security/device_cgroup.c

   1 /*
   2  * dev_cgroup.c - device cgroup subsystem
   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>
  12 #include <linux/seq_file.h>
  13
  14 #define ACC_MKNOD 1
  15 #define ACC_READ  2
  16 #define ACC_WRITE 4
  17 #define ACC_MASK (ACC_MKNOD | ACC_READ | ACC_WRITE)
  18
  19 #define DEV_BLOCK 1
  20 #define DEV_CHAR  2
  21 #define DEV_ALL   4  /* this represents all devices */
  22
  23 /*
  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().
  27  *
  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().
  32  *
  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.
  37  */
  38
  39 struct dev_whitelist_item {
  40         u32 major, minor;
  41         short type;
  42         short access;
  43         struct list_head list;
  44         struct rcu_head rcu;
  45 };
  46
  47 struct dev_cgroup {
  48         struct cgroup_subsys_state css;
  49         struct list_head whitelist;
  50         spinlock_t lock;
  51 };
  52
  53 static inline struct dev_cgroup *css_to_devcgroup(struct cgroup_subsys_state *s)
  54 {
  55         return container_of(s, struct dev_cgroup, css);
  56 }
  57
  58 static inline struct dev_cgroup *cgroup_to_devcgroup(struct cgroup *cgroup)
  59 {
  60         return css_to_devcgroup(cgroup_subsys_state(cgroup, devices_subsys_id));
  61 }
  62
  63 static inline struct dev_cgroup *task_devcgroup(struct task_struct *task)
  64 {
  65         return css_to_devcgroup(task_subsys_state(task, devices_subsys_id));
  66 }
  67
  68 struct cgroup_subsys devices_subsys;
  69
  70 static int devcgroup_can_attach(struct cgroup_subsys *ss,
  71                 struct cgroup *new_cgroup, struct task_struct *task)
  72 {
  73         if (current != task && !capable(CAP_SYS_ADMIN))
  74                         return -EPERM;
  75
  76         return 0;
  77 }
  78
  79 /*
  80  * called under cgroup_lock()
  81  */
  82 static int dev_whitelist_copy(struct list_head *dest, struct list_head *orig)
  83 {
  84         struct dev_whitelist_item *wh, *tmp, *new;
  85
  86         list_for_each_entry(wh, orig, list) {
  87                 new = kmalloc(sizeof(*wh), GFP_KERNEL);
  88                 if (!new)
  89                         goto free_and_exit;
  90                 new->major = wh->major;
  91                 new->minor = wh->minor;
  92                 new->type = wh->type;
  93                 new->access = wh->access;
  94                 list_add_tail(&new->list, dest);
  95         }
  96
  97         return 0;
  98
  99 free_and_exit:
 100         list_for_each_entry_safe(wh, tmp, dest, list) {
 101                 list_del(&wh->list);
 102                 kfree(wh);
 103         }
 104         return -ENOMEM;
 105 }
 106
 107 /* Stupid prototype - don't bother combining existing entries */
 108 /*
 109  * called under cgroup_lock()
 110  * since the list is visible to other tasks, we need the spinlock also
 111  */
 112 static int dev_whitelist_add(struct dev_cgroup *dev_cgroup,
 113                         struct dev_whitelist_item *wh)
 114 {
 115         struct dev_whitelist_item *whcopy, *walk;
 116
 117         whcopy = kmalloc(sizeof(*whcopy), GFP_KERNEL);
 118         if (!whcopy)
 119                 return -ENOMEM;
 120
 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)
 125                         continue;
 126                 if (walk->major != wh->major)
 127                         continue;
 128                 if (walk->minor != wh->minor)
 129                         continue;
 130
 131                 walk->access |= wh->access;
 132                 kfree(whcopy);
 133                 whcopy = NULL;
 134         }
 135
 136         if (whcopy != NULL)
 137                 list_add_tail_rcu(&whcopy->list, &dev_cgroup->whitelist);
 138         spin_unlock(&dev_cgroup->lock);
 139         return 0;
 140 }
 141
 142 static void whitelist_item_free(struct rcu_head *rcu)
 143 {
 144         struct dev_whitelist_item *item;
 145
 146         item = container_of(rcu, struct dev_whitelist_item, rcu);
 147         kfree(item);
 148 }
 149
 150 /*
 151  * called under cgroup_lock()
 152  * since the list is visible to other tasks, we need the spinlock also
 153  */
 154 static void dev_whitelist_rm(struct dev_cgroup *dev_cgroup,
 155                         struct dev_whitelist_item *wh)
 156 {
 157         struct dev_whitelist_item *walk, *tmp;
 158
 159         spin_lock(&dev_cgroup->lock);
 160         list_for_each_entry_safe(walk, tmp, &dev_cgroup->whitelist, list) {
 161                 if (walk->type == DEV_ALL)
 162                         goto remove;
 163                 if (walk->type != wh->type)
 164                         continue;
 165                 if (walk->major != ~0 && walk->major != wh->major)
 166                         continue;
 167                 if (walk->minor != ~0 && walk->minor != wh->minor)
 168                         continue;
 169
 170 remove:
 171                 walk->access &= ~wh->access;
 172                 if (!walk->access) {
 173                         list_del_rcu(&walk->list);
 174                         call_rcu(&walk->rcu, whitelist_item_free);
 175                 }
 176         }
 177         spin_unlock(&dev_cgroup->lock);
 178 }
 179
 180 /*
 181  * called from kernel/cgroup.c with cgroup_lock() held.
