Commit | Line | Data |
---|---|---|
08ce5f16 | 1 | /* |
47c59803 | 2 | * device_cgroup.c - device cgroup subsystem |
08ce5f16 SH |
3 | * |
4 | * Copyright 2007 IBM Corp | |
5 | */ | |
6 | ||
7 | #include <linux/device_cgroup.h> | |
8 | #include <linux/cgroup.h> | |
9 | #include <linux/ctype.h> | |
10 | #include <linux/list.h> | |
11 | #include <linux/uaccess.h> | |
29486df3 | 12 | #include <linux/seq_file.h> |
5a0e3ad6 | 13 | #include <linux/slab.h> |
47c59803 | 14 | #include <linux/rcupdate.h> |
b4046f00 | 15 | #include <linux/mutex.h> |
08ce5f16 SH |
16 | |
17 | #define ACC_MKNOD 1 | |
18 | #define ACC_READ 2 | |
19 | #define ACC_WRITE 4 | |
20 | #define ACC_MASK (ACC_MKNOD | ACC_READ | ACC_WRITE) | |
21 | ||
22 | #define DEV_BLOCK 1 | |
23 | #define DEV_CHAR 2 | |
24 | #define DEV_ALL 4 /* this represents all devices */ | |
25 | ||
b4046f00 LZ |
26 | static DEFINE_MUTEX(devcgroup_mutex); |
27 | ||
c39a2a30 AR |
28 | enum devcg_behavior { |
29 | DEVCG_DEFAULT_NONE, | |
30 | DEVCG_DEFAULT_ALLOW, | |
31 | DEVCG_DEFAULT_DENY, | |
32 | }; | |
33 | ||
08ce5f16 | 34 | /* |
db9aeca9 | 35 | * exception list locking rules: |
b4046f00 | 36 | * hold devcgroup_mutex for update/read. |
47c59803 | 37 | * hold rcu_read_lock() for read. |
08ce5f16 SH |
38 | */ |
39 | ||
db9aeca9 | 40 | struct dev_exception_item { |
08ce5f16 SH |
41 | u32 major, minor; |
42 | short type; | |
43 | short access; | |
44 | struct list_head list; | |
4efd1a1b | 45 | struct rcu_head rcu; |
08ce5f16 SH |
46 | }; |
47 | ||
48 | struct dev_cgroup { | |
49 | struct cgroup_subsys_state css; | |
db9aeca9 | 50 | struct list_head exceptions; |
c39a2a30 | 51 | enum devcg_behavior behavior; |
08ce5f16 SH |
52 | }; |
53 | ||
b66862f7 PE |
54 | static inline struct dev_cgroup *css_to_devcgroup(struct cgroup_subsys_state *s) |
55 | { | |
a7c6d554 | 56 | return s ? container_of(s, struct dev_cgroup, css) : NULL; |
b66862f7 PE |
57 | } |
58 | ||
f92523e3 PM |
59 | static inline struct dev_cgroup *task_devcgroup(struct task_struct *task) |
60 | { | |
073219e9 | 61 | return css_to_devcgroup(task_css(task, devices_cgrp_id)); |
f92523e3 PM |
62 | } |
63 | ||
08ce5f16 | 64 | /* |
b4046f00 | 65 | * called under devcgroup_mutex |
08ce5f16 | 66 | */ |
db9aeca9 | 67 | static int dev_exceptions_copy(struct list_head *dest, struct list_head *orig) |
08ce5f16 | 68 | { |
db9aeca9 | 69 | struct dev_exception_item *ex, *tmp, *new; |
08ce5f16 | 70 | |
4b1c7840 TH |
71 | lockdep_assert_held(&devcgroup_mutex); |
72 | ||
db9aeca9 AR |
73 | list_for_each_entry(ex, orig, list) { |
74 | new = kmemdup(ex, sizeof(*ex), GFP_KERNEL); | |
08ce5f16 SH |
75 | if (!new) |
76 | goto free_and_exit; | |
08ce5f16 SH |
77 | list_add_tail(&new->list, dest); |
78 | } | |
79 | ||
80 | return 0; | |
81 | ||
82 | free_and_exit: | |
db9aeca9 AR |
83 | list_for_each_entry_safe(ex, tmp, dest, list) { |
84 | list_del(&ex->list); | |
85 | kfree(ex); | |
08ce5f16 SH |
86 | } |
87 | return -ENOMEM; | |
88 | } | |
89 | ||
08ce5f16 | 90 | /* |
b4046f00 | 91 | * called under devcgroup_mutex |
08ce5f16 | 92 | */ |
db9aeca9 AR |
93 | static int dev_exception_add(struct dev_cgroup *dev_cgroup, |
94 | struct dev_exception_item *ex) | |
08ce5f16 | 95 | { |
db9aeca9 | 96 | struct dev_exception_item *excopy, *walk; |
08ce5f16 | 97 | |
4b1c7840 TH |
98 | lockdep_assert_held(&devcgroup_mutex); |
99 | ||
db9aeca9 AR |
100 | excopy = kmemdup(ex, sizeof(*ex), GFP_KERNEL); |
101 | if (!excopy) | |
08ce5f16 SH |
102 | return -ENOMEM; |
103 | ||
db9aeca9 AR |
104 | list_for_each_entry(walk, &dev_cgroup->exceptions, list) { |
105 | if (walk->type != ex->type) | |
d1ee2971 | 106 | continue; |
db9aeca9 | 107 | if (walk->major != ex->major) |
d1ee2971 | 108 | continue; |
db9aeca9 | 109 | if (walk->minor != ex->minor) |
d1ee2971 PE |
110 | continue; |
111 | ||
db9aeca9 AR |
112 | walk->access |= ex->access; |
113 | kfree(excopy); | |
114 | excopy = NULL; | |
d1ee2971 PE |
115 | } |
116 | ||
db9aeca9 AR |
117 | if (excopy != NULL) |
118 | list_add_tail_rcu(&excopy->list, &dev_cgroup->exceptions); | |
