kernel/capability.c

   1 /*
   2  * linux/kernel/capability.c
   3  *
   4  * Copyright (C) 1997  Andrew Main <zefram@fysh.org>
   5  *
   6  * Integrated into 2.1.97+,  Andrew G. Morgan <morgan@kernel.org>
   7  * 30 May 2002: Cleanup, Robert M. Love <rml@tech9.net>
   8  */
   9
  10 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  11
  12 #include <linux/audit.h>
  13 #include <linux/capability.h>
  14 #include <linux/mm.h>
  15 #include <linux/export.h>
  16 #include <linux/security.h>
  17 #include <linux/syscalls.h>
  18 #include <linux/pid_namespace.h>
  19 #include <linux/user_namespace.h>
  20 #include <asm/uaccess.h>
  21
  22 /*
  23  * Leveraged for setting/resetting capabilities
  24  */
  25
  26 const kernel_cap_t __cap_empty_set = CAP_EMPTY_SET;
  27 EXPORT_SYMBOL(__cap_empty_set);
  28
  29 int file_caps_enabled = 1;
  30
  31 static int __init file_caps_disable(char *str)
  32 {
  33         file_caps_enabled = 0;
  34         return 1;
  35 }
  36 __setup("no_file_caps", file_caps_disable);
  37
  38 #ifdef CONFIG_MULTIUSER
  39 /*
  40  * More recent versions of libcap are available from:
  41  *
  42  *   http://www.kernel.org/pub/linux/libs/security/linux-privs/
  43  */
  44
  45 static void warn_legacy_capability_use(void)
  46 {
  47         char name[sizeof(current->comm)];
  48
  49         pr_info_once("warning: `%s' uses 32-bit capabilities (legacy support in use)\n",
  50                      get_task_comm(name, current));
  51 }
  52
  53 /*
  54  * Version 2 capabilities worked fine, but the linux/capability.h file
  55  * that accompanied their introduction encouraged their use without
  56  * the necessary user-space source code changes. As such, we have
  57  * created a version 3 with equivalent functionality to version 2, but
  58  * with a header change to protect legacy source code from using
  59  * version 2 when it wanted to use version 1. If your system has code
  60  * that trips the following warning, it is using version 2 specific
  61  * capabilities and may be doing so insecurely.
  62  *
  63  * The remedy is to either upgrade your version of libcap (to 2.10+,
  64  * if the application is linked against it), or recompile your
  65  * application with modern kernel headers and this warning will go
  66  * away.
  67  */
  68
  69 static void warn_deprecated_v2(void)
  70 {
  71         char name[sizeof(current->comm)];
  72
  73         pr_info_once("warning: `%s' uses deprecated v2 capabilities in a way that may be insecure\n",
  74                      get_task_comm(name, current));
  75 }
  76
  77 /*
  78  * Version check. Return the number of u32s in each capability flag
  79  * array, or a negative value on error.
  80  */
  81 static int cap_validate_magic(cap_user_header_t header, unsigned *tocopy)
  82 {
  83         __u32 version;
  84
  85         if (get_user(version, &header->version))
  86                 return -EFAULT;
  87
  88         switch (version) {
  89         case _LINUX_CAPABILITY_VERSION_1:
  90                 warn_legacy_capability_use();
  91                 *tocopy = _LINUX_CAPABILITY_U32S_1;
  92                 break;
  93         case _LINUX_CAPABILITY_VERSION_2:
  94                 warn_deprecated_v2();
  95                 /*
  96                  * fall through - v3 is otherwise equivalent to v2.
  97                  */
  98         case _LINUX_CAPABILITY_VERSION_3:
  99                 *tocopy = _LINUX_CAPABILITY_U32S_3;
 100                 break;
 101         default:
 102                 if (put_user((u32)_KERNEL_CAPABILITY_VERSION, &header->version))
 103                         return -EFAULT;
 104                 return -EINVAL;
 105         }
 106
 107         return 0;
 108 }
 109
 110 /*
 111  * The only thing that can change the capabilities of the current
 112  * process is the current process. As such, we can't be in this code
 113  * at the same time as we are in the process of setting capabilities
 114  * in this process. The net result is that we can limit our use of
 115  * locks to when we are reading the caps of another process.
 116  */
 117 static inline int cap_get_target_pid(pid_t pid, kernel_cap_t *pEp,
 118                                      kernel_cap_t *pIp, kernel_cap_t *pPp)
 119 {
 120         int ret;
 121
 122         if (pid && (pid != task_pid_vnr(current))) {
 123                 struct task_struct *target;
 124
 125                 rcu_read_lock();
 126
 127                 target = find_task_by_vpid(pid);
 128                 if (!target)
 129                         ret = -ESRCH;
 130                 else
 131                         ret = security_capget(target, pEp, pIp, pPp);
 132
 133                 rcu_read_unlock();
 134         } else
 135                 ret = security_capget(current, pEp, pIp, pPp);
 136
 137         return ret;
 138 }
 139
 140 /**
 141  * sys_capget - get the capabilities of a given process.
