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/personality.h>
  37 #include <linux/ptrace.h>
  38 #include <linux/fs_struct.h>
  39
  40 #include <linux/compat.h>
  41 #include <linux/syscalls.h>
  42 #include <linux/kprobes.h>
  43 #include <linux/user_namespace.h>
  44
  45 #include <asm/uaccess.h>
  46 #include <asm/io.h>
  47 #include <asm/unistd.h>
  48
  49 #ifndef SET_UNALIGN_CTL
  50 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  51 #endif
  52 #ifndef GET_UNALIGN_CTL
  53 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  54 #endif
  55 #ifndef SET_FPEMU_CTL
  56 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
  57 #endif
  58 #ifndef GET_FPEMU_CTL
  59 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
  60 #endif
  61 #ifndef SET_FPEXC_CTL
  62 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
  63 #endif
  64 #ifndef GET_FPEXC_CTL
  65 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
  66 #endif
  67 #ifndef GET_ENDIAN
  68 # define GET_ENDIAN(a,b)        (-EINVAL)
  69 #endif
  70 #ifndef SET_ENDIAN
  71 # define SET_ENDIAN(a,b)        (-EINVAL)
  72 #endif
  73 #ifndef GET_TSC_CTL
  74 # define GET_TSC_CTL(a)         (-EINVAL)
  75 #endif
  76 #ifndef SET_TSC_CTL
  77 # define SET_TSC_CTL(a)         (-EINVAL)
  78 #endif
  79
  80 /*
  81  * this is where the system-wide overflow UID and GID are defined, for
  82  * architectures that now have 32-bit UID/GID but didn't in the past
  83  */
  84
  85 int overflowuid = DEFAULT_OVERFLOWUID;
  86 int overflowgid = DEFAULT_OVERFLOWGID;
  87
  88 #ifdef CONFIG_UID16
  89 EXPORT_SYMBOL(overflowuid);
  90 EXPORT_SYMBOL(overflowgid);
  91 #endif
  92
  93 /*
  94  * the same as above, but for filesystems which can only store a 16-bit
  95  * UID and GID. as such, this is needed on all architectures
  96  */
  97
  98 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
  99 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
 100
 101 EXPORT_SYMBOL(fs_overflowuid);
 102 EXPORT_SYMBOL(fs_overflowgid);
 103
 104 /*
 105  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
 106  */
 107
 108 int C_A_D = 1;
 109 struct pid *cad_pid;
 110 EXPORT_SYMBOL(cad_pid);
 111
 112 /*
 113  * If set, this is used for preparing the system to power off.
 114  */
 115
 116 void (*pm_power_off_prepare)(void);
 117
 118 /*
 119  * set the priority of a task
 120  * - the caller must hold the RCU read lock
 121  */
 122 static int set_one_prio(struct task_struct *p, int niceval, int error)
 123 {
 124         const struct cred *cred = current_cred(), *pcred = __task_cred(p);
 125         int no_nice;
 126
 127         if (pcred->uid  != cred->euid &&
 128             pcred->euid != cred->euid && !capable(CAP_SYS_NICE)) {
 129                 error = -EPERM;
 130                 goto out;
 131         }
 132         if (niceval < task_nice(p) && !can_nice(p, niceval)) {
 133                 error = -EACCES;
 134                 goto out;
 135         }
 136         no_nice = security_task_setnice(p, niceval);
 137         if (no_nice) {
 138                 error = no_nice;
 139                 goto out;
 140         }
 141         if (error == -ESRCH)
 142                 error = 0;
 143         set_user_nice(p, niceval);
 144 out:
 145         return error;
 146 }
 147
 148 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
 149 {
 150         struct task_struct *g, *p;
 151         struct user_struct *user;
 152         const struct cred *cred = current_cred();
 153         int error = -EINVAL;
 154         struct pid *pgrp;
 155
 156         if (which > PRIO_USER || which < PRIO_PROCESS)
 157                 goto out;
 158
 159         /* normalize: avoid signed division (rounding problems) */
 160         error = -ESRCH;
 161         if (niceval < -20)
 162                 niceval = -20;
 163         if (niceval > 19)
 164                 niceval = 19;
 165
 166         rcu_read_lock();
 167         read_lock(&tasklist_lock);
 168         switch (which) {
 169                 case PRIO_PROCESS:
 170                         if (who)
 171                                 p = find_task_by_vpid(who);
 172                         else
 173                                 p = current;
 174                         if (p)
 175                                 error = set_one_prio(p, niceval, error);
 176                         break;
 177                 case PRIO_PGRP:
 178                         if (who)
 179                                 pgrp = find_vpid(who);
 180                         else
 181                                 pgrp = task_pgrp(current);
 182                         do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
 183                                 error = set_one_prio(p, niceval, error);
 184                         } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
 185                         break;
 186                 case PRIO_USER:
 187                         user = (struct user_struct *) cred->user;
 188                         if (!who)
 189                                 who = cred->uid;
 190                         else if ((who != cred->uid) &&
 191                                  !(user = find_user(who)))
 192                                 goto out_unlock;        /* No processes for this user */
 193
 194                         do_each_thread(g, p) {
 195                                 if (__task_cred(p)->uid == who)
 196                                         error = set_one_prio(p, niceval, error);
 197                         } while_each_thread(g, p);
 198                         if (who != cred->uid)
 199                                 free_uid(user);         /* For find_user() */
 200                         break;
 201         }
 202 out_unlock:
 203         read_unlock(&tasklist_lock);
 204         rcu_read_unlock();
 205 out:
 206         return error;
 207 }
 208
 209 /*
 210  * Ugh. To avoid negative return values, "getpriority()" will
 211  * not return the normal nice-value, but a negated value that
 212  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 213  * to stay compatible.
