1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* Common capabilities, needed by capability.o. 3 */ 4 5 #include <linux/capability.h> 6 #include <linux/audit.h> 7 #include <linux/init.h> 8 #include <linux/kernel.h> 9 #include <linux/lsm_hooks.h> 10 #include <linux/file.h> 11 #include <linux/mm.h> 12 #include <linux/mman.h> 13 #include <linux/pagemap.h> 14 #include <linux/swap.h> 15 #include <linux/skbuff.h> 16 #include <linux/netlink.h> 17 #include <linux/ptrace.h> 18 #include <linux/xattr.h> 19 #include <linux/hugetlb.h> 20 #include <linux/mount.h> 21 #include <linux/sched.h> 22 #include <linux/prctl.h> 23 #include <linux/securebits.h> 24 #include <linux/user_namespace.h> 25 #include <linux/binfmts.h> 26 #include <linux/personality.h> 27 28 /* 29 * If a non-root user executes a setuid-root binary in 30 * !secure(SECURE_NOROOT) mode, then we raise capabilities. 31 * However if fE is also set, then the intent is for only 32 * the file capabilities to be applied, and the setuid-root 33 * bit is left on either to change the uid (plausible) or 34 * to get full privilege on a kernel without file capabilities 35 * support. So in that case we do not raise capabilities. 36 * 37 * Warn if that happens, once per boot. 38 */ 39 static void warn_setuid_and_fcaps_mixed(const char *fname) 40 { 41 static int warned; 42 if (!warned) { 43 printk(KERN_INFO "warning: `%s' has both setuid-root and" 44 " effective capabilities. Therefore not raising all" 45 " capabilities.\n", fname); 46 warned = 1; 47 } 48 } 49 50 /** 51 * cap_capable - Determine whether a task has a particular effective capability 52 * @cred: The credentials to use 53 * @ns: The user namespace in which we need the capability 54 * @cap: The capability to check for 55 * @opts: Bitmask of options defined in include/linux/security.h 56 * 57 * Determine whether the nominated task has the specified capability amongst 58 * its effective set, returning 0 if it does, -ve if it does not. 59 * 60 * NOTE WELL: cap_has_capability() cannot be used like the kernel's capable() 61 * and has_capability() functions. That is, it has the reverse semantics: 62 * cap_has_capability() returns 0 when a task has a capability, but the 63 * kernel's capable() and has_capability() returns 1 for this case. 64 */ 65 int cap_capable(const struct cred *cred, struct user_namespace *targ_ns, 66 int cap, unsigned int opts) 67 { 68 struct user_namespace *ns = targ_ns; 69 70 /* See if cred has the capability in the target user namespace 71 * by examining the target user namespace and all of the target 72 * user namespace's parents. 73 */ 74 for (;;) { 75 /* Do we have the necessary capabilities? */ 76 if (ns == cred->user_ns) 77 return cap_raised(cred->cap_effective, cap) ? 0 : -EPERM; 78 79 /* 80 * If we're already at a lower level than we're looking for, 81 * we're done searching. 82 */ 83 if (ns->level <= cred->user_ns->level) 84 return -EPERM; 85 86 /* 87 * The owner of the user namespace in the parent of the 88 * user namespace has all caps. 89 */ 90 if ((ns->parent == cred->user_ns) && uid_eq(ns->owner, cred->euid)) 91 return 0; 92 93 /* 94 * If you have a capability in a parent user ns, then you have 95 * it over all children user namespaces as well. 96 */ 97 ns = ns->parent; 98 } 99 100 /* We never get here */ 101 } 102 103 /** 104 * cap_settime - Determine whether the current process may set the system clock 105 * @ts: The time to set 106 * @tz: The timezone to set 107 * 108 * Determine whether the current process may set the system clock and timezone 109 * information, returning 0 if permission granted, -ve if denied. 110 */ 111 int cap_settime(const struct timespec64 *ts, const struct timezone *tz) 112 { 113 if (!capable(CAP_SYS_TIME)) 114 return -EPERM; 115 return 0; 116 } 117 118 /** 119 * cap_ptrace_access_check - Determine whether the current process may access 120 * another 121 * @child: The process to be accessed 122 * @mode: The mode of attachment. 123 * 124 * If we are in the same or an ancestor user_ns and have all the target 125 * task's capabilities, then ptrace access is allowed. 126 * If we have the ptrace capability to the target user_ns, then ptrace 127 * access is allowed. 128 * Else denied. 129 * 130 * Determine whether a process may access another, returning 0 if permission 131 * granted, -ve if denied. 132 */ 133 int cap_ptrace_access_check(struct task_struct *child, unsigned int mode) 134 { 135 int ret = 0; 136 const struct cred *cred, *child_cred; 137 const kernel_cap_t *caller_caps; 138 139 rcu_read_lock(); 140 cred = current_cred(); 141 child_cred = __task_cred(child); 142 if (mode & PTRACE_MODE_FSCREDS) 143 caller_caps = &cred->cap_effective; 144 else 145 caller_caps = &cred->cap_permitted; 146 if (cred->user_ns == child_cred->user_ns && 147 cap_issubset(child_cred->cap_permitted, *caller_caps)) 148 goto out; 149 if (ns_capable(child_cred->user_ns, CAP_SYS_PTRACE)) 150 goto out; 151 ret = -EPERM; 152 out: 153 rcu_read_unlock(); 154 return ret; 155 } 156 157 /** 158 * cap_ptrace_traceme - Determine whether another process may trace the current 159 * @parent: The task proposed to be the tracer 160 * 161 * If parent is in the same or an ancestor user_ns and has all current's 162 * capabilities, then ptrace access is allowed. 163 * If parent has the ptrace capability to current's user_ns, then ptrace 164 * access is allowed. 165 * Else denied. 166 * 167 * Determine whether the nominated task is permitted to trace the current 168 * process, returning 0 if permission is granted, -ve if denied. 