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 #include <linux/gfp.h>
  40
  41 #include <linux/compat.h>
  42 #include <linux/syscalls.h>
  43 #include <linux/kprobes.h>
  44 #include <linux/user_namespace.h>
  45
  46 #include <asm/uaccess.h>
  47 #include <asm/io.h>
  48 #include <asm/unistd.h>
  49
  50 #ifndef SET_UNALIGN_CTL
  51 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  52 #endif
  53 #ifndef GET_UNALIGN_CTL
  54 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  55 #endif
  56 #ifndef SET_FPEMU_CTL
  57 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
  58 #endif
  59 #ifndef GET_FPEMU_CTL
  60 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
  61 #endif
  62 #ifndef SET_FPEXC_CTL
  63 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
  64 #endif
  65 #ifndef GET_FPEXC_CTL
  66 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
  67 #endif
  68 #ifndef GET_ENDIAN
  69 # define GET_ENDIAN(a,b)        (-EINVAL)
  70 #endif
  71 #ifndef SET_ENDIAN
  72 # define SET_ENDIAN(a,b)        (-EINVAL)
  73 #endif
  74 #ifndef GET_TSC_CTL
  75 # define GET_TSC_CTL(a)         (-EINVAL)
  76 #endif
  77 #ifndef SET_TSC_CTL
  78 # define SET_TSC_CTL(a)         (-EINVAL)
  79 #endif
  80
  81 /*
  82  * this is where the system-wide overflow UID and GID are defined, for
  83  * architectures that now have 32-bit UID/GID but didn't in the past
  84  */
  85
  86 int overflowuid = DEFAULT_OVERFLOWUID;
  87 int overflowgid = DEFAULT_OVERFLOWGID;
  88
  89 #ifdef CONFIG_UID16
  90 EXPORT_SYMBOL(overflowuid);
  91 EXPORT_SYMBOL(overflowgid);
  92 #endif
  93
  94 /*
  95  * the same as above, but for filesystems which can only store a 16-bit
  96  * UID and GID. as such, this is needed on all architectures
  97  */
  98
  99 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
 100 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
 101
 102 EXPORT_SYMBOL(fs_overflowuid);
 103 EXPORT_SYMBOL(fs_overflowgid);
 104
 105 /*
 106  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
 107  */
 108
 109 int C_A_D = 1;
 110 struct pid *cad_pid;
 111 EXPORT_SYMBOL(cad_pid);
 112
 113 /*
 114  * If set, this is used for preparing the system to power off.
 115  */
 116
 117 void (*pm_power_off_prepare)(void);
 118
 119 /*
 120  * set the priority of a task
 121  * - the caller must hold the RCU read lock
 122  */
 123 static int set_one_prio(struct task_struct *p, int niceval, int error)
 124 {
 125         const struct cred *cred = current_cred(), *pcred = __task_cred(p);
 126         int no_nice;
 127
 128         if (pcred->uid  != cred->euid &&
 129             pcred->euid != cred->euid && !capable(CAP_SYS_NICE)) {
 130                 error = -EPERM;
 131                 goto out;
 132         }
 133         if (niceval < task_nice(p) && !can_nice(p, niceval)) {
 134                 error = -EACCES;
 135                 goto out;
 136         }
 137         no_nice = security_task_setnice(p, niceval);
 138         if (no_nice) {
 139                 error = no_nice;
 140                 goto out;
 141         }
 142         if (error == -ESRCH)
 143                 error = 0;
 144         set_user_nice(p, niceval);
 145 out:
 146         return error;
 147 }
 148
 149 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
 150 {
 151         struct task_struct *g, *p;
 152         struct user_struct *user;
 153         const struct cred *cred = current_cred();
 154         int error = -EINVAL;
 155         struct pid *pgrp;
 156
 157         if (which > PRIO_USER || which < PRIO_PROCESS)
 158                 goto out;
 159
 160         /* normalize: avoid signed division (rounding problems) */
 161         error = -ESRCH;
 162         if (niceval < -20)
 163                 niceval = -20;
 164         if (niceval > 19)
 165                 niceval = 19;
 166
 167         rcu_read_lock();
 168         read_lock(&tasklist_lock);
 169         switch (which) {
 170                 case PRIO_PROCESS:
 171                         if (who)
 172                                 p = find_task_by_vpid(who);
 173                         else
 174                                 p = current;
 175                         if (p)
 176                                 error = set_one_prio(p, niceval, error);
 177                         break;
 178                 case PRIO_PGRP:
 179                         if (who)
 180                                 pgrp = find_vpid(who);
 181                         else
 182                                 pgrp = task_pgrp(current);
 183                         do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
