
1856 | total_len = left > size ? size: left; | |
1857 | memcpy(buf, cache + offset, total_len); | |
1858 | return total_len; | |
1859 | } | |
2dc17609 | 1860 | |
2c51f8dd | 1861 | cg = get_pid_cgroup(fc->pid, "memory"); |
2dc17609 | 1862 | if (!cg) |
53b43826 | 1863 | return read_file("/proc/meminfo", buf, size, d); |
2dc17609 SH |
1864 | |
1865 | if (!cgm_get_value("memory", cg, "memory.limit_in_bytes", &memlimit_str)) | |
2c51f8dd | 1866 | goto err; |
2dc17609 | 1867 | if (!cgm_get_value("memory", cg, "memory.usage_in_bytes", &memusage_str)) |
2c51f8dd | 1868 | goto err; |
2dc17609 | 1869 | if (!cgm_get_value("memory", cg, "memory.stat", &memstat_str)) |
2c51f8dd | 1870 | goto err; |
2dc17609 SH |
1871 | memlimit = strtoul(memlimit_str, NULL, 10); |
1872 | memusage = strtoul(memusage_str, NULL, 10); | |
1873 | memlimit /= 1024; | |
1874 | memusage /= 1024; | |
1875 | get_mem_cached(memstat_str, &cached); | |
1876 | ||
1877 | f = fopen("/proc/meminfo", "r"); | |
1878 | if (!f) | |
2c51f8dd | 1879 | goto err; |
2dc17609 SH |
1880 | |
1881 | while (getline(&line, &linelen, f) != -1) { | |
1882 | size_t l; | |
1883 | char *printme, lbuf[100]; | |
1884 | ||
1885 | memset(lbuf, 0, 100); | |
1886 | if (startswith(line, "MemTotal:")) { | |
1887 | sscanf(line+14, "%lu", &hosttotal); | |
1888 | if (hosttotal < memlimit) | |
1889 | memlimit = hosttotal; | |
1890 | snprintf(lbuf, 100, "MemTotal: %8lu kB\n", memlimit); | |
1891 | printme = lbuf; | |
1892 | } else if (startswith(line, "MemFree:")) { | |
1893 | snprintf(lbuf, 100, "MemFree: %8lu kB\n", memlimit - memusage); | |
1894 | printme = lbuf; | |
1895 | } else if (startswith(line, "MemAvailable:")) { | |
1896 | snprintf(lbuf, 100, "MemAvailable: %8lu kB\n", memlimit - memusage); | |
1897 | printme = lbuf; | |
1898 | } else if (startswith(line, "Buffers:")) { | |
1899 | snprintf(lbuf, 100, "Buffers: %8lu kB\n", 0UL); | |
1900 | printme = lbuf; | |
1901 | } else if (startswith(line, "Cached:")) { | |
1902 | snprintf(lbuf, 100, "Cached: %8lu kB\n", cached); | |
1903 | printme = lbuf; | |
1904 | } else if (startswith(line, "SwapCached:")) { | |
1905 | snprintf(lbuf, 100, "SwapCached: %8lu kB\n", 0UL); | |
1906 | printme = lbuf; | |
1907 | } else | |
1908 | printme = line; | |
97f1f27b YY |
1909 | |
1910 | l = snprintf(cache, cache_size, "%s", printme); | |
e1068397 MM |
1911 | if (l < 0) { |
1912 | perror("Error writing to cache"); | |
1913 | rv = 0; | |
1914 | goto err; | |
1915 | ||
1916 | } | |
1917 | if (l >= cache_size) { | |
1918 | fprintf(stderr, "Internal error: truncated write to cache\n"); | |
1919 | rv = 0; | |
1920 | goto err; | |
1921 | } | |
1922 | ||
97f1f27b YY |
1923 | cache += l; |
1924 | cache_size -= l; | |
2f919d9d | 1925 | total_len += l; |
2dc17609 SH |
1926 | } |
1927 | ||
97f1f27b YY |
1928 | d->size = total_len; |
1929 | if (total_len > size ) total_len = size; | |
1930 | memcpy(buf, d->buf, total_len); | |
1931 | ||
e1068397 | 1932 | rv = total_len; |
2c51f8dd SH |
1933 | err: |
1934 | if (f) | |
1935 | fclose(f); | |
92c84dc4 | 1936 | free(line); |
2c51f8dd SH |
1937 | free(cg); |
1938 | free(memlimit_str); | |
1939 | free(memusage_str); | |
1940 | free(memstat_str); | |
e1068397 | 1941 | return rv; |
23ce2127 SH |
1942 | } |
1943 | ||
1944 | /* | |
1945 | * Read the cpuset.cpus for cg | |
2c51f8dd | 1946 | * Return the answer in a newly allocated string which must be freed |
23ce2127 SH |
1947 | */ |
1948 | static char *get_cpuset(const char *cg) | |
1949 | { | |
1950 | char *answer; | |
1951 | ||
1952 | if (!cgm_get_value("cpuset", cg, "cpuset.cpus", &answer)) | |
1953 | return NULL; | |
1954 | return answer; | |
1955 | } | |
1956 | ||
fa47bb52 | 1957 | bool cpu_in_cpuset(int cpu, const char *cpuset); |
23ce2127 | 1958 | |
aeb56147 SH |
1959 | static bool cpuline_in_cpuset(const char *line, const char *cpuset) |
1960 | { | |
1961 | int cpu; | |
1962 | ||
1963 | if (sscanf(line, "processor : %d", &cpu) != 1) | |
1964 | return false; | |
1965 | return cpu_in_cpuset(cpu, cpuset); | |
1966 | } | |
1967 | ||
23ce2127 SH |
1968 | /* |
1969 | * check whether this is a '^processor" line in /proc/cpuinfo | |
1970 | */ | |
1971 | static bool is_processor_line(const char *line) | |
1972 | { | |
1973 | int cpu; | |
1974 | ||
1975 | if (sscanf(line, "processor : %d", &cpu) == 1) | |
1976 | return true; | |
1977 | return false; | |
1978 | } | |
1979 | ||
23ce2127 SH |
1980 | static int proc_cpuinfo_read(char *buf, size_t size, off_t offset, |
1981 | struct fuse_file_info *fi) | |
1982 | { | |
1983 | struct fuse_context *fc = fuse_get_context(); | |
97f1f27b | 1984 | struct file_info *d = (struct file_info *)fi->fh; |
2c51f8dd SH |
1985 | char *cg; |
1986 | char *cpuset = NULL; | |
23ce2127 | 1987 | char *line = NULL; |
e1068397 | 1988 | size_t linelen = 0, total_len = 0, rv = 0; |
23ce2127 SH |
1989 | bool am_printing = false; |
1990 | int curcpu = -1; | |
97f1f27b YY |
1991 | char *cache = d->buf; |
1992 | size_t cache_size = d->buflen; | |
