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