08ce5f16 SH |
119 | return 0; |
120 | } | |
121 | ||
122 | /* | |
b4046f00 | 123 | * called under devcgroup_mutex |
08ce5f16 | 124 | */ |
db9aeca9 AR |
125 | static void dev_exception_rm(struct dev_cgroup *dev_cgroup, |
126 | struct dev_exception_item *ex) | |
08ce5f16 | 127 | { |
db9aeca9 | 128 | struct dev_exception_item *walk, *tmp; |
08ce5f16 | 129 | |
4b1c7840 TH |
130 | lockdep_assert_held(&devcgroup_mutex); |
131 | ||
db9aeca9 AR |
132 | list_for_each_entry_safe(walk, tmp, &dev_cgroup->exceptions, list) { |
133 | if (walk->type != ex->type) | |
08ce5f16 | 134 | continue; |
db9aeca9 | 135 | if (walk->major != ex->major) |
08ce5f16 | 136 | continue; |
db9aeca9 | 137 | if (walk->minor != ex->minor) |
08ce5f16 SH |
138 | continue; |
139 | ||
db9aeca9 | 140 | walk->access &= ~ex->access; |
08ce5f16 | 141 | if (!walk->access) { |
4efd1a1b | 142 | list_del_rcu(&walk->list); |
6034f7e6 | 143 | kfree_rcu(walk, rcu); |
08ce5f16 SH |
144 | } |
145 | } | |
08ce5f16 SH |
146 | } |
147 | ||
53eb8c82 JS |
148 | static void __dev_exception_clean(struct dev_cgroup *dev_cgroup) |
149 | { | |
150 | struct dev_exception_item *ex, *tmp; | |
151 | ||
152 | list_for_each_entry_safe(ex, tmp, &dev_cgroup->exceptions, list) { | |
153 | list_del_rcu(&ex->list); | |
154 | kfree_rcu(ex, rcu); | |
155 | } | |
156 | } | |
157 | ||
868539a3 | 158 | /** |
db9aeca9 AR |
159 | * dev_exception_clean - frees all entries of the exception list |
160 | * @dev_cgroup: dev_cgroup with the exception list to be cleaned | |
868539a3 AR |
161 | * |
162 | * called under devcgroup_mutex | |
163 | */ | |
db9aeca9 | 164 | static void dev_exception_clean(struct dev_cgroup *dev_cgroup) |
868539a3 | 165 | { |
4b1c7840 TH |
166 | lockdep_assert_held(&devcgroup_mutex); |
167 | ||
53eb8c82 | 168 | __dev_exception_clean(dev_cgroup); |
868539a3 AR |
169 | } |
170 | ||
bd2953eb AR |
171 | static inline bool is_devcg_online(const struct dev_cgroup *devcg) |
172 | { | |
173 | return (devcg->behavior != DEVCG_DEFAULT_NONE); | |
174 | } | |
175 | ||
1909554c AR |
176 | /** |
177 | * devcgroup_online - initializes devcgroup's behavior and exceptions based on | |
178 | * parent's | |
eb95419b | 179 | * @css: css getting online |
1909554c AR |
180 | * returns 0 in case of success, error code otherwise |
181 | */ | |
eb95419b | 182 | static int devcgroup_online(struct cgroup_subsys_state *css) |
1909554c | 183 | { |
eb95419b | 184 | struct dev_cgroup *dev_cgroup = css_to_devcgroup(css); |
5c9d535b | 185 | struct dev_cgroup *parent_dev_cgroup = css_to_devcgroup(css->parent); |
1909554c AR |
186 | int ret = 0; |
187 | ||
188 | mutex_lock(&devcgroup_mutex); | |
1909554c AR |
189 | |
190 | if (parent_dev_cgroup == NULL) | |
191 | dev_cgroup->behavior = DEVCG_DEFAULT_ALLOW; | |
192 | else { | |
193 | ret = dev_exceptions_copy(&dev_cgroup->exceptions, | |
194 | &parent_dev_cgroup->exceptions); | |
195 | if (!ret) | |
196 | dev_cgroup->behavior = parent_dev_cgroup->behavior; | |
197 | } | |
198 | mutex_unlock(&devcgroup_mutex); | |
199 | ||
200 | return ret; | |
201 | } | |
202 | ||
eb95419b | 203 | static void devcgroup_offline(struct cgroup_subsys_state *css) |
1909554c | 204 | { |
eb95419b | 205 | struct dev_cgroup *dev_cgroup = css_to_devcgroup(css); |
1909554c AR |
206 | |
207 | mutex_lock(&devcgroup_mutex); | |
208 | dev_cgroup->behavior = DEVCG_DEFAULT_NONE; | |
209 | mutex_unlock(&devcgroup_mutex); | |
210 | } | |
211 | ||
08ce5f16 SH |
212 | /* |
213 | * called from kernel/cgroup.c with cgroup_lock() held. | |
214 | */ | |
eb95419b TH |
215 | static struct cgroup_subsys_state * |
216 | devcgroup_css_alloc(struct cgroup_subsys_state *parent_css) | |
08ce5f16 | 217 | { |
1909554c | 218 | struct dev_cgroup *dev_cgroup; |
08ce5f16 SH |
219 | |
220 | dev_cgroup = kzalloc(sizeof(*dev_cgroup), GFP_KERNEL); | |
221 | if (!dev_cgroup) | |
222 | return ERR_PTR(-ENOMEM); | |
db9aeca9 | 223 | INIT_LIST_HEAD(&dev_cgroup->exceptions); |
1909554c | 224 | dev_cgroup->behavior = DEVCG_DEFAULT_NONE; |
08ce5f16 | 225 | |
08ce5f16 SH |
226 | return &dev_cgroup->css; |