2c51f8dd | 1993 | FILE *f = NULL; |
23ce2127 | 1994 | |
97f1f27b YY |
1995 | if (offset){ |
1996 | if (offset > d->size) | |
1997 | return -EINVAL; | |
1998 | int left = d->size - offset; | |
1999 | total_len = left > size ? size: left; | |
2000 | memcpy(buf, cache + offset, total_len); | |
2f919d9d | 2001 | return total_len; |
97f1f27b | 2002 | } |
23ce2127 | 2003 | |
2c51f8dd | 2004 | cg = get_pid_cgroup(fc->pid, "cpuset"); |
23ce2127 | 2005 | if (!cg) |
53b43826 | 2006 | return read_file("proc/cpuinfo", buf, size, d); |
23ce2127 SH |
2007 | |
2008 | cpuset = get_cpuset(cg); | |
2009 | if (!cpuset) | |
2c51f8dd | 2010 | goto err; |
23ce2127 SH |
2011 | |
2012 | f = fopen("/proc/cpuinfo", "r"); | |
2013 | if (!f) | |
2c51f8dd | 2014 | goto err; |
23ce2127 SH |
2015 | |
2016 | while (getline(&line, &linelen, f) != -1) { | |
2017 | size_t l; | |
2018 | if (is_processor_line(line)) { | |
aeb56147 | 2019 | am_printing = cpuline_in_cpuset(line, cpuset); |
23ce2127 SH |
2020 | if (am_printing) { |
2021 | curcpu ++; | |
97f1f27b | 2022 | l = snprintf(cache, cache_size, "processor : %d\n", curcpu); |
e1068397 MM |
2023 | if (l < 0) { |
2024 | perror("Error writing to cache"); | |
2025 | rv = 0; | |
2026 | goto err; | |
2027 | } | |
2028 | if (l >= cache_size) { | |
2029 | fprintf(stderr, "Internal error: truncated write to cache\n"); | |
2030 | rv = 0; | |
2031 | goto err; | |
2032 | } | |
97f1f27b YY |
2033 | if (l < cache_size){ |
2034 | cache += l; | |
2035 | cache_size -= l; | |
2036 | total_len += l; | |
2037 | }else{ | |
2038 | cache += cache_size; | |
2039 | total_len += cache_size; | |
2040 | cache_size = 0; | |
2041 | break; | |
2042 | } | |
23ce2127 SH |
2043 | } |
2044 | continue; | |
2045 | } | |
2046 | if (am_printing) { | |
97f1f27b | 2047 | l = snprintf(cache, cache_size, "%s", line); |
e1068397 MM |
2048 | if (l < 0) { |
2049 | perror("Error writing to cache"); | |
2050 | rv = 0; | |
2051 | goto err; | |
2052 | } | |
2053 | if (l >= cache_size) { | |
2054 | fprintf(stderr, "Internal error: truncated write to cache\n"); | |
2055 | rv = 0; | |
2056 | goto err; | |
2057 | } | |
97f1f27b YY |
2058 | if (l < cache_size) { |
2059 | cache += l; | |
2060 | cache_size -= l; | |
2061 | total_len += l; | |
2062 | } else { | |
2063 | cache += cache_size; | |
2064 | total_len += cache_size; | |
2065 | cache_size = 0; | |
2066 | break; | |
2067 | } | |
23ce2127 SH |
2068 | } |
2069 | } | |
2070 | ||
97f1f27b YY |
2071 | d->size = total_len; |
2072 | if (total_len > size ) total_len = size; | |
2073 | ||
2074 | /* read from off 0 */ | |
2075 | memcpy(buf, d->buf, total_len); | |
e1068397 | 2076 | rv = total_len; |
2c51f8dd SH |
2077 | err: |
2078 | if (f) | |
2079 | fclose(f); | |
92c84dc4 | 2080 | free(line); |
2c51f8dd SH |
2081 | free(cpuset); |
2082 | free(cg); | |
e1068397 | 2083 | return rv; |
23ce2127 SH |
2084 | } |
2085 | ||
2086 | static int proc_stat_read(char *buf, size_t size, off_t offset, | |
2087 | struct fuse_file_info *fi) | |
2088 | { | |
aeb56147 | 2089 | struct fuse_context *fc = fuse_get_context(); |
97f1f27b | 2090 | struct file_info *d = (struct file_info *)fi->fh; |
2c51f8dd SH |
2091 | char *cg; |
2092 | char *cpuset = NULL; | |
aeb56147 | 2093 | char *line = NULL; |
e1068397 | 2094 | size_t linelen = 0, total_len = 0, rv = 0; |
2a0fde62 | 2095 | int curcpu = -1; /* cpu numbering starts at 0 */ |
97f1f27b YY |
2096 | unsigned long user = 0, nice = 0, system = 0, idle = 0, iowait = 0, irq = 0, softirq = 0, steal = 0, guest = 0; |
2097 | unsigned long user_sum = 0, nice_sum = 0, system_sum = 0, idle_sum = 0, iowait_sum = 0, | |
2098 | irq_sum = 0, softirq_sum = 0, steal_sum = 0, guest_sum = 0; | |
2099 | #define CPUALL_MAX_SIZE BUF_RESERVE_SIZE | |
2100 | char cpuall[CPUALL_MAX_SIZE]; | |
2101 | /* reserve for cpu all */ | |
2102 | char *cache = d->buf + CPUALL_MAX_SIZE; | |
2103 | size_t cache_size = d->buflen - CPUALL_MAX_SIZE; | |
2c51f8dd | 2104 | FILE *f = NULL; |
aeb56147 | 2105 | |
97f1f27b YY |
2106 | if (offset){ |
2107 | if (offset > d->size) | |
2108 | return -EINVAL; | |
2109 | int left = d->size - offset; | |
2110 | total_len = left > size ? size: left; | |
2111 | memcpy(buf, d->buf + offset, total_len); | |
2f919d9d | 2112 | return total_len; |
97f1f27b | 2113 | } |
aeb56147 | 2114 | |
2c51f8dd | 2115 | cg = get_pid_cgroup(fc->pid, "cpuset"); |
aeb56147 | 2116 | if (!cg) |
53b43826 | 2117 | return read_file("/proc/stat", buf, size, d); |
aeb56147 SH |
2118 | |
2119 | cpuset = get_cpuset(cg); | |
2120 | if (!cpuset) | |
2c51f8dd | 2121 | goto err; |
aeb56147 SH |
2122 | |
2123 | f = fopen("/proc/stat", "r"); | |
2124 | if (!f) | |
2c51f8dd | 2125 | goto err; |
aeb56147 | 2126 | |
97f1f27b YY |
2127 | //skip first line |
2128 | if (getline(&line, &linelen, f) < 0) { | |
2129 | fprintf(stderr, "proc_stat_read read first line failed\n"); | |
2c51f8dd | 2130 | goto err; |
97f1f27b YY |
2131 | } |
2132 | ||
aeb56147 SH |
2133 | while (getline(&line, &linelen, f) != -1) { |