 142  * @header: pointer to struct that contains capability version and
 143  *      target pid data
 144  * @dataptr: pointer to struct that contains the effective, permitted,
 145  *      and inheritable capabilities that are returned
 146  *
 147  * Returns 0 on success and < 0 on error.
 148  */
 149 SYSCALL_DEFINE2(capget, cap_user_header_t, header, cap_user_data_t, dataptr)
 150 {
 151         int ret = 0;
 152         pid_t pid;
 153         unsigned tocopy;
 154         kernel_cap_t pE, pI, pP;
 155
 156         ret = cap_validate_magic(header, &tocopy);
 157         if ((dataptr == NULL) || (ret != 0))
 158                 return ((dataptr == NULL) && (ret == -EINVAL)) ? 0 : ret;
 159
 160         if (get_user(pid, &header->pid))
 161                 return -EFAULT;
 162
 163         if (pid < 0)
 164                 return -EINVAL;
 165
 166         ret = cap_get_target_pid(pid, &pE, &pI, &pP);
 167         if (!ret) {
 168                 struct __user_cap_data_struct kdata[_KERNEL_CAPABILITY_U32S];
 169                 unsigned i;
 170
 171                 for (i = 0; i < tocopy; i++) {
 172                         kdata[i].effective = pE.cap[i];
 173                         kdata[i].permitted = pP.cap[i];
 174                         kdata[i].inheritable = pI.cap[i];
 175                 }
 176
 177                 /*
 178                  * Note, in the case, tocopy < _KERNEL_CAPABILITY_U32S,
 179                  * we silently drop the upper capabilities here. This
 180                  * has the effect of making older libcap
 181                  * implementations implicitly drop upper capability
 182                  * bits when they perform a: capget/modify/capset
 183                  * sequence.
 184                  *
 185                  * This behavior is considered fail-safe
 186                  * behavior. Upgrading the application to a newer
 187                  * version of libcap will enable access to the newer
 188                  * capabilities.
 189                  *
 190                  * An alternative would be to return an error here
 191                  * (-ERANGE), but that causes legacy applications to
 192                  * unexpectedly fail; the capget/modify/capset aborts
 193                  * before modification is attempted and the application
 194                  * fails.
 195                  */
 196                 if (copy_to_user(dataptr, kdata, tocopy
 197                                  * sizeof(struct __user_cap_data_struct))) {
 198                         return -EFAULT;
 199                 }
 200         }
 201
 202         return ret;
 203 }
 204
 205 /**
 206  * sys_capset - set capabilities for a process or (*) a group of processes
 207  * @header: pointer to struct that contains capability version and
 208  *      target pid data
 209  * @data: pointer to struct that contains the effective, permitted,
 210  *      and inheritable capabilities
 211  *
 212  * Set capabilities for the current process only.  The ability to any other
 213  * process(es) has been deprecated and removed.
 214  *
 215  * The restrictions on setting capabilities are specified as:
 216  *
 217  * I: any raised capabilities must be a subset of the old permitted
 218  * P: any raised capabilities must be a subset of the old permitted
 219  * E: must be set to a subset of new permitted
 220  *
 221  * Returns 0 on success and < 0 on error.