227 | } | |
228 | ||
eb95419b | 229 | static void devcgroup_css_free(struct cgroup_subsys_state *css) |
08ce5f16 | 230 | { |
eb95419b | 231 | struct dev_cgroup *dev_cgroup = css_to_devcgroup(css); |
08ce5f16 | 232 | |
53eb8c82 | 233 | __dev_exception_clean(dev_cgroup); |
08ce5f16 SH |
234 | kfree(dev_cgroup); |
235 | } | |
236 | ||
237 | #define DEVCG_ALLOW 1 | |
238 | #define DEVCG_DENY 2 | |
29486df3 SH |
239 | #define DEVCG_LIST 3 |
240 | ||
17d213f8 | 241 | #define MAJMINLEN 13 |
29486df3 | 242 | #define ACCLEN 4 |
08ce5f16 SH |
243 | |
244 | static void set_access(char *acc, short access) | |
245 | { | |
246 | int idx = 0; | |
29486df3 | 247 | memset(acc, 0, ACCLEN); |
08ce5f16 SH |
248 | if (access & ACC_READ) |
249 | acc[idx++] = 'r'; | |
250 | if (access & ACC_WRITE) | |
251 | acc[idx++] = 'w'; | |
252 | if (access & ACC_MKNOD) | |
253 | acc[idx++] = 'm'; | |
254 | } | |
255 | ||
256 | static char type_to_char(short type) | |
257 | { | |
258 | if (type == DEV_ALL) | |
259 | return 'a'; | |
260 | if (type == DEV_CHAR) | |
261 | return 'c'; | |
262 | if (type == DEV_BLOCK) | |
263 | return 'b'; | |
264 | return 'X'; | |
265 | } | |
266 | ||
29486df3 | 267 | static void set_majmin(char *str, unsigned m) |
08ce5f16 | 268 | { |
08ce5f16 | 269 | if (m == ~0) |
7759fc9d | 270 | strcpy(str, "*"); |
08ce5f16 | 271 | else |
7759fc9d | 272 | sprintf(str, "%u", m); |
08ce5f16 SH |
273 | } |
274 | ||
2da8ca82 | 275 | static int devcgroup_seq_show(struct seq_file *m, void *v) |
08ce5f16 | 276 | { |
2da8ca82 | 277 | struct dev_cgroup *devcgroup = css_to_devcgroup(seq_css(m)); |
db9aeca9 | 278 | struct dev_exception_item *ex; |
29486df3 | 279 | char maj[MAJMINLEN], min[MAJMINLEN], acc[ACCLEN]; |
08ce5f16 | 280 | |
4efd1a1b | 281 | rcu_read_lock(); |
ad676077 AR |
282 | /* |
283 | * To preserve the compatibility: | |
284 | * - Only show the "all devices" when the default policy is to allow | |
285 | * - List the exceptions in case the default policy is to deny | |
286 | * This way, the file remains as a "whitelist of devices" | |
287 | */ | |
5b7aa7d5 | 288 | if (devcgroup->behavior == DEVCG_DEFAULT_ALLOW) { |
ad676077 AR |
289 | set_access(acc, ACC_MASK); |
290 | set_majmin(maj, ~0); | |
291 | set_majmin(min, ~0); | |
292 | seq_printf(m, "%c %s:%s %s\n", type_to_char(DEV_ALL), | |
29486df3 | 293 | maj, min, acc); |
ad676077 | 294 | } else { |
db9aeca9 AR |
295 | list_for_each_entry_rcu(ex, &devcgroup->exceptions, list) { |
296 | set_access(acc, ex->access); | |
297 | set_majmin(maj, ex->major); | |
298 | set_majmin(min, ex->minor); | |
299 | seq_printf(m, "%c %s:%s %s\n", type_to_char(ex->type), | |
ad676077 AR |
300 | maj, min, acc); |
301 | } | |
08ce5f16 | 302 | } |
4efd1a1b | 303 | rcu_read_unlock(); |
08ce5f16 | 304 | |
29486df3 | 305 | return 0; |
08ce5f16 SH |
306 | } |
307 | ||
ad676077 | 308 | /** |
f5f3cf6f | 309 | * match_exception - iterates the exception list trying to find a complete match |
79d71974 AR |
310 | * @exceptions: list of exceptions |
311 | * @type: device type (DEV_BLOCK or DEV_CHAR) | |
312 | * @major: device file major number, ~0 to match all | |
313 | * @minor: device file minor number, ~0 to match all | |
314 | * @access: permission mask (ACC_READ, ACC_WRITE, ACC_MKNOD) | |
315 | * | |
f5f3cf6f AR |
316 | * It is considered a complete match if an exception is found that will |
317 | * contain the entire range of provided parameters. | |
318 | * | |
319 | * Return: true in case it matches an exception completely | |
08ce5f16 | 320 | */ |
79d71974 AR |
321 | static bool match_exception(struct list_head *exceptions, short type, |
322 | u32 major, u32 minor, short access) | |
08ce5f16 | 323 | { |
db9aeca9 | 324 | struct dev_exception_item *ex; |
08ce5f16 | 325 | |
79d71974 AR |
326 | list_for_each_entry_rcu(ex, exceptions, list) { |
327 | if ((type & DEV_BLOCK) && !(ex->type & DEV_BLOCK)) | |
328 | continue; | |
329 | if ((type & DEV_CHAR) && !(ex->type & DEV_CHAR)) | |
330 | continue; | |
331 | if (ex->major != ~0 && ex->major != major) | |
332 | continue; | |