2134 | size_t l; | |
2135 | int cpu; | |
2a0fde62 | 2136 | char cpu_char[10]; /* That's a lot of cores */ |
aeb56147 SH |
2137 | char *c; |
2138 | ||
2a0fde62 CB |
2139 | if (sscanf(line, "cpu%9[^ ]", cpu_char) != 1) { |
2140 | /* not a ^cpuN line containing a number N, just print it */ | |
97f1f27b | 2141 | l = snprintf(cache, cache_size, "%s", line); |
e1068397 MM |
2142 | if (l < 0) { |
2143 | perror("Error writing to cache"); | |
2144 | rv = 0; | |
2145 | goto err; | |
2146 | } | |
2147 | if (l >= cache_size) { | |
2148 | fprintf(stderr, "Internal error: truncated write to cache\n"); | |
2149 | rv = 0; | |
2150 | goto err; | |
2151 | } | |
2152 | if (l < cache_size) { | |
97f1f27b YY |
2153 | cache += l; |
2154 | cache_size -= l; | |
2155 | total_len += l; | |
2156 | continue; | |
e1068397 | 2157 | } else { |
97f1f27b YY |
2158 | //no more space, break it |
2159 | cache += cache_size; | |
2160 | total_len += cache_size; | |
2161 | cache_size = 0; | |
2162 | break; | |
2163 | } | |
aeb56147 | 2164 | } |
2a0fde62 CB |
2165 | |
2166 | if (sscanf(cpu_char, "%d", &cpu) != 1) | |
2167 | continue; | |
aeb56147 SH |
2168 | if (!cpu_in_cpuset(cpu, cpuset)) |
2169 | continue; | |
2170 | curcpu ++; | |
2171 | ||
2172 | c = strchr(line, ' '); | |
2173 | if (!c) | |
2174 | continue; | |
25c5e8fb | 2175 | l = snprintf(cache, cache_size, "cpu%d%s", curcpu, c); |
e1068397 MM |
2176 | if (l < 0) { |
2177 | perror("Error writing to cache"); | |
2178 | rv = 0; | |
2179 | goto err; | |
2180 | ||
2181 | } | |
2182 | if (l >= cache_size) { | |
2183 | fprintf(stderr, "Internal error: truncated write to cache\n"); | |
2184 | rv = 0; | |
2185 | goto err; | |
2186 | } | |
2187 | ||
97f1f27b YY |
2188 | cache += l; |
2189 | cache_size -= l; | |
aeb56147 | 2190 | total_len += l; |
2f919d9d | 2191 | |
97f1f27b YY |
2192 | if (sscanf(line, "%*s %lu %lu %lu %lu %lu %lu %lu %lu %lu", &user, &nice, &system, &idle, &iowait, &irq, |
2193 | &softirq, &steal, &guest) != 9) | |
2194 | continue; | |
2195 | user_sum += user; | |
2196 | nice_sum += nice; | |
2197 | system_sum += system; | |
2198 | idle_sum += idle; | |
2199 | iowait_sum += iowait; | |
2200 | irq_sum += irq; | |
2201 | softirq_sum += softirq; | |
2202 | steal_sum += steal; | |
2f919d9d | 2203 | guest_sum += guest; |
97f1f27b YY |
2204 | } |
2205 | ||
2206 | cache = d->buf; | |
2207 | ||
2f919d9d | 2208 | int cpuall_len = snprintf(cpuall, CPUALL_MAX_SIZE, "%s %lu %lu %lu %lu %lu %lu %lu %lu %lu\n", |
97f1f27b YY |
2209 | "cpu ", user_sum, nice_sum, system_sum, idle_sum, iowait_sum, irq_sum, softirq_sum, steal_sum, guest_sum); |
2210 | if (cpuall_len > 0 && cpuall_len < CPUALL_MAX_SIZE){ | |
2211 | memcpy(cache, cpuall, cpuall_len); | |
2f919d9d | 2212 | cache += cpuall_len; |
2c51f8dd | 2213 | } else{ |
97f1f27b YY |
2214 | /* shouldn't happen */ |
2215 | fprintf(stderr, "proc_stat_read copy cpuall failed, cpuall_len=%d\n", cpuall_len); | |
2216 | cpuall_len = 0; | |
2217 | } | |
2218 | ||
2219 | memmove(cache, d->buf + CPUALL_MAX_SIZE, total_len); | |
2220 | total_len += cpuall_len; | |
2221 | d->size = total_len; | |
2222 | if (total_len > size ) total_len = size; | |
2223 | ||
2224 | memcpy(buf, d->buf, total_len); | |
e1068397 | 2225 | rv = total_len; |
2c51f8dd SH |
2226 | |
2227 | err: | |
2228 | if (f) | |
2229 | fclose(f); | |
92c84dc4 | 2230 | free(line); |
2c51f8dd SH |
2231 | free(cpuset); |
2232 | free(cg); | |
e1068397 | 2233 | return rv; |
23ce2127 SH |
2234 | } |
2235 | ||
7bbf2246 SH |
2236 | /* |
2237 | * How to guess what to present for uptime? | |
2238 | * One thing we could do would be to take the date on the caller's | |
2239 | * memory.usage_in_bytes file, which should equal the time of creation | |
2240 | * of his cgroup. However, a task could be in a sub-cgroup of the | |
2241 | * container. The same problem exists if we try to look at the ages | |
2242 | * of processes in the caller's cgroup. | |
2243 | * | |
2244 | * So we'll fork a task that will enter the caller's pidns, mount a | |
2245 | * fresh procfs, get the age of /proc/1, and pass that back over a pipe. | |
2246 | * | |
2247 | * For the second uptime #, we'll do as Stéphane had done, just copy | |
2248 | * the number from /proc/uptime. Not sure how to best emulate 'idle' | |
2249 | * time. Maybe someone can come up with a good algorithm and submit a | |
2250 | * patch. Maybe something based on cpushare info? | |
2251 | */ | |
41bb9357 SH |
2252 | |
2253 | /* return age of the reaper for $pid, taken from ctime of its procdir */ | |
2254 | static long int get_pid1_time(pid_t pid) | |
2255 | { | |
2256 | char fnam[100]; | |
ea56f722 | 2257 | int fd, cpipe[2], ret; |
41bb9357 | 2258 | struct stat sb; |
ea56f722 SH |
2259 | pid_t cpid; |
2260 | struct timeval tv; | |
2261 | fd_set s; | |
2262 | char v; | |
41bb9357 SH |
2263 | |
2264 | if (unshare(CLONE_NEWNS)) | |
2265 | return 0; | |
2266 | ||
5ca64c2a SG |
2267 | if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL)) { |
