kernel/sys.c

   1 /*
   2  *  linux/kernel/sys.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6
   7 #include <linux/module.h>
   8 #include <linux/mm.h>
   9 #include <linux/utsname.h>
  10 #include <linux/mman.h>
  11 #include <linux/notifier.h>
  12 #include <linux/reboot.h>
  13 #include <linux/prctl.h>
  14 #include <linux/highuid.h>
  15 #include <linux/fs.h>
  16 #include <linux/perf_event.h>
  17 #include <linux/resource.h>
  18 #include <linux/kernel.h>
  19 #include <linux/kexec.h>
  20 #include <linux/workqueue.h>
  21 #include <linux/capability.h>
  22 #include <linux/device.h>
  23 #include <linux/key.h>
  24 #include <linux/times.h>
  25 #include <linux/posix-timers.h>
  26 #include <linux/security.h>
  27 #include <linux/dcookies.h>
  28 #include <linux/suspend.h>
  29 #include <linux/tty.h>
  30 #include <linux/signal.h>
  31 #include <linux/cn_proc.h>
  32 #include <linux/getcpu.h>
  33 #include <linux/task_io_accounting_ops.h>
  34 #include <linux/seccomp.h>
  35 #include <linux/cpu.h>
  36 #include <linux/ptrace.h>
  37 #include <linux/fs_struct.h>
  38
  39 #include <linux/compat.h>
  40 #include <linux/syscalls.h>
  41 #include <linux/kprobes.h>
  42 #include <linux/user_namespace.h>
  43
  44 #include <asm/uaccess.h>
  45 #include <asm/io.h>
  46 #include <asm/unistd.h>
  47
  48 #ifndef SET_UNALIGN_CTL
  49 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  50 #endif
  51 #ifndef GET_UNALIGN_CTL
  52 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  53 #endif
  54 #ifndef SET_FPEMU_CTL
  55 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
  56 #endif
  57 #ifndef GET_FPEMU_CTL
  58 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
  59 #endif
  60 #ifndef SET_FPEXC_CTL
  61 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
  62 #endif
  63 #ifndef GET_FPEXC_CTL
  64 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
  65 #endif
  66 #ifndef GET_ENDIAN
  67 # define GET_ENDIAN(a,b)        (-EINVAL)
  68 #endif
  69 #ifndef SET_ENDIAN
  70 # define SET_ENDIAN(a,b)        (-EINVAL)
  71 #endif
  72 #ifndef GET_TSC_CTL
  73 # define GET_TSC_CTL(a)         (-EINVAL)
  74 #endif
  75 #ifndef SET_TSC_CTL
  76 # define SET_TSC_CTL(a)         (-EINVAL)
  77 #endif
  78
  79 /*
  80  * this is where the system-wide overflow UID and GID are defined, for
  81  * architectures that now have 32-bit UID/GID but didn't in the past
  82  */
  83
  84 int overflowuid = DEFAULT_OVERFLOWUID;
  85 int overflowgid = DEFAULT_OVERFLOWGID;
  86
  87 #ifdef CONFIG_UID16
  88 EXPORT_SYMBOL(overflowuid);
  89 EXPORT_SYMBOL(overflowgid);
  90 #endif
  91
  92 /*
  93  * the same as above, but for filesystems which can only store a 16-bit
  94  * UID and GID. as such, this is needed on all architectures
  95  */
  96
  97 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
  98 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
  99
 100 EXPORT_SYMBOL(fs_overflowuid);
 101 EXPORT_SYMBOL(fs_overflowgid);
 102
 103 /*
 104  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
 105  */
 106
 107 int C_A_D = 1;
 108 struct pid *cad_pid;
 109 EXPORT_SYMBOL(cad_pid);
 110
 111 /*
 112  * If set, this is used for preparing the system to power off.
 113  */
 114
 115 void (*pm_power_off_prepare)(void);
 116
 117 /*
 118  * set the priority of a task
 119  * - the caller must hold the RCU read lock
 120  */
 121 static int set_one_prio(struct task_struct *p, int niceval, int error)
 122 {
 123         const struct cred *cred = current_cred(), *pcred = __task_cred(p);
 124         int no_nice;
 125
 126         if (pcred->uid  != cred->euid &&
 127             pcred->euid != cred->euid && !capable(CAP_SYS_NICE)) {
 128                 error = -EPERM;
 129                 goto out;
 130         }
 131         if (niceval < task_nice(p) && !can_nice(p, niceval)) {
 132                 error = -EACCES;
 133                 goto out;
 134         }
 135         no_nice = security_task_setnice(p, niceval);
 136         if (no_nice) {
 137                 error = no_nice;
 138                 goto out;
 139         }
 140         if (error == -ESRCH)
 141                 error = 0;
 142         set_user_nice(p, niceval);
 143 out:
 144         return error;
 145 }
 146
 147 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
 148 {
 149         struct task_struct *g, *p;
 150         struct user_struct *user;
 151         const struct cred *cred = current_cred();
 152         int error = -EINVAL;
 153         struct pid *pgrp;
 154
 155         if (which > PRIO_USER || which < PRIO_PROCESS)
 156                 goto out;
 157
 158         /* normalize: avoid signed division (rounding problems) */
 159         error = -ESRCH;
 160         if (niceval < -20)
 161                 niceval = -20;
 162         if (niceval > 19)
 163                 niceval = 19;
 164
 165         rcu_read_lock();
 166         read_lock(&tasklist_lock);
 167         switch (which) {
 168                 case PRIO_PROCESS:
 169                         if (who)
 170                                 p = find_task_by_vpid(who);
 171                         else
 172                                 p = current;
 173                         if (p)
 174                                 error = set_one_prio(p, niceval, error);
 175                         break;
 176                 case PRIO_PGRP:
 177                         if (who)
 178                                 pgrp = find_vpid(who);
 179                         else
 180                                 pgrp = task_pgrp(current);
 181                         do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
 182                                 error = set_one_prio(p, niceval, error);
 183                         } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
 184                         break;
 185                 case PRIO_USER:
 186                         user = (struct user_struct *) cred->user;
 187                         if (!who)
 188                                 who = cred->uid;
 189                         else if ((who != cred->uid) &&
 190                                  !(user = find_user(who)))
 191                                 goto out_unlock;        /* No processes for this user */
 192
 193                         do_each_thread(g, p) {
 194                                 if (__task_cred(p)->uid == who)
 195                                         error = set_one_prio(p, niceval, error);
 196                         } while_each_thread(g, p);
 197                         if (who != cred->uid)
 198                                 free_uid(user);         /* For find_user() */
 199                         break;
 200         }
 201 out_unlock:
 202         read_unlock(&tasklist_lock);
 203         rcu_read_unlock();
 204 out:
 205         return error;
 206 }
 207
 208 /*
 209  * Ugh. To avoid negative return values, "getpriority()" will
 210  * not return the normal nice-value, but a negated value that
 211  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 212  * to stay compatible.