333 | if (ex->minor != ~0 && ex->minor != minor) | |
334 | continue; | |
335 | /* provided access cannot have more than the exception rule */ | |
336 | if (access & (~ex->access)) | |
337 | continue; | |
338 | return true; | |
339 | } | |
340 | return false; | |
341 | } | |
342 | ||
343 | /** | |
f5f3cf6f | 344 | * match_exception_partial - iterates the exception list trying to find a partial match |
79d71974 AR |
345 | * @exceptions: list of exceptions |
346 | * @type: device type (DEV_BLOCK or DEV_CHAR) | |
347 | * @major: device file major number, ~0 to match all | |
348 | * @minor: device file minor number, ~0 to match all | |
349 | * @access: permission mask (ACC_READ, ACC_WRITE, ACC_MKNOD) | |
350 | * | |
f5f3cf6f AR |
351 | * It is considered a partial match if an exception's range is found to |
352 | * contain *any* of the devices specified by provided parameters. This is | |
353 | * used to make sure no extra access is being granted that is forbidden by | |
354 | * any of the exception list. | |
355 | * | |
356 | * Return: true in case the provided range mat matches an exception completely | |
79d71974 AR |
357 | */ |
358 | static bool match_exception_partial(struct list_head *exceptions, short type, | |
359 | u32 major, u32 minor, short access) | |
360 | { | |
361 | struct dev_exception_item *ex; | |
4b1c7840 | 362 | |
79d71974 AR |
363 | list_for_each_entry_rcu(ex, exceptions, list) { |
364 | if ((type & DEV_BLOCK) && !(ex->type & DEV_BLOCK)) | |
08ce5f16 | 365 | continue; |
79d71974 | 366 | if ((type & DEV_CHAR) && !(ex->type & DEV_CHAR)) |
08ce5f16 | 367 | continue; |
79d71974 AR |
368 | /* |
369 | * We must be sure that both the exception and the provided | |
370 | * range aren't masking all devices | |
371 | */ | |
372 | if (ex->major != ~0 && major != ~0 && ex->major != major) | |
08ce5f16 | 373 | continue; |
79d71974 | 374 | if (ex->minor != ~0 && minor != ~0 && ex->minor != minor) |
08ce5f16 | 375 | continue; |
79d71974 AR |
376 | /* |
377 | * In order to make sure the provided range isn't matching | |
378 | * an exception, all its access bits shouldn't match the | |
379 | * exception's access bits | |
380 | */ | |
381 | if (!(access & ex->access)) | |
08ce5f16 | 382 | continue; |
79d71974 | 383 | return true; |
08ce5f16 | 384 | } |
79d71974 AR |
385 | return false; |
386 | } | |
387 | ||
388 | /** | |
f5f3cf6f | 389 | * verify_new_ex - verifies if a new exception is allowed by parent cgroup's permissions |
79d71974 AR |
390 | * @dev_cgroup: dev cgroup to be tested against |
391 | * @refex: new exception | |
392 | * @behavior: behavior of the exception's dev_cgroup | |
f5f3cf6f AR |
393 | * |
394 | * This is used to make sure a child cgroup won't have more privileges | |
395 | * than its parent | |
79d71974 AR |
396 | */ |
397 | static bool verify_new_ex(struct dev_cgroup *dev_cgroup, | |
398 | struct dev_exception_item *refex, | |
399 | enum devcg_behavior behavior) | |
400 | { | |
401 | bool match = false; | |
402 | ||
f78f5b90 | 403 | RCU_LOCKDEP_WARN(!rcu_read_lock_held() && |
dc3a04d5 | 404 | !lockdep_is_held(&devcgroup_mutex), |
f78f5b90 | 405 | "device_cgroup:verify_new_ex called without proper synchronization"); |
ad676077 | 406 | |
c39a2a30 AR |
407 | if (dev_cgroup->behavior == DEVCG_DEFAULT_ALLOW) { |
408 | if (behavior == DEVCG_DEFAULT_ALLOW) { | |
79d71974 AR |
409 | /* |
410 | * new exception in the child doesn't matter, only | |
411 | * adding extra restrictions | |
412 | */ | |
c39a2a30 AR |
413 | return true; |
414 | } else { | |
79d71974 AR |
415 | /* |
416 | * new exception in the child will add more devices | |
417 | * that can be acessed, so it can't match any of | |
418 | * parent's exceptions, even slightly | |
419 | */ | |
420 | match = match_exception_partial(&dev_cgroup->exceptions, | |
421 | refex->type, | |
422 | refex->major, | |
423 | refex->minor, | |
424 | refex->access); | |
425 | ||
c39a2a30 | 426 | if (match) |
c39a2a30 | 427 | return false; |
26898fdf | 428 | return true; |
c39a2a30 | 429 | } |
26898fdf | 430 | } else { |
79d71974 AR |