 184                                 error = set_one_prio(p, niceval, error);
 185                         } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
 186                         break;
 187                 case PRIO_USER:
 188                         user = (struct user_struct *) cred->user;
 189                         if (!who)
 190                                 who = cred->uid;
 191                         else if ((who != cred->uid) &&
 192                                  !(user = find_user(who)))
 193                                 goto out_unlock;        /* No processes for this user */
 194
 195                         do_each_thread(g, p) {
 196                                 if (__task_cred(p)->uid == who)
 197                                         error = set_one_prio(p, niceval, error);
 198                         } while_each_thread(g, p);
 199                         if (who != cred->uid)
 200                                 free_uid(user);         /* For find_user() */
 201                         break;
 202         }
 203 out_unlock:
 204         read_unlock(&tasklist_lock);
 205         rcu_read_unlock();
 206 out:
 207         return error;
 208 }
 209
 210 /*
 211  * Ugh. To avoid negative return values, "getpriority()" will
 212  * not return the normal nice-value, but a negated value that
 213  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 214  * to stay compatible.
 215  */
 216 SYSCALL_DEFINE2(getpriority, int, which, int, who)
 217 {
 218         struct task_struct *g, *p;
 219         struct user_struct *user;
 220         const struct cred *cred = current_cred();
 221         long niceval, retval = -ESRCH;
 222         struct pid *pgrp;
 223
 224         if (which > PRIO_USER || which < PRIO_PROCESS)
 225                 return -EINVAL;
 226
 227         rcu_read_lock();
 228         read_lock(&tasklist_lock);
 229         switch (which) {
 230                 case PRIO_PROCESS:
 231                         if (who)
 232                                 p = find_task_by_vpid(who);
 233                         else
 234                                 p = current;
 235                         if (p) {
 236                                 niceval = 20 - task_nice(p);
 237                                 if (niceval > retval)
 238                                         retval = niceval;
 239                         }
 240                         break;
 241                 case PRIO_PGRP:
 242                         if (who)
 243                                 pgrp = find_vpid(who);
 244                         else
 245                                 pgrp = task_pgrp(current);
 246                         do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
 247                                 niceval = 20 - task_nice(p);
 248                                 if (niceval > retval)
 249                                         retval = niceval;
 250                         } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
 251                         break;
 252                 case PRIO_USER:
 253                         user = (struct user_struct *) cred->user;
 254                         if (!who)
 255                                 who = cred->uid;
 256                         else if ((who != cred->uid) &&
 257                                  !(user = find_user(who)))
 258                                 goto out_unlock;        /* No processes for this user */
 259
 260                         do_each_thread(g, p) {
 261                                 if (__task_cred(p)->uid == who) {
 262                                         niceval = 20 - task_nice(p);
 263                                         if (niceval > retval)
 264                                                 retval = niceval;
 265                                 }
 266                         } while_each_thread(g, p);
 267                         if (who != cred->uid)
 268                                 free_uid(user);         /* for find_user() */
 269                         break;
 270         }
 271 out_unlock:
 272         read_unlock(&tasklist_lock);
 273         rcu_read_unlock();
 274
 275         return retval;
 276 }
 277
 278 /**
 279  *      emergency_restart - reboot the system
 280  *
 281  *      Without shutting down any hardware or taking any locks
 282  *      reboot the system.  This is called when we know we are in
 283  *      trouble so this is our best effort to reboot.  This is
 284  *      safe to call in interrupt context.