 213  */
 214 SYSCALL_DEFINE2(getpriority, int, which, int, who)
 215 {
 216         struct task_struct *g, *p;
 217         struct user_struct *user;
 218         const struct cred *cred = current_cred();
 219         long niceval, retval = -ESRCH;
 220         struct pid *pgrp;
 221
 222         if (which > PRIO_USER || which < PRIO_PROCESS)
 223                 return -EINVAL;
 224
 225         read_lock(&tasklist_lock);
 226         switch (which) {
 227                 case PRIO_PROCESS:
 228                         if (who)
 229                                 p = find_task_by_vpid(who);
 230                         else
 231                                 p = current;
 232                         if (p) {
 233                                 niceval = 20 - task_nice(p);
 234                                 if (niceval > retval)
 235                                         retval = niceval;
 236                         }
 237                         break;
 238                 case PRIO_PGRP:
 239                         if (who)
 240                                 pgrp = find_vpid(who);
 241                         else
 242                                 pgrp = task_pgrp(current);
 243                         do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
 244                                 niceval = 20 - task_nice(p);
 245                                 if (niceval > retval)
 246                                         retval = niceval;
 247                         } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
 248                         break;
 249                 case PRIO_USER:
 250                         user = (struct user_struct *) cred->user;
 251                         if (!who)
 252                                 who = cred->uid;
 253                         else if ((who != cred->uid) &&
 254                                  !(user = find_user(who)))
 255                                 goto out_unlock;        /* No processes for this user */
 256
 257                         do_each_thread(g, p) {
 258                                 if (__task_cred(p)->uid == who) {
 259                                         niceval = 20 - task_nice(p);
 260                                         if (niceval > retval)
 261                                                 retval = niceval;
 262                                 }
 263                         } while_each_thread(g, p);
 264                         if (who != cred->uid)
 265                                 free_uid(user);         /* for find_user() */
 266                         break;
 267         }
 268 out_unlock:
 269         read_unlock(&tasklist_lock);
 270
 271         return retval;
 272 }
 273
 274 /**
 275  *      emergency_restart - reboot the system
 276  *
 277  *      Without shutting down any hardware or taking any locks
 278  *      reboot the system.  This is called when we know we are in
 279  *      trouble so this is our best effort to reboot.  This is
 280  *      safe to call in interrupt context.
 281  */
 282 void emergency_restart(void)
 283 {
 284         machine_emergency_restart();
 285 }
 286 EXPORT_SYMBOL_GPL(emergency_restart);
 287
 288 void kernel_restart_prepare(char *cmd)
 289 {
 290         blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
 291         system_state = SYSTEM_RESTART;
 292         device_shutdown();
 293         sysdev_shutdown();
 294 }
 295
 296 /**
 297  *      kernel_restart - reboot the system
 298  *      @cmd: pointer to buffer containing command to execute for restart
 299  *              or %NULL
 300  *
 301  *      Shutdown everything and perform a clean reboot.
 302  *      This is not safe to call in interrupt context.
 303  */
 304 void kernel_restart(char *cmd)
 305 {
 306         kernel_restart_prepare(cmd);
 307         if (!cmd)
 308                 printk(KERN_EMERG "Restarting system.\n");
 309         else
 310                 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
 311         machine_restart(cmd);
 312 }
 313 EXPORT_SYMBOL_GPL(kernel_restart);
 314
 315 static void kernel_shutdown_prepare(enum system_states state)
 316 {
 317         blocking_notifier_call_chain(&reboot_notifier_list,
 318                 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
 319         system_state = state;
 320         device_shutdown();
 321 }
 322 /**
 323  *      kernel_halt - halt the system
 324  *
 325  *      Shutdown everything and perform a clean system halt.
 326  */
 327 void kernel_halt(void)
 328 {
 329         kernel_shutdown_prepare(SYSTEM_HALT);
 330         sysdev_shutdown();
 331         printk(KERN_EMERG "System halted.\n");
 332         machine_halt();
 333 }
 334
 335 EXPORT_SYMBOL_GPL(kernel_halt);
 336
 337 /**
 338  *      kernel_power_off - power_off the system
 339  *
 340  *      Shutdown everything and perform a clean system power_off.
 341  */
 342 void kernel_power_off(void)
 343 {
 344         kernel_shutdown_prepare(SYSTEM_POWER_OFF);
 345         if (pm_power_off_prepare)
 346                 pm_power_off_prepare();
 347         disable_nonboot_cpus();
 348         sysdev_shutdown();
 349         printk(KERN_EMERG "Power down.\n");
 350         machine_power_off();
 351 }
 352 EXPORT_SYMBOL_GPL(kernel_power_off);
 353
 354 static DEFINE_MUTEX(reboot_mutex);
 355
 356 /*
 357  * Reboot system call: for obvious reasons only root may call it,
 358  * and even root needs to set up some magic numbers in the registers
 359  * so that some mistake won't make this reboot the whole machine.
 360  * You can also set the meaning of the ctrl-alt-del-key here.
 361  *
 362  * reboot doesn't sync: do that yourself before calling this.
 363  */
 364 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
 365                 void __user *, arg)
 366 {
 367         char buffer[256];
 368         int ret = 0;
 369
 370         /* We only trust the superuser with rebooting the system. */
 371         if (!capable(CAP_SYS_BOOT))
 372                 return -EPERM;
 373
 374         /* For safety, we require "magic" arguments. */
 375         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 376             (magic2 != LINUX_REBOOT_MAGIC2 &&
 377                         magic2 != LINUX_REBOOT_MAGIC2A &&
 378                         magic2 != LINUX_REBOOT_MAGIC2B &&
 379                         magic2 != LINUX_REBOOT_MAGIC2C))
 380                 return -EINVAL;
 381
 382         /* Instead of trying to make the power_off code look like
 383          * halt when pm_power_off is not set do it the easy way.
