linux/security/commoncap.c
<<
>>
Prefs
   1/* Common capabilities, needed by capability.o.
   2 *
   3 *      This program is free software; you can redistribute it and/or modify
   4 *      it under the terms of the GNU General Public License as published by
   5 *      the Free Software Foundation; either version 2 of the License, or
   6 *      (at your option) any later version.
   7 *
   8 */
   9
  10#include <linux/capability.h>
  11#include <linux/audit.h>
  12#include <linux/module.h>
  13#include <linux/init.h>
  14#include <linux/kernel.h>
  15#include <linux/security.h>
  16#include <linux/file.h>
  17#include <linux/mm.h>
  18#include <linux/mman.h>
  19#include <linux/pagemap.h>
  20#include <linux/swap.h>
  21#include <linux/skbuff.h>
  22#include <linux/netlink.h>
  23#include <linux/ptrace.h>
  24#include <linux/xattr.h>
  25#include <linux/hugetlb.h>
  26#include <linux/mount.h>
  27#include <linux/sched.h>
  28#include <linux/prctl.h>
  29#include <linux/securebits.h>
  30#include <linux/user_namespace.h>
  31#include <linux/binfmts.h>
  32#include <linux/personality.h>
  33
  34/*
  35 * If a non-root user executes a setuid-root binary in
  36 * !secure(SECURE_NOROOT) mode, then we raise capabilities.
  37 * However if fE is also set, then the intent is for only
  38 * the file capabilities to be applied, and the setuid-root
  39 * bit is left on either to change the uid (plausible) or
  40 * to get full privilege on a kernel without file capabilities
  41 * support.  So in that case we do not raise capabilities.
  42 *
  43 * Warn if that happens, once per boot.
  44 */
  45static void warn_setuid_and_fcaps_mixed(const char *fname)
  46{
  47        static int warned;
  48        if (!warned) {
  49                printk(KERN_INFO "warning: `%s' has both setuid-root and"
  50                        " effective capabilities. Therefore not raising all"
  51                        " capabilities.\n", fname);
  52                warned = 1;
  53        }
  54}
  55
  56int cap_netlink_send(struct sock *sk, struct sk_buff *skb)
  57{
  58        return 0;
  59}
  60
  61/**
  62 * cap_capable - Determine whether a task has a particular effective capability
  63 * @cred: The credentials to use
  64 * @ns:  The user namespace in which we need the capability
  65 * @cap: The capability to check for
  66 * @audit: Whether to write an audit message or not
  67 *
  68 * Determine whether the nominated task has the specified capability amongst
  69 * its effective set, returning 0 if it does, -ve if it does not.
  70 *
  71 * NOTE WELL: cap_has_capability() cannot be used like the kernel's capable()
  72 * and has_capability() functions.  That is, it has the reverse semantics:
  73 * cap_has_capability() returns 0 when a task has a capability, but the
  74 * kernel's capable() and has_capability() returns 1 for this case.
  75 */
  76int cap_capable(const struct cred *cred, struct user_namespace *targ_ns,
  77                int cap, int audit)
  78{
  79        for (;;) {
  80                /* The owner of the user namespace has all caps. */
  81                if (targ_ns != &init_user_ns && uid_eq(targ_ns->owner, cred->euid))
  82                        return 0;
  83
  84                /* Do we have the necessary capabilities? */
  85                if (targ_ns == cred->user_ns)
  86                        return cap_raised(cred->cap_effective, cap) ? 0 : -EPERM;
  87
  88                /* Have we tried all of the parent namespaces? */
  89                if (targ_ns == &init_user_ns)
  90                        return -EPERM;
  91
  92                /*
  93                 *If you have a capability in a parent user ns, then you have
  94                 * it over all children user namespaces as well.
  95                 */
  96                targ_ns = targ_ns->parent;
  97        }
  98
  99        /* We never get here */
 100}
 101
 102/**
 103 * cap_settime - Determine whether the current process may set the system clock
 104 * @ts: The time to set
 105 * @tz: The timezone to set
 106 *
 107 * Determine whether the current process may set the system clock and timezone
 108 * information, returning 0 if permission granted, -ve if denied.
 109 */
 110int cap_settime(const struct timespec *ts, const struct timezone *tz)
 111{
 112        if (!capable(CAP_SYS_TIME))
 113                return -EPERM;
 114        return 0;
 115}
 116
 117/**
 118 * cap_ptrace_access_check - Determine whether the current process may access
 119 *                         another
 120 * @child: The process to be accessed
 121 * @mode: The mode of attachment.
 122 *
 123 * If we are in the same or an ancestor user_ns and have all the target
 124 * task's capabilities, then ptrace access is allowed.
 125 * If we have the ptrace capability to the target user_ns, then ptrace
 126 * access is allowed.