2268 | perror("rslave mount failed"); | |
2269 | return 0; | |
2270 | } | |
2271 | ||
c0adec85 SH |
2272 | ret = snprintf(fnam, sizeof(fnam), "/proc/%d/ns/pid", pid); |
2273 | if (ret < 0 || ret >= sizeof(fnam)) | |
2274 | return 0; | |
2275 | ||
41bb9357 SH |
2276 | fd = open(fnam, O_RDONLY); |
2277 | if (fd < 0) { | |
2278 | perror("get_pid1_time open of ns/pid"); | |
2279 | return 0; | |
2280 | } | |
2281 | if (setns(fd, 0)) { | |
2282 | perror("get_pid1_time setns 1"); | |
2283 | close(fd); | |
2284 | return 0; | |
2285 | } | |
2286 | close(fd); | |
41bb9357 | 2287 | |
ea56f722 SH |
2288 | if (pipe(cpipe) < 0) |
2289 | exit(1); | |
41bb9357 | 2290 | |
ea56f722 SH |
2291 | loop: |
2292 | cpid = fork(); | |
2293 | if (cpid < 0) | |
41bb9357 | 2294 | return 0; |
ea56f722 SH |
2295 | |
2296 | if (!cpid) { | |
2297 | char b = '1'; | |
2298 | close(cpipe[0]); | |
2299 | if (write(cpipe[1], &b, sizeof(char)) < 0) { | |
2300 | fprintf(stderr, "%s (child): erorr on write: %s\n", | |
2301 | __func__, strerror(errno)); | |
2302 | } | |
2303 | close(cpipe[1]); | |
2304 | umount2("/proc", MNT_DETACH); | |
2305 | if (mount("proc", "/proc", "proc", 0, NULL)) { | |
2306 | perror("get_pid1_time mount"); | |
2307 | return 0; | |
2308 | } | |
2309 | ret = lstat("/proc/1", &sb); | |
2310 | if (ret) { | |
2311 | perror("get_pid1_time lstat"); | |
2312 | return 0; | |
2313 | } | |
2314 | return time(NULL) - sb.st_ctime; | |
41bb9357 | 2315 | } |
ea56f722 SH |
2316 | |
2317 | // give the child 1 second to be done forking and | |
2318 | // write it's ack | |
2319 | FD_ZERO(&s); | |
2320 | FD_SET(cpipe[0], &s); | |
2321 | tv.tv_sec = 1; | |
2322 | tv.tv_usec = 0; | |
2323 | ret = select(cpipe[0]+1, &s, NULL, NULL, &tv); | |
2324 | if (ret <= 0) | |
2325 | goto again; | |
2326 | ret = read(cpipe[0], &v, 1); | |
2327 | if (ret != sizeof(char) || v != '1') { | |
2328 | goto again; | |
41bb9357 | 2329 | } |
ea56f722 SH |
2330 | |
2331 | wait_for_pid(cpid); | |
67bd113f | 2332 | _exit(0); |
ea56f722 SH |
2333 | |
2334 | again: | |
2335 | kill(cpid, SIGKILL); | |
2336 | wait_for_pid(cpid); | |
2337 | goto loop; | |
41bb9357 SH |
2338 | } |
2339 | ||
2340 | static long int getreaperage(pid_t qpid) | |
2341 | { | |
2342 | int pid, mypipe[2], ret; | |
2343 | struct timeval tv; | |
2344 | fd_set s; | |
2345 | long int mtime, answer = 0; | |
2346 | ||
2347 | if (pipe(mypipe)) { | |
2348 | return 0; | |
2349 | } | |
2350 | ||
2351 | pid = fork(); | |
2352 | ||
2353 | if (!pid) { // child | |
2354 | mtime = get_pid1_time(qpid); | |
2355 | if (write(mypipe[1], &mtime, sizeof(mtime)) != sizeof(mtime)) | |
2356 | fprintf(stderr, "Warning: bad write from getreaperage\n"); | |
67bd113f | 2357 | _exit(0); |
41bb9357 SH |
2358 | } |
2359 | ||
2360 | close(mypipe[1]); | |
2361 | FD_ZERO(&s); | |
2362 | FD_SET(mypipe[0], &s); | |
2363 | tv.tv_sec = 1; | |
2364 | tv.tv_usec = 0; | |
2365 | ret = select(mypipe[0]+1, &s, NULL, NULL, &tv); | |
ea56f722 | 2366 | if (ret <= 0) { |
41bb9357 SH |
2367 | perror("select"); |
2368 | goto out; | |
2369 | } | |
2370 | if (!ret) { | |
1420baf8 | 2371 | fprintf(stderr, "timed out\n"); |
41bb9357 SH |
2372 | goto out; |
2373 | } | |
2374 | if (read(mypipe[0], &mtime, sizeof(mtime)) != sizeof(mtime)) { | |
2375 | perror("read"); | |
2376 | goto out; | |
2377 | } | |
2378 | answer = mtime; | |
2379 | ||
2380 | out: | |
2381 | wait_for_pid(pid); | |
2382 | close(mypipe[0]); | |
2383 | return answer; | |
2384 | } | |
2385 | ||
2386 | static long int getprocidle(void) | |
2387 | { | |
2388 | FILE *f = fopen("/proc/uptime", "r"); | |
2389 | long int age, idle; | |
92c84dc4 | 2390 | int ret; |
41bb9357 SH |
2391 | if (!f) |
2392 | return 0; | |
92c84dc4 SH |
2393 | ret = fscanf(f, "%ld %ld", &age, &idle); |
2394 | fclose(f); | |
2395 | if (ret != 2) | |
41bb9357 SH |
2396 | return 0; |
2397 | return idle; | |
2398 | } | |
2399 | ||
2400 | /* | |
2401 | * We read /proc/uptime and reuse its second field. | |
2402 | * For the first field, we use the mtime for the reaper for | |
2403 | * the calling pid as returned by getreaperage | |
2404 | */ | |
23ce2127 SH |
2405 | static int proc_uptime_read(char *buf, size_t size, off_t offset, |
2406 | struct fuse_file_info *fi) | |
2407 | { | |
41bb9357 | 2408 | struct fuse_context *fc = fuse_get_context(); |
97f1f27b | 2409 | struct file_info *d = (struct file_info *)fi->fh; |
41bb9357 SH |
2410 | long int reaperage = getreaperage(fc->pid);; |
2411 | long int idletime = getprocidle(); | |
97f1f27b | 2412 | size_t total_len = 0; |
41bb9357 | 2413 | |
97f1f27b YY |
2414 | if (offset){ |
2415 | if (offset > d->size) | |
2416 | return -EINVAL; | |
2417 | return 0; | |
2418 | } | |
2419 | ||
2420 | total_len = snprintf(buf, size, "%ld %ld\n", reaperage, idletime); | |
e1068397 MM |
2421 | if (total_len < 0){ |
2422 | perror("Error writing to cache"); | |
2423 | return 0; | |
2424 | } | |
cdcdb29b MM |
2425 | if (total_len >= size){ |
2426 | d->size = size; | |
2427 | return size; | |
2428 | } | |
