2 * device_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>
13 #include <linux/slab.h>
14 #include <linux/rcupdate.h>
15 #include <linux/mutex.h>
20 #define ACC_MASK (ACC_MKNOD | ACC_READ | ACC_WRITE)
24 #define DEV_ALL 4 /* this represents all devices */
26 static DEFINE_MUTEX(devcgroup_mutex);
29 * whitelist locking rules:
30 * hold devcgroup_mutex for update/read.
31 * hold rcu_read_lock() for read.
34 struct dev_whitelist_item {
38 struct list_head list;
43 struct cgroup_subsys_state css;
44 struct list_head whitelist;
47 static inline struct dev_cgroup *css_to_devcgroup(struct cgroup_subsys_state *s)
49 return container_of(s, struct dev_cgroup, css);
52 static inline struct dev_cgroup *cgroup_to_devcgroup(struct cgroup *cgroup)
54 return css_to_devcgroup(cgroup_subsys_state(cgroup, devices_subsys_id));
57 static inline struct dev_cgroup *task_devcgroup(struct task_struct *task)
59 return css_to_devcgroup(task_subsys_state(task, devices_subsys_id));
62 struct cgroup_subsys devices_subsys;
64 static int devcgroup_can_attach(struct cgroup_subsys *ss,
65 struct cgroup *new_cgroup, struct task_struct *task)
67 if (current != task && !capable(CAP_SYS_ADMIN))
74 * called under devcgroup_mutex
76 static int dev_whitelist_copy(struct list_head *dest, struct list_head *orig)
78 struct dev_whitelist_item *wh, *tmp, *new;
80 list_for_each_entry(wh, orig, list) {
81 new = kmemdup(wh, sizeof(*wh), GFP_KERNEL);
84 list_add_tail(&new->list, dest);
90 list_for_each_entry_safe(wh, tmp, dest, list) {
97 /* Stupid prototype - don't bother combining existing entries */
99 * called under devcgroup_mutex
101 static int dev_whitelist_add(struct dev_cgroup *dev_cgroup,
102 struct dev_whitelist_item *wh)
104 struct dev_whitelist_item *whcopy, *walk;
106 whcopy = kmemdup(wh, sizeof(*wh), GFP_KERNEL);
110 list_for_each_entry(walk, &dev_cgroup->whitelist, list) {
111 if (walk->type != wh->type)
113 if (walk->major != wh->major)
115 if (walk->minor != wh->minor)
118 walk->access |= wh->access;
124 list_add_tail_rcu(&whcopy->list, &dev_cgroup->whitelist);
128 static void whitelist_item_free(struct rcu_head *rcu)
130 struct dev_whitelist_item *item;
132 item = container_of(rcu, struct dev_whitelist_item, rcu);
137 * called under devcgroup_mutex
139 static void dev_whitelist_rm(struct dev_cgroup *dev_cgroup,
140 struct dev_whitelist_item *wh)
142 struct dev_whitelist_item *walk, *tmp;
144 list_for_each_entry_safe(walk, tmp, &dev_cgroup->whitelist, list) {
145 if (walk->type == DEV_ALL)
147 if (walk->type != wh->type)
149 if (walk->major != ~0 && walk->major != wh->major)
151 if (walk->minor != ~0 && walk->minor != wh->minor)
155 walk->access &= ~wh->access;
157 list_del_rcu(&walk->list);
158 call_rcu(&walk->rcu, whitelist_item_free);
164 * called from kernel/cgroup.c with cgroup_lock() held.