431 | /* |
432 | * Only behavior == DEVCG_DEFAULT_DENY allowed here, therefore | |
433 | * the new exception will add access to more devices and must | |
434 | * be contained completely in an parent's exception to be | |
435 | * allowed | |
436 | */ | |
437 | match = match_exception(&dev_cgroup->exceptions, refex->type, | |
438 | refex->major, refex->minor, | |
439 | refex->access); | |
440 | ||
c39a2a30 AR |
441 | if (match) |
442 | /* parent has an exception that matches the proposed */ | |
26898fdf | 443 | return true; |
c39a2a30 AR |
444 | else |
445 | return false; | |
26898fdf AR |
446 | } |
447 | return false; | |
08ce5f16 SH |
448 | } |
449 | ||
450 | /* | |
451 | * parent_has_perm: | |
db9aeca9 | 452 | * when adding a new allow rule to a device exception list, the rule |
08ce5f16 SH |
453 | * must be allowed in the parent device |
454 | */ | |
f92523e3 | 455 | static int parent_has_perm(struct dev_cgroup *childcg, |
db9aeca9 | 456 | struct dev_exception_item *ex) |
08ce5f16 | 457 | { |
5c9d535b | 458 | struct dev_cgroup *parent = css_to_devcgroup(childcg->css.parent); |
08ce5f16 | 459 | |
63876986 | 460 | if (!parent) |
08ce5f16 | 461 | return 1; |
79d71974 | 462 | return verify_new_ex(parent, ex, childcg->behavior); |
08ce5f16 SH |
463 | } |
464 | ||
d2c2b11c AR |
465 | /** |
466 | * parent_allows_removal - verify if it's ok to remove an exception | |
467 | * @childcg: child cgroup from where the exception will be removed | |
468 | * @ex: exception being removed | |
469 | * | |
470 | * When removing an exception in cgroups with default ALLOW policy, it must | |
471 | * be checked if removing it will give the child cgroup more access than the | |
472 | * parent. | |
473 | * | |
474 | * Return: true if it's ok to remove exception, false otherwise | |
475 | */ | |
476 | static bool parent_allows_removal(struct dev_cgroup *childcg, | |
477 | struct dev_exception_item *ex) | |
478 | { | |
5c9d535b | 479 | struct dev_cgroup *parent = css_to_devcgroup(childcg->css.parent); |
d2c2b11c AR |
480 | |
481 | if (!parent) | |
482 | return true; | |
483 | ||
484 | /* It's always allowed to remove access to devices */ | |
485 | if (childcg->behavior == DEVCG_DEFAULT_DENY) | |
486 | return true; | |
487 | ||
488 | /* | |
489 | * Make sure you're not removing part or a whole exception existing in | |
490 | * the parent cgroup | |
491 | */ | |
492 | return !match_exception_partial(&parent->exceptions, ex->type, | |
493 | ex->major, ex->minor, ex->access); | |
494 | } | |
495 | ||
4cef7299 AR |
496 | /** |
497 | * may_allow_all - checks if it's possible to change the behavior to | |
498 | * allow based on parent's rules. | |
499 | * @parent: device cgroup's parent | |
500 | * returns: != 0 in case it's allowed, 0 otherwise | |
501 | */ | |
502 | static inline int may_allow_all(struct dev_cgroup *parent) | |
503 | { | |
64e10477 AR |
504 | if (!parent) |
505 | return 1; | |
4cef7299 AR |
506 | return parent->behavior == DEVCG_DEFAULT_ALLOW; |
507 | } | |
508 | ||
bd2953eb AR |
509 | /** |
510 | * revalidate_active_exceptions - walks through the active exception list and | |
511 | * revalidates the exceptions based on parent's | |
512 | * behavior and exceptions. The exceptions that | |
513 | * are no longer valid will be removed. | |
514 | * Called with devcgroup_mutex held. | |
515 | * @devcg: cgroup which exceptions will be checked | |
516 | * | |
517 | * This is one of the three key functions for hierarchy implementation. | |
518 | * This function is responsible for re-evaluating all the cgroup's active | |
519 | * exceptions due to a parent's exception change. | |
520 | * Refer to Documentation/cgroups/devices.txt for more details. | |
521 | */ | |
522 | static void revalidate_active_exceptions(struct dev_cgroup *devcg) | |
523 | { | |
524 | struct dev_exception_item *ex; | |
525 | struct list_head *this, *tmp; | |
526 | ||
527 | list_for_each_safe(this, tmp, &devcg->exceptions) { | |
528 | ex = container_of(this, struct dev_exception_item, list); | |
529 | if (!parent_has_perm(devcg, ex)) | |