2429 | ||
97f1f27b YY |
2430 | d->size = total_len; |
2431 | return total_len; | |
23ce2127 SH |
2432 | } |
2433 | ||
49878439 YY |
2434 | static int proc_diskstats_read(char *buf, size_t size, off_t offset, |
2435 | struct fuse_file_info *fi) | |
2436 | { | |
2437 | char dev_name[72]; | |
2438 | struct fuse_context *fc = fuse_get_context(); | |
97f1f27b | 2439 | struct file_info *d = (struct file_info *)fi->fh; |
2c51f8dd SH |
2440 | char *cg; |
2441 | char *io_serviced_str = NULL, *io_merged_str = NULL, *io_service_bytes_str = NULL, | |
49878439 YY |
2442 | *io_wait_time_str = NULL, *io_service_time_str = NULL; |
2443 | unsigned long read = 0, write = 0; | |
2444 | unsigned long read_merged = 0, write_merged = 0; | |
2445 | unsigned long read_sectors = 0, write_sectors = 0; | |
2446 | unsigned long read_ticks = 0, write_ticks = 0; | |
2447 | unsigned long ios_pgr = 0, tot_ticks = 0, rq_ticks = 0; | |
2448 | unsigned long rd_svctm = 0, wr_svctm = 0, rd_wait = 0, wr_wait = 0; | |
2449 | char *line = NULL; | |
e1068397 | 2450 | size_t linelen = 0, total_len = 0, rv = 0; |
49878439 YY |
2451 | unsigned int major = 0, minor = 0; |
2452 | int i = 0; | |
2c51f8dd | 2453 | FILE *f = NULL; |
49878439 | 2454 | |
97f1f27b YY |
2455 | if (offset){ |
2456 | if (offset > d->size) | |
2457 | return -EINVAL; | |
2458 | return 0; | |
2459 | } | |
49878439 | 2460 | |
2c51f8dd | 2461 | cg = get_pid_cgroup(fc->pid, "blkio"); |
49878439 | 2462 | if (!cg) |
53b43826 | 2463 | return read_file("/proc/diskstats", buf, size, d); |
49878439 YY |
2464 | |
2465 | if (!cgm_get_value("blkio", cg, "blkio.io_serviced", &io_serviced_str)) | |
2c51f8dd | 2466 | goto err; |
49878439 | 2467 | if (!cgm_get_value("blkio", cg, "blkio.io_merged", &io_merged_str)) |
2c51f8dd | 2468 | goto err; |
49878439 | 2469 | if (!cgm_get_value("blkio", cg, "blkio.io_service_bytes", &io_service_bytes_str)) |
2c51f8dd | 2470 | goto err; |
49878439 | 2471 | if (!cgm_get_value("blkio", cg, "blkio.io_wait_time", &io_wait_time_str)) |
2c51f8dd | 2472 | goto err; |
49878439 | 2473 | if (!cgm_get_value("blkio", cg, "blkio.io_service_time", &io_service_time_str)) |
2c51f8dd | 2474 | goto err; |
49878439 YY |
2475 | |
2476 | ||
2477 | f = fopen("/proc/diskstats", "r"); | |
2478 | if (!f) | |
2c51f8dd | 2479 | goto err; |
49878439 YY |
2480 | |
2481 | while (getline(&line, &linelen, f) != -1) { | |
2482 | size_t l; | |
2483 | char *printme, lbuf[256]; | |
2484 | ||
c0adec85 | 2485 | i = sscanf(line, "%u %u %71s", &major, &minor, dev_name); |
49878439 YY |
2486 | if(i == 3){ |
2487 | get_blkio_io_value(io_serviced_str, major, minor, "Read", &read); | |
2488 | get_blkio_io_value(io_serviced_str, major, minor, "Write", &write); | |
2489 | get_blkio_io_value(io_merged_str, major, minor, "Read", &read_merged); | |
2490 | get_blkio_io_value(io_merged_str, major, minor, "Write", &write_merged); | |
2491 | get_blkio_io_value(io_service_bytes_str, major, minor, "Read", &read_sectors); | |
2492 | read_sectors = read_sectors/512; | |
2493 | get_blkio_io_value(io_service_bytes_str, major, minor, "Write", &write_sectors); | |
2494 | write_sectors = write_sectors/512; | |
2f919d9d | 2495 | |
49878439 YY |
2496 | get_blkio_io_value(io_service_time_str, major, minor, "Read", &rd_svctm); |
2497 | rd_svctm = rd_svctm/1000000; | |
2498 | get_blkio_io_value(io_wait_time_str, major, minor, "Read", &rd_wait); | |
2499 | rd_wait = rd_wait/1000000; | |
2500 | read_ticks = rd_svctm + rd_wait; | |
2501 | ||
2502 | get_blkio_io_value(io_service_time_str, major, minor, "Write", &wr_svctm); | |
2503 | wr_svctm = wr_svctm/1000000; | |
2504 | get_blkio_io_value(io_wait_time_str, major, minor, "Write", &wr_wait); | |
2505 | wr_wait = wr_wait/1000000; | |
2506 | write_ticks = wr_svctm + wr_wait; | |
2507 | ||
2508 | get_blkio_io_value(io_service_time_str, major, minor, "Total", &tot_ticks); | |
2509 | tot_ticks = tot_ticks/1000000; | |
2510 | }else{ | |
2511 | continue; | |
2512 | } | |
2513 | ||
2514 | memset(lbuf, 0, 256); | |
2515 | if (read || write || read_merged || write_merged || read_sectors || write_sectors || read_ticks || write_ticks) { | |
2f919d9d | 2516 | snprintf(lbuf, 256, "%u %u %s %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu\n", |
49878439 YY |
2517 | major, minor, dev_name, read, read_merged, read_sectors, read_ticks, |
2518 | write, write_merged, write_sectors, write_ticks, ios_pgr, tot_ticks, rq_ticks); | |
2519 | printme = lbuf; | |
2520 | } else | |
2521 | continue; | |
2522 | ||
2523 | l = snprintf(buf, size, "%s", printme); | |
e1068397 MM |
2524 | if (l < 0) { |
2525 | perror("Error writing to fuse buf"); | |
2526 | rv = 0; | |
2527 | goto err; | |
2528 | } | |
2529 | if (l >= size) { | |
2530 | fprintf(stderr, "Internal error: truncated write to cache\n"); | |
2531 | rv = 0; | |
2532 | goto err; | |
2533 | } | |
49878439 YY |
2534 | buf += l; |
2535 | size -= l; | |
2536 | total_len += l; | |
2537 | } | |
2538 | ||
97f1f27b | 2539 | d->size = total_len; |