 384          */
 385         if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
 386                 cmd = LINUX_REBOOT_CMD_HALT;
 387
 388         mutex_lock(&reboot_mutex);
 389         switch (cmd) {
 390         case LINUX_REBOOT_CMD_RESTART:
 391                 kernel_restart(NULL);
 392                 break;
 393
 394         case LINUX_REBOOT_CMD_CAD_ON:
 395                 C_A_D = 1;
 396                 break;
 397
 398         case LINUX_REBOOT_CMD_CAD_OFF:
 399                 C_A_D = 0;
 400                 break;
 401
 402         case LINUX_REBOOT_CMD_HALT:
 403                 kernel_halt();
 404                 do_exit(0);
 405                 panic("cannot halt");
 406
 407         case LINUX_REBOOT_CMD_POWER_OFF:
 408                 kernel_power_off();
 409                 do_exit(0);
 410                 break;
 411
 412         case LINUX_REBOOT_CMD_RESTART2:
 413                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 414                         ret = -EFAULT;
 415                         break;
 416                 }
 417                 buffer[sizeof(buffer) - 1] = '\0';
 418
 419                 kernel_restart(buffer);
 420                 break;
 421
 422 #ifdef CONFIG_KEXEC
 423         case LINUX_REBOOT_CMD_KEXEC:
 424                 ret = kernel_kexec();
 425                 break;
 426 #endif
 427
 428 #ifdef CONFIG_HIBERNATION
 429         case LINUX_REBOOT_CMD_SW_SUSPEND:
 430                 ret = hibernate();
 431                 break;
 432 #endif
 433
 434         default:
 435                 ret = -EINVAL;
 436                 break;
 437         }
 438         mutex_unlock(&reboot_mutex);
 439         return ret;
 440 }
 441
 442 static void deferred_cad(struct work_struct *dummy)
 443 {
 444         kernel_restart(NULL);
 445 }
 446
 447 /*
 448  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 449  * As it's called within an interrupt, it may NOT sync: the only choice
 450  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 451  */
 452 void ctrl_alt_del(void)
 453 {
 454         static DECLARE_WORK(cad_work, deferred_cad);
 455
 456         if (C_A_D)
 457                 schedule_work(&cad_work);
 458         else
 459                 kill_cad_pid(SIGINT, 1);
 460 }
 461
 462 /*
 463  * Unprivileged users may change the real gid to the effective gid
 464  * or vice versa.  (BSD-style)
 465  *
 466  * If you set the real gid at all, or set the effective gid to a value not
 467  * equal to the real gid, then the saved gid is set to the new effective gid.
 468  *
 469  * This makes it possible for a setgid program to completely drop its
 470  * privileges, which is often a useful assertion to make when you are doing
 471  * a security audit over a program.
 472  *
 473  * The general idea is that a program which uses just setregid() will be
 474  * 100% compatible with BSD.  A program which uses just setgid() will be
 475  * 100% compatible with POSIX with saved IDs.
 476  *
 477  * SMP: There are not races, the GIDs are checked only by filesystem
 478  *      operations (as far as semantic preservation is concerned).
 479  */
 480 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
 481 {
 482         const struct cred *old;
 483         struct cred *new;
 484         int retval;
 485
 486         new = prepare_creds();
 487         if (!new)
 488                 return -ENOMEM;
 489         old = current_cred();
 490
 491         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 492         if (retval)
 493                 goto error;
 494
 495         retval = -EPERM;
 496         if (rgid != (gid_t) -1) {
 497                 if (old->gid == rgid ||
 498                     old->egid == rgid ||
 499                     capable(CAP_SETGID))
 500                         new->gid = rgid;
 501                 else
 502                         goto error;
 503         }
 504         if (egid != (gid_t) -1) {
 505                 if (old->gid == egid ||
 506                     old->egid == egid ||
 507                     old->sgid == egid ||
 508                     capable(CAP_SETGID))
 509                         new->egid = egid;
 510                 else
 511                         goto error;
 512         }
 513
 514         if (rgid != (gid_t) -1 ||
 515             (egid != (gid_t) -1 && egid != old->gid))
 516                 new->sgid = new->egid;
 517         new->fsgid = new->egid;
 518
 519         return commit_creds(new);
 520
 521 error:
 522         abort_creds(new);
 523         return retval;
 524 }
 525
 526 /*
 527  * setgid() is implemented like SysV w/ SAVED_IDS
 528  *
 529  * SMP: Same implicit races as above.
 530  */
 531 SYSCALL_DEFINE1(setgid, gid_t, gid)
 532 {
 533         const struct cred *old;
 534         struct cred *new;
 535         int retval;
 536
 537         new = prepare_creds();
 538         if (!new)
 539                 return -ENOMEM;
 540         old = current_cred();
 541
 542         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 543         if (retval)
 544                 goto error;
 545
 546         retval = -EPERM;
 547         if (capable(CAP_SETGID))
 548                 new->gid = new->egid = new->sgid = new->fsgid = gid;
 549         else if (gid == old->gid || gid == old->sgid)
 550                 new->egid = new->fsgid = gid;
 551         else
 552                 goto error;
 553
 554         return commit_creds(new);
 555
 556 error:
 557         abort_creds(new);
 558         return retval;
 559 }
 560
 561 /*
 562  * change the user struct in a credentials set to match the new UID
 563  */
 564 static int set_user(struct cred *new)
 565 {
 566         struct user_struct *new_user;
 567
 568         new_user = alloc_uid(current_user_ns(), new->uid);
 569         if (!new_user)
 570                 return -EAGAIN;
 571
 572         if (!task_can_switch_user(new_user, current)) {
 573                 free_uid(new_user);
 574                 return -EINVAL;
 575         }
 576
 577         if (atomic_read(&new_user->processes) >=
 578                                 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
 579                         new_user != INIT_USER) {
 580                 free_uid(new_user);
 581                 return -EAGAIN;
 582         }
 583
 584         free_uid(new->user);
 585         new->user = new_user;
 586         return 0;
 587 }
 588
 589 /*
 590  * Unprivileged users may change the real uid to the effective uid
 591  * or vice versa.  (BSD-style)
 592  *
 593  * If you set the real uid at all, or set the effective uid to a value not
 594  * equal to the real uid, then the saved uid is set to the new effective uid.
 595  *
 596  * This makes it possible for a setuid program to completely drop its
 597  * privileges, which is often a useful assertion to make when you are doing
 598  * a security audit over a program.
 599  *
 600  * The general idea is that a program which uses just setreuid() will be
 601  * 100% compatible with BSD.  A program which uses just setuid() will be
 602  * 100% compatible with POSIX with saved IDs.