530 | dev_exception_rm(devcg, ex); | |
531 | } | |
532 | } | |
533 | ||
bd2953eb AR |
534 | /** |
535 | * propagate_exception - propagates a new exception to the children | |
536 | * @devcg_root: device cgroup that added a new exception | |
537 | * @ex: new exception to be propagated | |
538 | * | |
539 | * returns: 0 in case of success, != 0 in case of error | |
540 | */ | |
541 | static int propagate_exception(struct dev_cgroup *devcg_root, | |
542 | struct dev_exception_item *ex) | |
543 | { | |
492eb21b | 544 | struct cgroup_subsys_state *pos; |
bd2953eb | 545 | int rc = 0; |
bd2953eb | 546 | |
d591fb56 | 547 | rcu_read_lock(); |
bd2953eb | 548 | |
492eb21b TH |
549 | css_for_each_descendant_pre(pos, &devcg_root->css) { |
550 | struct dev_cgroup *devcg = css_to_devcgroup(pos); | |
d591fb56 TH |
551 | |
552 | /* | |
553 | * Because devcgroup_mutex is held, no devcg will become | |
554 | * online or offline during the tree walk (see on/offline | |
555 | * methods), and online ones are safe to access outside RCU | |
556 | * read lock without bumping refcnt. | |
557 | */ | |
bd8815a6 | 558 | if (pos == &devcg_root->css || !is_devcg_online(devcg)) |
d591fb56 TH |
559 | continue; |
560 | ||
561 | rcu_read_unlock(); | |
bd2953eb AR |
562 | |
563 | /* | |
564 | * in case both root's behavior and devcg is allow, a new | |
565 | * restriction means adding to the exception list | |
566 | */ | |
567 | if (devcg_root->behavior == DEVCG_DEFAULT_ALLOW && | |
568 | devcg->behavior == DEVCG_DEFAULT_ALLOW) { | |
569 | rc = dev_exception_add(devcg, ex); | |
570 | if (rc) | |
571 | break; | |
572 | } else { | |
573 | /* | |
574 | * in the other possible cases: | |
575 | * root's behavior: allow, devcg's: deny | |
576 | * root's behavior: deny, devcg's: deny | |
577 | * the exception will be removed | |
578 | */ | |
579 | dev_exception_rm(devcg, ex); | |
580 | } | |
581 | revalidate_active_exceptions(devcg); | |
582 | ||
d591fb56 | 583 | rcu_read_lock(); |
bd2953eb | 584 | } |
d591fb56 TH |
585 | |
586 | rcu_read_unlock(); | |
bd2953eb AR |
587 | return rc; |
588 | } | |
589 | ||
08ce5f16 | 590 | /* |
db9aeca9 | 591 | * Modify the exception list using allow/deny rules. |
08ce5f16 SH |
592 | * CAP_SYS_ADMIN is needed for this. It's at least separate from CAP_MKNOD |
593 | * so we can give a container CAP_MKNOD to let it create devices but not | |
db9aeca9 | 594 | * modify the exception list. |
08ce5f16 SH |
595 | * It seems likely we'll want to add a CAP_CONTAINER capability to allow |
596 | * us to also grant CAP_SYS_ADMIN to containers without giving away the | |
db9aeca9 | 597 | * device exception list controls, but for now we'll stick with CAP_SYS_ADMIN |
08ce5f16 SH |
598 | * |
599 | * Taking rules away is always allowed (given CAP_SYS_ADMIN). Granting | |
600 | * new access is only allowed if you're in the top-level cgroup, or your | |
601 | * parent cgroup has the access you're asking for. | |
602 | */ | |
f92523e3 | 603 | static int devcgroup_update_access(struct dev_cgroup *devcgroup, |
4d3bb511 | 604 | int filetype, char *buffer) |
08ce5f16 | 605 | { |
f92523e3 | 606 | const char *b; |
26fd8405 | 607 | char temp[12]; /* 11 + 1 characters needed for a u32 */ |
c39a2a30 | 608 | int count, rc = 0; |
db9aeca9 | 609 | struct dev_exception_item ex; |
5c9d535b | 610 | struct dev_cgroup *parent = css_to_devcgroup(devcgroup->css.parent); |
08ce5f16 SH |
611 | |
612 | if (!capable(CAP_SYS_ADMIN)) | |
613 | return -EPERM; | |
614 | ||
db9aeca9 | 615 | memset(&ex, 0, sizeof(ex)); |
08ce5f16 SH |
616 | b = buffer; |
617 | ||
618 | switch (*b) { | |
619 | case 'a': | |
ad676077 AR |
620 | switch (filetype) { |
621 | case DEVCG_ALLOW: | |
7a3bb24f | 622 | if (css_has_online_children(&devcgroup->css)) |
bd2953eb AR |
623 | return -EINVAL; |
624 | ||
4cef7299 | 625 | if (!may_allow_all(parent)) |
ad676077 | 626 | return -EPERM; |
db9aeca9 | 627 | dev_exception_clean(devcgroup); |
64e10477 AR |
628 | devcgroup->behavior = DEVCG_DEFAULT_ALLOW; |
629 | if (!parent) | |
630 | break; | |
631 | ||
4cef7299 AR |