e1068397 | 2540 | rv = total_len; |
2c51f8dd SH |
2541 | err: |
2542 | free(cg); | |
2543 | if (f) | |
2544 | fclose(f); | |
49878439 | 2545 | free(line); |
2c51f8dd SH |
2546 | free(io_serviced_str); |
2547 | free(io_merged_str); | |
2548 | free(io_service_bytes_str); | |
2549 | free(io_wait_time_str); | |
2550 | free(io_service_time_str); | |
e1068397 | 2551 | return rv; |
49878439 YY |
2552 | } |
2553 | ||
23ce2127 SH |
2554 | static off_t get_procfile_size(const char *which) |
2555 | { | |
2556 | FILE *f = fopen(which, "r"); | |
2557 | char *line = NULL; | |
2558 | size_t len = 0; | |
2559 | ssize_t sz, answer = 0; | |
2560 | if (!f) | |
2561 | return 0; | |
2562 | ||
2563 | while ((sz = getline(&line, &len, f)) != -1) | |
2564 | answer += sz; | |
2565 | fclose (f); | |
92c84dc4 | 2566 | free(line); |
23ce2127 SH |
2567 | |
2568 | return answer; | |
2569 | } | |
2570 | ||
758ad80c SH |
2571 | static int proc_getattr(const char *path, struct stat *sb) |
2572 | { | |
35629743 SH |
2573 | struct timespec now; |
2574 | ||
2575 | memset(sb, 0, sizeof(struct stat)); | |
2576 | if (clock_gettime(CLOCK_REALTIME, &now) < 0) | |
2577 | return -EINVAL; | |
2578 | sb->st_uid = sb->st_gid = 0; | |
2579 | sb->st_atim = sb->st_mtim = sb->st_ctim = now; | |
2580 | if (strcmp(path, "/proc") == 0) { | |
2581 | sb->st_mode = S_IFDIR | 00555; | |
2582 | sb->st_nlink = 2; | |
2583 | return 0; | |
2584 | } | |
2585 | if (strcmp(path, "/proc/meminfo") == 0 || | |
2586 | strcmp(path, "/proc/cpuinfo") == 0 || | |
2587 | strcmp(path, "/proc/uptime") == 0 || | |
49878439 YY |
2588 | strcmp(path, "/proc/stat") == 0 || |
2589 | strcmp(path, "/proc/diskstats") == 0) { | |
7253e0a4 | 2590 | sb->st_size = 0; |
35629743 SH |
2591 | sb->st_mode = S_IFREG | 00444; |
2592 | sb->st_nlink = 1; | |
2593 | return 0; | |
2594 | } | |
2595 | ||
2596 | return -ENOENT; | |
2597 | } | |
2598 | ||
2599 | static int proc_readdir(const char *path, void *buf, fuse_fill_dir_t filler, off_t offset, | |
2600 | struct fuse_file_info *fi) | |
2601 | { | |
2602 | if (filler(buf, "cpuinfo", NULL, 0) != 0 || | |
2603 | filler(buf, "meminfo", NULL, 0) != 0 || | |
2604 | filler(buf, "stat", NULL, 0) != 0 || | |
49878439 YY |
2605 | filler(buf, "uptime", NULL, 0) != 0 || |
2606 | filler(buf, "diskstats", NULL, 0) != 0) | |
758ad80c | 2607 | return -EINVAL; |
758ad80c SH |
2608 | return 0; |
2609 | } | |
2610 | ||
35629743 SH |
2611 | static int proc_open(const char *path, struct fuse_file_info *fi) |
2612 | { | |
96fc5ee6 SH |
2613 | int type = -1; |
2614 | struct file_info *info; | |
2615 | ||
2616 | if (strcmp(path, "/proc/meminfo") == 0) | |
2617 | type = LXC_TYPE_PROC_MEMINFO; | |
2618 | else if (strcmp(path, "/proc/cpuinfo") == 0) | |
2619 | type = LXC_TYPE_PROC_CPUINFO; | |
2620 | else if (strcmp(path, "/proc/uptime") == 0) | |
2621 | type = LXC_TYPE_PROC_UPTIME; | |
2622 | else if (strcmp(path, "/proc/stat") == 0) | |
2623 | type = LXC_TYPE_PROC_STAT; | |
2624 | else if (strcmp(path, "/proc/diskstats") == 0) | |
2625 | type = LXC_TYPE_PROC_DISKSTATS; | |
2626 | if (type == -1) | |
2627 | return -ENOENT; | |
2628 | ||
2c51f8dd SH |
2629 | info = malloc(sizeof(*info)); |
2630 | if (!info) | |
2631 | return -ENOMEM; | |
2632 | ||
96fc5ee6 SH |
2633 | memset(info, 0, sizeof(*info)); |
2634 | info->type = type; | |
2635 | ||
97f1f27b | 2636 | info->buflen = get_procfile_size(path) + BUF_RESERVE_SIZE; |
2c51f8dd SH |
2637 | do { |
2638 | info->buf = malloc(info->buflen); | |
2639 | } while (!info->buf); | |
97f1f27b YY |
2640 | memset(info->buf, 0, info->buflen); |
2641 | /* set actual size to buffer size */ | |
2f919d9d | 2642 | info->size = info->buflen; |
97f1f27b | 2643 | |
96fc5ee6 SH |
2644 | fi->fh = (unsigned long)info; |
2645 | return 0; | |
2646 | } | |
2647 | ||
2648 | static int proc_release(const char *path, struct fuse_file_info *fi) | |
2649 | { | |
2650 | struct file_info *f = (struct file_info *)fi->fh; | |
2651 | ||
2652 | do_release_file_info(f); | |
2653 | return 0; | |
35629743 SH |
2654 | } |
2655 | ||
35629743 SH |
2656 | static int proc_read(const char *path, char *buf, size_t size, off_t offset, |
2657 | struct fuse_file_info *fi) | |
2658 | { | |
96fc5ee6 SH |
2659 | struct file_info *f = (struct file_info *) fi->fh; |
2660 | ||
2661 | switch (f->type) { | |
2f919d9d | 2662 | case LXC_TYPE_PROC_MEMINFO: |
23ce2127 | 2663 | return proc_meminfo_read(buf, size, offset, fi); |
96fc5ee6 | 2664 | case LXC_TYPE_PROC_CPUINFO: |
23ce2127 | 2665 | return proc_cpuinfo_read(buf, size, offset, fi); |
96fc5ee6 | 2666 | case LXC_TYPE_PROC_UPTIME: |
23ce2127 | 2667 | return proc_uptime_read(buf, size, offset, fi); |
96fc5ee6 | 2668 | case LXC_TYPE_PROC_STAT: |
23ce2127 | 2669 | return proc_stat_read(buf, size, offset, fi); |
96fc5ee6 | 2670 | case LXC_TYPE_PROC_DISKSTATS: |
49878439 | 2671 | return proc_diskstats_read(buf, size, offset, fi); |
96fc5ee6 SH |
2672 | default: |
2673 | return -EINVAL; | |
2674 | } | |
35629743 SH |
2675 | } |
2676 | ||
