device_cgroup.c 14.4 KB
Newer Older
1
/*
2
 * device_cgroup.c - device cgroup subsystem
3 4 5 6 7 8 9 10 11
 *
 * Copyright 2007 IBM Corp
 */

#include <linux/device_cgroup.h>
#include <linux/cgroup.h>
#include <linux/ctype.h>
#include <linux/list.h>
#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>
16 17 18 19 20 21 22 23 24 25

#define ACC_MKNOD 1
#define ACC_READ  2
#define ACC_WRITE 4
#define ACC_MASK (ACC_MKNOD | ACC_READ | ACC_WRITE)

#define DEV_BLOCK 1
#define DEV_CHAR  2
#define DEV_ALL   4  /* this represents all devices */

26 27
static DEFINE_MUTEX(devcgroup_mutex);

28
/*
29
 * exception list locking rules:
30
 * hold devcgroup_mutex for update/read.
31
 * hold rcu_read_lock() for read.
32 33
 */

34
struct dev_exception_item {
35 36 37 38
	u32 major, minor;
	short type;
	short access;
	struct list_head list;
39
	struct rcu_head rcu;
40 41 42 43
};

struct dev_cgroup {
	struct cgroup_subsys_state css;
44
	struct list_head exceptions;
45 46 47 48
	enum {
		DEVCG_DEFAULT_ALLOW,
		DEVCG_DEFAULT_DENY,
	} behavior;
49 50
};

51 52 53 54 55
static inline struct dev_cgroup *css_to_devcgroup(struct cgroup_subsys_state *s)
{
	return container_of(s, struct dev_cgroup, css);
}

56 57
static inline struct dev_cgroup *cgroup_to_devcgroup(struct cgroup *cgroup)
{
58
	return css_to_devcgroup(cgroup_subsys_state(cgroup, devices_subsys_id));
59 60
}

61 62 63 64 65
static inline struct dev_cgroup *task_devcgroup(struct task_struct *task)
{
	return css_to_devcgroup(task_subsys_state(task, devices_subsys_id));
}

66 67
struct cgroup_subsys devices_subsys;

68 69
static int devcgroup_can_attach(struct cgroup *new_cgrp,
				struct cgroup_taskset *set)
70
{
71
	struct task_struct *task = cgroup_taskset_first(set);
72

73 74
	if (current != task && !capable(CAP_SYS_ADMIN))
		return -EPERM;
75 76 77 78
	return 0;
}

/*
79
 * called under devcgroup_mutex
80
 */
81
static int dev_exceptions_copy(struct list_head *dest, struct list_head *orig)
82
{
83
	struct dev_exception_item *ex, *tmp, *new;
84

85 86
	list_for_each_entry(ex, orig, list) {
		new = kmemdup(ex, sizeof(*ex), GFP_KERNEL);
87 88 89 90 91 92 93 94
		if (!new)
			goto free_and_exit;
		list_add_tail(&new->list, dest);
	}

	return 0;

free_and_exit:
95 96 97
	list_for_each_entry_safe(ex, tmp, dest, list) {
		list_del(&ex->list);
		kfree(ex);
98 99 100 101 102
	}
	return -ENOMEM;
}

/*
103
 * called under devcgroup_mutex
104
 */
105 106
static int dev_exception_add(struct dev_cgroup *dev_cgroup,
			     struct dev_exception_item *ex)
107
{
108
	struct dev_exception_item *excopy, *walk;
109

110 111
	excopy = kmemdup(ex, sizeof(*ex), GFP_KERNEL);
	if (!excopy)
112 113
		return -ENOMEM;

114 115
	list_for_each_entry(walk, &dev_cgroup->exceptions, list) {
		if (walk->type != ex->type)
116
			continue;
117
		if (walk->major != ex->major)
118
			continue;
119
		if (walk->minor != ex->minor)
120 121
			continue;

122 123 124
		walk->access |= ex->access;
		kfree(excopy);
		excopy = NULL;
125 126
	}

127 128
	if (excopy != NULL)
		list_add_tail_rcu(&excopy->list, &dev_cgroup->exceptions);
129 130 131 132
	return 0;
}

/*
133
 * called under devcgroup_mutex
134
 */
135 136
static void dev_exception_rm(struct dev_cgroup *dev_cgroup,
			     struct dev_exception_item *ex)
137
{
138
	struct dev_exception_item *walk, *tmp;
139