632 | rc = dev_exceptions_copy(&devcgroup->exceptions, |
633 | &parent->exceptions); | |
634 | if (rc) | |
635 | return rc; | |
ad676077 AR |
636 | break; |
637 | case DEVCG_DENY: | |
7a3bb24f | 638 | if (css_has_online_children(&devcgroup->css)) |
bd2953eb AR |
639 | return -EINVAL; |
640 | ||
db9aeca9 | 641 | dev_exception_clean(devcgroup); |
5b7aa7d5 | 642 | devcgroup->behavior = DEVCG_DEFAULT_DENY; |
ad676077 AR |
643 | break; |
644 | default: | |
645 | return -EINVAL; | |
646 | } | |
647 | return 0; | |
08ce5f16 | 648 | case 'b': |
db9aeca9 | 649 | ex.type = DEV_BLOCK; |
08ce5f16 SH |
650 | break; |
651 | case 'c': | |
db9aeca9 | 652 | ex.type = DEV_CHAR; |
08ce5f16 SH |
653 | break; |
654 | default: | |
f92523e3 | 655 | return -EINVAL; |
08ce5f16 SH |
656 | } |
657 | b++; | |
f92523e3 PM |
658 | if (!isspace(*b)) |
659 | return -EINVAL; | |
08ce5f16 SH |
660 | b++; |
661 | if (*b == '*') { | |
db9aeca9 | 662 | ex.major = ~0; |
08ce5f16 SH |
663 | b++; |
664 | } else if (isdigit(*b)) { | |
26fd8405 AR |
665 | memset(temp, 0, sizeof(temp)); |
666 | for (count = 0; count < sizeof(temp) - 1; count++) { | |
667 | temp[count] = *b; | |
668 | b++; | |
669 | if (!isdigit(*b)) | |
670 | break; | |
671 | } | |
672 | rc = kstrtou32(temp, 10, &ex.major); | |
673 | if (rc) | |
674 | return -EINVAL; | |
08ce5f16 | 675 | } else { |
f92523e3 | 676 | return -EINVAL; |
08ce5f16 | 677 | } |
f92523e3 PM |
678 | if (*b != ':') |
679 | return -EINVAL; | |
08ce5f16 SH |
680 | b++; |
681 | ||
682 | /* read minor */ | |
683 | if (*b == '*') { | |
db9aeca9 | 684 | ex.minor = ~0; |
08ce5f16 SH |
685 | b++; |
686 | } else if (isdigit(*b)) { | |
26fd8405 AR |
687 | memset(temp, 0, sizeof(temp)); |
688 | for (count = 0; count < sizeof(temp) - 1; count++) { | |
689 | temp[count] = *b; | |
690 | b++; | |
691 | if (!isdigit(*b)) | |
692 | break; | |
693 | } | |
694 | rc = kstrtou32(temp, 10, &ex.minor); | |
695 | if (rc) | |
696 | return -EINVAL; | |
08ce5f16 | 697 | } else { |
f92523e3 | 698 | return -EINVAL; |
08ce5f16 | 699 | } |
f92523e3 PM |
700 | if (!isspace(*b)) |
701 | return -EINVAL; | |
08ce5f16 SH |
702 | for (b++, count = 0; count < 3; count++, b++) { |
703 | switch (*b) { | |
704 | case 'r': | |
db9aeca9 | 705 | ex.access |= ACC_READ; |
08ce5f16 SH |
706 | break; |
707 | case 'w': | |
db9aeca9 | 708 | ex.access |= ACC_WRITE; |
08ce5f16 SH |
709 | break; |
710 | case 'm': | |
db9aeca9 | 711 | ex.access |= ACC_MKNOD; |
08ce5f16 SH |
712 | break; |
713 | case '\n': | |
714 | case '\0': | |
715 | count = 3; | |
716 | break; | |
717 | default: | |
f92523e3 | 718 | return -EINVAL; |
08ce5f16 SH |
719 | } |
720 | } | |
721 | ||
08ce5f16 SH |
722 | switch (filetype) { |
723 | case DEVCG_ALLOW: | |
ad676077 AR |
724 | /* |
725 | * If the default policy is to allow by default, try to remove | |
726 | * an matching exception instead. And be silent about it: we | |
727 | * don't want to break compatibility | |
728 | */ | |
5b7aa7d5 | 729 | if (devcgroup->behavior == DEVCG_DEFAULT_ALLOW) { |
d2c2b11c AR |
730 | /* Check if the parent allows removing it first */ |
731 | if (!parent_allows_removal(devcgroup, &ex)) | |
732 | return -EPERM; | |
db9aeca9 | 733 | dev_exception_rm(devcgroup, &ex); |
d2c2b11c | 734 | break; |
ad676077 | 735 | } |
d2c2b11c AR |
736 | |
737 | if (!parent_has_perm(devcgroup, &ex)) | |
738 | return -EPERM; | |
bd2953eb AR |
739 | rc = dev_exception_add(devcgroup, &ex); |
740 | break; | |
08ce5f16 | 741 | case DEVCG_DENY: |
ad676077 AR |
742 | /* |
743 | * If the default policy is to deny by default, try to remove | |
744 | * an matching exception instead. And be silent about it: we | |
745 | * don't want to break compatibility | |
746 | */ | |
bd2953eb | 747 | if (devcgroup->behavior == DEVCG_DEFAULT_DENY) |
db9aeca9 | 748 | dev_exception_rm(devcgroup, &ex); |
bd2953eb AR |
749 | else |
750 | rc = dev_exception_add(devcgroup, &ex); | |
751 | ||
752 | if (rc) | |
753 | break; | |
754 | /* we only propagate new restrictions */ | |
755 | rc = propagate_exception(devcgroup, &ex); | |