 182  */
 183 static struct cgroup_subsys_state *devcgroup_create(struct cgroup_subsys *ss,
 184                                                 struct cgroup *cgroup)
 185 {
 186         struct dev_cgroup *dev_cgroup, *parent_dev_cgroup;
 187         struct cgroup *parent_cgroup;
 188         int ret;
 189
 190         dev_cgroup = kzalloc(sizeof(*dev_cgroup), GFP_KERNEL);
 191         if (!dev_cgroup)
 192                 return ERR_PTR(-ENOMEM);
 193         INIT_LIST_HEAD(&dev_cgroup->whitelist);
 194         parent_cgroup = cgroup->parent;
 195
 196         if (parent_cgroup == NULL) {
 197                 struct dev_whitelist_item *wh;
 198                 wh = kmalloc(sizeof(*wh), GFP_KERNEL);
 199                 if (!wh) {
 200                         kfree(dev_cgroup);
 201                         return ERR_PTR(-ENOMEM);
 202                 }
 203                 wh->minor = wh->major = ~0;
 204                 wh->type = DEV_ALL;
 205                 wh->access = ACC_MKNOD | ACC_READ | ACC_WRITE;
 206                 list_add(&wh->list, &dev_cgroup->whitelist);
 207         } else {
 208                 parent_dev_cgroup = cgroup_to_devcgroup(parent_cgroup);
 209                 ret = dev_whitelist_copy(&dev_cgroup->whitelist,
 210                                 &parent_dev_cgroup->whitelist);
 211                 if (ret) {
 212                         kfree(dev_cgroup);
 213                         return ERR_PTR(ret);
 214                 }
 215         }
 216
 217         spin_lock_init(&dev_cgroup->lock);
 218         return &dev_cgroup->css;
 219 }
 220
 221 static void devcgroup_destroy(struct cgroup_subsys *ss,
 222                         struct cgroup *cgroup)
 223 {
 224         struct dev_cgroup *dev_cgroup;
 225         struct dev_whitelist_item *wh, *tmp;
 226
 227         dev_cgroup = cgroup_to_devcgroup(cgroup);
 228         list_for_each_entry_safe(wh, tmp, &dev_cgroup->whitelist, list) {
 229                 list_del(&wh->list);
 230                 kfree(wh);
 231         }
 232         kfree(dev_cgroup);
 233 }
 234
 235 #define DEVCG_ALLOW 1
 236 #define DEVCG_DENY 2
 237 #define DEVCG_LIST 3
 238
 239 #define MAJMINLEN 13
 240 #define ACCLEN 4
 241
 242 static void set_access(char *acc, short access)
 243 {
 244         int idx = 0;
 245         memset(acc, 0, ACCLEN);
 246         if (access & ACC_READ)
 247                 acc[idx++] = 'r';
 248         if (access & ACC_WRITE)
 249                 acc[idx++] = 'w';
 250         if (access & ACC_MKNOD)
 251                 acc[idx++] = 'm';
 252 }
 253
 254 static char type_to_char(short type)
 255 {
 256         if (type == DEV_ALL)
 257                 return 'a';
 258         if (type == DEV_CHAR)
 259                 return 'c';
 260         if (type == DEV_BLOCK)
 261                 return 'b';
 262         return 'X';
 263 }
 264
 265 static void set_majmin(char *str, unsigned m)
 266 {
 267         memset(str, 0, MAJMINLEN);
 268         if (m == ~0)
 269                 sprintf(str, "*");
 270         else
 271                 snprintf(str, MAJMINLEN, "%u", m);
 272 }
 273
 274 static int devcgroup_seq_read(struct cgroup *cgroup, struct cftype *cft,
 275                                 struct seq_file *m)
 276 {
 277         struct dev_cgroup *devcgroup = cgroup_to_devcgroup(cgroup);
 278         struct dev_whitelist_item *wh;
 279         char maj[MAJMINLEN], min[MAJMINLEN], acc[ACCLEN];
 280
 281         rcu_read_lock();
 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),
 287                            maj, min, acc);
 288         }
 289         rcu_read_unlock();
 290
 291         return 0;
 292 }
 293
 294 /*
 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
 300  */
 301 static int may_access_whitelist(struct dev_cgroup *c,
 302                                        struct dev_whitelist_item *refwh)
 303 {
 304         struct dev_whitelist_item *whitem;
 305
 306         list_for_each_entry(whitem, &c->whitelist, list) {
 307                 if (whitem->type & DEV_ALL)
 308                         return 1;
 309                 if ((refwh->type & DEV_BLOCK) && !(whitem->type & DEV_BLOCK))
 310                         continue;
 311                 if ((refwh->type & DEV_CHAR) && !(whitem->type & DEV_CHAR))
 312                         continue;
 313                 if (whitem->major != ~0 && whitem->major != refwh->major)
 314                         continue;