 127 * Else denied.
 128 *
 129 * Determine whether a process may access another, returning 0 if permission
 130 * granted, -ve if denied.
 131 */
 132int cap_ptrace_access_check(struct task_struct *child, unsigned int mode)
 133{
 134        int ret = 0;
 135        const struct cred *cred, *child_cred;
 136
 137        rcu_read_lock();
 138        cred = current_cred();
 139        child_cred = __task_cred(child);
 140        if (cred->user_ns == child_cred->user_ns &&
 141            cap_issubset(child_cred->cap_permitted, cred->cap_permitted))
 142                goto out;
 143        if (ns_capable(child_cred->user_ns, CAP_SYS_PTRACE))
 144                goto out;
 145        ret = -EPERM;
 146out:
 147        rcu_read_unlock();
 148        return ret;
 149}
 150
 151/**
 152 * cap_ptrace_traceme - Determine whether another process may trace the current
 153 * @parent: The task proposed to be the tracer
 154 *
 155 * If parent is in the same or an ancestor user_ns and has all current's
 156 * capabilities, then ptrace access is allowed.
 157 * If parent has the ptrace capability to current's user_ns, then ptrace
 158 * access is allowed.
 159 * Else denied.
 160 *
 161 * Determine whether the nominated task is permitted to trace the current
 162 * process, returning 0 if permission is granted, -ve if denied.
 163 */
 164int cap_ptrace_traceme(struct task_struct *parent)
 165{
 166        int ret = 0;
 167        const struct cred *cred, *child_cred;
 168
 169        rcu_read_lock();
 170        cred = __task_cred(parent);
 171        child_cred = current_cred();
 172        if (cred->user_ns == child_cred->user_ns &&
 173            cap_issubset(child_cred->cap_permitted, cred->cap_permitted))
 174                goto out;
 175        if (has_ns_capability(parent, child_cred->user_ns, CAP_SYS_PTRACE))
 176                goto out;
 177        ret = -EPERM;
 178out:
 179        rcu_read_unlock();
 180        return ret;
 181}
 182
 183/**
 184 * cap_capget - Retrieve a task's capability sets
 185 * @target: The task from which to retrieve the capability sets
 186 * @effective: The place to record the effective set
 187 * @inheritable: The place to record the inheritable set
 188 * @permitted: The place to record the permitted set
 189 *
 190 * This function retrieves the capabilities of the nominated task and returns
 191 * them to the caller.
 192 */
 193int cap_capget(struct task_struct *target, kernel_cap_t *effective,
 194               kernel_cap_t *inheritable, kernel_cap_t *permitted)
 195{
 196        const struct cred *cred;
 197
 198        /* Derived from kernel/capability.c:sys_capget. */
 199        rcu_read_lock();
 200        cred = __task_cred(target);
 201        *effective   = cred->cap_effective;
 202        *inheritable = cred->cap_inheritable;
 203        *permitted   = cred->cap_permitted;
 204        rcu_read_unlock();
 205        return 0;
 206}
 207
 208/*
 209 * Determine whether the inheritable capabilities are limited to the old
 210 * permitted set.  Returns 1 if they are limited, 0 if they are not.
 211 */
 212static inline int cap_inh_is_capped(void)
 213{
 214
 215        /* they are so limited unless the current task has the CAP_SETPCAP
 216         * capability
 217         */
 218        if (cap_capable(current_cred(), current_cred()->user_ns,
 219                        CAP_SETPCAP, SECURITY_CAP_AUDIT) == 0)
 220                return 0;
 221        return 1;
 222}
 223
 224/**
 225 * cap_capset - Validate and apply proposed changes to current's capabilities
 226 * @new: The proposed new credentials; alterations should be made here
 227 * @old: The current task's current credentials
 228 * @effective: A pointer to the proposed new effective capabilities set
 229 * @inheritable: A pointer to the proposed new inheritable capabilities set
 230 * @permitted: A pointer to the proposed new permitted capabilities set
 231 *
 232 * This function validates and applies a proposed mass change to the current
 233 * process's capability sets.  The changes are made to the proposed new
 234 * credentials, and assuming no error, will be committed by the caller of LSM.