169 */ 170 int cap_ptrace_traceme(struct task_struct *parent) 171 { 172 int ret = 0; 173 const struct cred *cred, *child_cred; 174 175 rcu_read_lock(); 176 cred = __task_cred(parent); 177 child_cred = current_cred(); 178 if (cred->user_ns == child_cred->user_ns && 179 cap_issubset(child_cred->cap_permitted, cred->cap_permitted)) 180 goto out; 181 if (has_ns_capability(parent, child_cred->user_ns, CAP_SYS_PTRACE)) 182 goto out; 183 ret = -EPERM; 184 out: 185 rcu_read_unlock(); 186 return ret; 187 } 188 189 /** 190 * cap_capget - Retrieve a task's capability sets 191 * @target: The task from which to retrieve the capability sets 192 * @effective: The place to record the effective set 193 * @inheritable: The place to record the inheritable set 194 * @permitted: The place to record the permitted set 195 * 196 * This function retrieves the capabilities of the nominated task and returns 197 * them to the caller. 198 */ 199 int cap_capget(struct task_struct *target, kernel_cap_t *effective, 200 kernel_cap_t *inheritable, kernel_cap_t *permitted) 201 { 202 const struct cred *cred; 203 204 /* Derived from kernel/capability.c:sys_capget. */ 205 rcu_read_lock(); 206 cred = __task_cred(target); 207 *effective = cred->cap_effective; 208 *inheritable = cred->cap_inheritable; 209 *permitted = cred->cap_permitted; 210 rcu_read_unlock(); 211 return 0; 212 } 213 214 /* 215 * Determine whether the inheritable capabilities are limited to the old 216 * permitted set. Returns 1 if they are limited, 0 if they are not. 217 */ 218 static inline int cap_inh_is_capped(void) 219 { 220 /* they are so limited unless the current task has the CAP_SETPCAP 221 * capability 222 */ 223 if (cap_capable(current_cred(), current_cred()->user_ns, 224 CAP_SETPCAP, CAP_OPT_NONE) == 0) 225 return 0; 226 return 1; 227 } 228 229 /** 230 * cap_capset - Validate and apply proposed changes to current's capabilities 231 * @new: The proposed new credentials; alterations should be made here 232 * @old: The current task's current credentials 233 * @effective: A pointer to the proposed new effective capabilities set 234 * @inheritable: A pointer to the proposed new inheritable capabilities set 235 * @permitted: A pointer to the proposed new permitted capabilities set 236 * 237 * This function validates and applies a proposed mass change to the current 238 * process's capability sets. The changes are made to the proposed new 239 * credentials, and assuming no error, will be committed by the caller of LSM. 240 */ 241 int cap_capset(struct cred *new, 242 const struct cred *old, 243 const kernel_cap_t *effective, 244 const kernel_cap_t *inheritable, 245 const kernel_cap_t *permitted) 246 { 247 if (cap_inh_is_capped() && 248 !cap_issubset(*inheritable, 249 cap_combine(old->cap_inheritable, 250 old->cap_permitted))) 251 /* incapable of using this inheritable set */ 252 return -EPERM; 253 254 if (!cap_issubset(*inheritable, 255 cap_combine(old->cap_inheritable, 256 old->cap_bset))) 257 /* no new pI capabilities outside bounding set */ 258 return -EPERM; 259 260 /* verify restrictions on target's new Permitted set */ 261 if (!cap_issubset(*permitted, old->cap_permitted)) 262 return -EPERM; 263 264 /* verify the _new_Effective_ is a subset of the _new_Permitted_ */ 265 if (!cap_issubset(*effective, *permitted)) 266 return -EPERM; 267 268 new->cap_effective = *effective; 269 new->cap_inheritable = *inheritable; 270 new->cap_permitted = *permitted; 271 272 /* 273 * Mask off ambient bits that are no longer both permitted and 274 * inheritable. 275 */ 276 new->cap_ambient = cap_intersect(new->cap_ambient, 277 cap_intersect(*permitted, 278 *inheritable)); 279 if (WARN_ON(!cap_ambient_invariant_ok(new))) 280 return -EINVAL; 281 return 0; 282 } 283 284 /** 285 * cap_inode_need_killpriv - Determine if inode change affects privileges 286 * @dentry: The inode/dentry in being changed with change marked ATTR_KILL_PRIV 287 * 288 * Determine if an inode having a change applied that's marked ATTR_KILL_PRIV 289 * affects the security markings on that inode, and if it is, should 290 * inode_killpriv() be invoked or the change rejected. 291 * 292 * Returns 1 if security.capability has a value, meaning inode_killpriv() 293 * is required, 0 otherwise, meaning inode_killpriv() is not required. 294 */ 295 int cap_inode_need_killpriv(struct dentry *dentry) 296 { 297 struct inode *inode = d_backing_inode(dentry); 298 int error; 299 300 error = __vfs_getxattr(dentry, inode, XATTR_NAME_CAPS, NULL, 0); 301 return error > 0; 302 } 303 304 /** 305 * cap_inode_killpriv - Erase the security markings on an inode 306 * @dentry: The inode/dentry to alter 307 * 308 * Erase the privilege-enhancing security markings on an inode. 309 * 310 * Returns 0 if successful, -ve on error. 311 */ 312 int cap_inode_killpriv(struct dentry *dentry) 313 { 314 int error; 315 316 error = __vfs_removexattr(dentry, XATTR_NAME_CAPS); 317 if (error == -EOPNOTSUPP) 318 error = 0; 319 return error; 320 } 321 322 static bool rootid_owns_currentns(kuid_t kroot) 323 { 324 struct user_namespace *ns; 325 326 if (!uid_valid(kroot)) 327 return false; 328 329 for (ns = current_user_ns(); ; ns = ns->parent) { 330 if (from_kuid(ns, kroot) == 0) 331 return true; 332 if (ns == &init_user_ns) 333 break; 334 } 335 336 return false; 337 } 338 339 static __u32 sansflags(__u32 m) 340 { 341 return m & ~VFS_CAP_FLAGS_EFFECTIVE; 342 } 343 344 static bool is_v2header(size_t size, const struct vfs_cap_data *cap) 345 { 346 if (size != XATTR_CAPS_SZ_2) 347 return false; 348 return sansflags(le32_to_cpu(cap->magic_etc)) == VFS_CAP_REVISION_2; 349 } 350 351 static bool is_v3header(size_t size, const struct vfs_cap_data *cap) 352 { 353 if (size != XATTR_CAPS_SZ_3) 354 return false; 355 return sansflags(le32_to_cpu(cap->magic_etc)) == VFS_CAP_REVISION_3; 356 } 357 358 /* 359 * getsecurity: We are called for security.