756 | break; | |
08ce5f16 | 757 | default: |
bd2953eb | 758 | rc = -EINVAL; |
08ce5f16 | 759 | } |
bd2953eb | 760 | return rc; |
f92523e3 | 761 | } |
08ce5f16 | 762 | |
451af504 TH |
763 | static ssize_t devcgroup_access_write(struct kernfs_open_file *of, |
764 | char *buf, size_t nbytes, loff_t off) | |
f92523e3 PM |
765 | { |
766 | int retval; | |
b4046f00 LZ |
767 | |
768 | mutex_lock(&devcgroup_mutex); | |
451af504 TH |
769 | retval = devcgroup_update_access(css_to_devcgroup(of_css(of)), |
770 | of_cft(of)->private, strstrip(buf)); | |
b4046f00 | 771 | mutex_unlock(&devcgroup_mutex); |
451af504 | 772 | return retval ?: nbytes; |
08ce5f16 SH |
773 | } |
774 | ||
775 | static struct cftype dev_cgroup_files[] = { | |
776 | { | |
777 | .name = "allow", | |
451af504 | 778 | .write = devcgroup_access_write, |
08ce5f16 SH |
779 | .private = DEVCG_ALLOW, |
780 | }, | |
781 | { | |
782 | .name = "deny", | |
451af504 | 783 | .write = devcgroup_access_write, |
08ce5f16 SH |
784 | .private = DEVCG_DENY, |
785 | }, | |
29486df3 SH |
786 | { |
787 | .name = "list", | |
2da8ca82 | 788 | .seq_show = devcgroup_seq_show, |
29486df3 SH |
789 | .private = DEVCG_LIST, |
790 | }, | |
4baf6e33 | 791 | { } /* terminate */ |
08ce5f16 SH |
792 | }; |
793 | ||
073219e9 | 794 | struct cgroup_subsys devices_cgrp_subsys = { |
92fb9748 TH |
795 | .css_alloc = devcgroup_css_alloc, |
796 | .css_free = devcgroup_css_free, | |
1909554c AR |
797 | .css_online = devcgroup_online, |
798 | .css_offline = devcgroup_offline, | |
5577964e | 799 | .legacy_cftypes = dev_cgroup_files, |
08ce5f16 SH |
800 | }; |
801 | ||
ad676077 AR |
802 | /** |
803 | * __devcgroup_check_permission - checks if an inode operation is permitted | |
804 | * @dev_cgroup: the dev cgroup to be tested against | |
805 | * @type: device type | |
806 | * @major: device major number | |
807 | * @minor: device minor number | |
808 | * @access: combination of ACC_WRITE, ACC_READ and ACC_MKNOD | |
809 | * | |
810 | * returns 0 on success, -EPERM case the operation is not permitted | |
811 | */ | |
8c9506d1 | 812 | static int __devcgroup_check_permission(short type, u32 major, u32 minor, |
ad676077 | 813 | short access) |
08ce5f16 | 814 | { |
8c9506d1 | 815 | struct dev_cgroup *dev_cgroup; |
79d71974 | 816 | bool rc; |
36fd71d2 | 817 | |
ad676077 | 818 | rcu_read_lock(); |
8c9506d1 | 819 | dev_cgroup = task_devcgroup(current); |
79d71974 AR |
820 | if (dev_cgroup->behavior == DEVCG_DEFAULT_ALLOW) |
821 | /* Can't match any of the exceptions, even partially */ | |
822 | rc = !match_exception_partial(&dev_cgroup->exceptions, | |
823 | type, major, minor, access); | |
824 | else | |
825 | /* Need to match completely one exception to be allowed */ | |
826 | rc = match_exception(&dev_cgroup->exceptions, type, major, | |
827 | minor, access); | |
ad676077 | 828 | rcu_read_unlock(); |
cd500819 | 829 | |
ad676077 AR |
830 | if (!rc) |
831 | return -EPERM; | |
36fd71d2 | 832 | |
ad676077 AR |
833 | return 0; |
834 | } | |
08ce5f16 | 835 | |
ad676077 AR |
836 | int __devcgroup_inode_permission(struct inode *inode, int mask) |
837 | { | |
ad676077 AR |
838 | short type, access = 0; |
839 | ||
840 | if (S_ISBLK(inode->i_mode)) | |
841 | type = DEV_BLOCK; | |
842 | if (S_ISCHR(inode->i_mode)) | |
843 | type = DEV_CHAR; | |
844 | if (mask & MAY_WRITE) | |
845 | access |= ACC_WRITE; | |
846 | if (mask & MAY_READ) | |
847 | access |= ACC_READ; | |
848 | ||
8c9506d1 JS |
849 | return __devcgroup_check_permission(type, imajor(inode), iminor(inode), |
850 | access); | |
08ce5f16 SH |
851 | } |
852 | ||
853 | int devcgroup_inode_mknod(int mode, dev_t dev) | |
854 | { | |
ad676077 | 855 | short type; |
08ce5f16 | 856 | |
0b82ac37 SH |
857 | if (!S_ISBLK(mode) && !S_ISCHR(mode)) |
858 | return 0; | |
859 | ||
ad676077 AR |
860 | if (S_ISBLK(mode)) |
861 | type = DEV_BLOCK; | |
862 | else | |
863 | type = DEV_CHAR; | |
36fd71d2 | 864 | |
8c9506d1 JS |
865 | return __devcgroup_check_permission(type, MAJOR(dev), MINOR(dev), |
866 | ACC_MKNOD); | |
36fd71d2 | 867 | |
08ce5f16 | 868 | } |