 222  */
 223 SYSCALL_DEFINE2(capset, cap_user_header_t, header, const cap_user_data_t, data)
 224 {
 225         struct __user_cap_data_struct kdata[_KERNEL_CAPABILITY_U32S];
 226         unsigned i, tocopy, copybytes;
 227         kernel_cap_t inheritable, permitted, effective;
 228         struct cred *new;
 229         int ret;
 230         pid_t pid;
 231
 232         ret = cap_validate_magic(header, &tocopy);
 233         if (ret != 0)
 234                 return ret;
 235
 236         if (get_user(pid, &header->pid))
 237                 return -EFAULT;
 238
 239         /* may only affect current now */
 240         if (pid != 0 && pid != task_pid_vnr(current))
 241                 return -EPERM;
 242
 243         copybytes = tocopy * sizeof(struct __user_cap_data_struct);
 244         if (copybytes > sizeof(kdata))
 245                 return -EFAULT;
 246
 247         if (copy_from_user(&kdata, data, copybytes))
 248                 return -EFAULT;
 249
 250         for (i = 0; i < tocopy; i++) {
 251                 effective.cap[i] = kdata[i].effective;
 252                 permitted.cap[i] = kdata[i].permitted;
 253                 inheritable.cap[i] = kdata[i].inheritable;
 254         }
 255         while (i < _KERNEL_CAPABILITY_U32S) {
 256                 effective.cap[i] = 0;
 257                 permitted.cap[i] = 0;
 258                 inheritable.cap[i] = 0;
 259                 i++;
 260         }
 261
 262         effective.cap[CAP_LAST_U32] &= CAP_LAST_U32_VALID_MASK;
 263         permitted.cap[CAP_LAST_U32] &= CAP_LAST_U32_VALID_MASK;
 264         inheritable.cap[CAP_LAST_U32] &= CAP_LAST_U32_VALID_MASK;
 265
 266         new = prepare_creds();
 267         if (!new)
 268                 return -ENOMEM;
 269
 270         ret = security_capset(new, current_cred(),
 271                               &effective, &inheritable, &permitted);
 272         if (ret < 0)
 273                 goto error;
 274
 275         audit_log_capset(new, current_cred());
 276
 277         return commit_creds(new);
 278
 279 error:
 280         abort_creds(new);
 281         return ret;
 282 }
 283
 284 /**
 285  * has_ns_capability - Does a task have a capability in a specific user ns
 286  * @t: The task in question
 287  * @ns: target user namespace
 288  * @cap: The capability to be tested for
 289  *
 290  * Return true if the specified task has the given superior capability
 291  * currently in effect to the specified user namespace, false if not.
 292  *
 293  * Note that this does not set PF_SUPERPRIV on the task.
 294  */
 295 bool has_ns_capability(struct task_struct *t,
 296                        struct user_namespace *ns, int cap)
 297 {
 298         int ret;
 299
 300         rcu_read_lock();
 301         ret = security_capable(__task_cred(t), ns, cap);
 302         rcu_read_unlock();
 303
 304         return (ret == 0);
 305 }
 306
 307 /**
 308  * has_capability - Does a task have a capability in init_user_ns
 309  * @t: The task in question
 310  * @cap: The capability to be tested for
 311  *
 312  * Return true if the specified task has the given superior capability
 313  * currently in effect to the initial user namespace, false if not.
 314  *
 315  * Note that this does not set PF_SUPERPRIV on the task.
 316  */
 317 bool has_capability(struct task_struct *t, int cap)
 318 {
 319         return has_ns_capability(t, &init_user_ns, cap);
 320 }
 321
 322 /**
 323  * has_ns_capability_noaudit - Does a task have a capability (unaudited)
 324  * in a specific user ns.
 325  * @t: The task in question
 326  * @ns: target user namespace
 327  * @cap: The capability to be tested for
 328  *
 329  * Return true if the specified task has the given superior capability
 330  * currently in effect to the specified user namespace, false if not.
 331  * Do not write an audit message for the check.
 332  *
 333  * Note that this does not set PF_SUPERPRIV on the task.
 334  */
 335 bool has_ns_capability_noaudit(struct task_struct *t,
 336                                struct user_namespace *ns, int cap)
 337 {
 338         int ret;
 339
 340         rcu_read_lock();
 341         ret = security_capable_noaudit(__task_cred(t), ns, cap);
 342         rcu_read_unlock();
 343
 344         return (ret == 0);
 345 }
 346
 347 /**
 348  * has_capability_noaudit - Does a task have a capability (unaudited) in the
 349  * initial user ns
 350  * @t: The task in question
 351  * @cap: The capability to be tested for
 352  *
 353  * Return true if the specified task has the given superior capability
 354  * currently in effect to init_user_ns, false if not.  Don't write an
 355  * audit message for the check.
 356  *
 357  * Note that this does not set PF_SUPERPRIV on the task.
 358  */
 359 bool has_capability_noaudit(struct task_struct *t, int cap)
 360 {
 361         return has_ns_capability_noaudit(t, &init_user_ns, cap);
 362 }
 363
 364 static bool ns_capable_common(struct user_namespace *ns, int cap, bool audit)
 365 {
 366         int capable;
 367
 368         if (unlikely(!cap_valid(cap))) {
 369                 pr_crit("capable() called with invalid cap=%u\n", cap);
 370                 BUG();
 371         }
 372
 373         capable = audit ? security_capable(current_cred(), ns, cap) :
 374                           security_capable_noaudit(current_cred(), ns, cap);
 375         if (capable == 0) {
 376                 current->flags |= PF_SUPERPRIV;
 377                 return true;
 378         }
 379         return false;
 380 }
 381
 382 /**
 383  * ns_capable - Determine if the current task has a superior capability in effect
 384  * @ns:  The usernamespace we want the capability in
 385  * @cap: The capability to be tested for
 386  *
 387  * Return true if the current task has the given superior capability currently
 388  * available for use, false if not.