 603  */
 604 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
 605 {
 606         const struct cred *old;
 607         struct cred *new;
 608         int retval;
 609
 610         new = prepare_creds();
 611         if (!new)
 612                 return -ENOMEM;
 613         old = current_cred();
 614
 615         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 616         if (retval)
 617                 goto error;
 618
 619         retval = -EPERM;
 620         if (ruid != (uid_t) -1) {
 621                 new->uid = ruid;
 622                 if (old->uid != ruid &&
 623                     old->euid != ruid &&
 624                     !capable(CAP_SETUID))
 625                         goto error;
 626         }
 627
 628         if (euid != (uid_t) -1) {
 629                 new->euid = euid;
 630                 if (old->uid != euid &&
 631                     old->euid != euid &&
 632                     old->suid != euid &&
 633                     !capable(CAP_SETUID))
 634                         goto error;
 635         }
 636
 637         if (new->uid != old->uid) {
 638                 retval = set_user(new);
 639                 if (retval < 0)
 640                         goto error;
 641         }
 642         if (ruid != (uid_t) -1 ||
 643             (euid != (uid_t) -1 && euid != old->uid))
 644                 new->suid = new->euid;
 645         new->fsuid = new->euid;
 646
 647         retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
 648         if (retval < 0)
 649                 goto error;
 650
 651         return commit_creds(new);
 652
 653 error:
 654         abort_creds(new);
 655         return retval;
 656 }
 657
 658 /*
 659  * setuid() is implemented like SysV with SAVED_IDS
 660  *
 661  * Note that SAVED_ID's is deficient in that a setuid root program
 662  * like sendmail, for example, cannot set its uid to be a normal
 663  * user and then switch back, because if you're root, setuid() sets
 664  * the saved uid too.  If you don't like this, blame the bright people
 665  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 666  * will allow a root program to temporarily drop privileges and be able to
 667  * regain them by swapping the real and effective uid.
 668  */
 669 SYSCALL_DEFINE1(setuid, uid_t, uid)
 670 {
 671         const struct cred *old;
 672         struct cred *new;
 673         int retval;
 674
 675         new = prepare_creds();
 676         if (!new)
 677                 return -ENOMEM;
 678         old = current_cred();
 679
 680         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 681         if (retval)
 682                 goto error;
 683
 684         retval = -EPERM;
 685         if (capable(CAP_SETUID)) {
 686                 new->suid = new->uid = uid;
 687                 if (uid != old->uid) {
 688                         retval = set_user(new);
 689                         if (retval < 0)
 690                                 goto error;
 691                 }
 692         } else if (uid != old->uid && uid != new->suid) {
 693                 goto error;
 694         }
 695
 696         new->fsuid = new->euid = uid;
 697
 698         retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
 699         if (retval < 0)
 700                 goto error;
 701
 702         return commit_creds(new);
 703
 704 error:
 705         abort_creds(new);
 706         return retval;
 707 }
 708
 709
 710 /*
 711  * This function implements a generic ability to update ruid, euid,
 712  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 713  */
 714 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
 715 {
 716         const struct cred *old;
 717         struct cred *new;
 718         int retval;
 719
 720         new = prepare_creds();
 721         if (!new)
 722                 return -ENOMEM;
 723
 724         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 725         if (retval)
 726                 goto error;
 727         old = current_cred();
 728
 729         retval = -EPERM;
 730         if (!capable(CAP_SETUID)) {
 731                 if (ruid != (uid_t) -1 && ruid != old->uid &&
 732                     ruid != old->euid  && ruid != old->suid)
 733                         goto error;
 734                 if (euid != (uid_t) -1 && euid != old->uid &&
 735                     euid != old->euid  && euid != old->suid)
 736                         goto error;
 737                 if (suid != (uid_t) -1 && suid != old->uid &&
 738                     suid != old->euid  && suid != old->suid)
 739                         goto error;
 740         }
 741
 742         if (ruid != (uid_t) -1) {
 743                 new->uid = ruid;
 744                 if (ruid != old->uid) {
 745                         retval = set_user(new);
 746                         if (retval < 0)
 747                                 goto error;
 748                 }
 749         }
 750         if (euid != (uid_t) -1)
 751                 new->euid = euid;
 752         if (suid != (uid_t) -1)
 753                 new->suid = suid;
 754         new->fsuid = new->euid;
 755
 756         retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
 757         if (retval < 0)
 758                 goto error;
 759
 760         return commit_creds(new);
 761
 762 error:
 763         abort_creds(new);
 764         return retval;
 765 }
 766
 767 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruid, uid_t __user *, euid, uid_t __user *, suid)
 768 {
 769         const struct cred *cred = current_cred();
 770         int retval;
 771
 772         if (!(retval   = put_user(cred->uid,  ruid)) &&
 773             !(retval   = put_user(cred->euid, euid)))
 774                 retval = put_user(cred->suid, suid);
 775
 776         return retval;
 777 }
 778
 779 /*
 780  * Same as above, but for rgid, egid, sgid.
 781  */
 782 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
 783 {
 784         const struct cred *old;
 785         struct cred *new;
 786         int retval;
 787
 788         new = prepare_creds();
 789         if (!new)
 790                 return -ENOMEM;
 791         old = current_cred();
 792
 793         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 794         if (retval)
 795                 goto error;
 796
 797         retval = -EPERM;
 798         if (!capable(CAP_SETGID)) {
 799                 if (rgid != (gid_t) -1 && rgid != old->gid &&
 800                     rgid != old->egid  && rgid != old->sgid)
 801                         goto error;
 802                 if (egid != (gid_t) -1 && egid != old->gid &&
 803                     egid != old->egid  && egid != old->sgid)
 804                         goto error;
 805                 if (sgid != (gid_t) -1 && sgid != old->gid &&
 806                     sgid != old->egid  && sgid != old->sgid)
 807                         goto error;
 808         }
 809
 810         if (rgid != (gid_t) -1)
 811                 new->gid = rgid;
 812         if (egid != (gid_t) -1)
 813                 new->egid = egid;
 814         if (sgid != (gid_t) -1)
 815                 new->sgid = sgid;
 816         new->fsgid = new->egid;
 817
 818         return commit_creds(new);
 819
 820 error:
 821         abort_creds(new);
 822         return retval;
 823 }
 824
 825 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgid, gid_t __user *, egid, gid_t __user *, sgid)
 826 {
 827         const struct cred *cred = current_cred();
 828         int retval;
 829
 830         if (!(retval   = put_user(cred->gid,  rgid)) &&
 831             !(retval   = put_user(cred->egid, egid)))
 832                 retval = put_user(cred->sgid, sgid);
 833
 834         return retval;
 835 }
 836
 837
 838 /*
 839  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 840  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 841  * whatever uid it wants to). It normally shadows "euid", except when
 842  * explicitly set by setfsuid() or for access..