 214  */
 215 SYSCALL_DEFINE2(getpriority, int, which, int, who)
 216 {
 217         struct task_struct *g, *p;
 218         struct user_struct *user;
 219         const struct cred *cred = current_cred();
 220         long niceval, retval = -ESRCH;
 221         struct pid *pgrp;
 222
 223         if (which > PRIO_USER || which < PRIO_PROCESS)
 224                 return -EINVAL;
 225
 226         rcu_read_lock();
 227         read_lock(&tasklist_lock);
 228         switch (which) {
 229                 case PRIO_PROCESS:
 230                         if (who)
 231                                 p = find_task_by_vpid(who);
 232                         else
 233                                 p = current;
 234                         if (p) {
 235                                 niceval = 20 - task_nice(p);
 236                                 if (niceval > retval)
 237                                         retval = niceval;
 238                         }
 239                         break;
 240                 case PRIO_PGRP:
 241                         if (who)
 242                                 pgrp = find_vpid(who);
 243                         else
 244                                 pgrp = task_pgrp(current);
 245                         do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
 246                                 niceval = 20 - task_nice(p);
 247                                 if (niceval > retval)
 248                                         retval = niceval;
 249                         } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
 250                         break;
 251                 case PRIO_USER:
 252                         user = (struct user_struct *) cred->user;
 253                         if (!who)
 254                                 who = cred->uid;
 255                         else if ((who != cred->uid) &&
 256                                  !(user = find_user(who)))
 257                                 goto out_unlock;        /* No processes for this user */
 258
 259                         do_each_thread(g, p) {
 260                                 if (__task_cred(p)->uid == who) {
 261                                         niceval = 20 - task_nice(p);
 262                                         if (niceval > retval)
 263                                                 retval = niceval;
 264                                 }
 265                         } while_each_thread(g, p);
 266                         if (who != cred->uid)
 267                                 free_uid(user);         /* for find_user() */
 268                         break;
 269         }
 270 out_unlock:
 271         read_unlock(&tasklist_lock);
 272         rcu_read_unlock();
 273
 274         return retval;
 275 }
 276
 277 /**
 278  *      emergency_restart - reboot the system
 279  *
 280  *      Without shutting down any hardware or taking any locks
 281  *      reboot the system.  This is called when we know we are in
 282  *      trouble so this is our best effort to reboot.  This is
 283  *      safe to call in interrupt context.
 284  */
 285 void emergency_restart(void)
 286 {
 287         machine_emergency_restart();
 288 }
 289 EXPORT_SYMBOL_GPL(emergency_restart);
 290
 291 void kernel_restart_prepare(char *cmd)
 292 {
 293         blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
 294         system_state = SYSTEM_RESTART;
 295         device_shutdown();
 296         sysdev_shutdown();
 297 }
 298
 299 /**
 300  *      kernel_restart - reboot the system
 301  *      @cmd: pointer to buffer containing command to execute for restart
 302  *              or %NULL
 303  *
 304  *      Shutdown everything and perform a clean reboot.
 305  *      This is not safe to call in interrupt context.
 306  */
 307 void kernel_restart(char *cmd)
 308 {
 309         kernel_restart_prepare(cmd);
 310         if (!cmd)
 311                 printk(KERN_EMERG "Restarting system.\n");
 312         else
 313                 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
 314         machine_restart(cmd);
 315 }
 316 EXPORT_SYMBOL_GPL(kernel_restart);
 317
 318 static void kernel_shutdown_prepare(enum system_states state)
 319 {
 320         blocking_notifier_call_chain(&reboot_notifier_list,
 321                 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
 322         system_state = state;
 323         device_shutdown();
 324 }
 325 /**
 326  *      kernel_halt - halt the system
 327  *
 328  *      Shutdown everything and perform a clean system halt.