140 141
	list_for_each_entry_safe(walk, tmp, &dev_cgroup->exceptions, list) {
		if (walk->type != ex->type)
142
			continue;
143
		if (walk->major != ex->major)
144
			continue;
145
		if (walk->minor != ex->minor)
146 147
			continue;

148
		walk->access &= ~ex->access;
149
		if (!walk->access) {
150
			list_del_rcu(&walk->list);
151
			kfree_rcu(walk, rcu);
152 153 154 155
		}
	}
}

156
/**
157 158
 * dev_exception_clean - frees all entries of the exception list
 * @dev_cgroup: dev_cgroup with the exception list to be cleaned
159 160 161
 *
 * called under devcgroup_mutex
 */
162
static void dev_exception_clean(struct dev_cgroup *dev_cgroup)
163
{
164
	struct dev_exception_item *ex, *tmp;
165

166
	list_for_each_entry_safe(ex, tmp, &dev_cgroup->exceptions, list) {
Tejun Heo's avatar
Tejun Heo committed
167 168
		list_del_rcu(&ex->list);
		kfree_rcu(ex, rcu);
169 170 171
	}
}

172 173 174
/*
 * called from kernel/cgroup.c with cgroup_lock() held.
 */
175
static struct cgroup_subsys_state *devcgroup_create(struct cgroup *cgroup)
176 177 178 179 180 181 182 183
{
	struct dev_cgroup *dev_cgroup, *parent_dev_cgroup;
	struct cgroup *parent_cgroup;
	int ret;

	dev_cgroup = kzalloc(sizeof(*dev_cgroup), GFP_KERNEL);
	if (!dev_cgroup)
		return ERR_PTR(-ENOMEM);
184
	INIT_LIST_HEAD(&dev_cgroup->exceptions);
185 186
	parent_cgroup = cgroup->parent;

187
	if (parent_cgroup == NULL)
188
		dev_cgroup->behavior = DEVCG_DEFAULT_ALLOW;
189
	else {
190
		parent_dev_cgroup = cgroup_to_devcgroup(parent_cgroup);
191
		mutex_lock(&devcgroup_mutex);
192 193
		ret = dev_exceptions_copy(&dev_cgroup->exceptions,
					  &parent_dev_cgroup->exceptions);
194
		dev_cgroup->behavior = parent_dev_cgroup->behavior;
195
		mutex_unlock(&devcgroup_mutex);
196 197 198 199 200 201 202 203 204
		if (ret) {
			kfree(dev_cgroup);
			return ERR_PTR(ret);
		}
	}

	return &dev_cgroup->css;
}

205
static void devcgroup_destroy(struct cgroup *cgroup)
206 207 208 209
{
	struct dev_cgroup *dev_cgroup;

	dev_cgroup = cgroup_to_devcgroup(cgroup);
210
	dev_exception_clean(dev_cgroup);
211 212 213 214 215
	kfree(dev_cgroup);
}

#define DEVCG_ALLOW 1
#define DEVCG_DENY 2
216 217
#define DEVCG_LIST 3

218
#define MAJMINLEN 13
219
#define ACCLEN 4
220 221 222 223

static void set_access(char *acc, short access)
{
	int idx = 0;
224
	memset(acc, 0, ACCLEN);
225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243
	if (access & ACC_READ)
		acc[idx++] = 'r';
	if (access & ACC_WRITE)
		acc[idx++] = 'w';
	if (access & ACC_MKNOD)
		acc[idx++] = 'm';
}

static char type_to_char(short type)
{
	if (type == DEV_ALL)
		return 'a';
	if (type == DEV_CHAR)
		return 'c';
	if (type == DEV_BLOCK)
		return 'b';
	return 'X';
}

244
static void set_majmin(char *str, unsigned m)
245 246
{
	if (m == ~0)
Li Zefan's avatar
Li Zefan committed
247
		strcpy(str, "*");
248
	else
Li Zefan's avatar
Li Zefan committed
249
		sprintf(str, "%u", m);
250 251
}

252 253
static int devcgroup_seq_read(struct cgroup *cgroup, struct cftype *cft,
				struct seq_file *m)
254
{
255
	struct dev_cgroup *devcgroup = cgroup_to_devcgroup(cgroup);
256
	struct dev_exception_item *ex;
257
	char maj[MAJMINLEN], min[MAJMINLEN], acc[ACCLEN];
258