 285  */
 286 void emergency_restart(void)
 287 {
 288         machine_emergency_restart();
 289 }
 290 EXPORT_SYMBOL_GPL(emergency_restart);
 291
 292 void kernel_restart_prepare(char *cmd)
 293 {
 294         blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
 295         system_state = SYSTEM_RESTART;
 296         device_shutdown();
 297         sysdev_shutdown();
 298 }
 299
 300 /**
 301  *      kernel_restart - reboot the system
 302  *      @cmd: pointer to buffer containing command to execute for restart
 303  *              or %NULL
 304  *
 305  *      Shutdown everything and perform a clean reboot.
 306  *      This is not safe to call in interrupt context.
 307  */
 308 void kernel_restart(char *cmd)
 309 {
 310         kernel_restart_prepare(cmd);
 311         if (!cmd)
 312                 printk(KERN_EMERG "Restarting system.\n");
 313         else
 314                 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
 315         machine_restart(cmd);
 316 }
 317 EXPORT_SYMBOL_GPL(kernel_restart);
 318
 319 static void kernel_shutdown_prepare(enum system_states state)
 320 {
 321         blocking_notifier_call_chain(&reboot_notifier_list,
 322                 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
 323         system_state = state;
 324         device_shutdown();
 325 }
 326 /**
 327  *      kernel_halt - halt the system
 328  *
 329  *      Shutdown everything and perform a clean system halt.
 330  */
 331 void kernel_halt(void)
 332 {
 333         kernel_shutdown_prepare(SYSTEM_HALT);
 334         sysdev_shutdown();
 335         printk(KERN_EMERG "System halted.\n");
 336         machine_halt();
 337 }
 338
 339 EXPORT_SYMBOL_GPL(kernel_halt);
 340
 341 /**
 342  *      kernel_power_off - power_off the system
 343  *
 344  *      Shutdown everything and perform a clean system power_off.
 345  */
 346 void kernel_power_off(void)
 347 {
 348         kernel_shutdown_prepare(SYSTEM_POWER_OFF);
 349         if (pm_power_off_prepare)
 350                 pm_power_off_prepare();
 351         disable_nonboot_cpus();
 352         sysdev_shutdown();
 353         printk(KERN_EMERG "Power down.\n");
 354         machine_power_off();
 355 }
 356 EXPORT_SYMBOL_GPL(kernel_power_off);
 357
 358 static DEFINE_MUTEX(reboot_mutex);
 359
 360 /*
 361  * Reboot system call: for obvious reasons only root may call it,
 362  * and even root needs to set up some magic numbers in the registers
 363  * so that some mistake won't make this reboot the whole machine.
 364  * You can also set the meaning of the ctrl-alt-del-key here.
 365  *
 366  * reboot doesn't sync: do that yourself before calling this.
 367  */
 368 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
 369                 void __user *, arg)
 370 {
 371         char buffer[256];
 372         int ret = 0;
 373
 374         /* We only trust the superuser with rebooting the system. */
 375         if (!capable(CAP_SYS_BOOT))
 376                 return -EPERM;
 377
 378         /* For safety, we require "magic" arguments. */
 379         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 380             (magic2 != LINUX_REBOOT_MAGIC2 &&
 381                         magic2 != LINUX_REBOOT_MAGIC2A &&
 382                         magic2 != LINUX_REBOOT_MAGIC2B &&
 383                         magic2 != LINUX_REBOOT_MAGIC2C))
 384                 return -EINVAL;
 385
 386         /* Instead of trying to make the power_off code look like
 387          * halt when pm_power_off is not set do it the easy way.