 843  */
 844 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
 845 {
 846         const struct cred *old;
 847         struct cred *new;
 848         uid_t old_fsuid;
 849
 850         new = prepare_creds();
 851         if (!new)
 852                 return current_fsuid();
 853         old = current_cred();
 854         old_fsuid = old->fsuid;
 855
 856         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS) < 0)
 857                 goto error;
 858
 859         if (uid == old->uid  || uid == old->euid  ||
 860             uid == old->suid || uid == old->fsuid ||
 861             capable(CAP_SETUID)) {
 862                 if (uid != old_fsuid) {
 863                         new->fsuid = uid;
 864                         if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
 865                                 goto change_okay;
 866                 }
 867         }
 868
 869 error:
 870         abort_creds(new);
 871         return old_fsuid;
 872
 873 change_okay:
 874         commit_creds(new);
 875         return old_fsuid;
 876 }
 877
 878 /*
 879  * Samma på svenska..
 880  */
 881 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
 882 {
 883         const struct cred *old;
 884         struct cred *new;
 885         gid_t old_fsgid;
 886
 887         new = prepare_creds();
 888         if (!new)
 889                 return current_fsgid();
 890         old = current_cred();
 891         old_fsgid = old->fsgid;
 892
 893         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 894                 goto error;
 895
 896         if (gid == old->gid  || gid == old->egid  ||
 897             gid == old->sgid || gid == old->fsgid ||
 898             capable(CAP_SETGID)) {
 899                 if (gid != old_fsgid) {
 900                         new->fsgid = gid;
 901                         goto change_okay;
 902                 }
 903         }
 904
 905 error:
 906         abort_creds(new);
 907         return old_fsgid;
 908
 909 change_okay:
 910         commit_creds(new);
 911         return old_fsgid;
 912 }
 913
 914 void do_sys_times(struct tms *tms)
 915 {
 916         cputime_t tgutime, tgstime, cutime, cstime;
 917
 918         spin_lock_irq(&current->sighand->siglock);
 919         thread_group_times(current, &tgutime, &tgstime);
 920         cutime = current->signal->cutime;
 921         cstime = current->signal->cstime;
 922         spin_unlock_irq(&current->sighand->siglock);
 923         tms->tms_utime = cputime_to_clock_t(tgutime);
 924         tms->tms_stime = cputime_to_clock_t(tgstime);
 925         tms->tms_cutime = cputime_to_clock_t(cutime);
 926         tms->tms_cstime = cputime_to_clock_t(cstime);
 927 }
 928
 929 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
 930 {
 931         if (tbuf) {
 932                 struct tms tmp;
 933
 934                 do_sys_times(&tmp);
 935                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
 936                         return -EFAULT;
 937         }
 938         force_successful_syscall_return();
 939         return (long) jiffies_64_to_clock_t(get_jiffies_64());
 940 }
 941
 942 /*
 943  * This needs some heavy checking ...
 944  * I just haven't the stomach for it. I also don't fully
 945  * understand sessions/pgrp etc. Let somebody who does explain it.
 946  *
 947  * OK, I think I have the protection semantics right.... this is really
 948  * only important on a multi-user system anyway, to make sure one user
 949  * can't send a signal to a process owned by another.  -TYT, 12/12/91
 950  *
 951  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
 952  * LBT 04.03.94
 953  */
 954 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
 955 {
 956         struct task_struct *p;
 957         struct task_struct *group_leader = current->group_leader;
 958         struct pid *pgrp;
 959         int err;
 960
 961         if (!pid)
 962                 pid = task_pid_vnr(group_leader);
 963         if (!pgid)
 964                 pgid = pid;
 965         if (pgid < 0)
 966                 return -EINVAL;
 967
 968         /* From this point forward we keep holding onto the tasklist lock
 969          * so that our parent does not change from under us. -DaveM
 970          */
 971         write_lock_irq(&tasklist_lock);
 972
 973         err = -ESRCH;
 974         p = find_task_by_vpid(pid);
 975         if (!p)
 976                 goto out;
 977
 978         err = -EINVAL;
 979         if (!thread_group_leader(p))
 980                 goto out;
 981
 982         if (same_thread_group(p->real_parent, group_leader)) {
 983                 err = -EPERM;
 984                 if (task_session(p) != task_session(group_leader))
 985                         goto out;
 986                 err = -EACCES;
 987                 if (p->did_exec)
 988                         goto out;
 989         } else {
 990                 err = -ESRCH;
 991                 if (p != group_leader)
 992                         goto out;
 993         }
 994
 995         err = -EPERM;
 996         if (p->signal->leader)
 997                 goto out;
 998
 999         pgrp = task_pid(p);
1000         if (pgid != pid) {
1001                 struct task_struct *g;
1002
1003                 pgrp = find_vpid(pgid);
1004                 g = pid_task(pgrp, PIDTYPE_PGID);
1005                 if (!g || task_session(g) != task_session(group_leader))
1006                         goto out;
1007         }
1008
1009         err = security_task_setpgid(p, pgid);
1010         if (err)
1011                 goto out;
1012
1013         if (task_pgrp(p) != pgrp)
1014                 change_pid(p, PIDTYPE_PGID, pgrp);
1015
1016         err = 0;
1017 out:
1018         /* All paths lead to here, thus we are safe. -DaveM */
1019         write_unlock_irq(&tasklist_lock);
1020         return err;
1021 }
1022
1023 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1024 {
1025         struct task_struct *p;
1026         struct pid *grp;
1027         int retval;
1028
1029         rcu_read_lock();
1030         if (!pid)
1031                 grp = task_pgrp(current);
1032         else {
1033                 retval = -ESRCH;
1034                 p = find_task_by_vpid(pid);
1035                 if (!p)
1036                         goto out;
1037                 grp = task_pgrp(p);
1038                 if (!grp)
1039                         goto out;
1040
1041                 retval = security_task_getpgid(p);
1042                 if (retval)
1043                         goto out;
1044         }
1045         retval = pid_vnr(grp);
1046 out:
1047         rcu_read_unlock();
1048         return retval;
1049 }
1050
1051 #ifdef __ARCH_WANT_SYS_GETPGRP
1052
1053 SYSCALL_DEFINE0(getpgrp)
1054 {
1055         return sys_getpgid(0);
1056 }
1057
1058 #endif
1059
1060 SYSCALL_DEFINE1(getsid, pid_t, pid)
1061 {
1062         struct task_struct *p;
1063         struct pid *sid;
1064         int retval;
1065
1066         rcu_read_lock();
1067         if (!pid)
1068                 sid = task_session(current);
1069         else {
1070                 retval = -ESRCH;
1071                 p = find_task_by_vpid(pid);
1072                 if (!p)
1073                         goto out;
1074                 sid = task_session(p);
1075                 if (!sid)
1076                         goto out;
1077
1078                 retval = security_task_getsid(p);
1079                 if (retval)
1080                         goto out;
1081         }
1082         retval = pid_vnr(sid);
1083 out:
1084         rcu_read_unlock();
1085         return retval;
1086 }
1087
1088 SYSCALL_DEFINE0(setsid)
1089 {
1090         struct task_struct *group_leader = current->group_leader;
1091         struct pid *sid = task_pid(group_leader);
1092         pid_t session = pid_vnr(sid);
1093         int err = -EPERM;
1094
1095         write_lock_irq(&tasklist_lock);
1096         /* Fail if I am already a session leader */
1097         if (group_leader->signal->leader)
1098                 goto out;
1099
1100         /* Fail if a process group id already exists that equals the
1101          * proposed session id.