259
	rcu_read_lock();
260 261 262 263 264 265
	/*
	 * To preserve the compatibility:
	 * - Only show the "all devices" when the default policy is to allow
	 * - List the exceptions in case the default policy is to deny
	 * This way, the file remains as a "whitelist of devices"
	 */
266
	if (devcgroup->behavior == DEVCG_DEFAULT_ALLOW) {
267 268 269 270
		set_access(acc, ACC_MASK);
		set_majmin(maj, ~0);
		set_majmin(min, ~0);
		seq_printf(m, "%c %s:%s %s\n", type_to_char(DEV_ALL),
271
			   maj, min, acc);
272
	} else {
273 274 275 276 277
		list_for_each_entry_rcu(ex, &devcgroup->exceptions, list) {
			set_access(acc, ex->access);
			set_majmin(maj, ex->major);
			set_majmin(min, ex->minor);
			seq_printf(m, "%c %s:%s %s\n", type_to_char(ex->type),
278 279
				   maj, min, acc);
		}
280
	}
281
	rcu_read_unlock();
282

283
	return 0;
284 285
}

286
/**
287 288 289 290 291
 * may_access - verifies if a new exception is part of what is allowed
 *		by a dev cgroup based on the default policy +
 *		exceptions. This is used to make sure a child cgroup
 *		won't have more privileges than its parent or to
 *		verify if a certain access is allowed.
292
 * @dev_cgroup: dev cgroup to be tested against
293
 * @refex: new exception
294
 */
295 296
static int may_access(struct dev_cgroup *dev_cgroup,
		      struct dev_exception_item *refex)
297
{
298
	struct dev_exception_item *ex;
299
	bool match = false;
300

Tejun Heo's avatar
Tejun Heo committed
301
	list_for_each_entry_rcu(ex, &dev_cgroup->exceptions, list) {
302
		if ((refex->type & DEV_BLOCK) && !(ex->type & DEV_BLOCK))
303
			continue;
304
		if ((refex->type & DEV_CHAR) && !(ex->type & DEV_CHAR))
305
			continue;
306
		if (ex->major != ~0 && ex->major != refex->major)
307
			continue;
308
		if (ex->minor != ~0 && ex->minor != refex->minor)
309
			continue;
310
		if (refex->access & (~ex->access))
311
			continue;
312 313
		match = true;
		break;
314
	}
315 316

	/*
317
	 * In two cases we'll consider this new exception valid:
318
	 * - the dev cgroup has its default policy to allow + exception list:
319
	 *   the new exception should *not* match any of the exceptions
320
	 *   (behavior == DEVCG_DEFAULT_ALLOW, !match)
321
	 * - the dev cgroup has its default policy to deny + exception list:
322
	 *   the new exception *should* match the exceptions
323
	 *   (behavior == DEVCG_DEFAULT_DENY, match)
324
	 */
325
	if ((dev_cgroup->behavior == DEVCG_DEFAULT_DENY) == match)
326
		return 1;
327 328 329 330 331
	return 0;
}

/*
 * parent_has_perm:
332
 * when adding a new allow rule to a device exception list, the rule
333 334
 * must be allowed in the parent device
 */
335
static int parent_has_perm(struct dev_cgroup *childcg,
336
				  struct dev_exception_item *ex)
337
{
338
	struct cgroup *pcg = childcg->css.cgroup->parent;
339 340 341 342 343
	struct dev_cgroup *parent;

	if (!pcg)
		return 1;
	parent = cgroup_to_devcgroup(pcg);
344
	return may_access(parent, ex);
345 346
}

347 348 349 350 351 352 353 354
/**
 * may_allow_all - checks if it's possible to change the behavior to
 *		   allow based on parent's rules.
 * @parent: device cgroup's parent
 * returns: != 0 in case it's allowed, 0 otherwise
 */
static inline int may_allow_all(struct dev_cgroup *parent)
{
355 356
	if (!parent)
		return 1;
357 358 359
	return parent->behavior == DEVCG_DEFAULT_ALLOW;
}