 315                 if (whitem->minor != ~0 && whitem->minor != refwh->minor)
 316                         continue;
 317                 if (refwh->access & (~whitem->access))
 318                         continue;
 319                 return 1;
 320         }
 321         return 0;
 322 }
 323
 324 /*
 325  * parent_has_perm:
 326  * when adding a new allow rule to a device whitelist, the rule
 327  * must be allowed in the parent device
 328  */
 329 static int parent_has_perm(struct dev_cgroup *childcg,
 330                                   struct dev_whitelist_item *wh)
 331 {
 332         struct cgroup *pcg = childcg->css.cgroup->parent;
 333         struct dev_cgroup *parent;
 334         int ret;
 335
 336         if (!pcg)
 337                 return 1;
 338         parent = cgroup_to_devcgroup(pcg);
 339         spin_lock(&parent->lock);
 340         ret = may_access_whitelist(parent, wh);
 341         spin_unlock(&parent->lock);
 342         return ret;
 343 }
 344
 345 /*
 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
 353  *
 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.
 357  */
 358 static int devcgroup_update_access(struct dev_cgroup *devcgroup,
 359                                    int filetype, const char *buffer)
 360 {
 361         struct dev_cgroup *cur_devcgroup;
 362         const char *b;
 363         int retval = 0, count;
 364         struct dev_whitelist_item wh;
 365
 366         if (!capable(CAP_SYS_ADMIN))
 367                 return -EPERM;
 368
 369         cur_devcgroup = task_devcgroup(current);
 370
 371         memset(&wh, 0, sizeof(wh));
 372         b = buffer;
 373
 374         switch (*b) {
 375         case 'a':
 376                 wh.type = DEV_ALL;
 377                 wh.access = ACC_MASK;
 378                 wh.major = ~0;
 379                 wh.minor = ~0;
 380                 goto handle;
 381         case 'b':
 382                 wh.type = DEV_BLOCK;
 383                 break;
 384         case 'c':
 385                 wh.type = DEV_CHAR;
 386                 break;
 387         default:
 388                 return -EINVAL;
 389         }
 390         b++;
 391         if (!isspace(*b))
 392                 return -EINVAL;
 393         b++;
 394         if (*b == '*') {
 395                 wh.major = ~0;
 396                 b++;
 397         } else if (isdigit(*b)) {
 398                 wh.major = 0;
 399                 while (isdigit(*b)) {
 400                         wh.major = wh.major*10+(*b-'0');
 401                         b++;
 402                 }
 403         } else {
 404                 return -EINVAL;
 405         }
 406         if (*b != ':')
 407                 return -EINVAL;
 408         b++;
 409
 410         /* read minor */
 411         if (*b == '*') {
 412                 wh.minor = ~0;
 413                 b++;
 414         } else if (isdigit(*b)) {
 415                 wh.minor = 0;
 416                 while (isdigit(*b)) {
 417                         wh.minor = wh.minor*10+(*b-'0');
 418                         b++;
 419                 }
 420         } else {
 421                 return -EINVAL;
 422         }
 423         if (!isspace(*b))
 424                 return -EINVAL;
 425         for (b++, count = 0; count < 3; count++, b++) {
 426                 switch (*b) {
 427                 case 'r':
 428                         wh.access |= ACC_READ;
 429                         break;
 430                 case 'w':
 431                         wh.access |= ACC_WRITE;
 432                         break;
 433                 case 'm':
 434                         wh.access |= ACC_MKNOD;
 435                         break;
 436                 case '\n':
 437                 case '\0':
 438                         count = 3;
 439                         break;
 440                 default:
 441                         return -EINVAL;
 442                 }
 443         }
 444
 445 handle:
 446         retval = 0;
 447         switch (filetype) {
 448         case DEVCG_ALLOW:
 449                 if (!parent_has_perm(devcgroup, &wh))
 450                         return -EPERM;
 451                 return dev_whitelist_add(devcgroup, &wh);
 452         case DEVCG_DENY:
 453                 dev_whitelist_rm(devcgroup, &wh);
 454                 break;
 455         default:
 456                 return -EINVAL;
 457         }
 458         return 0;
 459 }
 460