1102          */
1103         if (pid_task(sid, PIDTYPE_PGID))
1104                 goto out;
1105
1106         group_leader->signal->leader = 1;
1107         __set_special_pids(sid);
1108
1109         proc_clear_tty(group_leader);
1110
1111         err = session;
1112 out:
1113         write_unlock_irq(&tasklist_lock);
1114         if (err > 0)
1115                 proc_sid_connector(group_leader);
1116         return err;
1117 }
1118
1119 DECLARE_RWSEM(uts_sem);
1120
1121 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1122 {
1123         int errno = 0;
1124
1125         down_read(&uts_sem);
1126         if (copy_to_user(name, utsname(), sizeof *name))
1127                 errno = -EFAULT;
1128         up_read(&uts_sem);
1129         return errno;
1130 }
1131
1132 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1133 {
1134         int errno;
1135         char tmp[__NEW_UTS_LEN];
1136
1137         if (!capable(CAP_SYS_ADMIN))
1138                 return -EPERM;
1139         if (len < 0 || len > __NEW_UTS_LEN)
1140                 return -EINVAL;
1141         down_write(&uts_sem);
1142         errno = -EFAULT;
1143         if (!copy_from_user(tmp, name, len)) {
1144                 struct new_utsname *u = utsname();
1145
1146                 memcpy(u->nodename, tmp, len);
1147                 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1148                 errno = 0;
1149         }
1150         up_write(&uts_sem);
1151         return errno;
1152 }
1153
1154 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1155
1156 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1157 {
1158         int i, errno;
1159         struct new_utsname *u;
1160
1161         if (len < 0)
1162                 return -EINVAL;
1163         down_read(&uts_sem);
1164         u = utsname();
1165         i = 1 + strlen(u->nodename);
1166         if (i > len)
1167                 i = len;
1168         errno = 0;
1169         if (copy_to_user(name, u->nodename, i))
1170                 errno = -EFAULT;
1171         up_read(&uts_sem);
1172         return errno;
1173 }
1174
1175 #endif
1176
1177 /*
1178  * Only setdomainname; getdomainname can be implemented by calling
1179  * uname()
1180  */
1181 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1182 {
1183         int errno;
1184         char tmp[__NEW_UTS_LEN];
1185
1186         if (!capable(CAP_SYS_ADMIN))
1187                 return -EPERM;
1188         if (len < 0 || len > __NEW_UTS_LEN)
1189                 return -EINVAL;
1190
1191         down_write(&uts_sem);
1192         errno = -EFAULT;
1193         if (!copy_from_user(tmp, name, len)) {
1194                 struct new_utsname *u = utsname();
1195
1196                 memcpy(u->domainname, tmp, len);
1197                 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1198                 errno = 0;
1199         }
1200         up_write(&uts_sem);
1201         return errno;
1202 }
1203
1204 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1205 {
1206         if (resource >= RLIM_NLIMITS)
1207                 return -EINVAL;
1208         else {
1209                 struct rlimit value;
1210                 task_lock(current->group_leader);
1211                 value = current->signal->rlim[resource];
1212                 task_unlock(current->group_leader);
1213                 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1214         }
1215 }
1216
1217 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1218
1219 /*
1220  *      Back compatibility for getrlimit. Needed for some apps.
1221  */
1222
1223 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1224                 struct rlimit __user *, rlim)
1225 {
1226         struct rlimit x;
1227         if (resource >= RLIM_NLIMITS)
1228                 return -EINVAL;
1229
1230         task_lock(current->group_leader);
1231         x = current->signal->rlim[resource];
1232         task_unlock(current->group_leader);
1233         if (x.rlim_cur > 0x7FFFFFFF)
1234                 x.rlim_cur = 0x7FFFFFFF;
1235         if (x.rlim_max > 0x7FFFFFFF)
1236                 x.rlim_max = 0x7FFFFFFF;
1237         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1238 }
1239
1240 #endif
1241
1242 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1243 {
1244         struct rlimit new_rlim, *old_rlim;
1245         int retval;
1246
1247         if (resource >= RLIM_NLIMITS)
1248                 return -EINVAL;
1249         if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1250                 return -EFAULT;
1251         if (new_rlim.rlim_cur > new_rlim.rlim_max)
1252                 return -EINVAL;
1253         old_rlim = current->signal->rlim + resource;
1254         if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1255             !capable(CAP_SYS_RESOURCE))
1256                 return -EPERM;
1257         if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > sysctl_nr_open)
1258                 return -EPERM;
1259
1260         retval = security_task_setrlimit(resource, &new_rlim);
1261         if (retval)
1262                 return retval;
1263
1264         if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1265                 /*
1266                  * The caller is asking for an immediate RLIMIT_CPU
1267                  * expiry.  But we use the zero value to mean "it was
1268                  * never set".  So let's cheat and make it one second
1269                  * instead
1270                  */
1271                 new_rlim.rlim_cur = 1;
1272         }
1273
1274         task_lock(current->group_leader);
1275         *old_rlim = new_rlim;
1276         task_unlock(current->group_leader);
1277
1278         if (resource != RLIMIT_CPU)
1279                 goto out;
1280
1281         /*
1282          * RLIMIT_CPU handling.   Note that the kernel fails to return an error
1283          * code if it rejected the user's attempt to set RLIMIT_CPU.  This is a
1284          * very long-standing error, and fixing it now risks breakage of
1285          * applications, so we live with it
1286          */
1287         if (new_rlim.rlim_cur == RLIM_INFINITY)
1288                 goto out;
1289
1290         update_rlimit_cpu(new_rlim.rlim_cur);
1291 out:
1292         return 0;
1293 }
1294
1295 /*
1296  * It would make sense to put struct rusage in the task_struct,
1297  * except that would make the task_struct be *really big*.  After
1298  * task_struct gets moved into malloc'ed memory, it would
1299  * make sense to do this.  It will make moving the rest of the information
1300  * a lot simpler!  (Which we're not doing right now because we're not
1301  * measuring them yet).