 235 */
 236int cap_capset(struct cred *new,
 237               const struct cred *old,
 238               const kernel_cap_t *effective,
 239               const kernel_cap_t *inheritable,
 240               const kernel_cap_t *permitted)
 241{
 242        if (cap_inh_is_capped() &&
 243            !cap_issubset(*inheritable,
 244                          cap_combine(old->cap_inheritable,
 245                                      old->cap_permitted)))
 246                /* incapable of using this inheritable set */
 247                return -EPERM;
 248
 249        if (!cap_issubset(*inheritable,
 250                          cap_combine(old->cap_inheritable,
 251                                      old->cap_bset)))
 252                /* no new pI capabilities outside bounding set */
 253                return -EPERM;
 254
 255        /* verify restrictions on target's new Permitted set */
 256        if (!cap_issubset(*permitted, old->cap_permitted))
 257                return -EPERM;
 258
 259        /* verify the _new_Effective_ is a subset of the _new_Permitted_ */
 260        if (!cap_issubset(*effective, *permitted))
 261                return -EPERM;
 262
 263        new->cap_effective   = *effective;
 264        new->cap_inheritable = *inheritable;
 265        new->cap_permitted   = *permitted;
 266        return 0;
 267}
 268
 269/*
 270 * Clear proposed capability sets for execve().
 271 */
 272static inline void bprm_clear_caps(struct linux_binprm *bprm)
 273{
 274        cap_clear(bprm->cred->cap_permitted);
 275        bprm->cap_effective = false;
 276}
 277
 278/**
 279 * cap_inode_need_killpriv - Determine if inode change affects privileges
 280 * @dentry: The inode/dentry in being changed with change marked ATTR_KILL_PRIV
 281 *
 282 * Determine if an inode having a change applied that's marked ATTR_KILL_PRIV
 283 * affects the security markings on that inode, and if it is, should
 284 * inode_killpriv() be invoked or the change rejected?
 285 *
 286 * Returns 0 if granted; +ve if granted, but inode_killpriv() is required; and
 287 * -ve to deny the change.
 288 */
 289int cap_inode_need_killpriv(struct dentry *dentry)
 290{
 291        struct inode *inode = dentry->d_inode;
 292        int error;
 293
 294        if (!inode->i_op->getxattr)
 295               return 0;
 296
 297        error = inode->i_op->getxattr(dentry, XATTR_NAME_CAPS, NULL, 0);
 298        if (error <= 0)
 299                return 0;
 300        return 1;
 301}
 302
 303/**
 304 * cap_inode_killpriv - Erase the security markings on an inode
 305 * @dentry: The inode/dentry to alter
 306 *
 307 * Erase the privilege-enhancing security markings on an inode.
 308 *
 309 * Returns 0 if successful, -ve on error.
 310 */
 311int cap_inode_killpriv(struct dentry *dentry)
 312{
 313        struct inode *inode = dentry->d_inode;
 314
 315        if (!inode->i_op->removexattr)
 316               return 0;
 317
 318        return inode->i_op->removexattr(dentry, XATTR_NAME_CAPS);
 319}
 320
 321/*
 322 * Calculate the new process capability sets from the capability sets attached
 323 * to a file.
 324 */
 325static inline int bprm_caps_from_vfs_caps(struct cpu_vfs_cap_data *caps,
 326                                          struct linux_binprm *bprm,
 327                                          bool *effective,
 328                                          bool *has_cap)
 329{
 330        struct cred *new = bprm->cred;
 331        unsigned i;
 332        int ret = 0;
 333
 334        if (caps->magic_etc & VFS_CAP_FLAGS_EFFECTIVE)
 335                *effective = true;
 336
 337        if (caps->magic_etc & VFS_CAP_REVISION_MASK)
 338                *has_cap = true;
 339
 340        CAP_FOR_EACH_U32(i) {
 341                __u32 permitted = caps->permitted.cap[i];
 342                __u32 inheritable = caps->inheritable.cap[i];
 343
 344                /*
 345                 * pP' = (X & fP) | (pI & fI)
 346                 */
 347                new->cap_permitted.cap[i] =
 348                        (new->cap_bset.cap[i] & permitted) |
 349                        (new->cap_inheritable.cap[i] & inheritable);
 350
 351                if (permitted & ~new->cap_permitted.cap[i])
 352                        /* insufficient to execute correctly */
 353                        ret = -EPERM;
 354        }
 355
 356        /*
 357         * For legacy apps, with no internal support for recognizing they
 358         * do not have enough capabilities, we return an error if they are
 359         * missing some "forced" (aka file-permitted) capabilities.
 360         */
 361        return *effective ? ret : 0;
 362}
 363
 364/*
 365 * Extract the on-exec-apply capability sets for an executable file.