166 static struct cgroup_subsys_state *devcgroup_create(struct cgroup_subsys *ss,
167 struct cgroup *cgroup)
169 struct dev_cgroup *dev_cgroup, *parent_dev_cgroup;
170 struct cgroup *parent_cgroup;
173 dev_cgroup = kzalloc(sizeof(*dev_cgroup), GFP_KERNEL);
175 return ERR_PTR(-ENOMEM);
176 INIT_LIST_HEAD(&dev_cgroup->whitelist);
177 parent_cgroup = cgroup->parent;
179 if (parent_cgroup == NULL) {
180 struct dev_whitelist_item *wh;
181 wh = kmalloc(sizeof(*wh), GFP_KERNEL);
184 return ERR_PTR(-ENOMEM);
186 wh->minor = wh->major = ~0;
188 wh->access = ACC_MASK;
189 list_add(&wh->list, &dev_cgroup->whitelist);
191 parent_dev_cgroup = cgroup_to_devcgroup(parent_cgroup);
192 mutex_lock(&devcgroup_mutex);
193 ret = dev_whitelist_copy(&dev_cgroup->whitelist,
194 &parent_dev_cgroup->whitelist);
195 mutex_unlock(&devcgroup_mutex);
202 return &dev_cgroup->css;
205 static void devcgroup_destroy(struct cgroup_subsys *ss,
206 struct cgroup *cgroup)
208 struct dev_cgroup *dev_cgroup;
209 struct dev_whitelist_item *wh, *tmp;
211 dev_cgroup = cgroup_to_devcgroup(cgroup);
212 list_for_each_entry_safe(wh, tmp, &dev_cgroup->whitelist, list) {
219 #define DEVCG_ALLOW 1
226 static void set_access(char *acc, short access)
229 memset(acc, 0, ACCLEN);
230 if (access & ACC_READ)
232 if (access & ACC_WRITE)
234 if (access & ACC_MKNOD)
238 static char type_to_char(short type)
242 if (type == DEV_CHAR)
244 if (type == DEV_BLOCK)
249 static void set_majmin(char *str, unsigned m)
254 sprintf(str, "%u", m);
257 static int devcgroup_seq_read(struct cgroup *cgroup, struct cftype *cft,
260 struct dev_cgroup *devcgroup = cgroup_to_devcgroup(cgroup);
261 struct dev_whitelist_item *wh;
262 char maj[MAJMINLEN], min[MAJMINLEN], acc[ACCLEN];
265 list_for_each_entry_rcu(wh, &devcgroup->whitelist, list) {
266 set_access(acc, wh->access);
267 set_majmin(maj, wh->major);
268 set_majmin(min, wh->minor);
269 seq_printf(m, "%c %s:%s %s\n", type_to_char(wh->type),
278 * may_access_whitelist:
279 * does the access granted to dev_cgroup c contain the access
280 * requested in whitelist item refwh.
281 * return 1 if yes, 0 if no.
282 * call with devcgroup_mutex held
284 static int may_access_whitelist(struct dev_cgroup *c,
285 struct dev_whitelist_item *refwh)
287 struct dev_whitelist_item *whitem;
289 list_for_each_entry(whitem, &c->whitelist, list) {
290 if (whitem->type & DEV_ALL)
292 if ((refwh->type & DEV_BLOCK) && !(whitem->type & DEV_BLOCK))
294 if ((refwh->type & DEV_CHAR) && !(whitem->type & DEV_CHAR))
296 if (whitem->major != ~0 && whitem->major != refwh->major)
298 if (whitem->minor != ~0 && whitem->minor != refwh->minor)
300 if (refwh->access & (~whitem->access))
309 * when adding a new allow rule to a device whitelist, the rule
310 * must be allowed in the parent device
312 static int parent_has_perm(struct dev_cgroup *childcg,
313 struct dev_whitelist_item *wh)
315 struct cgroup *pcg = childcg->css.cgroup->parent;
316 struct dev_cgroup *parent;
320 parent = cgroup_to_devcgroup(pcg);
321 return may_access_whitelist(parent, wh);
325 * Modify the whitelist using allow/deny rules.
326 * CAP_SYS_ADMIN is needed for this. It's at least separate from CAP_MKNOD
327 * so we can give a container CAP_MKNOD to let it create devices but not
328 * modify the whitelist.
329 * It seems likely we'll want to add a CAP_CONTAINER capability to allow
330 * us to also grant CAP_SYS_ADMIN to containers without giving away the
331 * device whitelist controls, but for now we'll stick with CAP_SYS_ADMIN
333 * Taking rules away is always allowed (given CAP_SYS_ADMIN). Granting
334 * new access is only allowed if you're in the top-level cgroup, or your
335 * parent cgroup has the access you're asking for.