* before any attempt to read the 360 * xattr from the inode itself. 361 * 362 * This gives us a chance to read the on-disk value and convert it. If we 363 * return -EOPNOTSUPP, then vfs_getxattr() will call the i_op handler. 364 * 365 * Note we are not called by vfs_getxattr_alloc(), but that is only called 366 * by the integrity subsystem, which really wants the unconverted values - 367 * so that's good. 368 */ 369 int cap_inode_getsecurity(struct inode *inode, const char *name, void **buffer, 370 bool alloc) 371 { 372 int size, ret; 373 kuid_t kroot; 374 u32 nsmagic, magic; 375 uid_t root, mappedroot; 376 char *tmpbuf = NULL; 377 struct vfs_cap_data *cap; 378 struct vfs_ns_cap_data *nscap = NULL; 379 struct dentry *dentry; 380 struct user_namespace *fs_ns; 381 382 if (strcmp(name, "capability") != 0) 383 return -EOPNOTSUPP; 384 385 dentry = d_find_any_alias(inode); 386 if (!dentry) 387 return -EINVAL; 388 389 size = sizeof(struct vfs_ns_cap_data); 390 ret = (int) vfs_getxattr_alloc(dentry, XATTR_NAME_CAPS, 391 &tmpbuf, size, GFP_NOFS); 392 dput(dentry); 393 394 if (ret < 0) 395 return ret; 396 397 fs_ns = inode->i_sb->s_user_ns; 398 cap = (struct vfs_cap_data *) tmpbuf; 399 if (is_v2header((size_t) ret, cap)) { 400 root = 0; 401 } else if (is_v3header((size_t) ret, cap)) { 402 nscap = (struct vfs_ns_cap_data *) tmpbuf; 403 root = le32_to_cpu(nscap->rootid); 404 } else { 405 size = -EINVAL; 406 goto out_free; 407 } 408 409 kroot = make_kuid(fs_ns, root); 410 411 /* If the root kuid maps to a valid uid in current ns, then return 412 * this as a nscap. */ 413 mappedroot = from_kuid(current_user_ns(), kroot); 414 if (mappedroot != (uid_t)-1 && mappedroot != (uid_t)0) { 415 size = sizeof(struct vfs_ns_cap_data); 416 if (alloc) { 417 if (!nscap) { 418 /* v2 -> v3 conversion */ 419 nscap = kzalloc(size, GFP_ATOMIC); 420 if (!nscap) { 421 size = -ENOMEM; 422 goto out_free; 423 } 424 nsmagic = VFS_CAP_REVISION_3; 425 magic = le32_to_cpu(cap->magic_etc); 426 if (magic & VFS_CAP_FLAGS_EFFECTIVE) 427 nsmagic |= VFS_CAP_FLAGS_EFFECTIVE; 428 memcpy(&nscap->data, &cap->data, sizeof(__le32) * 2 * VFS_CAP_U32); 429 nscap->magic_etc = cpu_to_le32(nsmagic); 430 } else { 431 /* use allocated v3 buffer */ 432 tmpbuf = NULL; 433 } 434 nscap->rootid = cpu_to_le32(mappedroot); 435 *buffer = nscap; 436 } 437 goto out_free; 438 } 439 440 if (!rootid_owns_currentns(kroot)) { 441 size = -EOVERFLOW; 442 goto out_free; 443 } 444 445 /* This comes from a parent namespace. Return as a v2 capability */ 446 size = sizeof(struct vfs_cap_data); 447 if (alloc) { 448 if (nscap) { 449 /* v3 -> v2 conversion */ 450 cap = kzalloc(size, GFP_ATOMIC); 451 if (!cap) { 452 size = -ENOMEM; 453 goto out_free; 454 } 455 magic = VFS_CAP_REVISION_2; 456 nsmagic = le32_to_cpu(nscap->magic_etc); 457 if (nsmagic & VFS_CAP_FLAGS_EFFECTIVE) 458 magic |= VFS_CAP_FLAGS_EFFECTIVE; 459 memcpy(&cap->data, &nscap->data, sizeof(__le32) * 2 * VFS_CAP_U32); 460 cap->magic_etc = cpu_to_le32(magic); 461 } else { 462 /* use unconverted v2 */ 463 tmpbuf = NULL; 464 } 465 *buffer = cap; 466 } 467 out_free: 468 kfree(tmpbuf); 469 return size; 470 } 471 472 static kuid_t rootid_from_xattr(const void *value, size_t size, 473 struct user_namespace *task_ns) 474 { 475 const struct vfs_ns_cap_data *nscap = value; 476 uid_t rootid = 0; 477 478 if (size == XATTR_CAPS_SZ_3) 479 rootid = le32_to_cpu(nscap->rootid); 480 481 return make_kuid(task_ns, rootid); 482 } 483 484 static bool validheader(size_t size, const struct vfs_cap_data *cap) 485 { 486 return is_v2header(size, cap) || is_v3header(size, cap); 487 } 488 489 /* 490 * User requested a write of security.capability. If needed, update the 491 * xattr to change from v2 to v3, or to fixup the v3 rootid. 492 * 493 * If all is ok, we return the new size, on error return < 0. 494 */ 495 int cap_convert_nscap(struct dentry *dentry, const void **ivalue, size_t size) 496 { 497 struct vfs_ns_cap_data *nscap; 498 uid_t nsrootid; 499 const struct vfs_cap_data *cap = *ivalue; 500 __u32 magic, nsmagic; 501 struct inode *inode = d_backing_inode(dentry); 502 struct user_namespace *task_ns = current_user_ns(), 503 *fs_ns = inode->i_sb->s_user_ns, 504 *ancestor; 505 kuid_t rootid; 506 size_t newsize; 507 508 if (!*ivalue) 509 return -EINVAL; 510 if (!validheader(size, cap)) 511 return -EINVAL; 512 if (!capable_wrt_inode_uidgid(inode, CAP_SETFCAP)) 513 return -EPERM; 514 if (size == XATTR_CAPS_SZ_2) 515 if (ns_capable(inode->i_sb->s_user_ns, CAP_SETFCAP)) 516 /* user is privileged, just write the v2 */ 517 return size; 518 519 rootid = rootid_from_xattr(*ivalue, size, task_ns); 520 if (!uid_valid(rootid)) 521 return -EINVAL; 522 523 nsrootid = from_kuid(fs_ns, rootid); 524 if (nsrootid == -1) 525 return -EINVAL; 526 527 /* 528 * Do not allow allow adding a v3 filesystem capability xattr 529 * if the rootid field is ambiguous. 