 366 */
 367int get_vfs_caps_from_disk(const struct dentry *dentry, struct cpu_vfs_cap_data *cpu_caps)
 368{
 369        struct inode *inode = dentry->d_inode;
 370        __u32 magic_etc;
 371        unsigned tocopy, i;
 372        int size;
 373        struct vfs_cap_data caps;
 374
 375        memset(cpu_caps, 0, sizeof(struct cpu_vfs_cap_data));
 376
 377        if (!inode || !inode->i_op->getxattr)
 378                return -ENODATA;
 379
 380        size = inode->i_op->getxattr((struct dentry *)dentry, XATTR_NAME_CAPS, &caps,
 381                                   XATTR_CAPS_SZ);
 382        if (size == -ENODATA || size == -EOPNOTSUPP)
 383                /* no data, that's ok */
 384                return -ENODATA;
 385        if (size < 0)
 386                return size;
 387
 388        if (size < sizeof(magic_etc))
 389                return -EINVAL;
 390
 391        cpu_caps->magic_etc = magic_etc = le32_to_cpu(caps.magic_etc);
 392
 393        switch (magic_etc & VFS_CAP_REVISION_MASK) {
 394        case VFS_CAP_REVISION_1:
 395                if (size != XATTR_CAPS_SZ_1)
 396                        return -EINVAL;
 397                tocopy = VFS_CAP_U32_1;
 398                break;
 399        case VFS_CAP_REVISION_2:
 400                if (size != XATTR_CAPS_SZ_2)
 401                        return -EINVAL;
 402                tocopy = VFS_CAP_U32_2;
 403                break;
 404        default:
 405                return -EINVAL;
 406        }
 407
 408        CAP_FOR_EACH_U32(i) {
 409                if (i >= tocopy)
 410                        break;
 411                cpu_caps->permitted.cap[i] = le32_to_cpu(caps.data[i].permitted);
 412                cpu_caps->inheritable.cap[i] = le32_to_cpu(caps.data[i].inheritable);
 413        }
 414
 415        return 0;
 416}
 417
 418/*
 419 * Attempt to get the on-exec apply capability sets for an executable file from
 420 * its xattrs and, if present, apply them to the proposed credentials being
 421 * constructed by execve().
 422 */
 423static int get_file_caps(struct linux_binprm *bprm, bool *effective, bool *has_cap)
 424{
 425        struct dentry *dentry;
 426        int rc = 0;
 427        struct cpu_vfs_cap_data vcaps;
 428
 429        bprm_clear_caps(bprm);
 430
 431        if (!file_caps_enabled)
 432                return 0;
 433
 434        if (bprm->file->f_vfsmnt->mnt_flags & MNT_NOSUID)
 435                return 0;
 436
 437        dentry = dget(bprm->file->f_dentry);
 438
 439        rc = get_vfs_caps_from_disk(dentry, &vcaps);
 440        if (rc < 0) {
 441                if (rc == -EINVAL)
 442                        printk(KERN_NOTICE "%s: get_vfs_caps_from_disk returned %d for %s\n",
 443                                __func__, rc, bprm->filename);
 444                else if (rc == -ENODATA)
 445                        rc = 0;
 446                goto out;
 447        }
 448
 449        rc = bprm_caps_from_vfs_caps(&vcaps, bprm, effective, has_cap);
 450        if (rc == -EINVAL)
 451                printk(KERN_NOTICE "%s: cap_from_disk returned %d for %s\n",
 452                       __func__, rc, bprm->filename);
 453
 454out:
 455        dput(dentry);
 456        if (rc)
 457                bprm_clear_caps(bprm);
 458
 459        return rc;
 460}
 461
 462/**
 463 * cap_bprm_set_creds - Set up the proposed credentials for execve().
 464 * @bprm: The execution parameters, including the proposed creds
 465 *
 466 * Set up the proposed credentials for a new execution context being
 467 * constructed by execve().  The proposed creds in @bprm->cred is altered,
 468 * which won't take effect immediately.  Returns 0 if successful, -ve on error.
 469 */
 470int cap_bprm_set_creds(struct linux_binprm *bprm)
 471{
 472        const struct cred *old = current_cred();
 473        struct cred *new = bprm->cred;
 474        bool effective, has_cap = false;
 475        int ret;
 476        kuid_t root_uid;
 477
 478        effective = false;
 479        ret = get_file_caps(bprm, &effective, &has_cap);
 480        if (ret < 0)
 481                return ret;
 482
 483        root_uid = make_kuid(new->user_ns, 0);
 484
 485        if (!issecure(SECURE_NOROOT)) {
 486                /*
 487                 * If the legacy file capability is set, then don't set privs
 488                 * for a setuid root binary run by a non-root user.  Do set it
 489                 * for a root user just to cause least surprise to an admin.
 490                 */
 491                if (has_cap && !uid_eq(new->uid, root_uid) && uid_eq(new->euid, root_uid)) {
 492                        warn_setuid_and_fcaps_mixed(bprm->filename);
 493                        goto skip;
 494                }
 495                /*
 496                 * To support inheritance of root-permissions and suid-root
 497                 * executables under compatibility mode, we override the
 498                 * capability sets for the file.
 499                 *
 500                 * If only the real uid is 0, we do not set the effective bit.