 388          */
 389         if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
 390                 cmd = LINUX_REBOOT_CMD_HALT;
 391
 392         mutex_lock(&reboot_mutex);
 393         switch (cmd) {
 394         case LINUX_REBOOT_CMD_RESTART:
 395                 kernel_restart(NULL);
 396                 break;
 397
 398         case LINUX_REBOOT_CMD_CAD_ON:
 399                 C_A_D = 1;
 400                 break;
 401
 402         case LINUX_REBOOT_CMD_CAD_OFF:
 403                 C_A_D = 0;
 404                 break;
 405
 406         case LINUX_REBOOT_CMD_HALT:
 407                 kernel_halt();
 408                 do_exit(0);
 409                 panic("cannot halt");
 410
 411         case LINUX_REBOOT_CMD_POWER_OFF:
 412                 kernel_power_off();
 413                 do_exit(0);
 414                 break;
 415
 416         case LINUX_REBOOT_CMD_RESTART2:
 417                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 418                         ret = -EFAULT;
 419                         break;
 420                 }
 421                 buffer[sizeof(buffer) - 1] = '\0';
 422
 423                 kernel_restart(buffer);
 424                 break;
 425
 426 #ifdef CONFIG_KEXEC
 427         case LINUX_REBOOT_CMD_KEXEC:
 428                 ret = kernel_kexec();
 429                 break;
 430 #endif
 431
 432 #ifdef CONFIG_HIBERNATION
 433         case LINUX_REBOOT_CMD_SW_SUSPEND:
 434                 ret = hibernate();
 435                 break;
 436 #endif
 437
 438         default:
 439                 ret = -EINVAL;
 440                 break;
 441         }
 442         mutex_unlock(&reboot_mutex);
 443         return ret;
 444 }
 445
 446 static void deferred_cad(struct work_struct *dummy)
 447 {
 448         kernel_restart(NULL);
 449 }
 450
 451 /*
 452  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 453  * As it's called within an interrupt, it may NOT sync: the only choice
 454  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 455  */
 456 void ctrl_alt_del(void)
 457 {
 458         static DECLARE_WORK(cad_work, deferred_cad);
 459
 460         if (C_A_D)
 461                 schedule_work(&cad_work);
 462         else
 463                 kill_cad_pid(SIGINT, 1);
 464 }
 465
 466 /*
 467  * Unprivileged users may change the real gid to the effective gid
 468  * or vice versa.  (BSD-style)
 469  *
 470  * If you set the real gid at all, or set the effective gid to a value not
 471  * equal to the real gid, then the saved gid is set to the new effective gid.
 472  *
 473  * This makes it possible for a setgid program to completely drop its
 474  * privileges, which is often a useful assertion to make when you are doing
 475  * a security audit over a program.
 476  *
 477  * The general idea is that a program which uses just setregid() will be
 478  * 100% compatible with BSD.  A program which uses just setgid() will be
 479  * 100% compatible with POSIX with saved IDs.
 480  *
 481  * SMP: There are not races, the GIDs are checked only by filesystem
 482  *      operations (as far as semantic preservation is concerned).
 483  */
 484 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
 485 {
 486         const struct cred *old;
 487         struct cred *new;
 488         int retval;
 489
 490         new = prepare_creds();
 491         if (!new)
 492                 return -ENOMEM;
 493         old = current_cred();
 494
 495         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 496         if (retval)
 497                 goto error;
 498
 499         retval = -EPERM;
 500         if (rgid != (gid_t) -1) {
 501                 if (old->gid == rgid ||
 502                     old->egid == rgid ||
 503                     capable(CAP_SETGID))
 504                         new->gid = rgid;
 505                 else
 506                         goto error;
 507         }
 508         if (egid != (gid_t) -1) {
 509                 if (old->gid == egid ||
 510                     old->egid == egid ||
 511                     old->sgid == egid ||
 512                     capable(CAP_SETGID))
 513                         new->egid = egid;
 514                 else
 515                         goto error;
 516         }
 517
 518         if (rgid != (gid_t) -1 ||
 519             (egid != (gid_t) -1 && egid != old->gid))
 520                 new->sgid = new->egid;
 521         new->fsgid = new->egid;
 522
 523         return commit_creds(new);
 524
 525 error:
 526         abort_creds(new);
 527         return retval;
 528 }
 529
 530 /*
 531  * setgid() is implemented like SysV w/ SAVED_IDS
 532  *
 533  * SMP: Same implicit races as above.