 329  */
 330 void kernel_halt(void)
 331 {
 332         kernel_shutdown_prepare(SYSTEM_HALT);
 333         sysdev_shutdown();
 334         printk(KERN_EMERG "System halted.\n");
 335         machine_halt();
 336 }
 337
 338 EXPORT_SYMBOL_GPL(kernel_halt);
 339
 340 /**
 341  *      kernel_power_off - power_off the system
 342  *
 343  *      Shutdown everything and perform a clean system power_off.
 344  */
 345 void kernel_power_off(void)
 346 {
 347         kernel_shutdown_prepare(SYSTEM_POWER_OFF);
 348         if (pm_power_off_prepare)
 349                 pm_power_off_prepare();
 350         disable_nonboot_cpus();
 351         sysdev_shutdown();
 352         printk(KERN_EMERG "Power down.\n");
 353         machine_power_off();
 354 }
 355 EXPORT_SYMBOL_GPL(kernel_power_off);
 356
 357 static DEFINE_MUTEX(reboot_mutex);
 358
 359 /*
 360  * Reboot system call: for obvious reasons only root may call it,
 361  * and even root needs to set up some magic numbers in the registers
 362  * so that some mistake won't make this reboot the whole machine.
 363  * You can also set the meaning of the ctrl-alt-del-key here.
 364  *
 365  * reboot doesn't sync: do that yourself before calling this.
 366  */
 367 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
 368                 void __user *, arg)
 369 {
 370         char buffer[256];
 371         int ret = 0;
 372
 373         /* We only trust the superuser with rebooting the system. */
 374         if (!capable(CAP_SYS_BOOT))
 375                 return -EPERM;
 376
 377         /* For safety, we require "magic" arguments. */
 378         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 379             (magic2 != LINUX_REBOOT_MAGIC2 &&
 380                         magic2 != LINUX_REBOOT_MAGIC2A &&
 381                         magic2 != LINUX_REBOOT_MAGIC2B &&
 382                         magic2 != LINUX_REBOOT_MAGIC2C))
 383                 return -EINVAL;
 384
 385         /* Instead of trying to make the power_off code look like
 386          * halt when pm_power_off is not set do it the easy way.
 387          */
 388         if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
 389                 cmd = LINUX_REBOOT_CMD_HALT;
 390
 391         mutex_lock(&reboot_mutex);
 392         switch (cmd) {
 393         case LINUX_REBOOT_CMD_RESTART:
 394                 kernel_restart(NULL);
 395                 break;
 396
 397         case LINUX_REBOOT_CMD_CAD_ON:
 398                 C_A_D = 1;
 399                 break;
 400
 401         case LINUX_REBOOT_CMD_CAD_OFF:
 402                 C_A_D = 0;
 403                 break;
 404
 405         case LINUX_REBOOT_CMD_HALT:
 406                 kernel_halt();
 407                 do_exit(0);
 408                 panic("cannot halt");
 409
 410         case LINUX_REBOOT_CMD_POWER_OFF:
 411                 kernel_power_off();
 412                 do_exit(0);
 413                 break;
 414
 415         case LINUX_REBOOT_CMD_RESTART2:
 416                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 417                         ret = -EFAULT;
 418                         break;
 419                 }
 420                 buffer[sizeof(buffer) - 1] = '\0';
 421
 422                 kernel_restart(buffer);
 423                 break;
 424
 425 #ifdef CONFIG_KEXEC
 426         case LINUX_REBOOT_CMD_KEXEC:
 427                 ret = kernel_kexec();
 428                 break;
 429 #endif
 430
 431 #ifdef CONFIG_HIBERNATION
 432         case LINUX_REBOOT_CMD_SW_SUSPEND:
 433                 ret = hibernate();
 434                 break;
 435 #endif
 436
 437         default:
 438                 ret = -EINVAL;
 439                 break;
 440         }
 441         mutex_unlock(&reboot_mutex);
 442         return ret;
 443 }
 444
 445 static void deferred_cad(struct work_struct *dummy)
 446 {
 447         kernel_restart(NULL);
 448 }
 449
 450 /*
 451  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 452  * As it's called within an interrupt, it may NOT sync: the only choice
 453  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 454  */
 455 void ctrl_alt_del(void)
 456 {
 457         static DECLARE_WORK(cad_work, deferred_cad);
 458
 459         if (C_A_D)
 460                 schedule_work(&cad_work);
 461         else
 462                 kill_cad_pid(SIGINT, 1);
 463 }
 464
 465 /*
 466  * Unprivileged users may change the real gid to the effective gid
 467  * or vice versa.  (BSD-style)
 468  *
 469  * If you set the real gid at all, or set the effective gid to a value not
 470  * equal to the real gid, then the saved gid is set to the new effective gid.