 501                 */
 502                if (uid_eq(new->euid, root_uid) || uid_eq(new->uid, root_uid)) {
 503                        /* pP' = (cap_bset & ~0) | (pI & ~0) */
 504                        new->cap_permitted = cap_combine(old->cap_bset,
 505                                                         old->cap_inheritable);
 506                }
 507                if (uid_eq(new->euid, root_uid))
 508                        effective = true;
 509        }
 510skip:
 511
 512        /* if we have fs caps, clear dangerous personality flags */
 513        if (!cap_issubset(new->cap_permitted, old->cap_permitted))
 514                bprm->per_clear |= PER_CLEAR_ON_SETID;
 515
 516
 517        /* Don't let someone trace a set[ug]id/setpcap binary with the revised
 518         * credentials unless they have the appropriate permit.
 519         *
 520         * In addition, if NO_NEW_PRIVS, then ensure we get no new privs.
 521         */
 522        if ((!uid_eq(new->euid, old->uid) ||
 523             !gid_eq(new->egid, old->gid) ||
 524             !cap_issubset(new->cap_permitted, old->cap_permitted)) &&
 525            bprm->unsafe & ~LSM_UNSAFE_PTRACE_CAP) {
 526                /* downgrade; they get no more than they had, and maybe less */
 527                if (!capable(CAP_SETUID) ||
 528                    (bprm->unsafe & LSM_UNSAFE_NO_NEW_PRIVS)) {
 529                        new->euid = new->uid;
 530                        new->egid = new->gid;
 531                }
 532                new->cap_permitted = cap_intersect(new->cap_permitted,
 533                                                   old->cap_permitted);
 534        }
 535
 536        new->suid = new->fsuid = new->euid;
 537        new->sgid = new->fsgid = new->egid;
 538
 539        if (effective)
 540                new->cap_effective = new->cap_permitted;
 541        else
 542                cap_clear(new->cap_effective);
 543        bprm->cap_effective = effective;
 544
 545        /*
 546         * Audit candidate if current->cap_effective is set
 547         *
 548         * We do not bother to audit if 3 things are true:
 549         *   1) cap_effective has all caps
 550         *   2) we are root
 551         *   3) root is supposed to have all caps (SECURE_NOROOT)
 552         * Since this is just a normal root execing a process.
 553         *
 554         * Number 1 above might fail if you don't have a full bset, but I think
 555         * that is interesting information to audit.
 556         */
 557        if (!cap_isclear(new->cap_effective)) {
 558                if (!cap_issubset(CAP_FULL_SET, new->cap_effective) ||
 559                    !uid_eq(new->euid, root_uid) || !uid_eq(new->uid, root_uid) ||
 560                    issecure(SECURE_NOROOT)) {
 561                        ret = audit_log_bprm_fcaps(bprm, new, old);
 562                        if (ret < 0)
 563                                return ret;
 564                }
 565        }
 566
 567        new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS);
 568        return 0;
 569}
 570
 571/**
 572 * cap_bprm_secureexec - Determine whether a secure execution is required
 573 * @bprm: The execution parameters
 574 *
 575 * Determine whether a secure execution is required, return 1 if it is, and 0
 576 * if it is not.
 577 *
 578 * The credentials have been committed by this point, and so are no longer
 579 * available through @bprm->cred.
 580 */
 581int cap_bprm_secureexec(struct linux_binprm *bprm)
 582{
 583        const struct cred *cred = current_cred();
 584        kuid_t root_uid = make_kuid(cred->user_ns, 0);
 585
 586        if (!uid_eq(cred->uid, root_uid)) {
 587                if (bprm->cap_effective)
 588                        return 1;
 589                if (!cap_isclear(cred->cap_permitted))
 590                        return 1;
 591        }
 592
 593        return (!uid_eq(cred->euid, cred->uid) ||
 594                !gid_eq(cred->egid, cred->gid));
 595}
 596
 597/**
 598 * cap_inode_setxattr - Determine whether an xattr may be altered
 599 * @dentry: The inode/dentry being altered
 600 * @name: The name of the xattr to be changed
 601 * @value: The value that the xattr will be changed to
 602 * @size: The size of value
 603 * @flags: The replacement flag
 604 *
 605 * Determine whether an xattr may be altered or set on an inode, returning 0 if
 606 * permission is granted, -ve if denied.
 607 *
 608 * This is used to make sure security xattrs don't get updated or set by those
 609 * who aren't privileged to do so.
 610 */
 611int cap_inode_setxattr(struct dentry *dentry, const char *name,
 612                       const void *value, size_t size, int flags)
 613{
 614        if (!strcmp(name, XATTR_NAME_CAPS)) {
 615                if (!capable(CAP_SETFCAP))
 616                        return -EPERM;
 617                return 0;
 618        }
 619
 620        if (!strncmp(name, XATTR_SECURITY_PREFIX,
 621                     sizeof(XATTR_SECURITY_PREFIX) - 1) &&
 622            !capable(CAP_SYS_ADMIN))
 623                return -EPERM;
 624        return 0;
 625}
 626
 627/**
 628 * cap_inode_removexattr - Determine whether an xattr may be removed
 629 * @dentry: The inode/dentry being altered
 630 * @name: The name of the xattr to be changed
 631 *
 632 * Determine whether an xattr may be removed from an inode, returning 0 if
 633 * permission is granted, -ve if denied.