 534  */
 535 SYSCALL_DEFINE1(setgid, gid_t, gid)
 536 {
 537         const struct cred *old;
 538         struct cred *new;
 539         int retval;
 540
 541         new = prepare_creds();
 542         if (!new)
 543                 return -ENOMEM;
 544         old = current_cred();
 545
 546         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 547         if (retval)
 548                 goto error;
 549
 550         retval = -EPERM;
 551         if (capable(CAP_SETGID))
 552                 new->gid = new->egid = new->sgid = new->fsgid = gid;
 553         else if (gid == old->gid || gid == old->sgid)
 554                 new->egid = new->fsgid = gid;
 555         else
 556                 goto error;
 557
 558         return commit_creds(new);
 559
 560 error:
 561         abort_creds(new);
 562         return retval;
 563 }
 564
 565 /*
 566  * change the user struct in a credentials set to match the new UID
 567  */
 568 static int set_user(struct cred *new)
 569 {
 570         struct user_struct *new_user;
 571
 572         new_user = alloc_uid(current_user_ns(), new->uid);
 573         if (!new_user)
 574                 return -EAGAIN;
 575
 576         if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
 577                         new_user != INIT_USER) {
 578                 free_uid(new_user);
 579                 return -EAGAIN;
 580         }
 581
 582         free_uid(new->user);
 583         new->user = new_user;
 584         return 0;
 585 }
 586
 587 /*
 588  * Unprivileged users may change the real uid to the effective uid
 589  * or vice versa.  (BSD-style)
 590  *
 591  * If you set the real uid at all, or set the effective uid to a value not
 592  * equal to the real uid, then the saved uid is set to the new effective uid.
 593  *
 594  * This makes it possible for a setuid program to completely drop its
 595  * privileges, which is often a useful assertion to make when you are doing
 596  * a security audit over a program.
 597  *
 598  * The general idea is that a program which uses just setreuid() will be
 599  * 100% compatible with BSD.  A program which uses just setuid() will be
 600  * 100% compatible with POSIX with saved IDs.
 601  */
 602 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
 603 {
 604         const struct cred *old;
 605         struct cred *new;
 606         int retval;
 607
 608         new = prepare_creds();
 609         if (!new)
 610                 return -ENOMEM;
 611         old = current_cred();
 612
 613         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 614         if (retval)
 615                 goto error;
 616
 617         retval = -EPERM;
 618         if (ruid != (uid_t) -1) {
 619                 new->uid = ruid;
 620                 if (old->uid != ruid &&
 621                     old->euid != ruid &&
 622                     !capable(CAP_SETUID))
 623                         goto error;
 624         }
 625
 626         if (euid != (uid_t) -1) {
 627                 new->euid = euid;
 628                 if (old->uid != euid &&
 629                     old->euid != euid &&
 630                     old->suid != euid &&
 631                     !capable(CAP_SETUID))
 632                         goto error;
 633         }
 634
 635         if (new->uid != old->uid) {
 636                 retval = set_user(new);
 637                 if (retval < 0)
 638                         goto error;
 639         }
 640         if (ruid != (uid_t) -1 ||
 641             (euid != (uid_t) -1 && euid != old->uid))
 642                 new->suid = new->euid;
 643         new->fsuid = new->euid;
 644
 645         retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
 646         if (retval < 0)
 647                 goto error;
 648
 649         return commit_creds(new);
 650
 651 error:
 652         abort_creds(new);
 653         return retval;
 654 }
 655
 656 /*
 657  * setuid() is implemented like SysV with SAVED_IDS
 658  *
 659  * Note that SAVED_ID's is deficient in that a setuid root program
 660  * like sendmail, for example, cannot set its uid to be a normal
 661  * user and then switch back, because if you're root, setuid() sets
 662  * the saved uid too.  If you don't like this, blame the bright people
 663  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 664  * will allow a root program to temporarily drop privileges and be able to
 665  * regain them by swapping the real and effective uid.
