2 * dev_cgroup.c - device cgroup subsystem
4 * Copyright 2007 IBM Corp
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>
12 #include <linux/seq_file.h>
17 #define ACC_MASK (ACC_MKNOD | ACC_READ | ACC_WRITE)
21 #define DEV_ALL 4 /* this represents all devices */
24 * whitelist locking rules:
25 * cgroup_lock() cannot be taken under dev_cgroup->lock.
26 * dev_cgroup->lock can be taken with or without cgroup_lock().
28 * modifications always require cgroup_lock
29 * modifications to a list which is visible require the
30 * dev_cgroup->lock *and* cgroup_lock()
31 * walking the list requires dev_cgroup->lock or cgroup_lock().
33 * reasoning: dev_whitelist_copy() needs to kmalloc, so needs
34 * a mutex, which the cgroup_lock() is. Since modifying
35 * a visible list requires both locks, either lock can be
36 * taken for walking the list.
39 struct dev_whitelist_item
{
43 struct list_head list
;
48 struct cgroup_subsys_state css
;
49 struct list_head whitelist
;
53 static inline struct dev_cgroup
*css_to_devcgroup(struct cgroup_subsys_state
*s
)
55 return container_of(s
, struct dev_cgroup
, css
);
58 static inline struct dev_cgroup
*cgroup_to_devcgroup(struct cgroup
*cgroup
)
60 return css_to_devcgroup(cgroup_subsys_state(cgroup
, devices_subsys_id
));
63 static inline struct dev_cgroup
*task_devcgroup(struct task_struct
*task
)
65 return css_to_devcgroup(task_subsys_state(task
, devices_subsys_id
));
68 struct cgroup_subsys devices_subsys
;
70 static int devcgroup_can_attach(struct cgroup_subsys
*ss
,
71 struct cgroup
*new_cgroup
, struct task_struct
*task
)
73 if (current
!= task
&& !capable(CAP_SYS_ADMIN
))
80 * called under cgroup_lock()
82 static int dev_whitelist_copy(struct list_head
*dest
, struct list_head
*orig
)
84 struct dev_whitelist_item
*wh
, *tmp
, *new;
86 list_for_each_entry(wh
, orig
, list
) {
87 new = kmalloc(sizeof(*wh
), GFP_KERNEL
);
90 new->major
= wh
->major
;
91 new->minor
= wh
->minor
;
93 new->access
= wh
->access
;
94 list_add_tail(&new->list
, dest
);
100 list_for_each_entry_safe(wh
, tmp
, dest
, list
) {
107 /* Stupid prototype - don't bother combining existing entries */
109 * called under cgroup_lock()
110 * since the list is visible to other tasks, we need the spinlock also
112 static int dev_whitelist_add(struct dev_cgroup
*dev_cgroup
,
113 struct dev_whitelist_item
*wh
)
115 struct dev_whitelist_item
*whcopy
, *walk
;
117 whcopy
= kmalloc(sizeof(*whcopy
), GFP_KERNEL
);
121 memcpy(whcopy
, wh
, sizeof(*whcopy
));
122 spin_lock(&dev_cgroup
->lock
);
123 list_for_each_entry(walk
, &dev_cgroup
->whitelist
, list
) {
124 if (walk
->type
!= wh
->type
)
126 if (walk
->major
!= wh
->major
)
128 if (walk
->minor
!= wh
->minor
)
131 walk
->access
|= wh
->access
;
137 list_add_tail_rcu(&whcopy
->list
, &dev_cgroup
->whitelist
);
138 spin_unlock(&dev_cgroup
->lock
);
142 static void whitelist_item_free(struct rcu_head
*rcu
)
144 struct dev_whitelist_item
*item
;
146 item
= container_of(rcu
, struct dev_whitelist_item
, rcu
);
151 * called under cgroup_lock()
152 * since the list is visible to other tasks, we need the spinlock also
154 static void dev_whitelist_rm(struct dev_cgroup
*dev_cgroup
,
155 struct dev_whitelist_item
*wh
)
157 struct dev_whitelist_item
*walk
, *tmp
;
159 spin_lock(&dev_cgroup
->lock
);
160 list_for_each_entry_safe(walk
, tmp
, &dev_cgroup
->whitelist
, list
) {
161 if (walk
->type
== DEV_ALL
)
163 if (walk
->type
!= wh
->type
)
165 if (walk
->major
!= ~0 && walk
->major
!= wh
->major
)
167 if (walk
->minor
!= ~0 && walk
->minor
!= wh
->minor
)
171 walk
->access
&= ~wh
->access
;
173 list_del_rcu(&walk
->list
);
174 call_rcu(&walk
->rcu
, whitelist_item_free
);
177 spin_unlock(&dev_cgroup
->lock
);
181 * called from kernel/cgroup.c with cgroup_lock() held.