530 */ 531 for (ancestor = task_ns->parent; ancestor; ancestor = ancestor->parent) { 532 if (from_kuid(ancestor, rootid) == 0) 533 return -EINVAL; 534 } 535 536 newsize = sizeof(struct vfs_ns_cap_data); 537 nscap = kmalloc(newsize, GFP_ATOMIC); 538 if (!nscap) 539 return -ENOMEM; 540 nscap->rootid = cpu_to_le32(nsrootid); 541 nsmagic = VFS_CAP_REVISION_3; 542 magic = le32_to_cpu(cap->magic_etc); 543 if (magic & VFS_CAP_FLAGS_EFFECTIVE) 544 nsmagic |= VFS_CAP_FLAGS_EFFECTIVE; 545 nscap->magic_etc = cpu_to_le32(nsmagic); 546 memcpy(&nscap->data, &cap->data, sizeof(__le32) * 2 * VFS_CAP_U32); 547 548 *ivalue = nscap; 549 return newsize; 550 } 551 552 /* 553 * Calculate the new process capability sets from the capability sets attached 554 * to a file. 555 */ 556 static inline int bprm_caps_from_vfs_caps(struct cpu_vfs_cap_data *caps, 557 struct linux_binprm *bprm, 558 bool *effective, 559 bool *has_fcap) 560 { 561 struct cred *new = bprm->cred; 562 unsigned i; 563 int ret = 0; 564 565 if (caps->magic_etc & VFS_CAP_FLAGS_EFFECTIVE) 566 *effective = true; 567 568 if (caps->magic_etc & VFS_CAP_REVISION_MASK) 569 *has_fcap = true; 570 571 CAP_FOR_EACH_U32(i) { 572 __u32 permitted = caps->permitted.cap[i]; 573 __u32 inheritable = caps->inheritable.cap[i]; 574 575 /* 576 * pP' = (X & fP) | (pI & fI) 577 * The addition of pA' is handled later. 578 */ 579 new->cap_permitted.cap[i] = 580 (new->cap_bset.cap[i] & permitted) | 581 (new->cap_inheritable.cap[i] & inheritable); 582 583 if (permitted & ~new->cap_permitted.cap[i]) 584 /* insufficient to execute correctly */ 585 ret = -EPERM; 586 } 587 588 /* 589 * For legacy apps, with no internal support for recognizing they 590 * do not have enough capabilities, we return an error if they are 591 * missing some "forced" (aka file-permitted) capabilities. 592 */ 593 return *effective ? ret : 0; 594 } 595 596 /* 597 * Extract the on-exec-apply capability sets for an executable file. 598 */ 599 int get_vfs_caps_from_disk(const struct dentry *dentry, struct cpu_vfs_cap_data *cpu_caps) 600 { 601 struct inode *inode = d_backing_inode(dentry); 602 __u32 magic_etc; 603 unsigned tocopy, i; 604 int size; 605 struct vfs_ns_cap_data data, *nscaps = &data; 606 struct vfs_cap_data *caps = (struct vfs_cap_data *) &data; 607 kuid_t rootkuid; 608 struct user_namespace *fs_ns; 609 610 memset(cpu_caps, 0, sizeof(struct cpu_vfs_cap_data)); 611 612 if (!inode) 613 return -ENODATA; 614 615 fs_ns = inode->i_sb->s_user_ns; 616 size = __vfs_getxattr((struct dentry *)dentry, inode, 617 XATTR_NAME_CAPS, &data, XATTR_CAPS_SZ); 618 if (size == -ENODATA || size == -EOPNOTSUPP) 619 /* no data, that's ok */ 620 return -ENODATA; 621 622 if (size < 0) 623 return size; 624 625 if (size < sizeof(magic_etc)) 626 return -EINVAL; 627 628 cpu_caps->magic_etc = magic_etc = le32_to_cpu(caps->magic_etc); 629 630 rootkuid = make_kuid(fs_ns, 0); 631 switch (magic_etc & VFS_CAP_REVISION_MASK) { 632 case VFS_CAP_REVISION_1: 633 if (size != XATTR_CAPS_SZ_1) 634 return -EINVAL; 635 tocopy = VFS_CAP_U32_1; 636 break; 637 case VFS_CAP_REVISION_2: 638 if (size != XATTR_CAPS_SZ_2) 639 return -EINVAL; 640 tocopy = VFS_CAP_U32_2; 641 break; 642 case VFS_CAP_REVISION_3: 643 if (size != XATTR_CAPS_SZ_3) 644 return -EINVAL; 645 tocopy = VFS_CAP_U32_3; 646 rootkuid = make_kuid(fs_ns, le32_to_cpu(nscaps->rootid)); 647 break; 648 649 default: 650 return -EINVAL; 651 } 652 /* Limit the caps to the mounter of the filesystem 653 * or the more limited uid specified in the xattr. 654 */ 655 if (!rootid_owns_currentns(rootkuid)) 656 return -ENODATA; 657 658 CAP_FOR_EACH_U32(i) { 659 if (i >= tocopy) 660 break; 661 cpu_caps->permitted.cap[i] = le32_to_cpu(caps->data[i].permitted); 662 cpu_caps->inheritable.cap[i] = le32_to_cpu(caps->data[i].inheritable); 663 } 664 665 cpu_caps->permitted.cap[CAP_LAST_U32] &= CAP_LAST_U32_VALID_MASK; 666 cpu_caps->inheritable.cap[CAP_LAST_U32] &= CAP_LAST_U32_VALID_MASK; 667 668 cpu_caps->rootid = rootkuid; 669 670 return 0; 671 } 672 673 /* 674 * Attempt to get the on-exec apply capability sets for an executable file from 675 * its xattrs and, if present, apply them to the proposed credentials being 676 * constructed by execve(). 677 */ 678 static int get_file_caps(struct linux_binprm *bprm, struct file *file, 679 bool *effective, bool *has_fcap) 680 { 681 int rc = 0; 682 struct cpu_vfs_cap_data vcaps; 683 684 cap_clear(bprm->cred->cap_permitted); 685 686 if (!file_caps_enabled) 687 return 0; 688 689 if (!mnt_may_suid(file->f_path.mnt)) 690 return 0; 691 692 /* 693 * This check is redundant with mnt_may_suid() but is kept to make 694 * explicit that capability bits are limited to s_user_ns and its 695 * descendants. 696 */ 697 if (!current_in_userns(file->f_path.mnt->mnt_sb->s_user_ns)) 698 return 0; 699 700 rc = get_vfs_caps_from_disk(file->f_path.dentry, &vcaps); 701 if (rc < 0) { 702 if (rc == -EINVAL) 703 printk(KERN_NOTICE "Invalid argument reading file caps for %s\n", 704 bprm->filename); 705 else if (rc == -ENODATA) 706 rc = 0; 707 goto out; 708 } 709 710 rc = bprm_caps_from_vfs_caps(&vcaps, bprm, effective, has_fcap); 711 712 out: 713 if (rc) 714 cap_clear(bprm->cred->cap_permitted); 715 716 return rc; 717 } 718 719 static inline bool root_privileged(void) { return !issecure(SECURE_NOROOT); } 720 721 static inline bool __is_real(kuid_t uid, struct cred *cred) 722 { return uid_eq(cred->uid, uid); } 723 724 static inline bool __is_eff(kuid_t uid, struct cred *cred) 725 { return uid_eq(cred->euid, uid); } 726 727 static inline bool __is_suid(kuid_t uid, struct cred *cred) 728 { return !__is_real(uid, cred) && __is_eff(uid, cred); } 729 730 /* 731 * handle_privileged_root - Handle case of privileged root 732 * @bprm: The execution parameters, including the proposed creds 733 * @has_fcap: Are any file capabilities set? 734 * @effective: Do we have effective root privilege? 