 389  *
 390  * This sets PF_SUPERPRIV on the task if the capability is available on the
 391  * assumption that it's about to be used.
 392  */
 393 bool ns_capable(struct user_namespace *ns, int cap)
 394 {
 395         return ns_capable_common(ns, cap, true);
 396 }
 397 EXPORT_SYMBOL(ns_capable);
 398
 399 /**
 400  * ns_capable_noaudit - Determine if the current task has a superior capability
 401  * (unaudited) in effect
 402  * @ns:  The usernamespace we want the capability in
 403  * @cap: The capability to be tested for
 404  *
 405  * Return true if the current task has the given superior capability currently
 406  * available for use, false if not.
 407  *
 408  * This sets PF_SUPERPRIV on the task if the capability is available on the
 409  * assumption that it's about to be used.
 410  */
 411 bool ns_capable_noaudit(struct user_namespace *ns, int cap)
 412 {
 413         return ns_capable_common(ns, cap, false);
 414 }
 415 EXPORT_SYMBOL(ns_capable_noaudit);
 416
 417 /**
 418  * capable - Determine if the current task has a superior capability in effect
 419  * @cap: The capability to be tested for
 420  *
 421  * Return true if the current task has the given superior capability currently
 422  * available for use, false if not.
 423  *
 424  * This sets PF_SUPERPRIV on the task if the capability is available on the
 425  * assumption that it's about to be used.
 426  */
 427 bool capable(int cap)
 428 {
 429         return ns_capable(&init_user_ns, cap);
 430 }
 431 EXPORT_SYMBOL(capable);
 432 #endif /* CONFIG_MULTIUSER */
 433
 434 /**
 435  * file_ns_capable - Determine if the file's opener had a capability in effect
 436  * @file:  The file we want to check
 437  * @ns:  The usernamespace we want the capability in
 438  * @cap: The capability to be tested for
 439  *
 440  * Return true if task that opened the file had a capability in effect
 441  * when the file was opened.
 442  *
 443  * This does not set PF_SUPERPRIV because the caller may not
 444  * actually be privileged.
 445  */
 446 bool file_ns_capable(const struct file *file, struct user_namespace *ns,
 447                      int cap)
 448 {
 449         if (WARN_ON_ONCE(!cap_valid(cap)))
 450                 return false;
 451
 452         if (security_capable(file->f_cred, ns, cap) == 0)
 453                 return true;
 454
 455         return false;
 456 }
 457 EXPORT_SYMBOL(file_ns_capable);
 458
 459 /**
 460  * privileged_wrt_inode_uidgid - Do capabilities in the namespace work over the inode?
 461  * @ns: The user namespace in question
 462  * @inode: The inode in question
 463  *
 464  * Return true if the inode uid and gid are within the namespace.
 465  */
 466 bool privileged_wrt_inode_uidgid(struct user_namespace *ns, const struct inode *inode)
 467 {
 468         return kuid_has_mapping(ns, inode->i_uid) &&
 469                 kgid_has_mapping(ns, inode->i_gid);
 470 }
 471
 472 /**
 473  * capable_wrt_inode_uidgid - Check nsown_capable and uid and gid mapped
 474  * @inode: The inode in question
 475  * @cap: The capability in question
 476  *
 477  * Return true if the current task has the given capability targeted at
 478  * its own user namespace and that the given inode's uid and gid are
 479  * mapped into the current user namespace.
 480  */
 481 bool capable_wrt_inode_uidgid(const struct inode *inode, int cap)
 482 {
 483         struct user_namespace *ns = current_user_ns();
 484
 485         return ns_capable(ns, cap) && privileged_wrt_inode_uidgid(ns, inode);
 486 }
 487 EXPORT_SYMBOL(capable_wrt_inode_uidgid);
 488
 489 /**
 490  * ptracer_capable - Determine if the ptracer holds CAP_SYS_PTRACE in the namespace
 491  * @tsk: The task that may be ptraced
 492  * @ns: The user namespace to search for CAP_SYS_PTRACE in
 493  *
 494  * Return true if the task that is ptracing the current task had CAP_SYS_PTRACE
 495  * in the specified user namespace.
 496  */
 497 bool ptracer_capable(struct task_struct *tsk, struct user_namespace *ns)
 498 {
 499         int ret = 0;  /* An absent tracer adds no restrictions */
 500         const struct cred *cred;
 501         rcu_read_lock();
 502         cred = rcu_dereference(tsk->ptracer_cred);
 503         if (cred)
 504                 ret = security_capable_noaudit(cred, ns, CAP_SYS_PTRACE);
 505         rcu_read_unlock();
 506         return (ret == 0);
 507 }