 471  *
 472  * This makes it possible for a setgid program to completely drop its
 473  * privileges, which is often a useful assertion to make when you are doing
 474  * a security audit over a program.
 475  *
 476  * The general idea is that a program which uses just setregid() will be
 477  * 100% compatible with BSD.  A program which uses just setgid() will be
 478  * 100% compatible with POSIX with saved IDs.
 479  *
 480  * SMP: There are not races, the GIDs are checked only by filesystem
 481  *      operations (as far as semantic preservation is concerned).
 482  */
 483 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
 484 {
 485         const struct cred *old;
 486         struct cred *new;
 487         int retval;
 488
 489         new = prepare_creds();
 490         if (!new)
 491                 return -ENOMEM;
 492         old = current_cred();
 493
 494         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 495         if (retval)
 496                 goto error;
 497
 498         retval = -EPERM;
 499         if (rgid != (gid_t) -1) {
 500                 if (old->gid == rgid ||
 501                     old->egid == rgid ||
 502                     capable(CAP_SETGID))
 503                         new->gid = rgid;
 504                 else
 505                         goto error;
 506         }
 507         if (egid != (gid_t) -1) {
 508                 if (old->gid == egid ||
 509                     old->egid == egid ||
 510                     old->sgid == egid ||
 511                     capable(CAP_SETGID))
 512                         new->egid = egid;
 513                 else
 514                         goto error;
 515         }
 516
 517         if (rgid != (gid_t) -1 ||
 518             (egid != (gid_t) -1 && egid != old->gid))
 519                 new->sgid = new->egid;
 520         new->fsgid = new->egid;
 521
 522         return commit_creds(new);
 523
 524 error:
 525         abort_creds(new);
 526         return retval;
 527 }
 528
 529 /*
 530  * setgid() is implemented like SysV w/ SAVED_IDS
 531  *
 532  * SMP: Same implicit races as above.
 533  */
 534 SYSCALL_DEFINE1(setgid, gid_t, gid)
 535 {
 536         const struct cred *old;
 537         struct cred *new;
 538         int retval;
 539
 540         new = prepare_creds();
 541         if (!new)
 542                 return -ENOMEM;
 543         old = current_cred();
 544
 545         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 546         if (retval)
 547                 goto error;
 548
 549         retval = -EPERM;
 550         if (capable(CAP_SETGID))
 551                 new->gid = new->egid = new->sgid = new->fsgid = gid;
 552         else if (gid == old->gid || gid == old->sgid)
 553                 new->egid = new->fsgid = gid;
 554         else
 555                 goto error;
 556
 557         return commit_creds(new);
 558
 559 error:
 560         abort_creds(new);
 561         return retval;
 562 }
 563
 564 /*
 565  * change the user struct in a credentials set to match the new UID
 566  */
 567 static int set_user(struct cred *new)
 568 {
 569         struct user_struct *new_user;
 570
 571         new_user = alloc_uid(current_user_ns(), new->uid);
 572         if (!new_user)
 573                 return -EAGAIN;
 574
 575         if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
 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 #ifdef COMPAT_UTS_MACHINE
1119 #define override_architecture(name) \
1120         (current->personality == PER_LINUX32 && \
1121          copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1122                       sizeof(COMPAT_UTS_MACHINE)))
1123 #else
1124 #define override_architecture(name)     0
1125 #endif
1126
1127 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1128 {
1129         int errno = 0;
1130
1131         down_read(&uts_sem);
1132         if (copy_to_user(name, utsname(), sizeof *name))
1133                 errno = -EFAULT;
1134         up_read(&uts_sem);
1135
1136         if (!errno && override_architecture(name))
1137                 errno = -EFAULT;
1138         return errno;
1139 }
1140
1141 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1142 /*
1143  * Old cruft
1144  */
1145 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1146 {
1147         int error = 0;
1148
1149         if (!name)
1150                 return -EFAULT;
1151
1152         down_read(&uts_sem);
1153         if (copy_to_user(name, utsname(), sizeof(*name)))
1154                 error = -EFAULT;
1155         up_read(&uts_sem);
1156
1157         if (!error && override_architecture(name))
1158                 error = -EFAULT;
1159         return error;
1160 }
1161
1162 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1163 {
1164         int error;
1165
1166         if (!name)
1167                 return -EFAULT;
1168         if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1169                 return -EFAULT;
1170
1171         down_read(&uts_sem);
1172         error = __copy_to_user(&name->sysname, &utsname()->sysname,
1173                                __OLD_UTS_LEN);
1174         error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1175         error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1176                                 __OLD_UTS_LEN);
1177         error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1178         error |= __copy_to_user(&name->release, &utsname()->release,
1179                                 __OLD_UTS_LEN);
1180         error |= __put_user(0, name->release + __OLD_UTS_LEN);
1181         error |= __copy_to_user(&name->version, &utsname()->version,
1182                                 __OLD_UTS_LEN);
1183         error |= __put_user(0, name->version + __OLD_UTS_LEN);
1184         error |= __copy_to_user(&name->machine, &utsname()->machine,
1185                                 __OLD_UTS_LEN);
1186         error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1187         up_read(&uts_sem);
1188