735 * @root_uid: This namespace' root UID WRT initial USER namespace 736 * 737 * Handle the case where root is privileged and hasn't been neutered by 738 * SECURE_NOROOT. If file capabilities are set, they won't be combined with 739 * set UID root and nothing is changed. If we are root, cap_permitted is 740 * updated. If we have become set UID root, the effective bit is set. 741 */ 742 static void handle_privileged_root(struct linux_binprm *bprm, bool has_fcap, 743 bool *effective, kuid_t root_uid) 744 { 745 const struct cred *old = current_cred(); 746 struct cred *new = bprm->cred; 747 748 if (!root_privileged()) 749 return; 750 /* 751 * If the legacy file capability is set, then don't set privs 752 * for a setuid root binary run by a non-root user. Do set it 753 * for a root user just to cause least surprise to an admin. 754 */ 755 if (has_fcap && __is_suid(root_uid, new)) { 756 warn_setuid_and_fcaps_mixed(bprm->filename); 757 return; 758 } 759 /* 760 * To support inheritance of root-permissions and suid-root 761 * executables under compatibility mode, we override the 762 * capability sets for the file. 763 */ 764 if (__is_eff(root_uid, new) || __is_real(root_uid, new)) { 765 /* pP' = (cap_bset & ~0) | (pI & ~0) */ 766 new->cap_permitted = cap_combine(old->cap_bset, 767 old->cap_inheritable); 768 } 769 /* 770 * If only the real uid is 0, we do not set the effective bit. 771 */ 772 if (__is_eff(root_uid, new)) 773 *effective = true; 774 } 775 776 #define __cap_gained(field, target, source) \ 777 !cap_issubset(target->cap_##field, source->cap_##field) 778 #define __cap_grew(target, source, cred) \ 779 !cap_issubset(cred->cap_##target, cred->cap_##source) 780 #define __cap_full(field, cred) \ 781 cap_issubset(CAP_FULL_SET, cred->cap_##field) 782 783 static inline bool __is_setuid(struct cred *new, const struct cred *old) 784 { return !uid_eq(new->euid, old->uid); } 785 786 static inline bool __is_setgid(struct cred *new, const struct cred *old) 787 { return !gid_eq(new->egid, old->gid); } 788 789 /* 790 * 1) Audit candidate if current->cap_effective is set 791 * 792 * We do not bother to audit if 3 things are true: 793 * 1) cap_effective has all caps 794 * 2) we became root *OR* are were already root 795 * 3) root is supposed to have all caps (SECURE_NOROOT) 796 * Since this is just a normal root execing a process. 797 * 798 * Number 1 above might fail if you don't have a full bset, but I think 799 * that is interesting information to audit. 800 * 801 * A number of other conditions require logging: 802 * 2) something prevented setuid root getting all caps 803 * 3) non-setuid root gets fcaps 804 * 4) non-setuid root gets ambient 805 */ 806 static inline bool nonroot_raised_pE(struct cred *new, const struct cred *old, 807 kuid_t root, bool has_fcap) 808 { 809 bool ret = false; 810 811 if ((__cap_grew(effective, ambient, new) && 812 !(__cap_full(effective, new) && 813 (__is_eff(root, new) || __is_real(root, new)) && 814 root_privileged())) || 815 (root_privileged() && 816 __is_suid(root, new) && 817 !__cap_full(effective, new)) || 818 (!__is_setuid(new, old) && 819 ((has_fcap && 820 __cap_gained(permitted, new, old)) || 821 __cap_gained(ambient, new, old)))) 822 823 ret = true; 824 825 return ret; 826 } 827 828 /** 829 * cap_bprm_creds_from_file - Set up the proposed credentials for execve(). 830 * @bprm: The execution parameters, including the proposed creds 831 * @file: The file to pull the credentials from 832 * 833 * Set up the proposed credentials for a new execution context being 834 * constructed by execve(). The proposed creds in @bprm->cred is altered, 835 * which won't take effect immediately. Returns 0 if successful, -ve on error. 836 */ 837 int cap_bprm_creds_from_file(struct linux_binprm *bprm, struct file *file) 838 { 839 /* Process setpcap binaries and capabilities for uid 0 */ 840 const struct cred *old = current_cred(); 841 struct cred *new = bprm->cred; 842 bool effective = false, has_fcap = false, is_setid; 843 int ret; 844 kuid_t root_uid; 845 846 if (WARN_ON(!cap_ambient_invariant_ok(old))) 847 return -EPERM; 848 849 ret = get_file_caps(bprm, file, &effective, &has_fcap); 850 if (ret < 0) 851 return ret; 852 853 root_uid = make_kuid(new->user_ns, 0); 854 855 handle_privileged_root(bprm, has_fcap, &effective, root_uid); 856 857 /* if we have fs caps, clear dangerous personality flags */ 858 if (__cap_gained(permitted, new, old)) 859 bprm->per_clear |= PER_CLEAR_ON_SETID; 860 861 /* Don't let someone trace a set[ug]id/setpcap binary with the revised 862 * credentials unless they have the appropriate permit. 863 * 864 * In addition, if NO_NEW_PRIVS, then ensure we get no new privs. 865 */ 866 is_setid = __is_setuid(new, old) || __is_setgid(new, old); 867 868 if ((is_setid || __cap_gained(permitted, new, old)) && 869 ((bprm->unsafe & ~LSM_UNSAFE_PTRACE) || 870 !ptracer_capable(current, new->user_ns))) { 871 /* downgrade; they get no more than they had, and maybe less */ 872 if (!ns_capable(new->user_ns, CAP_SETUID) || 873 (bprm->unsafe & LSM_UNSAFE_NO_NEW_PRIVS)) { 874 new->euid = new->uid; 875 new->egid = new->gid; 876 } 877 new->cap_permitted = cap_intersect(new->cap_permitted, 878 old->cap_permitted); 879 } 880 881 new->suid = new->fsuid = new->euid; 882 new->sgid = new->fsgid = new->egid; 883 884 /* File caps or setid cancels ambient. */ 885 if (has_fcap || is_setid) 886 cap_clear(new->cap_ambient); 887 888 /* 889 * Now that we've computed pA', update pP' to give: 890 * pP' = (X & fP) | (pI & fI) | pA' 891 */ 892 new->cap_permitted = cap_combine(new->cap_permitted, new->cap_ambient); 893 894 /* 895 * Set pE' = (fE ? pP' : pA'). Because pA' is zero if fE is set, 896 * this is the same as pE' = (fE ? pP' : 0) | pA'. 