337 static int devcgroup_update_access(struct dev_cgroup *devcgroup,
338 int filetype, const char *buffer)
343 struct dev_whitelist_item wh;
345 if (!capable(CAP_SYS_ADMIN))
348 memset(&wh, 0, sizeof(wh));
354 wh.access = ACC_MASK;
374 } else if (isdigit(*b)) {
375 wh.major = simple_strtoul(b, &endp, 10);
388 } else if (isdigit(*b)) {
389 wh.minor = simple_strtoul(b, &endp, 10);
396 for (b++, count = 0; count < 3; count++, b++) {
399 wh.access |= ACC_READ;
402 wh.access |= ACC_WRITE;
405 wh.access |= ACC_MKNOD;
419 if (!parent_has_perm(devcgroup, &wh))
421 return dev_whitelist_add(devcgroup, &wh);
423 dev_whitelist_rm(devcgroup, &wh);
431 static int devcgroup_access_write(struct cgroup *cgrp, struct cftype *cft,
436 mutex_lock(&devcgroup_mutex);
437 retval = devcgroup_update_access(cgroup_to_devcgroup(cgrp),
438 cft->private, buffer);
439 mutex_unlock(&devcgroup_mutex);
443 static struct cftype dev_cgroup_files[] = {
446 .write_string = devcgroup_access_write,
447 .private = DEVCG_ALLOW,
451 .write_string = devcgroup_access_write,
452 .private = DEVCG_DENY,
456 .read_seq_string = devcgroup_seq_read,
457 .private = DEVCG_LIST,
461 static int devcgroup_populate(struct cgroup_subsys *ss,
462 struct cgroup *cgroup)
464 return cgroup_add_files(cgroup, ss, dev_cgroup_files,
465 ARRAY_SIZE(dev_cgroup_files));
468 struct cgroup_subsys devices_subsys = {
470 .can_attach = devcgroup_can_attach,
471 .create = devcgroup_create,
472 .destroy = devcgroup_destroy,
473 .populate = devcgroup_populate,
474 .subsys_id = devices_subsys_id,
477 int devcgroup_inode_permission(struct inode *inode, int mask)
479 struct dev_cgroup *dev_cgroup;
480 struct dev_whitelist_item *wh;
482 dev_t device = inode->i_rdev;
485 if (!S_ISBLK(inode->i_mode) && !S_ISCHR(inode->i_mode))
490 dev_cgroup = task_devcgroup(current);
492 list_for_each_entry_rcu(wh, &dev_cgroup->whitelist, list) {
493 if (wh->type & DEV_ALL)
495 if ((wh->type & DEV_BLOCK) && !S_ISBLK(inode->i_mode))
497 if ((wh->type & DEV_CHAR) && !S_ISCHR(inode->i_mode))
499 if (wh->major != ~0 && wh->major != imajor(inode))
501 if (wh->minor != ~0 && wh->minor != iminor(inode))
504 if ((mask & MAY_WRITE) && !(wh->access & ACC_WRITE))
506 if ((mask & MAY_READ) && !(wh->access & ACC_READ))
518 int devcgroup_inode_mknod(int mode, dev_t dev)
520 struct dev_cgroup *dev_cgroup;
521 struct dev_whitelist_item *wh;
523 if (!S_ISBLK(mode) && !S_ISCHR(mode))
528 dev_cgroup = task_devcgroup(current);
530 list_for_each_entry_rcu(wh, &dev_cgroup->whitelist, list) {
531 if (wh->type & DEV_ALL)
533 if ((wh->type & DEV_BLOCK) && !S_ISBLK(mode))
535 if ((wh->type & DEV_CHAR) && !S_ISCHR(mode))
537 if (wh->major != ~0 && wh->major != MAJOR(dev))
539 if (wh->minor != ~0 && wh->minor != MINOR(dev))
542 if (!(wh->access & ACC_MKNOD))