360
/*
361
 * Modify the exception list using allow/deny rules.
362 363
 * CAP_SYS_ADMIN is needed for this.  It's at least separate from CAP_MKNOD
 * so we can give a container CAP_MKNOD to let it create devices but not
364
 * modify the exception list.
365 366
 * It seems likely we'll want to add a CAP_CONTAINER capability to allow
 * us to also grant CAP_SYS_ADMIN to containers without giving away the
367
 * device exception list controls, but for now we'll stick with CAP_SYS_ADMIN
368 369 370 371 372
 *
 * Taking rules away is always allowed (given CAP_SYS_ADMIN).  Granting
 * new access is only allowed if you're in the top-level cgroup, or your
 * parent cgroup has the access you're asking for.
 */
373 374
static int devcgroup_update_access(struct dev_cgroup *devcgroup,
				   int filetype, const char *buffer)
375
{
376
	const char *b;
377 378
	char temp[12];		/* 11 + 1 characters needed for a u32 */
	int count, rc;
379
	struct dev_exception_item ex;
380
	struct cgroup *p = devcgroup->css.cgroup;
381
	struct dev_cgroup *parent = NULL;
382 383 384 385

	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;

386 387 388
	if (p->parent)
		parent = cgroup_to_devcgroup(p->parent);

389
	memset(&ex, 0, sizeof(ex));
390 391 392 393
	b = buffer;

	switch (*b) {
	case 'a':
394 395
		switch (filetype) {
		case DEVCG_ALLOW:
396
			if (!may_allow_all(parent))
397
				return -EPERM;
398
			dev_exception_clean(devcgroup);
399 400 401 402
			devcgroup->behavior = DEVCG_DEFAULT_ALLOW;
			if (!parent)
				break;

403 404 405 406
			rc = dev_exceptions_copy(&devcgroup->exceptions,
						 &parent->exceptions);
			if (rc)
				return rc;
407 408
			break;
		case DEVCG_DENY:
409
			dev_exception_clean(devcgroup);
410
			devcgroup->behavior = DEVCG_DEFAULT_DENY;
411 412 413 414 415
			break;
		default:
			return -EINVAL;
		}
		return 0;
416
	case 'b':
417
		ex.type = DEV_BLOCK;
418 419
		break;
	case 'c':
420
		ex.type = DEV_CHAR;
421 422
		break;
	default:
423
		return -EINVAL;
424 425
	}
	b++;
426 427
	if (!isspace(*b))
		return -EINVAL;
428 429
	b++;
	if (*b == '*') {
430
		ex.major = ~0;
431 432
		b++;
	} else if (isdigit(*b)) {
433 434 435 436 437 438 439 440 441 442
		memset(temp, 0, sizeof(temp));
		for (count = 0; count < sizeof(temp) - 1; count++) {
			temp[count] = *b;
			b++;
			if (!isdigit(*b))
				break;
		}
		rc = kstrtou32(temp, 10, &ex.major);
		if (rc)
			return -EINVAL;
443
	} else {
444
		return -EINVAL;
445
	}
446 447
	if (*b != ':')
		return -EINVAL;
448 449 450 451
	b++;

	/* read minor */
	if (*b == '*') {
452
		ex.minor = ~0;
453 454
		b++;
	} else if (isdigit(*b)) {
455 456 457 458 459 460 461 462 463 464
		memset(temp, 0, sizeof(temp));
		for (count = 0; count < sizeof(temp) - 1; count++) {
			temp[count] = *b;
			b++;
			if (!isdigit(*b))
				break;
		}
		rc = kstrtou32(temp, 10, &ex.minor);
		if (rc)
			return -EINVAL;
465
	} else {
466
		return -EINVAL;
467
	}
468 469
	if (!isspace(*b))
		return -EINVAL;
470 471 472
	for (b++, count = 0; count < 3; count++, b++) {
		switch (*b) {
		case 'r':
473
			ex.access |= ACC_READ;
474 475
			break;
		case 'w':
476
			ex.access |= ACC_WRITE;
477 478
			break;
		case 'm':
479
			ex.access |= ACC_MKNOD;
480 481 482 483 484 485
			break;
		case '\n':
		case '\0':
			count = 3;
			break;
		default:
486
			return -EINVAL;
487 488 489 490 491
		}
	}

	switch (filetype) {
	case DEVCG_ALLOW:
492
		if (!parent_has_perm(devcgroup, &ex))
493
			return -EPERM;
494 495 496 497 498
		/*
		 * If the default policy is to allow by default, try to remove
		 * an matching exception instead. And be silent about it: we
		 * don't want to break compatibility
		 */
499
		if (devcgroup->behavior == DEVCG_DEFAULT_ALLOW) {
500
			dev_exception_rm(devcgroup, &ex);
501 502
			return 0;
		}
503
		return dev_exception_add(devcgroup, &ex);
504
	case DEVCG_DENY:
505 506 507 508 509
		/*
		 * If the default policy is to deny by default, try to remove
		 * an matching exception instead. And be silent about it: we
		 * don't want to break compatibility
		 */
510
		if (devcgroup->behavior == DEVCG_DEFAULT_DENY) {
511
			dev_exception_rm(devcgroup, &ex);
512 513
			return 0;
		}
514
		return dev_exception_add(devcgroup, &ex);
515
	default:
516
		return -EINVAL;
517
	}
518 519
	return 0;
}
520