 666  */
 667 SYSCALL_DEFINE1(setuid, uid_t, uid)
 668 {
 669         const struct cred *old;
 670         struct cred *new;
 671         int retval;
 672
 673         new = prepare_creds();
 674         if (!new)
 675                 return -ENOMEM;
 676         old = current_cred();
 677
 678         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 679         if (retval)
 680                 goto error;
 681
 682         retval = -EPERM;
 683         if (capable(CAP_SETUID)) {
 684                 new->suid = new->uid = uid;
 685                 if (uid != old->uid) {
 686                         retval = set_user(new);
 687                         if (retval < 0)
 688                                 goto error;
 689                 }
 690         } else if (uid != old->uid && uid != new->suid) {
 691                 goto error;
 692         }
 693
 694         new->fsuid = new->euid = uid;
 695
 696         retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
 697         if (retval < 0)
 698                 goto error;
 699
 700         return commit_creds(new);
 701
 702 error:
 703         abort_creds(new);
 704         return retval;
 705 }
 706
 707
 708 /*
 709  * This function implements a generic ability to update ruid, euid,
 710  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 711  */
 712 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
 713 {
 714         const struct cred *old;
 715         struct cred *new;
 716         int retval;
 717
 718         new = prepare_creds();
 719         if (!new)
 720                 return -ENOMEM;
 721
 722         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 723         if (retval)
 724                 goto error;
 725         old = current_cred();
 726
 727         retval = -EPERM;
 728         if (!capable(CAP_SETUID)) {
 729                 if (ruid != (uid_t) -1 && ruid != old->uid &&
 730                     ruid != old->euid  && ruid != old->suid)
 731                         goto error;
 732                 if (euid != (uid_t) -1 && euid != old->uid &&
 733                     euid != old->euid  && euid != old->suid)
 734                         goto error;
 735                 if (suid != (uid_t) -1 && suid != old->uid &&
 736                     suid != old->euid  && suid != old->suid)
 737                         goto error;
 738         }
 739
 740         if (ruid != (uid_t) -1) {
 741                 new->uid = ruid;
 742                 if (ruid != old->uid) {
 743                         retval = set_user(new);
 744                         if (retval < 0)
 745                                 goto error;
 746                 }
 747         }
 748         if (euid != (uid_t) -1)
 749                 new->euid = euid;
 750         if (suid != (uid_t) -1)
 751                 new->suid = suid;
 752         new->fsuid = new->euid;
 753
 754         retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
 755         if (retval < 0)
 756                 goto error;
 757
 758         return commit_creds(new);
 759
 760 error:
 761         abort_creds(new);
 762         return retval;
 763 }
 764
 765 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruid, uid_t __user *, euid, uid_t __user *, suid)
 766 {
 767         const struct cred *cred = current_cred();
 768         int retval;
 769
 770         if (!(retval   = put_user(cred->uid,  ruid)) &&
 771             !(retval   = put_user(cred->euid, euid)))
 772                 retval = put_user(cred->suid, suid);
 773
 774         return retval;
 775 }
 776
 777 /*
 778  * Same as above, but for rgid, egid, sgid.
 779  */
 780 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
 781 {
 782         const struct cred *old;
 783         struct cred *new;
 784         int retval;
 785
 786         new = prepare_creds();
 787         if (!new)
 788                 return -ENOMEM;
 789         old = current_cred();
 790
 791         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 792         if (retval)
 793                 goto error;
 794
 795         retval = -EPERM;
 796         if (!capable(CAP_SETGID)) {
 797                 if (rgid != (gid_t) -1 && rgid != old->gid &&
 798                     rgid != old->egid  && rgid != old->sgid)
 799                         goto error;
 800                 if (egid != (gid_t) -1 && egid != old->gid &&
 801                     egid != old->egid  && egid != old->sgid)
 802                         goto error;
 803                 if (sgid != (gid_t) -1 && sgid != old->gid &&
 804                     sgid != old->egid  && sgid != old->sgid)
 805                         goto error;
 806         }
 807
 808         if (rgid != (gid_t) -1)
 809                 new->gid = rgid;
 810         if (egid != (gid_t) -1)
 811                 new->egid = egid;
 812         if (sgid != (gid_t) -1)
 813                 new->sgid = sgid;
 814         new->fsgid = new->egid;
 815
 816         return commit_creds(new);
 817
 818 error:
 819         abort_creds(new);
 820         return retval;
 821 }
 822
 823 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgid, gid_t __user *, egid, gid_t __user *, sgid)
 824 {
 825         const struct cred *cred = current_cred();
 826         int retval;
 827
 828         if (!(retval   = put_user(cred->gid,  rgid)) &&
 829             !(retval   = put_user(cred->egid, egid)))
 830                 retval = put_user(cred->sgid, sgid);
 831
 832         return retval;
 833 }
 834
 835
 836 /*
 837  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 838  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 839  * whatever uid it wants to). It normally shadows "euid", except when
 840  * explicitly set by setfsuid() or for access..