897 */ 898 if (effective) 899 new->cap_effective = new->cap_permitted; 900 else 901 new->cap_effective = new->cap_ambient; 902 903 if (WARN_ON(!cap_ambient_invariant_ok(new))) 904 return -EPERM; 905 906 if (nonroot_raised_pE(new, old, root_uid, has_fcap)) { 907 ret = audit_log_bprm_fcaps(bprm, new, old); 908 if (ret < 0) 909 return ret; 910 } 911 912 new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS); 913 914 if (WARN_ON(!cap_ambient_invariant_ok(new))) 915 return -EPERM; 916 917 /* Check for privilege-elevated exec. */ 918 if (is_setid || 919 (!__is_real(root_uid, new) && 920 (effective || 921 __cap_grew(permitted, ambient, new)))) 922 bprm->secureexec = 1; 923 924 return 0; 925 } 926 927 /** 928 * cap_inode_setxattr - Determine whether an xattr may be altered 929 * @dentry: The inode/dentry being altered 930 * @name: The name of the xattr to be changed 931 * @value: The value that the xattr will be changed to 932 * @size: The size of value 933 * @flags: The replacement flag 934 * 935 * Determine whether an xattr may be altered or set on an inode, returning 0 if 936 * permission is granted, -ve if denied. 937 * 938 * This is used to make sure security xattrs don't get updated or set by those 939 * who aren't privileged to do so. 940 */ 941 int cap_inode_setxattr(struct dentry *dentry, const char *name, 942 const void *value, size_t size, int flags) 943 { 944 struct user_namespace *user_ns = dentry->d_sb->s_user_ns; 945 946 /* Ignore non-security xattrs */ 947 if (strncmp(name, XATTR_SECURITY_PREFIX, 948 XATTR_SECURITY_PREFIX_LEN) != 0) 949 return 0; 950 951 /* 952 * For XATTR_NAME_CAPS the check will be done in 953 * cap_convert_nscap(), called by setxattr() 954 */ 955 if (strcmp(name, XATTR_NAME_CAPS) == 0) 956 return 0; 957 958 if (!ns_capable(user_ns, CAP_SYS_ADMIN)) 959 return -EPERM; 960 return 0; 961 } 962 963 /** 964 * cap_inode_removexattr - Determine whether an xattr may be removed 965 * @dentry: The inode/dentry being altered 966 * @name: The name of the xattr to be changed 967 * 968 * Determine whether an xattr may be removed from an inode, returning 0 if 969 * permission is granted, -ve if denied. 970 * 971 * This is used to make sure security xattrs don't get removed by those who 972 * aren't privileged to remove them. 973 */ 974 int cap_inode_removexattr(struct dentry *dentry, const char *name) 975 { 976 struct user_namespace *user_ns = dentry->d_sb->s_user_ns; 977 978 /* Ignore non-security xattrs */ 979 if (strncmp(name, XATTR_SECURITY_PREFIX, 980 XATTR_SECURITY_PREFIX_LEN) != 0) 981 return 0; 982 983 if (strcmp(name, XATTR_NAME_CAPS) == 0) { 984 /* security.capability gets namespaced */ 985 struct inode *inode = d_backing_inode(dentry); 986 if (!inode) 987 return -EINVAL; 988 if (!capable_wrt_inode_uidgid(inode, CAP_SETFCAP)) 989 return -EPERM; 990 return 0; 991 } 992 993 if (!ns_capable(user_ns, CAP_SYS_ADMIN)) 994 return -EPERM; 995 return 0; 996 } 997 998 /* 999 * cap_emulate_setxuid() fixes the effective / permitted capabilities of 1000 * a process after a call to setuid, setreuid, or setresuid. 1001 * 1002 * 1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of 1003 * {r,e,s}uid != 0, the permitted and effective capabilities are 1004 * cleared. 1005 * 1006 * 2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective 1007 * capabilities of the process are cleared. 1008 * 1009 * 3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective 1010 * capabilities are set to the permitted capabilities. 1011 * 1012 * fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should 1013 * never happen. 1014 * 1015 * -astor 1016 * 1017 * cevans - New behaviour, Oct '99 1018 * A process may, via prctl(), elect to keep its capabilities when it 1019 * calls setuid() and switches away from uid==0. Both permitted and 1020 * effective sets will be retained. 1021 * Without this change, it was impossible for a daemon to drop only some 1022 * of its privilege. The call to setuid(!=0) would drop all privileges! 1023 * Keeping uid 0 is not an option because uid 0 owns too many vital 1024 * files.. 1025 * Thanks to Olaf Kirch and Peter Benie for spotting this. 1026 */ 1027 static inline void cap_emulate_setxuid(struct cred *new, const struct cred *old) 1028 { 1029 kuid_t root_uid = make_kuid(old->user_ns, 0); 1030 1031 if ((uid_eq(old->uid, root_uid) || 1032 uid_eq(old->euid, root_uid) || 1033 uid_eq(old->suid, root_uid)) && 1034 (!uid_eq(new->uid, root_uid) && 1035 !uid_eq(new->euid, root_uid) && 1036 !uid_eq(new->suid, root_uid))) { 1037 if (!issecure(SECURE_KEEP_CAPS)) { 1038 cap_clear(new->cap_permitted); 1039 cap_clear(new->cap_effective); 1040 } 1041 1042 /* 1043 * Pre-ambient programs expect setresuid to nonroot followed 1044 * by exec to drop capabilities. We should make sure that 1045 * this remains the case. 1046 */ 1047 cap_clear(new->cap_ambient); 1048 } 1049 if (uid_eq(old->euid, root_uid) && !uid_eq(new->euid, root_uid)) 1050 cap_clear(new->cap_effective); 1051 if (!uid_eq(old->euid, root_uid) && uid_eq(new->euid, root_uid)) 1052 new->cap_effective = new->cap_permitted; 1053 } 1054 1055 /** 1056 * cap_task_fix_setuid - Fix up the results of setuid() call 1057 * @new: The proposed credentials 1058 * @old: The current task's current credentials 1059 * @flags: Indications of what has changed 1060 * 1061 * Fix up the results of setuid() call before the credential changes are 1062 * actually applied, returning 0 to grant the changes, -ve to deny them. 1063 */ 1064 int cap_task_fix_setuid(struct cred *new, const struct cred *old, int flags) 1065 { 1066 switch (flags) { 1067 case LSM_SETID_RE: 1068 case LSM_SETID_ID: 1069 case LSM_SETID_RES: 1070 /* juggle the capabilities to follow [RES]UID changes unless 1071 * otherwise suppressed */ 1072 if (!issecure(SECURE_NO_SETUID_FIXUP)) 1073 cap_emulate_setxuid(new, old); 1074 break; 1075 1076 case LSM_SETID_FS: 1077 /* juggle the capabilties to follow FSUID changes, unless 1078 * otherwise suppressed 1079 * 1080 * FIXME - is fsuser used for all CAP_FS_MASK capabilities? 