1189         if (!error && override_architecture(name))
1190                 error = -EFAULT;
1191         return error ? -EFAULT : 0;
1192 }
1193 #endif
1194
1195 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1196 {
1197         int errno;
1198         char tmp[__NEW_UTS_LEN];
1199
1200         if (!capable(CAP_SYS_ADMIN))
1201                 return -EPERM;
1202         if (len < 0 || len > __NEW_UTS_LEN)
1203                 return -EINVAL;
1204         down_write(&uts_sem);
1205         errno = -EFAULT;
1206         if (!copy_from_user(tmp, name, len)) {
1207                 struct new_utsname *u = utsname();
1208
1209                 memcpy(u->nodename, tmp, len);
1210                 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1211                 errno = 0;
1212         }
1213         up_write(&uts_sem);
1214         return errno;
1215 }
1216
1217 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1218
1219 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1220 {
1221         int i, errno;
1222         struct new_utsname *u;
1223
1224         if (len < 0)
1225                 return -EINVAL;
1226         down_read(&uts_sem);
1227         u = utsname();
1228         i = 1 + strlen(u->nodename);
1229         if (i > len)
1230                 i = len;
1231         errno = 0;
1232         if (copy_to_user(name, u->nodename, i))
1233                 errno = -EFAULT;
1234         up_read(&uts_sem);
1235         return errno;
1236 }
1237
1238 #endif
1239
1240 /*
1241  * Only setdomainname; getdomainname can be implemented by calling
1242  * uname()
1243  */
1244 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1245 {
1246         int errno;
1247         char tmp[__NEW_UTS_LEN];
1248
1249         if (!capable(CAP_SYS_ADMIN))
1250                 return -EPERM;
1251         if (len < 0 || len > __NEW_UTS_LEN)
1252                 return -EINVAL;
1253
1254         down_write(&uts_sem);
1255         errno = -EFAULT;
1256         if (!copy_from_user(tmp, name, len)) {
1257                 struct new_utsname *u = utsname();
1258
1259                 memcpy(u->domainname, tmp, len);
1260                 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1261                 errno = 0;
1262         }
1263         up_write(&uts_sem);
1264         return errno;
1265 }
1266
1267 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1268 {
1269         if (resource >= RLIM_NLIMITS)
1270                 return -EINVAL;
1271         else {
1272                 struct rlimit value;
1273                 task_lock(current->group_leader);
1274                 value = current->signal->rlim[resource];
1275                 task_unlock(current->group_leader);
1276                 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1277         }
1278 }
1279
1280 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1281
1282 /*
1283  *      Back compatibility for getrlimit. Needed for some apps.
1284  */
1285
1286 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1287                 struct rlimit __user *, rlim)
1288 {
1289         struct rlimit x;
1290         if (resource >= RLIM_NLIMITS)
1291                 return -EINVAL;
1292
1293         task_lock(current->group_leader);
1294         x = current->signal->rlim[resource];
1295         task_unlock(current->group_leader);
1296         if (x.rlim_cur > 0x7FFFFFFF)
1297                 x.rlim_cur = 0x7FFFFFFF;
1298         if (x.rlim_max > 0x7FFFFFFF)
1299                 x.rlim_max = 0x7FFFFFFF;
1300         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1301 }
1302
1303 #endif
1304
1305 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1306 {
1307         struct rlimit new_rlim, *old_rlim;
1308         int retval;
1309
1310         if (resource >= RLIM_NLIMITS)
1311                 return -EINVAL;
1312         if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1313                 return -EFAULT;
1314         if (new_rlim.rlim_cur > new_rlim.rlim_max)
1315                 return -EINVAL;
1316         old_rlim = current->signal->rlim + resource;
1317         if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1318             !capable(CAP_SYS_RESOURCE))
1319                 return -EPERM;
1320         if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > sysctl_nr_open)
1321                 return -EPERM;
1322
1323         retval = security_task_setrlimit(resource, &new_rlim);
1324         if (retval)
1325                 return retval;
1326
1327         if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1328                 /*
1329                  * The caller is asking for an immediate RLIMIT_CPU
1330                  * expiry.  But we use the zero value to mean "it was
1331                  * never set".  So let's cheat and make it one second
1332                  * instead
1333                  */
1334                 new_rlim.rlim_cur = 1;
1335         }
1336
1337         task_lock(current->group_leader);
1338         *old_rlim = new_rlim;
1339         task_unlock(current->group_leader);
1340
1341         if (resource != RLIMIT_CPU)
1342                 goto out;
1343
1344         /*
1345          * RLIMIT_CPU handling.   Note that the kernel fails to return an error
1346          * code if it rejected the user's attempt to set RLIMIT_CPU.  This is a
1347          * very long-standing error, and fixing it now risks breakage of
1348          * applications, so we live with it
1349          */
1350         if (new_rlim.rlim_cur == RLIM_INFINITY)
1351                 goto out;
1352
1353         update_rlimit_cpu(new_rlim.rlim_cur);
1354 out:
1355         return 0;
1356 }
1357
1358 /*
1359  * It would make sense to put struct rusage in the task_struct,
1360  * except that would make the task_struct be *really big*.  After
1361  * task_struct gets moved into malloc'ed memory, it would
1362  * make sense to do this.  It will make moving the rest of the information
1363  * a lot simpler!  (Which we're not doing right now because we're not
1364  * measuring them yet).