 634 *
 635 * This is used to make sure security xattrs don't get removed by those who
 636 * aren't privileged to remove them.
 637 */
 638int cap_inode_removexattr(struct dentry *dentry, const char *name)
 639{
 640        if (!strcmp(name, XATTR_NAME_CAPS)) {
 641                if (!capable(CAP_SETFCAP))
 642                        return -EPERM;
 643                return 0;
 644        }
 645
 646        if (!strncmp(name, XATTR_SECURITY_PREFIX,
 647                     sizeof(XATTR_SECURITY_PREFIX) - 1) &&
 648            !capable(CAP_SYS_ADMIN))
 649                return -EPERM;
 650        return 0;
 651}
 652
 653/*
 654 * cap_emulate_setxuid() fixes the effective / permitted capabilities of
 655 * a process after a call to setuid, setreuid, or setresuid.
 656 *
 657 *  1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of
 658 *  {r,e,s}uid != 0, the permitted and effective capabilities are
 659 *  cleared.
 660 *
 661 *  2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective
 662 *  capabilities of the process are cleared.
 663 *
 664 *  3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective
 665 *  capabilities are set to the permitted capabilities.
 666 *
 667 *  fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should
 668 *  never happen.
 669 *
 670 *  -astor
 671 *
 672 * cevans - New behaviour, Oct '99
 673 * A process may, via prctl(), elect to keep its capabilities when it
 674 * calls setuid() and switches away from uid==0. Both permitted and
 675 * effective sets will be retained.
 676 * Without this change, it was impossible for a daemon to drop only some
 677 * of its privilege. The call to setuid(!=0) would drop all privileges!
 678 * Keeping uid 0 is not an option because uid 0 owns too many vital
 679 * files..
 680 * Thanks to Olaf Kirch and Peter Benie for spotting this.
 681 */
 682static inline void cap_emulate_setxuid(struct cred *new, const struct cred *old)
 683{
 684        kuid_t root_uid = make_kuid(old->user_ns, 0);
 685
 686        if ((uid_eq(old->uid, root_uid) ||
 687             uid_eq(old->euid, root_uid) ||
 688             uid_eq(old->suid, root_uid)) &&
 689            (!uid_eq(new->uid, root_uid) &&
 690             !uid_eq(new->euid, root_uid) &&
 691             !uid_eq(new->suid, root_uid)) &&
 692            !issecure(SECURE_KEEP_CAPS)) {
 693                cap_clear(new->cap_permitted);
 694                cap_clear(new->cap_effective);
 695        }
 696        if (uid_eq(old->euid, root_uid) && !uid_eq(new->euid, root_uid))
 697                cap_clear(new->cap_effective);
 698        if (!uid_eq(old->euid, root_uid) && uid_eq(new->euid, root_uid))
 699                new->cap_effective = new->cap_permitted;
 700}
 701
 702/**
 703 * cap_task_fix_setuid - Fix up the results of setuid() call
 704 * @new: The proposed credentials
 705 * @old: The current task's current credentials
 706 * @flags: Indications of what has changed
 707 *
 708 * Fix up the results of setuid() call before the credential changes are
 709 * actually applied, returning 0 to grant the changes, -ve to deny them.
 710 */
 711int cap_task_fix_setuid(struct cred *new, const struct cred *old, int flags)
 712{
 713        switch (flags) {
 714        case LSM_SETID_RE:
 715        case LSM_SETID_ID:
 716        case LSM_SETID_RES:
 717                /* juggle the capabilities to follow [RES]UID changes unless
 718                 * otherwise suppressed */
 719                if (!issecure(SECURE_NO_SETUID_FIXUP))
 720                        cap_emulate_setxuid(new, old);
 721                break;
 722
 723        case LSM_SETID_FS:
 724                /* juggle the capabilties to follow FSUID changes, unless
 725                 * otherwise suppressed
 726                 *
 727                 * FIXME - is fsuser used for all CAP_FS_MASK capabilities?
 728                 *          if not, we might be a bit too harsh here.