1081 * if not, we might be a bit too harsh here. 1082 */ 1083 if (!issecure(SECURE_NO_SETUID_FIXUP)) { 1084 kuid_t root_uid = make_kuid(old->user_ns, 0); 1085 if (uid_eq(old->fsuid, root_uid) && !uid_eq(new->fsuid, root_uid)) 1086 new->cap_effective = 1087 cap_drop_fs_set(new->cap_effective); 1088 1089 if (!uid_eq(old->fsuid, root_uid) && uid_eq(new->fsuid, root_uid)) 1090 new->cap_effective = 1091 cap_raise_fs_set(new->cap_effective, 1092 new->cap_permitted); 1093 } 1094 break; 1095 1096 default: 1097 return -EINVAL; 1098 } 1099 1100 return 0; 1101 } 1102 1103 /* 1104 * Rationale: code calling task_setscheduler, task_setioprio, and 1105 * task_setnice, assumes that 1106 * . if capable(cap_sys_nice), then those actions should be allowed 1107 * . if not capable(cap_sys_nice), but acting on your own processes, 1108 * then those actions should be allowed 1109 * This is insufficient now since you can call code without suid, but 1110 * yet with increased caps. 1111 * So we check for increased caps on the target process. 1112 */ 1113 static int cap_safe_nice(struct task_struct *p) 1114 { 1115 int is_subset, ret = 0; 1116 1117 rcu_read_lock(); 1118 is_subset = cap_issubset(__task_cred(p)->cap_permitted, 1119 current_cred()->cap_permitted); 1120 if (!is_subset && !ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) 1121 ret = -EPERM; 1122 rcu_read_unlock(); 1123 1124 return ret; 1125 } 1126 1127 /** 1128 * cap_task_setscheduler - Detemine if scheduler policy change is permitted 1129 * @p: The task to affect 1130 * 1131 * Detemine if the requested scheduler policy change is permitted for the 1132 * specified task, returning 0 if permission is granted, -ve if denied. 1133 */ 1134 int cap_task_setscheduler(struct task_struct *p) 1135 { 1136 return cap_safe_nice(p); 1137 } 1138 1139 /** 1140 * cap_task_ioprio - Detemine if I/O priority change is permitted 1141 * @p: The task to affect 1142 * @ioprio: The I/O priority to set 1143 * 1144 * Detemine if the requested I/O priority change is permitted for the specified 1145 * task, returning 0 if permission is granted, -ve if denied. 1146 */ 1147 int cap_task_setioprio(struct task_struct *p, int ioprio) 1148 { 1149 return cap_safe_nice(p); 1150 } 1151 1152 /** 1153 * cap_task_ioprio - Detemine if task priority change is permitted 1154 * @p: The task to affect 1155 * @nice: The nice value to set 1156 * 1157 * Detemine if the requested task priority change is permitted for the 1158 * specified task, returning 0 if permission is granted, -ve if denied. 1159 */ 1160 int cap_task_setnice(struct task_struct *p, int nice) 1161 { 1162 return cap_safe_nice(p); 1163 } 1164 1165 /* 1166 * Implement PR_CAPBSET_DROP. Attempt to remove the specified capability from 1167 * the current task's bounding set. Returns 0 on success, -ve on error. 1168 */ 1169 static int cap_prctl_drop(unsigned long cap) 1170 { 1171 struct cred *new; 1172 1173 if (!ns_capable(current_user_ns(), CAP_SETPCAP)) 1174 return -EPERM; 1175 if (!cap_valid(cap)) 1176 return -EINVAL; 1177 1178 new = prepare_creds(); 1179 if (!new) 1180 return -ENOMEM; 1181 cap_lower(new->cap_bset, cap); 1182 return commit_creds(new); 1183 } 1184 1185 /** 1186 * cap_task_prctl - Implement process control functions for this security module 1187 * @option: The process control function requested 1188 * @arg2, @arg3, @arg4, @arg5: The argument data for this function 1189 * 1190 * Allow process control functions (sys_prctl()) to alter capabilities; may 1191 * also deny access to other functions not otherwise implemented here. 1192 * 1193 * Returns 0 or +ve on success, -ENOSYS if this function is not implemented 1194 * here, other -ve on error. If -ENOSYS is returned, sys_prctl() and other LSM 1195 * modules will consider performing the function. 1196 */ 1197 int cap_task_prctl(int option, unsigned long arg2, unsigned long arg3, 1198 unsigned long arg4, unsigned long arg5) 1199 { 1200 const struct cred *old = current_cred(); 1201 struct cred *new; 1202 1203 switch (option) { 1204 case PR_CAPBSET_READ: 1205 if (!cap_valid(arg2)) 1206 return -EINVAL; 1207 return !!cap_raised(old->cap_bset, arg2); 1208 1209 case PR_CAPBSET_DROP: 1210 return cap_prctl_drop(arg2); 1211 1212 /* 1213 * The next four prctl's remain to assist with transitioning a 1214 * system from legacy UID=0 based privilege (when filesystem 1215 * capabilities are not in use) to a system using filesystem 1216 * capabilities only - as the POSIX.1e draft intended. 1217 * 1218 * Note: 1219 * 1220 * PR_SET_SECUREBITS = 1221 * issecure_mask(SECURE_KEEP_CAPS_LOCKED) 1222 * | issecure_mask(SECURE_NOROOT) 1223 * | issecure_mask(SECURE_NOROOT_LOCKED) 1224 * | issecure_mask(SECURE_NO_SETUID_FIXUP) 1225 * | issecure_mask(SECURE_NO_SETUID_FIXUP_LOCKED) 1226 * 1227 * will ensure that the current process and all of its 1228 * children will be locked into a pure 1229 * capability-based-privilege environment. 