1365  *
1366  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1367  * races with threads incrementing their own counters.  But since word
1368  * reads are atomic, we either get new values or old values and we don't
1369  * care which for the sums.  We always take the siglock to protect reading
1370  * the c* fields from p->signal from races with exit.c updating those
1371  * fields when reaping, so a sample either gets all the additions of a
1372  * given child after it's reaped, or none so this sample is before reaping.
1373  *
1374  * Locking:
1375  * We need to take the siglock for CHILDEREN, SELF and BOTH
1376  * for  the cases current multithreaded, non-current single threaded
1377  * non-current multithreaded.  Thread traversal is now safe with
1378  * the siglock held.
1379  * Strictly speaking, we donot need to take the siglock if we are current and
1380  * single threaded,  as no one else can take our signal_struct away, no one
1381  * else can  reap the  children to update signal->c* counters, and no one else
1382  * can race with the signal-> fields. If we do not take any lock, the
1383  * signal-> fields could be read out of order while another thread was just
1384  * exiting. So we should  place a read memory barrier when we avoid the lock.
1385  * On the writer side,  write memory barrier is implied in  __exit_signal
1386  * as __exit_signal releases  the siglock spinlock after updating the signal->
1387  * fields. But we don't do this yet to keep things simple.
1388  *
1389  */
1390
1391 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1392 {
1393         r->ru_nvcsw += t->nvcsw;
1394         r->ru_nivcsw += t->nivcsw;
1395         r->ru_minflt += t->min_flt;
1396         r->ru_majflt += t->maj_flt;
1397         r->ru_inblock += task_io_get_inblock(t);
1398         r->ru_oublock += task_io_get_oublock(t);
1399 }
1400
1401 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1402 {
1403         struct task_struct *t;
1404         unsigned long flags;
1405         cputime_t tgutime, tgstime, utime, stime;
1406         unsigned long maxrss = 0;
1407
1408         memset((char *) r, 0, sizeof *r);
1409         utime = stime = cputime_zero;
1410
1411         if (who == RUSAGE_THREAD) {
1412                 task_times(current, &utime, &stime);
1413                 accumulate_thread_rusage(p, r);
1414                 maxrss = p->signal->maxrss;
1415                 goto out;
1416         }
1417
1418         if (!lock_task_sighand(p, &flags))
1419                 return;
1420
1421         switch (who) {
1422                 case RUSAGE_BOTH:
1423                 case RUSAGE_CHILDREN:
1424                         utime = p->signal->cutime;
1425                         stime = p->signal->cstime;
1426                         r->ru_nvcsw = p->signal->cnvcsw;
1427                         r->ru_nivcsw = p->signal->cnivcsw;
1428                         r->ru_minflt = p->signal->cmin_flt;
1429                         r->ru_majflt = p->signal->cmaj_flt;
1430                         r->ru_inblock = p->signal->cinblock;
1431                         r->ru_oublock = p->signal->coublock;
1432                         maxrss = p->signal->cmaxrss;
1433
1434                         if (who == RUSAGE_CHILDREN)
1435                                 break;
1436
1437                 case RUSAGE_SELF:
1438                         thread_group_times(p, &tgutime, &tgstime);
1439                         utime = cputime_add(utime, tgutime);
1440                         stime = cputime_add(stime, tgstime);
1441                         r->ru_nvcsw += p->signal->nvcsw;
1442                         r->ru_nivcsw += p->signal->nivcsw;
1443                         r->ru_minflt += p->signal->min_flt;
1444                         r->ru_majflt += p->signal->maj_flt;
1445                         r->ru_inblock += p->signal->inblock;
1446                         r->ru_oublock += p->signal->oublock;
1447                         if (maxrss < p->signal->maxrss)
1448                                 maxrss = p->signal->maxrss;
1449                         t = p;
1450                         do {
1451                                 accumulate_thread_rusage(t, r);
1452                                 t = next_thread(t);
1453                         } while (t != p);
1454                         break;
1455
1456                 default:
1457                         BUG();
1458         }
1459         unlock_task_sighand(p, &flags);
1460
1461 out:
1462         cputime_to_timeval(utime, &r->ru_utime);
1463         cputime_to_timeval(stime, &r->ru_stime);
1464
1465         if (who != RUSAGE_CHILDREN) {
1466                 struct mm_struct *mm = get_task_mm(p);
1467                 if (mm) {
1468                         setmax_mm_hiwater_rss(&maxrss, mm);
1469                         mmput(mm);
1470                 }
1471         }
1472         r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1473 }
1474
1475 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1476 {
1477         struct rusage r;
1478         k_getrusage(p, who, &r);