 841  */
 842 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
 843 {
 844         const struct cred *old;
 845         struct cred *new;
 846         uid_t old_fsuid;
 847
 848         new = prepare_creds();
 849         if (!new)
 850                 return current_fsuid();
 851         old = current_cred();
 852         old_fsuid = old->fsuid;
 853
 854         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS) < 0)
 855                 goto error;
 856
 857         if (uid == old->uid  || uid == old->euid  ||
 858             uid == old->suid || uid == old->fsuid ||
 859             capable(CAP_SETUID)) {
 860                 if (uid != old_fsuid) {
 861                         new->fsuid = uid;
 862                         if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
 863                                 goto change_okay;
 864                 }
 865         }
 866
 867 error:
 868         abort_creds(new);
 869         return old_fsuid;
 870
 871 change_okay:
 872         commit_creds(new);
 873         return old_fsuid;
 874 }
 875
 876 /*
 877  * Samma på svenska..
 878  */
 879 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
 880 {
 881         const struct cred *old;
 882         struct cred *new;
 883         gid_t old_fsgid;
 884
 885         new = prepare_creds();
 886         if (!new)
 887                 return current_fsgid();
 888         old = current_cred();
 889         old_fsgid = old->fsgid;
 890
 891         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 892                 goto error;
 893
 894         if (gid == old->gid  || gid == old->egid  ||
 895             gid == old->sgid || gid == old->fsgid ||
 896             capable(CAP_SETGID)) {
 897                 if (gid != old_fsgid) {
 898                         new->fsgid = gid;
 899                         goto change_okay;
 900                 }
 901         }
 902
 903 error:
 904         abort_creds(new);
 905         return old_fsgid;
 906
 907 change_okay:
 908         commit_creds(new);
 909         return old_fsgid;
 910 }
 911
 912 void do_sys_times(struct tms *tms)
 913 {
 914         cputime_t tgutime, tgstime, cutime, cstime;
 915
 916         spin_lock_irq(&current->sighand->siglock);
 917         thread_group_times(current, &tgutime, &tgstime);
 918         cutime = current->signal->cutime;
 919         cstime = current->signal->cstime;
 920         spin_unlock_irq(&current->sighand->siglock);
 921         tms->tms_utime = cputime_to_clock_t(tgutime);
 922         tms->tms_stime = cputime_to_clock_t(tgstime);
 923         tms->tms_cutime = cputime_to_clock_t(cutime);
 924         tms->tms_cstime = cputime_to_clock_t(cstime);
 925 }
 926
 927 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
 928 {
 929         if (tbuf) {
 930                 struct tms tmp;
 931
 932                 do_sys_times(&tmp);
 933                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
 934                         return -EFAULT;
 935         }
 936         force_successful_syscall_return();
 937         return (long) jiffies_64_to_clock_t(get_jiffies_64());
 938 }
 939
 940 /*
 941  * This needs some heavy checking ...
 942  * I just haven't the stomach for it. I also don't fully
 943  * understand sessions/pgrp etc. Let somebody who does explain it.