1302  *
1303  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1304  * races with threads incrementing their own counters.  But since word
1305  * reads are atomic, we either get new values or old values and we don't
1306  * care which for the sums.  We always take the siglock to protect reading
1307  * the c* fields from p->signal from races with exit.c updating those
1308  * fields when reaping, so a sample either gets all the additions of a
1309  * given child after it's reaped, or none so this sample is before reaping.
1310  *
1311  * Locking:
1312  * We need to take the siglock for CHILDEREN, SELF and BOTH
1313  * for  the cases current multithreaded, non-current single threaded
1314  * non-current multithreaded.  Thread traversal is now safe with
1315  * the siglock held.
1316  * Strictly speaking, we donot need to take the siglock if we are current and
1317  * single threaded,  as no one else can take our signal_struct away, no one
1318  * else can  reap the  children to update signal->c* counters, and no one else
1319  * can race with the signal-> fields. If we do not take any lock, the
1320  * signal-> fields could be read out of order while another thread was just
1321  * exiting. So we should  place a read memory barrier when we avoid the lock.
1322  * On the writer side,  write memory barrier is implied in  __exit_signal
1323  * as __exit_signal releases  the siglock spinlock after updating the signal->
1324  * fields. But we don't do this yet to keep things simple.
1325  *
1326  */
1327
1328 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1329 {
1330         r->ru_nvcsw += t->nvcsw;
1331         r->ru_nivcsw += t->nivcsw;
1332         r->ru_minflt += t->min_flt;
1333         r->ru_majflt += t->maj_flt;
1334         r->ru_inblock += task_io_get_inblock(t);
1335         r->ru_oublock += task_io_get_oublock(t);
1336 }
1337
1338 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1339 {
1340         struct task_struct *t;
1341         unsigned long flags;
1342         cputime_t tgutime, tgstime, utime, stime;
1343         unsigned long maxrss = 0;
1344
1345         memset((char *) r, 0, sizeof *r);
1346         utime = stime = cputime_zero;
1347
1348         if (who == RUSAGE_THREAD) {
1349                 task_times(current, &utime, &stime);
1350                 accumulate_thread_rusage(p, r);
1351                 maxrss = p->signal->maxrss;
1352                 goto out;
1353         }
1354
1355         if (!lock_task_sighand(p, &flags))
1356                 return;
1357
1358         switch (who) {
1359                 case RUSAGE_BOTH:
1360                 case RUSAGE_CHILDREN:
1361                         utime = p->signal->cutime;
1362                         stime = p->signal->cstime;
1363                         r->ru_nvcsw = p->signal->cnvcsw;
1364                         r->ru_nivcsw = p->signal->cnivcsw;
1365                         r->ru_minflt = p->signal->cmin_flt;
1366                         r->ru_majflt = p->signal->cmaj_flt;
1367                         r->ru_inblock = p->signal->cinblock;
1368                         r->ru_oublock = p->signal->coublock;
1369                         maxrss = p->signal->cmaxrss;
1370
1371                         if (who == RUSAGE_CHILDREN)
1372                                 break;
1373
1374                 case RUSAGE_SELF:
1375                         thread_group_times(p, &tgutime, &tgstime);
1376                         utime = cputime_add(utime, tgutime);
1377                         stime = cputime_add(stime, tgstime);
1378                         r->ru_nvcsw += p->signal->nvcsw;
1379                         r->ru_nivcsw += p->signal->nivcsw;
1380                         r->ru_minflt += p->signal->min_flt;
1381                         r->ru_majflt += p->signal->maj_flt;
1382                         r->ru_inblock += p->signal->inblock;
1383                         r->ru_oublock += p->signal->oublock;
1384                         if (maxrss < p->signal->maxrss)
1385                                 maxrss = p->signal->maxrss;
1386                         t = p;
1387                         do {
1388                                 accumulate_thread_rusage(t, r);
1389                                 t = next_thread(t);
1390                         } while (t != p);
1391                         break;
1392
1393                 default:
1394                         BUG();
1395         }
1396         unlock_task_sighand(p, &flags);
1397
1398 out:
1399         cputime_to_timeval(utime, &r->ru_utime);
1400         cputime_to_timeval(stime, &r->ru_stime);
1401
1402         if (who != RUSAGE_CHILDREN) {
1403                 struct mm_struct *mm = get_task_mm(p);
1404                 if (mm) {
1405                         setmax_mm_hiwater_rss(&maxrss, mm);
1406                         mmput(mm);
1407                 }
1408         }
1409         r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1410 }
1411
1412 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1413 {
1414         struct rusage r;
1415         k_getrusage(p, who, &r);
1416         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1417 }
1418
1419 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1420 {
1421         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1422             who != RUSAGE_THREAD)
1423                 return -EINVAL;
1424         return getrusage(current, who, ru);
1425 }
1426
1427 SYSCALL_DEFINE1(umask, int, mask)
1428 {
1429         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1430         return mask;
1431 }
1432
1433 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1434                 unsigned long, arg4, unsigned long, arg5)
1435 {
1436         struct task_struct *me = current;
1437         unsigned char comm[sizeof(me->comm)];
1438         long error;
1439
1440         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1441         if (error != -ENOSYS)
1442                 return error;
1443
1444         error = 0;
1445         switch (option) {
1446                 case PR_SET_PDEATHSIG:
1447                         if (!valid_signal(arg2)) {
1448                                 error = -EINVAL;
1449                                 break;
1450                         }
1451                         me->pdeath_signal = arg2;
1452                         error = 0;
1453                         break;
1454                 case PR_GET_PDEATHSIG:
1455                         error = put_user(me->pdeath_signal, (int __user *)arg2);
1456                         break;
1457                 case PR_GET_DUMPABLE:
1458                         error = get_dumpable(me->mm);
1459                         break;
1460                 case PR_SET_DUMPABLE:
1461                         if (arg2 < 0 || arg2 > 1) {
1462                                 error = -EINVAL;
1463                                 break;
1464                         }
1465                         set_dumpable(me->mm, arg2);
1466                         error = 0;
1467                         break;
1468
1469                 case PR_SET_UNALIGN:
1470                         error = SET_UNALIGN_CTL(me, arg2);
1471                         break;
1472                 case PR_GET_UNALIGN:
1473                         error = GET_UNALIGN_CTL(me, arg2);
1474                         break;
1475                 case PR_SET_FPEMU:
1476                         error = SET_FPEMU_CTL(me, arg2);
1477                         break;
1478                 case PR_GET_FPEMU:
1479                         error = GET_FPEMU_CTL(me, arg2);
1480                         break;