1479         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1480 }
1481
1482 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1483 {
1484         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1485             who != RUSAGE_THREAD)
1486                 return -EINVAL;
1487         return getrusage(current, who, ru);
1488 }
1489
1490 SYSCALL_DEFINE1(umask, int, mask)
1491 {
1492         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1493         return mask;
1494 }
1495
1496 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1497                 unsigned long, arg4, unsigned long, arg5)
1498 {
1499         struct task_struct *me = current;
1500         unsigned char comm[sizeof(me->comm)];
1501         long error;
1502
1503         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1504         if (error != -ENOSYS)
1505                 return error;
1506
1507         error = 0;
1508         switch (option) {
1509                 case PR_SET_PDEATHSIG:
1510                         if (!valid_signal(arg2)) {
1511                                 error = -EINVAL;
1512                                 break;
1513                         }
1514                         me->pdeath_signal = arg2;
1515                         error = 0;
1516                         break;
1517                 case PR_GET_PDEATHSIG:
1518                         error = put_user(me->pdeath_signal, (int __user *)arg2);
1519                         break;
1520                 case PR_GET_DUMPABLE:
1521                         error = get_dumpable(me->mm);
1522                         break;
1523                 case PR_SET_DUMPABLE:
1524                         if (arg2 < 0 || arg2 > 1) {
1525                                 error = -EINVAL;
1526                                 break;
1527                         }
1528                         set_dumpable(me->mm, arg2);
1529                         error = 0;
1530                         break;
1531
1532                 case PR_SET_UNALIGN:
1533                         error = SET_UNALIGN_CTL(me, arg2);
1534                         break;
1535                 case PR_GET_UNALIGN:
1536                         error = GET_UNALIGN_CTL(me, arg2);
1537                         break;
1538                 case PR_SET_FPEMU:
1539                         error = SET_FPEMU_CTL(me, arg2);
1540                         break;
1541                 case PR_GET_FPEMU:
1542                         error = GET_FPEMU_CTL(me, arg2);
1543                         break;
1544                 case PR_SET_FPEXC:
1545                         error = SET_FPEXC_CTL(me, arg2);
1546                         break;
1547                 case PR_GET_FPEXC:
1548                         error = GET_FPEXC_CTL(me, arg2);
1549                         break;
1550                 case PR_GET_TIMING:
1551                         error = PR_TIMING_STATISTICAL;
1552                         break;
1553                 case PR_SET_TIMING:
1554                         if (arg2 != PR_TIMING_STATISTICAL)
1555                                 error = -EINVAL;
1556                         else
1557                                 error = 0;
1558                         break;
1559
1560                 case PR_SET_NAME:
1561                         comm[sizeof(me->comm)-1] = 0;
1562                         if (strncpy_from_user(comm, (char __user *)arg2,
1563                                               sizeof(me->comm) - 1) < 0)
1564                                 return -EFAULT;
1565                         set_task_comm(me, comm);
1566                         return 0;
1567                 case PR_GET_NAME:
1568                         get_task_comm(comm, me);
1569                         if (copy_to_user((char __user *)arg2, comm,
1570                                          sizeof(comm)))
1571                                 return -EFAULT;
1572                         return 0;
1573                 case PR_GET_ENDIAN:
1574                         error = GET_ENDIAN(me, arg2);
1575                         break;
1576                 case PR_SET_ENDIAN:
1577                         error = SET_ENDIAN(me, arg2);
1578                         break;
1579
1580                 case PR_GET_SECCOMP:
1581                         error = prctl_get_seccomp();
1582                         break;
1583                 case PR_SET_SECCOMP:
1584                         error = prctl_set_seccomp(arg2);
1585                         break;
1586                 case PR_GET_TSC:
1587                         error = GET_TSC_CTL(arg2);
1588                         break;
1589                 case PR_SET_TSC:
1590                         error = SET_TSC_CTL(arg2);
1591                         break;
1592                 case PR_TASK_PERF_EVENTS_DISABLE:
1593                         error = perf_event_task_disable();
1594                         break;
1595                 case PR_TASK_PERF_EVENTS_ENABLE:
1596                         error = perf_event_task_enable();
1597                         break;