183 static struct cgroup_subsys_state
*devcgroup_create(struct cgroup_subsys
*ss
,
184 struct cgroup
*cgroup
)
186 struct dev_cgroup
*dev_cgroup
, *parent_dev_cgroup
;
187 struct cgroup
*parent_cgroup
;
190 dev_cgroup
= kzalloc(sizeof(*dev_cgroup
), GFP_KERNEL
);
192 return ERR_PTR(-ENOMEM
);
193 INIT_LIST_HEAD(&dev_cgroup
->whitelist
);
194 parent_cgroup
= cgroup
->parent
;
196 if (parent_cgroup
== NULL
) {
197 struct dev_whitelist_item
*wh
;
198 wh
= kmalloc(sizeof(*wh
), GFP_KERNEL
);
201 return ERR_PTR(-ENOMEM
);
203 wh
->minor
= wh
->major
= ~0;
205 wh
->access
= ACC_MKNOD
| ACC_READ
| ACC_WRITE
;
206 list_add(&wh
->list
, &dev_cgroup
->whitelist
);
208 parent_dev_cgroup
= cgroup_to_devcgroup(parent_cgroup
);
209 ret
= dev_whitelist_copy(&dev_cgroup
->whitelist
,
210 &parent_dev_cgroup
->whitelist
);
217 spin_lock_init(&dev_cgroup
->lock
);
218 return &dev_cgroup
->css
;
221 static void devcgroup_destroy(struct cgroup_subsys
*ss
,
222 struct cgroup
*cgroup
)
224 struct dev_cgroup
*dev_cgroup
;
225 struct dev_whitelist_item
*wh
, *tmp
;
227 dev_cgroup
= cgroup_to_devcgroup(cgroup
);
228 list_for_each_entry_safe(wh
, tmp
, &dev_cgroup
->whitelist
, list
) {
235 #define DEVCG_ALLOW 1
242 static void set_access(char *acc
, short access
)
245 memset(acc
, 0, ACCLEN
);
246 if (access
& ACC_READ
)
248 if (access
& ACC_WRITE
)
250 if (access
& ACC_MKNOD
)
254 static char type_to_char(short type
)
258 if (type
== DEV_CHAR
)
260 if (type
== DEV_BLOCK
)
265 static void set_majmin(char *str
, unsigned m
)
267 memset(str
, 0, MAJMINLEN
);
271 snprintf(str
, MAJMINLEN
, "%u", m
);
274 static int devcgroup_seq_read(struct cgroup
*cgroup
, struct cftype
*cft
,
277 struct dev_cgroup
*devcgroup
= cgroup_to_devcgroup(cgroup
);
278 struct dev_whitelist_item
*wh
;
279 char maj
[MAJMINLEN
], min
[MAJMINLEN
], acc
[ACCLEN
];
282 list_for_each_entry_rcu(wh
, &devcgroup
->whitelist
, list
) {
283 set_access(acc
, wh
->access
);
284 set_majmin(maj
, wh
->major
);
285 set_majmin(min
, wh
->minor
);
286 seq_printf(m
, "%c %s:%s %s\n", type_to_char(wh
->type
),
295 * may_access_whitelist:
296 * does the access granted to dev_cgroup c contain the access
297 * requested in whitelist item refwh.
298 * return 1 if yes, 0 if no.
299 * call with c->lock held
301 static int may_access_whitelist(struct dev_cgroup
*c
,
302 struct dev_whitelist_item
*refwh
)
304 struct dev_whitelist_item
*whitem
;
306 list_for_each_entry(whitem
, &c
->whitelist
, list
) {
307 if (whitem
->type
& DEV_ALL
)
309 if ((refwh
->type
& DEV_BLOCK
) && !(whitem
->type
& DEV_BLOCK
))
311 if ((refwh
->type
& DEV_CHAR
) && !(whitem
->type
& DEV_CHAR
))
313 if (whitem
->major
!= ~0 && whitem
->major
!= refwh
->major
)
315 if (whitem
->minor
!= ~0 && whitem
->minor
!= refwh
->minor
)
317 if (refwh
->access
& (~whitem
->access
))
326 * when adding a new allow rule to a device whitelist, the rule
327 * must be allowed in the parent device
329 static int parent_has_perm(struct dev_cgroup
*childcg
,
330 struct dev_whitelist_item
*wh
)
332 struct cgroup
*pcg
= childcg
->css
.cgroup
->parent
;
333 struct dev_cgroup
*parent
;
338 parent
= cgroup_to_devcgroup(pcg
);
339 spin_lock(&parent
->lock
);
340 ret
= may_access_whitelist(parent
, wh
);
341 spin_unlock(&parent
->lock
);
346 * Modify the whitelist using allow/deny rules.
347 * CAP_SYS_ADMIN is needed for this. It's at least separate from CAP_MKNOD
348 * so we can give a container CAP_MKNOD to let it create devices but not
349 * modify the whitelist.