2ad6d2bd SH |
2677 | /* |
2678 | * FUSE ops for / | |
2679 | * these just delegate to the /proc and /cgroup ops as | |
2680 | * needed | |
2681 | */ | |
758ad80c SH |
2682 | |
2683 | static int lxcfs_getattr(const char *path, struct stat *sb) | |
2684 | { | |
2685 | if (strcmp(path, "/") == 0) { | |
2686 | sb->st_mode = S_IFDIR | 00755; | |
2687 | sb->st_nlink = 2; | |
2688 | return 0; | |
2689 | } | |
2690 | if (strncmp(path, "/cgroup", 7) == 0) { | |
2691 | return cg_getattr(path, sb); | |
2692 | } | |
35629743 | 2693 | if (strncmp(path, "/proc", 5) == 0) { |
758ad80c SH |
2694 | return proc_getattr(path, sb); |
2695 | } | |
2696 | return -EINVAL; | |
2697 | } | |
2698 | ||
2699 | static int lxcfs_opendir(const char *path, struct fuse_file_info *fi) | |
2700 | { | |
2701 | if (strcmp(path, "/") == 0) | |
2702 | return 0; | |
2703 | ||
2704 | if (strncmp(path, "/cgroup", 7) == 0) { | |
2705 | return cg_opendir(path, fi); | |
2706 | } | |
35629743 SH |
2707 | if (strcmp(path, "/proc") == 0) |
2708 | return 0; | |
2709 | return -ENOENT; | |
758ad80c SH |
2710 | } |
2711 | ||
2712 | static int lxcfs_readdir(const char *path, void *buf, fuse_fill_dir_t filler, off_t offset, | |
2713 | struct fuse_file_info *fi) | |
2714 | { | |
2715 | if (strcmp(path, "/") == 0) { | |
2716 | if (filler(buf, "proc", NULL, 0) != 0 || | |
2717 | filler(buf, "cgroup", NULL, 0) != 0) | |
2718 | return -EINVAL; | |
2719 | return 0; | |
2720 | } | |
35629743 | 2721 | if (strncmp(path, "/cgroup", 7) == 0) |
758ad80c | 2722 | return cg_readdir(path, buf, filler, offset, fi); |
35629743 SH |
2723 | if (strcmp(path, "/proc") == 0) |
2724 | return proc_readdir(path, buf, filler, offset, fi); | |
758ad80c SH |
2725 | return -EINVAL; |
2726 | } | |
2727 | ||
2728 | static int lxcfs_releasedir(const char *path, struct fuse_file_info *fi) | |
2729 | { | |
2730 | if (strcmp(path, "/") == 0) | |
2731 | return 0; | |
2732 | if (strncmp(path, "/cgroup", 7) == 0) { | |
2733 | return cg_releasedir(path, fi); | |
2734 | } | |
35629743 SH |
2735 | if (strcmp(path, "/proc") == 0) |
2736 | return 0; | |
758ad80c SH |
2737 | return -EINVAL; |
2738 | } | |
2739 | ||
99978832 SH |
2740 | static int lxcfs_open(const char *path, struct fuse_file_info *fi) |
2741 | { | |
35629743 | 2742 | if (strncmp(path, "/cgroup", 7) == 0) |
99978832 | 2743 | return cg_open(path, fi); |
35629743 SH |
2744 | if (strncmp(path, "/proc", 5) == 0) |
2745 | return proc_open(path, fi); | |
99978832 SH |
2746 | |
2747 | return -EINVAL; | |
2748 | } | |
2749 | ||
2750 | static int lxcfs_read(const char *path, char *buf, size_t size, off_t offset, | |
2751 | struct fuse_file_info *fi) | |
2752 | { | |
35629743 | 2753 | if (strncmp(path, "/cgroup", 7) == 0) |
99978832 | 2754 | return cg_read(path, buf, size, offset, fi); |
35629743 SH |
2755 | if (strncmp(path, "/proc", 5) == 0) |
2756 | return proc_read(path, buf, size, offset, fi); | |
99978832 SH |
2757 | |
2758 | return -EINVAL; | |
2759 | } | |
2760 | ||
2ad6d2bd SH |
2761 | int lxcfs_write(const char *path, const char *buf, size_t size, off_t offset, |
2762 | struct fuse_file_info *fi) | |
2763 | { | |
2764 | if (strncmp(path, "/cgroup", 7) == 0) { | |
2765 | return cg_write(path, buf, size, offset, fi); | |
2766 | } | |
2767 | ||
2768 | return -EINVAL; | |
2769 | } | |
2770 | ||
99978832 SH |
2771 | static int lxcfs_flush(const char *path, struct fuse_file_info *fi) |
2772 | { | |
2773 | return 0; | |
2774 | } | |
2775 | ||
2776 | static int lxcfs_release(const char *path, struct fuse_file_info *fi) | |
758ad80c | 2777 | { |
8f6e8f5e SH |
2778 | if (strncmp(path, "/cgroup", 7) == 0) |
2779 | return cg_release(path, fi); | |
8f6e8f5e | 2780 | if (strncmp(path, "/proc", 5) == 0) |
96fc5ee6 | 2781 | return proc_release(path, fi); |
8f6e8f5e SH |
2782 | |
2783 | return -EINVAL; | |
99978832 SH |
2784 | } |
2785 | ||
2786 | static int lxcfs_fsync(const char *path, int datasync, struct fuse_file_info *fi) | |
2787 | { | |
2788 | return 0; | |
758ad80c SH |
2789 | } |
2790 | ||
ab54b798 SH |
2791 | int lxcfs_mkdir(const char *path, mode_t mode) |
2792 | { | |
2793 | if (strncmp(path, "/cgroup", 7) == 0) | |
2794 | return cg_mkdir(path, mode); | |
2795 | ||
2796 | return -EINVAL; | |
2797 | } | |
2798 | ||
341b21ad SH |
2799 | int lxcfs_chown(const char *path, uid_t uid, gid_t gid) |
2800 | { | |
2801 | if (strncmp(path, "/cgroup", 7) == 0) | |
2802 | return cg_chown(path, uid, gid); | |
2803 | ||
2804 | return -EINVAL; | |
2805 | } | |
2806 | ||
2ad6d2bd SH |
2807 | /* |
2808 | * cat first does a truncate before doing ops->write. This doesn't | |
2809 | * really make sense for cgroups. So just return 0 always but do | |
2810 | * nothing. | |
2811 | */ | |
2812 | int lxcfs_truncate(const char *path, off_t newsize) | |
2813 | { | |
2814 | if (strncmp(path, "/cgroup", 7) == 0) | |
2815 | return 0; | |
2816 | return -EINVAL; | |
2817 | } | |
2818 | ||
50d8d5b5 SH |
2819 | int lxcfs_rmdir(const char *path) |
2820 | { | |