1598                 case PR_GET_TIMERSLACK:
1599                         error = current->timer_slack_ns;
1600                         break;
1601                 case PR_SET_TIMERSLACK:
1602                         if (arg2 <= 0)
1603                                 current->timer_slack_ns =
1604                                         current->default_timer_slack_ns;
1605                         else
1606                                 current->timer_slack_ns = arg2;
1607                         error = 0;
1608                         break;
1609                 case PR_MCE_KILL:
1610                         if (arg4 | arg5)
1611                                 return -EINVAL;
1612                         switch (arg2) {
1613                         case PR_MCE_KILL_CLEAR:
1614                                 if (arg3 != 0)
1615                                         return -EINVAL;
1616                                 current->flags &= ~PF_MCE_PROCESS;
1617                                 break;
1618                         case PR_MCE_KILL_SET:
1619                                 current->flags |= PF_MCE_PROCESS;
1620                                 if (arg3 == PR_MCE_KILL_EARLY)
1621                                         current->flags |= PF_MCE_EARLY;
1622                                 else if (arg3 == PR_MCE_KILL_LATE)
1623                                         current->flags &= ~PF_MCE_EARLY;
1624                                 else if (arg3 == PR_MCE_KILL_DEFAULT)
1625                                         current->flags &=
1626                                                 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
1627                                 else
1628                                         return -EINVAL;
1629                                 break;
1630                         default:
1631                                 return -EINVAL;
1632                         }
1633                         error = 0;
1634                         break;
1635                 case PR_MCE_KILL_GET:
1636                         if (arg2 | arg3 | arg4 | arg5)
1637                                 return -EINVAL;
1638                         if (current->flags & PF_MCE_PROCESS)
1639                                 error = (current->flags & PF_MCE_EARLY) ?
1640                                         PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
1641                         else
1642                                 error = PR_MCE_KILL_DEFAULT;
1643                         break;
1644                 default:
1645                         error = -EINVAL;
1646                         break;
1647         }
1648         return error;
1649 }
1650
1651 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1652                 struct getcpu_cache __user *, unused)
1653 {
1654         int err = 0;
1655         int cpu = raw_smp_processor_id();
1656         if (cpup)
1657                 err |= put_user(cpu, cpup);
1658         if (nodep)
1659                 err |= put_user(cpu_to_node(cpu), nodep);
1660         return err ? -EFAULT : 0;
1661 }
1662
1663 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1664
1665 static void argv_cleanup(char **argv, char **envp)
1666 {
1667         argv_free(argv);
1668 }
1669
1670 /**
1671  * orderly_poweroff - Trigger an orderly system poweroff
1672  * @force: force poweroff if command execution fails
1673  *
1674  * This may be called from any context to trigger a system shutdown.
1675  * If the orderly shutdown fails, it will force an immediate shutdown.
1676  */
1677 int orderly_poweroff(bool force)
1678 {
1679         int argc;
1680         char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1681         static char *envp[] = {
1682                 "HOME=/",
1683                 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1684                 NULL
1685         };
1686         int ret = -ENOMEM;
1687         struct subprocess_info *info;
1688
1689         if (argv == NULL) {
1690                 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1691                        __func__, poweroff_cmd);
1692                 goto out;
1693         }
1694
1695         info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1696         if (info == NULL) {
1697                 argv_free(argv);
1698                 goto out;
1699         }
1700
1701         call_usermodehelper_setcleanup(info, argv_cleanup);
1702
1703         ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1704
1705   out:
1706         if (ret && force) {
1707                 printk(KERN_WARNING "Failed to start orderly shutdown: "
1708                        "forcing the issue\n");
1709
1710                 /* I guess this should try to kick off some daemon to
1711                    sync and poweroff asap.  Or not even bother syncing
1712                    if we're doing an emergency shutdown? */
1713                 emergency_sync();
1714                 kernel_power_off();
1715         }
1716
1717         return ret;
1718 }
1719 EXPORT_SYMBOL_GPL(orderly_poweroff);