350 * It seems likely we'll want to add a CAP_CONTAINER capability to allow
351 * us to also grant CAP_SYS_ADMIN to containers without giving away the
352 * device whitelist controls, but for now we'll stick with CAP_SYS_ADMIN
354 * Taking rules away is always allowed (given CAP_SYS_ADMIN). Granting
355 * new access is only allowed if you're in the top-level cgroup, or your
356 * parent cgroup has the access you're asking for.
358 static int devcgroup_update_access(struct dev_cgroup
*devcgroup
,
359 int filetype
, const char *buffer
)
361 struct dev_cgroup
*cur_devcgroup
;
363 int retval
= 0, count
;
364 struct dev_whitelist_item wh
;
366 if (!capable(CAP_SYS_ADMIN
))
369 cur_devcgroup
= task_devcgroup(current
);
371 memset(&wh
, 0, sizeof(wh
));
377 wh
.access
= ACC_MASK
;
397 } else if (isdigit(*b
)) {
399 while (isdigit(*b
)) {
400 wh
.major
= wh
.major
*10+(*b
-'0');
414 } else if (isdigit(*b
)) {
416 while (isdigit(*b
)) {
417 wh
.minor
= wh
.minor
*10+(*b
-'0');
425 for (b
++, count
= 0; count
< 3; count
++, b
++) {
428 wh
.access
|= ACC_READ
;
431 wh
.access
|= ACC_WRITE
;
434 wh
.access
|= ACC_MKNOD
;
449 if (!parent_has_perm(devcgroup
, &wh
))
451 return dev_whitelist_add(devcgroup
, &wh
);
453 dev_whitelist_rm(devcgroup
, &wh
);
461 static int devcgroup_access_write(struct cgroup
*cgrp
, struct cftype
*cft
,
465 if (!cgroup_lock_live_group(cgrp
))
467 retval
= devcgroup_update_access(cgroup_to_devcgroup(cgrp
),
468 cft
->private, buffer
);
473 static struct cftype dev_cgroup_files
[] = {
476 .write_string
= devcgroup_access_write
,
477 .private = DEVCG_ALLOW
,
481 .write_string
= devcgroup_access_write
,
482 .private = DEVCG_DENY
,
486 .read_seq_string
= devcgroup_seq_read
,
487 .private = DEVCG_LIST
,
491 static int devcgroup_populate(struct cgroup_subsys
*ss
,
492 struct cgroup
*cgroup
)
494 return cgroup_add_files(cgroup
, ss
, dev_cgroup_files
,
495 ARRAY_SIZE(dev_cgroup_files
));
498 struct cgroup_subsys devices_subsys
= {
500 .can_attach
= devcgroup_can_attach
,
501 .create
= devcgroup_create
,
502 .destroy
= devcgroup_destroy
,
503 .populate
= devcgroup_populate
,
504 .subsys_id
= devices_subsys_id
,
507 int devcgroup_inode_permission(struct inode
*inode
, int mask
)
509 struct dev_cgroup
*dev_cgroup
;
510 struct dev_whitelist_item
*wh
;
512 dev_t device
= inode
->i_rdev
;
515 if (!S_ISBLK(inode
->i_mode
) && !S_ISCHR(inode
->i_mode
))
517 dev_cgroup
= css_to_devcgroup(task_subsys_state(current
,
523 list_for_each_entry_rcu(wh
, &dev_cgroup
->whitelist
, list
) {
524 if (wh
->type
& DEV_ALL
)
526 if ((wh
->type
& DEV_BLOCK
) && !S_ISBLK(inode
->i_mode
))
528 if ((wh
->type
& DEV_CHAR
) && !S_ISCHR(inode
->i_mode
))
530 if (wh
->major
!= ~0 && wh
->major
!= imajor(inode
))
532 if (wh
->minor
!= ~0 && wh
->minor
!= iminor(inode
))
535 if ((mask
& MAY_WRITE
) && !(wh
->access
& ACC_WRITE
))
537 if ((mask
& MAY_READ
) && !(wh
->access
& ACC_READ
))
547 int devcgroup_inode_mknod(int mode
, dev_t dev
)
549 struct dev_cgroup
*dev_cgroup
;
550 struct dev_whitelist_item
*wh
;
552 dev_cgroup
= css_to_devcgroup(task_subsys_state(current
,
558 list_for_each_entry(wh
, &dev_cgroup
->whitelist
, list
) {
559 if (wh
->type
& DEV_ALL
)
561 if ((wh
->type
& DEV_BLOCK
) && !S_ISBLK(mode
))
563 if ((wh
->type
& DEV_CHAR
) && !S_ISCHR(mode
))
565 if (wh
->major
!= ~0 && wh
->major
!= MAJOR(dev
))
567 if (wh
->minor
!= ~0 && wh
->minor
!= MINOR(dev
))
570 if (!(wh
->access
& ACC_MKNOD
))