521 522 523 524
static int devcgroup_access_write(struct cgroup *cgrp, struct cftype *cft,
				  const char *buffer)
{
	int retval;
525 526

	mutex_lock(&devcgroup_mutex);
527 528
	retval = devcgroup_update_access(cgroup_to_devcgroup(cgrp),
					 cft->private, buffer);
529
	mutex_unlock(&devcgroup_mutex);
530 531 532 533 534 535
	return retval;
}

static struct cftype dev_cgroup_files[] = {
	{
		.name = "allow",
536
		.write_string  = devcgroup_access_write,
537 538 539 540
		.private = DEVCG_ALLOW,
	},
	{
		.name = "deny",
541
		.write_string = devcgroup_access_write,
542 543
		.private = DEVCG_DENY,
	},
544 545 546 547 548
	{
		.name = "list",
		.read_seq_string = devcgroup_seq_read,
		.private = DEVCG_LIST,
	},
549
	{ }	/* terminate */
550 551 552 553 554 555
};

struct cgroup_subsys devices_subsys = {
	.name = "devices",
	.can_attach = devcgroup_can_attach,
	.create = devcgroup_create,
556
	.destroy = devcgroup_destroy,
557
	.subsys_id = devices_subsys_id,
558
	.base_cftypes = dev_cgroup_files,
559 560 561 562 563 564 565 566 567

	/*
	 * While devices cgroup has the rudimentary hierarchy support which
	 * checks the parent's restriction, it doesn't properly propagates
	 * config changes in ancestors to their descendents.  A child
	 * should only be allowed to add more restrictions to the parent's
	 * configuration.  Fix it and remove the following.
	 */
	.broken_hierarchy = true,
568 569
};

570 571 572 573 574 575 576 577 578 579
/**
 * __devcgroup_check_permission - checks if an inode operation is permitted
 * @dev_cgroup: the dev cgroup to be tested against
 * @type: device type
 * @major: device major number
 * @minor: device minor number
 * @access: combination of ACC_WRITE, ACC_READ and ACC_MKNOD
 *
 * returns 0 on success, -EPERM case the operation is not permitted
 */
580
static int __devcgroup_check_permission(short type, u32 major, u32 minor,
581
				        short access)
582
{
583
	struct dev_cgroup *dev_cgroup;
584
	struct dev_exception_item ex;
585
	int rc;
586

587 588 589 590 591
	memset(&ex, 0, sizeof(ex));
	ex.type = type;
	ex.major = major;
	ex.minor = minor;
	ex.access = access;
592

593
	rcu_read_lock();
594
	dev_cgroup = task_devcgroup(current);
595
	rc = may_access(dev_cgroup, &ex);
596
	rcu_read_unlock();
597

598 599
	if (!rc)
		return -EPERM;
600

601 602
	return 0;
}
603

604 605 606 607 608 609 610 611 612 613 614 615 616
int __devcgroup_inode_permission(struct inode *inode, int mask)
{
	short type, access = 0;

	if (S_ISBLK(inode->i_mode))
		type = DEV_BLOCK;
	if (S_ISCHR(inode->i_mode))
		type = DEV_CHAR;
	if (mask & MAY_WRITE)
		access |= ACC_WRITE;
	if (mask & MAY_READ)
		access |= ACC_READ;

617 618
	return __devcgroup_check_permission(type, imajor(inode), iminor(inode),
			access);
619 620 621 622
}

int devcgroup_inode_mknod(int mode, dev_t dev)
{
623
	short type;
624

625 626 627
	if (!S_ISBLK(mode) && !S_ISCHR(mode))
		return 0;

628 629 630 631
	if (S_ISBLK(mode))
		type = DEV_BLOCK;
	else
		type = DEV_CHAR;
632

633 634
	return __devcgroup_check_permission(type, MAJOR(dev), MINOR(dev),
			ACC_MKNOD);
635

636
}