 729                 */
 730                if (!issecure(SECURE_NO_SETUID_FIXUP)) {
 731                        kuid_t root_uid = make_kuid(old->user_ns, 0);
 732                        if (uid_eq(old->fsuid, root_uid) && !uid_eq(new->fsuid, root_uid))
 733                                new->cap_effective =
 734                                        cap_drop_fs_set(new->cap_effective);
 735
 736                        if (!uid_eq(old->fsuid, root_uid) && uid_eq(new->fsuid, root_uid))
 737                                new->cap_effective =
 738                                        cap_raise_fs_set(new->cap_effective,
 739                                                         new->cap_permitted);
 740                }
 741                break;
 742
 743        default:
 744                return -EINVAL;
 745        }
 746
 747        return 0;
 748}
 749
 750/*
 751 * Rationale: code calling task_setscheduler, task_setioprio, and
 752 * task_setnice, assumes that
 753 *   . if capable(cap_sys_nice), then those actions should be allowed
 754 *   . if not capable(cap_sys_nice), but acting on your own processes,
 755 *      then those actions should be allowed
 756 * This is insufficient now since you can call code without suid, but
 757 * yet with increased caps.
 758 * So we check for increased caps on the target process.
 759 */
 760static int cap_safe_nice(struct task_struct *p)
 761{
 762        int is_subset;
 763
 764        rcu_read_lock();
 765        is_subset = cap_issubset(__task_cred(p)->cap_permitted,
 766                                 current_cred()->cap_permitted);
 767        rcu_read_unlock();
 768
 769        if (!is_subset && !capable(CAP_SYS_NICE))
 770                return -EPERM;
 771        return 0;
 772}
 773
 774/**
 775 * cap_task_setscheduler - Detemine if scheduler policy change is permitted
 776 * @p: The task to affect
 777 *
 778 * Detemine if the requested scheduler policy change is permitted for the
 779 * specified task, returning 0 if permission is granted, -ve if denied.
 780 */
 781int cap_task_setscheduler(struct task_struct *p)
 782{
 783        return cap_safe_nice(p);
 784}
 785
 786/**
 787 * cap_task_ioprio - Detemine if I/O priority change is permitted
 788 * @p: The task to affect
 789 * @ioprio: The I/O priority to set
 790 *
 791 * Detemine if the requested I/O priority change is permitted for the specified
 792 * task, returning 0 if permission is granted, -ve if denied.
 793 */
 794int cap_task_setioprio(struct task_struct *p, int ioprio)
 795{
 796        return cap_safe_nice(p);
 797}
 798
 799/**
 800 * cap_task_ioprio - Detemine if task priority change is permitted
 801 * @p: The task to affect
 802 * @nice: The nice value to set
 803 *
 804 * Detemine if the requested task priority change is permitted for the
 805 * specified task, returning 0 if permission is granted, -ve if denied.
 806 */
 807int cap_task_setnice(struct task_struct *p, int nice)
 808{
 809        return cap_safe_nice(p);
 810}
 811
 812/*
 813 * Implement PR_CAPBSET_DROP.  Attempt to remove the specified capability from
 814 * the current task's bounding set.  Returns 0 on success, -ve on error.
 815 */
 816static long cap_prctl_drop(struct cred *new, unsigned long cap)
 817{
 818        if (!capable(CAP_SETPCAP))
 819                return -EPERM;
 820        if (!cap_valid(cap))
 821                return -EINVAL;
 822
 823        cap_lower(new->cap_bset, cap);
 824        return 0;
 825}
 826
 827/**
 828 * cap_task_prctl - Implement process control functions for this security module
 829 * @option: The process control function requested
 830 * @arg2, @arg3, @arg4, @arg5: The argument data for this function
 831 *
 832 * Allow process control functions (sys_prctl()) to alter capabilities; may
 833 * also deny access to other functions not otherwise implemented here.
 834 *
 835 * Returns 0 or +ve on success, -ENOSYS if this function is not implemented
 836 * here, other -ve on error.  If -ENOSYS is returned, sys_prctl() and other LSM
 837 * modules will consider performing the function.
 838 */
 839int cap_task_prctl(int option, unsigned long arg2, unsigned long arg3,
 840                   unsigned long arg4, unsigned long arg5)
 841{
 842        struct cred *new;
 843        long error = 0;
 844
 845        new = prepare_creds();
 846        if (!new)
 847                return -ENOMEM;
 848
 849        switch (option) {
 850        case PR_CAPBSET_READ:
 851                error = -EINVAL;
 852                if (!cap_valid(arg2))
 853                        goto error;
 854                error = !!cap_raised(new->cap_bset, arg2);
 855                goto no_change;
 856
 857        case PR_CAPBSET_DROP:
 858                error = cap_prctl_drop(new, arg2);
 859                if (error < 0)
 860                        goto error;
 861                goto changed;
 862
 863        /*
 864         * The next four prctl's remain to assist with transitioning a
 865         * system from legacy UID=0 based privilege (when filesystem
 866         * capabilities are not in use) to a system using filesystem
 867         * capabilities only - as the POSIX.1e draft intended.