2821 | if (strncmp(path, "/cgroup", 7) == 0) | |
2822 | return cg_rmdir(path); | |
2823 | return -EINVAL; | |
2824 | } | |
2825 | ||
fd2e4e03 SH |
2826 | int lxcfs_chmod(const char *path, mode_t mode) |
2827 | { | |
2828 | if (strncmp(path, "/cgroup", 7) == 0) | |
2829 | return cg_chmod(path, mode); | |
2830 | return -EINVAL; | |
2831 | } | |
2832 | ||
758ad80c SH |
2833 | const struct fuse_operations lxcfs_ops = { |
2834 | .getattr = lxcfs_getattr, | |
2835 | .readlink = NULL, | |
2836 | .getdir = NULL, | |
2837 | .mknod = NULL, | |
ab54b798 | 2838 | .mkdir = lxcfs_mkdir, |
758ad80c | 2839 | .unlink = NULL, |
50d8d5b5 | 2840 | .rmdir = lxcfs_rmdir, |
758ad80c SH |
2841 | .symlink = NULL, |
2842 | .rename = NULL, | |
2843 | .link = NULL, | |
fd2e4e03 | 2844 | .chmod = lxcfs_chmod, |
341b21ad | 2845 | .chown = lxcfs_chown, |
2ad6d2bd | 2846 | .truncate = lxcfs_truncate, |
758ad80c | 2847 | .utime = NULL, |
99978832 SH |
2848 | |
2849 | .open = lxcfs_open, | |
2850 | .read = lxcfs_read, | |
2851 | .release = lxcfs_release, | |
2ad6d2bd | 2852 | .write = lxcfs_write, |
99978832 | 2853 | |
758ad80c | 2854 | .statfs = NULL, |
99978832 SH |
2855 | .flush = lxcfs_flush, |
2856 | .fsync = lxcfs_fsync, | |
758ad80c SH |
2857 | |
2858 | .setxattr = NULL, | |
2859 | .getxattr = NULL, | |
2860 | .listxattr = NULL, | |
2861 | .removexattr = NULL, | |
2862 | ||
2863 | .opendir = lxcfs_opendir, | |
2864 | .readdir = lxcfs_readdir, | |
2865 | .releasedir = lxcfs_releasedir, | |
2866 | ||
2867 | .fsyncdir = NULL, | |
2868 | .init = NULL, | |
2869 | .destroy = NULL, | |
2870 | .access = NULL, | |
2871 | .create = NULL, | |
2872 | .ftruncate = NULL, | |
2873 | .fgetattr = NULL, | |
2874 | }; | |
2875 | ||
99978832 | 2876 | static void usage(const char *me) |
758ad80c SH |
2877 | { |
2878 | fprintf(stderr, "Usage:\n"); | |
2879 | fprintf(stderr, "\n"); | |
0b0f73db SH |
2880 | fprintf(stderr, "%s mountpoint\n", me); |
2881 | fprintf(stderr, "%s -h\n", me); | |
758ad80c SH |
2882 | exit(1); |
2883 | } | |
2884 | ||
99978832 | 2885 | static bool is_help(char *w) |
758ad80c SH |
2886 | { |
2887 | if (strcmp(w, "-h") == 0 || | |
2888 | strcmp(w, "--help") == 0 || | |
2889 | strcmp(w, "-help") == 0 || | |
2890 | strcmp(w, "help") == 0) | |
2891 | return true; | |
2892 | return false; | |
2893 | } | |
2894 | ||
0b0f73db SH |
2895 | void swallow_arg(int *argcp, char *argv[], char *which) |
2896 | { | |
2897 | int i; | |
2898 | ||
2899 | for (i = 1; argv[i]; i++) { | |
2900 | if (strcmp(argv[i], which) != 0) | |
2901 | continue; | |
2902 | for (; argv[i]; i++) { | |
2903 | argv[i] = argv[i+1]; | |
2904 | } | |
2905 | (*argcp)--; | |
2906 | return; | |
2907 | } | |
2908 | } | |
2909 | ||
2910 | void swallow_option(int *argcp, char *argv[], char *opt, char *v) | |
2911 | { | |
2912 | int i; | |
2913 | ||
2914 | for (i = 1; argv[i]; i++) { | |
2915 | if (!argv[i+1]) | |
2916 | continue; | |
2917 | if (strcmp(argv[i], opt) != 0) | |
2918 | continue; | |
2919 | if (strcmp(argv[i+1], v) != 0) { | |
2920 | fprintf(stderr, "Warning: unexpected fuse option %s\n", v); | |
2921 | exit(1); | |
2922 | } | |
2923 | for (; argv[i+1]; i++) { | |
2924 | argv[i] = argv[i+2]; | |
2925 | } | |
2926 | (*argcp) -= 2; | |
2927 | return; | |
2928 | } | |
2929 | } | |
2930 | ||
758ad80c SH |
2931 | int main(int argc, char *argv[]) |
2932 | { | |
c0adec85 | 2933 | int ret = -1; |
e5d26e0b | 2934 | struct lxcfs_state *d = NULL; |
0b0f73db SH |
2935 | /* |
2936 | * what we pass to fuse_main is: | |
2937 | * argv[0] -s -f -o allow_other,directio argv[1] NULL | |
2938 | */ | |
2c51f8dd SH |
2939 | int nargs = 5, cnt = 0; |
2940 | char *newargv[6]; | |
758ad80c | 2941 | |
0b0f73db SH |
2942 | /* accomodate older init scripts */ |
2943 | swallow_arg(&argc, argv, "-s"); | |
2944 | swallow_arg(&argc, argv, "-f"); | |
2945 | swallow_option(&argc, argv, "-o", "allow_other"); | |
2946 | ||
2e9c0b32 SH |
2947 | if (argc == 2 && strcmp(argv[1], "--version") == 0) { |
2948 | fprintf(stderr, "%s\n", VERSION); | |
2949 | exit(0); | |
2950 | } | |
0b0f73db | 2951 | if (argc != 2 || is_help(argv[1])) |
758ad80c SH |
2952 | usage(argv[0]); |
2953 | ||
2c51f8dd SH |
2954 | do { |
2955 | d = malloc(sizeof(*d)); | |
2956 | } while (!d); | |
0b0f73db | 2957 | |
38a76a91 | 2958 | newargv[cnt++] = argv[0]; |
38a76a91 SH |
2959 | newargv[cnt++] = "-f"; |
2960 | newargv[cnt++] = "-o"; | |
2961 | newargv[cnt++] = "allow_other,direct_io"; | |
2962 | newargv[cnt++] = argv[1]; | |
2963 | newargv[cnt++] = NULL; | |
758ad80c SH |
2964 | |
2965 | if (!cgm_escape_cgroup()) | |
2966 | fprintf(stderr, "WARNING: failed to escape to root cgroup\n"); | |
2967 | ||
2968 | if (!cgm_get_controllers(&d->subsystems)) | |
c0adec85 | 2969 | goto out; |
758ad80c | 2970 | |
38a76a91 | 2971 | ret = fuse_main(nargs, newargv, &lxcfs_ops, d); |
2c51f8dd | 2972 | cgm_dbus_disconnect(); |
758ad80c | 2973 | |
c0adec85 | 2974 | out: |
e5d26e0b | 2975 | free(d); |
758ad80c | 2976 | return ret; |
2183082c | 2977 | } |