1481                 case PR_SET_FPEXC:
1482                         error = SET_FPEXC_CTL(me, arg2);
1483                         break;
1484                 case PR_GET_FPEXC:
1485                         error = GET_FPEXC_CTL(me, arg2);
1486                         break;
1487                 case PR_GET_TIMING:
1488                         error = PR_TIMING_STATISTICAL;
1489                         break;
1490                 case PR_SET_TIMING:
1491                         if (arg2 != PR_TIMING_STATISTICAL)
1492                                 error = -EINVAL;
1493                         else
1494                                 error = 0;
1495                         break;
1496
1497                 case PR_SET_NAME:
1498                         comm[sizeof(me->comm)-1] = 0;
1499                         if (strncpy_from_user(comm, (char __user *)arg2,
1500                                               sizeof(me->comm) - 1) < 0)
1501                                 return -EFAULT;
1502                         set_task_comm(me, comm);
1503                         return 0;
1504                 case PR_GET_NAME:
1505                         get_task_comm(comm, me);
1506                         if (copy_to_user((char __user *)arg2, comm,
1507                                          sizeof(comm)))
1508                                 return -EFAULT;
1509                         return 0;
1510                 case PR_GET_ENDIAN:
1511                         error = GET_ENDIAN(me, arg2);
1512                         break;
1513                 case PR_SET_ENDIAN:
1514                         error = SET_ENDIAN(me, arg2);
1515                         break;
1516
1517                 case PR_GET_SECCOMP:
1518                         error = prctl_get_seccomp();
1519                         break;
1520                 case PR_SET_SECCOMP:
1521                         error = prctl_set_seccomp(arg2);
1522                         break;
1523                 case PR_GET_TSC:
1524                         error = GET_TSC_CTL(arg2);
1525                         break;
1526                 case PR_SET_TSC:
1527                         error = SET_TSC_CTL(arg2);
1528                         break;
1529                 case PR_TASK_PERF_EVENTS_DISABLE:
1530                         error = perf_event_task_disable();
1531                         break;
1532                 case PR_TASK_PERF_EVENTS_ENABLE:
1533                         error = perf_event_task_enable();
1534                         break;
1535                 case PR_GET_TIMERSLACK:
1536                         error = current->timer_slack_ns;
1537                         break;
1538                 case PR_SET_TIMERSLACK:
1539                         if (arg2 <= 0)
1540                                 current->timer_slack_ns =
1541                                         current->default_timer_slack_ns;
1542                         else
1543                                 current->timer_slack_ns = arg2;
1544                         error = 0;
1545                         break;
1546                 case PR_MCE_KILL:
1547                         if (arg4 | arg5)
1548                                 return -EINVAL;
1549                         switch (arg2) {
1550                         case PR_MCE_KILL_CLEAR:
1551                                 if (arg3 != 0)
1552                                         return -EINVAL;
1553                                 current->flags &= ~PF_MCE_PROCESS;
1554                                 break;
1555                         case PR_MCE_KILL_SET:
1556                                 current->flags |= PF_MCE_PROCESS;
1557                                 if (arg3 == PR_MCE_KILL_EARLY)
1558                                         current->flags |= PF_MCE_EARLY;
1559                                 else if (arg3 == PR_MCE_KILL_LATE)
1560                                         current->flags &= ~PF_MCE_EARLY;
1561                                 else if (arg3 == PR_MCE_KILL_DEFAULT)
1562                                         current->flags &=
1563                                                 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
1564                                 else
1565                                         return -EINVAL;
1566                                 break;
1567                         default:
1568                                 return -EINVAL;
1569                         }
1570                         error = 0;
1571                         break;
1572                 case PR_MCE_KILL_GET:
1573                         if (arg2 | arg3 | arg4 | arg5)
1574                                 return -EINVAL;
1575                         if (current->flags & PF_MCE_PROCESS)
1576                                 error = (current->flags & PF_MCE_EARLY) ?
1577                                         PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
1578                         else
1579                                 error = PR_MCE_KILL_DEFAULT;
1580                         break;
1581                 default:
1582                         error = -EINVAL;
1583                         break;
1584         }
1585         return error;
1586 }
1587
1588 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1589                 struct getcpu_cache __user *, unused)
1590 {
1591         int err = 0;
1592         int cpu = raw_smp_processor_id();
1593         if (cpup)
1594                 err |= put_user(cpu, cpup);
1595         if (nodep)
1596                 err |= put_user(cpu_to_node(cpu), nodep);
1597         return err ? -EFAULT : 0;
1598 }
1599
1600 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1601
1602 static void argv_cleanup(char **argv, char **envp)
1603 {
1604         argv_free(argv);
1605 }
1606
1607 /**
1608  * orderly_poweroff - Trigger an orderly system poweroff
1609  * @force: force poweroff if command execution fails
1610  *
1611  * This may be called from any context to trigger a system shutdown.
1612  * If the orderly shutdown fails, it will force an immediate shutdown.
1613  */
1614 int orderly_poweroff(bool force)
1615 {
1616         int argc;
1617         char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1618         static char *envp[] = {
1619                 "HOME=/",
1620                 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1621                 NULL
1622         };
1623         int ret = -ENOMEM;
1624         struct subprocess_info *info;
1625
1626         if (argv == NULL) {
1627                 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1628                        __func__, poweroff_cmd);
1629                 goto out;
1630         }
1631
1632         info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1633         if (info == NULL) {
1634                 argv_free(argv);
1635                 goto out;
1636         }
1637
1638         call_usermodehelper_setcleanup(info, argv_cleanup);
1639
1640         ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1641
1642   out:
1643         if (ret && force) {
1644                 printk(KERN_WARNING "Failed to start orderly shutdown: "
1645                        "forcing the issue\n");
1646
1647                 /* I guess this should try to kick off some daemon to
1648                    sync and poweroff asap.  Or not even bother syncing
1649                    if we're doing an emergency shutdown? */
1650                 emergency_sync();
1651                 kernel_power_off();
1652         }
1653
1654         return ret;
1655 }
1656 EXPORT_SYMBOL_GPL(orderly_poweroff);