 868         *
 869         * Note:
 870         *
 871         *  PR_SET_SECUREBITS =
 872         *      issecure_mask(SECURE_KEEP_CAPS_LOCKED)
 873         *    | issecure_mask(SECURE_NOROOT)
 874         *    | issecure_mask(SECURE_NOROOT_LOCKED)
 875         *    | issecure_mask(SECURE_NO_SETUID_FIXUP)
 876         *    | issecure_mask(SECURE_NO_SETUID_FIXUP_LOCKED)
 877         *
 878         * will ensure that the current process and all of its
 879         * children will be locked into a pure
 880         * capability-based-privilege environment.
 881         */
 882        case PR_SET_SECUREBITS:
 883                error = -EPERM;
 884                if ((((new->securebits & SECURE_ALL_LOCKS) >> 1)
 885                     & (new->securebits ^ arg2))                        /*[1]*/
 886                    || ((new->securebits & SECURE_ALL_LOCKS & ~arg2))   /*[2]*/
 887                    || (arg2 & ~(SECURE_ALL_LOCKS | SECURE_ALL_BITS))   /*[3]*/
 888                    || (cap_capable(current_cred(),
 889                                    current_cred()->user_ns, CAP_SETPCAP,
 890                                    SECURITY_CAP_AUDIT) != 0)           /*[4]*/
 891                        /*
 892                         * [1] no changing of bits that are locked
 893                         * [2] no unlocking of locks
 894                         * [3] no setting of unsupported bits
 895                         * [4] doing anything requires privilege (go read about
 896                         *     the "sendmail capabilities bug")
 897                         */
 898                    )
 899                        /* cannot change a locked bit */
 900                        goto error;
 901                new->securebits = arg2;
 902                goto changed;
 903
 904        case PR_GET_SECUREBITS:
 905                error = new->securebits;
 906                goto no_change;
 907
 908        case PR_GET_KEEPCAPS:
 909                if (issecure(SECURE_KEEP_CAPS))
 910                        error = 1;
 911                goto no_change;
 912
 913        case PR_SET_KEEPCAPS:
 914                error = -EINVAL;
 915                if (arg2 > 1) /* Note, we rely on arg2 being unsigned here */
 916                        goto error;
 917                error = -EPERM;
 918                if (issecure(SECURE_KEEP_CAPS_LOCKED))
 919                        goto error;
 920                if (arg2)
 921                        new->securebits |= issecure_mask(SECURE_KEEP_CAPS);
 922                else
 923                        new->securebits &= ~issecure_mask(SECURE_KEEP_CAPS);
 924                goto changed;
 925
 926        default:
 927                /* No functionality available - continue with default */
 928                error = -ENOSYS;
 929                goto error;
 930        }
 931
 932        /* Functionality provided */
 933changed:
 934        return commit_creds(new);
 935
 936no_change:
 937error:
 938        abort_creds(new);
 939        return error;
 940}
 941
 942/**
 943 * cap_vm_enough_memory - Determine whether a new virtual mapping is permitted
 944 * @mm: The VM space in which the new mapping is to be made
 945 * @pages: The size of the mapping
 946 *
 947 * Determine whether the allocation of a new virtual mapping by the current
 948 * task is permitted, returning 0 if permission is granted, -ve if not.
 949 */
 950int cap_vm_enough_memory(struct mm_struct *mm, long pages)
 951{
 952        int cap_sys_admin = 0;
 953
 954        if (cap_capable(current_cred(), &init_user_ns, CAP_SYS_ADMIN,
 955                        SECURITY_CAP_NOAUDIT) == 0)
 956                cap_sys_admin = 1;
 957        return __vm_enough_memory(mm, pages, cap_sys_admin);
 958}
 959
 960/*
 961 * cap_mmap_addr - check if able to map given addr
 962 * @addr: address attempting to be mapped
 963 *
 964 * If the process is attempting to map memory below dac_mmap_min_addr they need
 965 * CAP_SYS_RAWIO.  The other parameters to this function are unused by the
 966 * capability security module.  Returns 0 if this mapping should be allowed
 967 * -EPERM if not.
 968 */
 969int cap_mmap_addr(unsigned long addr)
 970{
 971        int ret = 0;
 972
 973        if (addr < dac_mmap_min_addr) {
 974                ret = cap_capable(current_cred(), &init_user_ns, CAP_SYS_RAWIO,
 975                                  SECURITY_CAP_AUDIT);
 976                /* set PF_SUPERPRIV if it turns out we allow the low mmap */
 977                if (ret == 0)
 978                        current->flags |= PF_SUPERPRIV;
 979        }
 980        return ret;
 981}
 982
 983int cap_mmap_file(struct file *file, unsigned long reqprot,
 984                  unsigned long prot, unsigned long flags)
 985{
 986        return 0;
 987}
 988
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.