1230 */ 1231 case PR_SET_SECUREBITS: 1232 if ((((old->securebits & SECURE_ALL_LOCKS) >> 1) 1233 & (old->securebits ^ arg2)) /*[1]*/ 1234 || ((old->securebits & SECURE_ALL_LOCKS & ~arg2)) /*[2]*/ 1235 || (arg2 & ~(SECURE_ALL_LOCKS | SECURE_ALL_BITS)) /*[3]*/ 1236 || (cap_capable(current_cred(), 1237 current_cred()->user_ns, 1238 CAP_SETPCAP, 1239 CAP_OPT_NONE) != 0) /*[4]*/ 1240 /* 1241 * [1] no changing of bits that are locked 1242 * [2] no unlocking of locks 1243 * [3] no setting of unsupported bits 1244 * [4] doing anything requires privilege (go read about 1245 * the "sendmail capabilities bug") 1246 */ 1247 ) 1248 /* cannot change a locked bit */ 1249 return -EPERM; 1250 1251 new = prepare_creds(); 1252 if (!new) 1253 return -ENOMEM; 1254 new->securebits = arg2; 1255 return commit_creds(new); 1256 1257 case PR_GET_SECUREBITS: 1258 return old->securebits; 1259 1260 case PR_GET_KEEPCAPS: 1261 return !!issecure(SECURE_KEEP_CAPS); 1262 1263 case PR_SET_KEEPCAPS: 1264 if (arg2 > 1) /* Note, we rely on arg2 being unsigned here */ 1265 return -EINVAL; 1266 if (issecure(SECURE_KEEP_CAPS_LOCKED)) 1267 return -EPERM; 1268 1269 new = prepare_creds(); 1270 if (!new) 1271 return -ENOMEM; 1272 if (arg2) 1273 new->securebits |= issecure_mask(SECURE_KEEP_CAPS); 1274 else 1275 new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS); 1276 return commit_creds(new); 1277 1278 case PR_CAP_AMBIENT: 1279 if (arg2 == PR_CAP_AMBIENT_CLEAR_ALL) { 1280 if (arg3 | arg4 | arg5) 1281 return -EINVAL; 1282 1283 new = prepare_creds(); 1284 if (!new) 1285 return -ENOMEM; 1286 cap_clear(new->cap_ambient); 1287 return commit_creds(new); 1288 } 1289 1290 if (((!cap_valid(arg3)) | arg4 | arg5)) 1291 return -EINVAL; 1292 1293 if (arg2 == PR_CAP_AMBIENT_IS_SET) { 1294 return !!cap_raised(current_cred()->cap_ambient, arg3); 1295 } else if (arg2 != PR_CAP_AMBIENT_RAISE && 1296 arg2 != PR_CAP_AMBIENT_LOWER) { 1297 return -EINVAL; 1298 } else { 1299 if (arg2 == PR_CAP_AMBIENT_RAISE && 1300 (!cap_raised(current_cred()->cap_permitted, arg3) || 1301 !cap_raised(current_cred()->cap_inheritable, 1302 arg3) || 1303 issecure(SECURE_NO_CAP_AMBIENT_RAISE))) 1304 return -EPERM; 1305 1306 new = prepare_creds(); 1307 if (!new) 1308 return -ENOMEM; 1309 if (arg2 == PR_CAP_AMBIENT_RAISE) 1310 cap_raise(new->cap_ambient, arg3); 1311 else 1312 cap_lower(new->cap_ambient, arg3); 1313 return commit_creds(new); 1314 } 1315 1316 default: 1317 /* No functionality available - continue with default */ 1318 return -ENOSYS; 1319 } 1320 } 1321 1322 /** 1323 * cap_vm_enough_memory - Determine whether a new virtual mapping is permitted 1324 * @mm: The VM space in which the new mapping is to be made 1325 * @pages: The size of the mapping 1326 * 1327 * Determine whether the allocation of a new virtual mapping by the current 1328 * task is permitted, returning 1 if permission is granted, 0 if not. 1329 */ 1330 int cap_vm_enough_memory(struct mm_struct *mm, long pages) 1331 { 1332 int cap_sys_admin = 0; 1333 1334 if (cap_capable(current_cred(), &init_user_ns, 1335 CAP_SYS_ADMIN, CAP_OPT_NOAUDIT) == 0) 1336 cap_sys_admin = 1; 1337 1338 return cap_sys_admin; 1339 } 1340 1341 /* 1342 * cap_mmap_addr - check if able to map given addr 1343 * @addr: address attempting to be mapped 1344 * 1345 * If the process is attempting to map memory below dac_mmap_min_addr they need 1346 * CAP_SYS_RAWIO. The other parameters to this function are unused by the 1347 * capability security module. Returns 0 if this mapping should be allowed 1348 * -EPERM if not. 1349 */ 1350 int cap_mmap_addr(unsigned long addr) 1351 { 1352 int ret = 0; 1353 1354 if (addr < dac_mmap_min_addr) { 1355 ret = cap_capable(current_cred(), &init_user_ns, CAP_SYS_RAWIO, 1356 CAP_OPT_NONE); 1357 /* set PF_SUPERPRIV if it turns out we allow the low mmap */ 1358 if (ret == 0) 1359 current->flags |= PF_SUPERPRIV; 1360 } 1361 return ret; 1362 } 1363 1364 int cap_mmap_file(struct file *file, unsigned long reqprot, 1365 unsigned long prot, unsigned long flags) 1366 { 1367 return 0; 1368 } 1369 1370 #ifdef CONFIG_SECURITY 1371 1372 static struct security_hook_list capability_hooks[] __lsm_ro_after_init = { 1373 LSM_HOOK_INIT(capable, cap_capable), 1374 LSM_HOOK_INIT(settime, cap_settime), 1375 LSM_HOOK_INIT(ptrace_access_check, cap_ptrace_access_check), 1376 LSM_HOOK_INIT(ptrace_traceme, cap_ptrace_traceme), 1377 LSM_HOOK_INIT(capget, cap_capget), 1378 LSM_HOOK_INIT(capset, cap_capset), 1379 LSM_HOOK_INIT(bprm_creds_from_file, cap_bprm_creds_from_file), 1380 LSM_HOOK_INIT(inode_need_killpriv, cap_inode_need_killpriv), 1381 LSM_HOOK_INIT(inode_killpriv, cap_inode_killpriv), 1382 LSM_HOOK_INIT(inode_getsecurity, cap_inode_getsecurity), 1383 LSM_HOOK_INIT(mmap_addr, cap_mmap_addr), 1384 LSM_HOOK_INIT(mmap_file, cap_mmap_file), 1385 LSM_HOOK_INIT(task_fix_setuid, cap_task_fix_setuid), 1386 LSM_HOOK_INIT(task_prctl, cap_task_prctl), 1387 LSM_HOOK_INIT(task_setscheduler, cap_task_setscheduler), 1388 LSM_HOOK_INIT(task_setioprio, cap_task_setioprio), 1389 LSM_HOOK_INIT(task_setnice, cap_task_setnice), 1390 LSM_HOOK_INIT(vm_enough_memory, cap_vm_enough_memory), 1391 }; 1392 1393 static int __init capability_init(void) 1394 { 1395 security_add_hooks(capability_hooks, ARRAY_SIZE(capability_hooks), 1396 "capability"); 1397 return 0; 1398 } 1399 1400 DEFINE_LSM(capability) = { 1401 .name = "capability", 1402 .order = LSM_ORDER_FIRST, 1403 .init = capability_init, 1404 }; 1405 1406 #endif /* CONFIG_SECURITY */ 1407