 461 static int devcgroup_access_write(struct cgroup *cgrp, struct cftype *cft,
 462                                   const char *buffer)
 463 {
 464         int retval;
 465         if (!cgroup_lock_live_group(cgrp))
 466                 return -ENODEV;
 467         retval = devcgroup_update_access(cgroup_to_devcgroup(cgrp),
 468                                          cft->private, buffer);
 469         cgroup_unlock();
 470         return retval;
 471 }
 472
 473 static struct cftype dev_cgroup_files[] = {
 474         {
 475                 .name = "allow",
 476                 .write_string  = devcgroup_access_write,
 477                 .private = DEVCG_ALLOW,
 478         },
 479         {
 480                 .name = "deny",
 481                 .write_string = devcgroup_access_write,
 482                 .private = DEVCG_DENY,
 483         },
 484         {
 485                 .name = "list",
 486                 .read_seq_string = devcgroup_seq_read,
 487                 .private = DEVCG_LIST,
 488         },
 489 };
 490
 491 static int devcgroup_populate(struct cgroup_subsys *ss,
 492                                 struct cgroup *cgroup)
 493 {
 494         return cgroup_add_files(cgroup, ss, dev_cgroup_files,
 495                                         ARRAY_SIZE(dev_cgroup_files));
 496 }
 497
 498 struct cgroup_subsys devices_subsys = {
 499         .name = "devices",
 500         .can_attach = devcgroup_can_attach,
 501         .create = devcgroup_create,
 502         .destroy  = devcgroup_destroy,
 503         .populate = devcgroup_populate,
 504         .subsys_id = devices_subsys_id,
 505 };
 506
 507 int devcgroup_inode_permission(struct inode *inode, int mask)
 508 {
 509         struct dev_cgroup *dev_cgroup;
 510         struct dev_whitelist_item *wh;
 511
 512         dev_t device = inode->i_rdev;
 513         if (!device)
 514                 return 0;
 515         if (!S_ISBLK(inode->i_mode) && !S_ISCHR(inode->i_mode))
 516                 return 0;
 517         dev_cgroup = css_to_devcgroup(task_subsys_state(current,
 518                                 devices_subsys_id));
 519         if (!dev_cgroup)
 520                 return 0;
 521
 522         rcu_read_lock();
 523         list_for_each_entry_rcu(wh, &dev_cgroup->whitelist, list) {
 524                 if (wh->type & DEV_ALL)
 525                         goto acc_check;
 526                 if ((wh->type & DEV_BLOCK) && !S_ISBLK(inode->i_mode))
 527                         continue;
 528                 if ((wh->type & DEV_CHAR) && !S_ISCHR(inode->i_mode))
 529                         continue;
 530                 if (wh->major != ~0 && wh->major != imajor(inode))
 531                         continue;
 532                 if (wh->minor != ~0 && wh->minor != iminor(inode))
 533                         continue;
 534 acc_check:
 535                 if ((mask & MAY_WRITE) && !(wh->access & ACC_WRITE))
 536                         continue;
 537                 if ((mask & MAY_READ) && !(wh->access & ACC_READ))
 538                         continue;
 539                 rcu_read_unlock();
 540                 return 0;
 541         }
 542         rcu_read_unlock();
 543
 544         return -EPERM;
 545 }
 546
 547 int devcgroup_inode_mknod(int mode, dev_t dev)
 548 {
 549         struct dev_cgroup *dev_cgroup;
 550         struct dev_whitelist_item *wh;
 551
 552         dev_cgroup = css_to_devcgroup(task_subsys_state(current,
 553                                 devices_subsys_id));
 554         if (!dev_cgroup)
 555                 return 0;
 556
 557         rcu_read_lock();
 558         list_for_each_entry(wh, &dev_cgroup->whitelist, list) {
 559                 if (wh->type & DEV_ALL)
 560                         goto acc_check;
 561                 if ((wh->type & DEV_BLOCK) && !S_ISBLK(mode))
 562                         continue;
 563                 if ((wh->type & DEV_CHAR) && !S_ISCHR(mode))
 564                         continue;
 565                 if (wh->major != ~0 && wh->major != MAJOR(dev))
 566                         continue;
 567                 if (wh->minor != ~0 && wh->minor != MINOR(dev))
 568                         continue;
 569 acc_check:
 570                 if (!(wh->access & ACC_MKNOD))
 571                         continue;
 572                 rcu_read_unlock();
 573                 return 0;
 574         }
 575         rcu_read_unlock();
 576         return -EPERM;
 577 }