 944  *
 945  * OK, I think I have the protection semantics right.... this is really
 946  * only important on a multi-user system anyway, to make sure one user
 947  * can't send a signal to a process owned by another.  -TYT, 12/12/91
 948  *
 949  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
 950  * LBT 04.03.94
 951  */
 952 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
 953 {
 954         struct task_struct *p;
 955         struct task_struct *group_leader = current->group_leader;
 956         struct pid *pgrp;
 957         int err;
 958
 959         if (!pid)
 960                 pid = task_pid_vnr(group_leader);
 961         if (!pgid)
 962                 pgid = pid;
 963         if (pgid < 0)
 964                 return -EINVAL;
 965
 966         /* From this point forward we keep holding onto the tasklist lock
 967          * so that our parent does not change from under us. -DaveM
 968          */
 969         write_lock_irq(&tasklist_lock);
 970
 971         err = -ESRCH;
 972         p = find_task_by_vpid(pid);
 973         if (!p)
 974                 goto out;
 975
 976         err = -EINVAL;
 977         if (!thread_group_leader(p))
 978                 goto out;
 979
 980         if (same_thread_group(p->real_parent, group_leader)) {
 981                 err = -EPERM;
 982                 if (task_session(p) != task_session(group_leader))
 983                         goto out;
 984                 err = -EACCES;
 985                 if (p->did_exec)
 986                         goto out;
 987         } else {
 988                 err = -ESRCH;
 989                 if (p != group_leader)
 990                         goto out;
 991         }
 992
 993         err = -EPERM;
 994         if (p->signal->leader)
 995                 goto out;
 996
 997         pgrp = task_pid(p);
 998         if (pgid != pid) {
 999                 struct task_struct *g;
1000
1001                 pgrp = find_vpid(pgid);
1002                 g = pid_task(pgrp, PIDTYPE_PGID);
1003                 if (!g || task_session(g) != task_session(group_leader))
1004                         goto out;
1005         }
1006
1007         err = security_task_setpgid(p, pgid);
1008         if (err)
1009                 goto out;
1010
1011         if (task_pgrp(p) != pgrp)
1012                 change_pid(p, PIDTYPE_PGID, pgrp);
1013
1014         err = 0;
1015 out:
1016         /* All paths lead to here, thus we are safe. -DaveM */
1017         write_unlock_irq(&tasklist_lock);
1018         return err;
1019 }
1020
1021 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1022 {
1023         struct task_struct *p;
1024         struct pid *grp;
1025         int retval;
1026
1027         rcu_read_lock();
1028         if (!pid)
1029                 grp = task_pgrp(current);
1030         else {
1031                 retval = -ESRCH;
1032                 p = find_task_by_vpid(pid);
1033                 if (!p)
1034                         goto out;
1035                 grp = task_pgrp(p);
1036                 if (!grp)
1037                         goto out;
1038
1039                 retval = security_task_getpgid(p);
1040                 if (retval)
1041                         goto out;
1042         }
1043         retval = pid_vnr(grp);
1044 out:
1045         rcu_read_unlock();
1046         return retval;
1047 }
1048
1049 #ifdef __ARCH_WANT_SYS_GETPGRP
1050
1051 SYSCALL_DEFINE0(getpgrp)
1052 {
1053         return sys_getpgid(0);
1054 }
1055
1056 #endif
1057
1058 SYSCALL_DEFINE1(getsid, pid_t, pid)
1059 {
1060         struct task_struct *p;
1061         struct pid *sid;
1062         int retval;
1063
1064         rcu_read_lock();
1065         if (!pid)
1066                 sid = task_session(current);
1067         else {
1068                 retval = -ESRCH;
1069                 p = find_task_by_vpid(pid);
1070                 if (!p)
1071                         goto out;
1072                 sid = task_session(p);
1073                 if (!sid)
1074                         goto out;
1075
1076                 retval = security_task_getsid(p);
1077                 if (retval)
1078                         goto out;
1079         }
1080         retval = pid_vnr(sid);
1081 out:
1082         rcu_read_unlock();
1083         return retval;
1084 }
1085
1086 SYSCALL_DEFINE0(setsid)
1087 {
1088         struct task_struct *group_leader = current->group_leader;
1089         struct pid *sid = task_pid(group_leader);
1090         pid_t session = pid_vnr(sid);
1091         int err = -EPERM;
1092
1093         write_lock_irq(&tasklist_lock);
1094         /* Fail if I am already a session leader */
1095         if (group_leader->signal->leader)
1096                 goto out;
1097
1098         /* Fail if a process group id already exists that equals the
1099          * proposed session id.
1100          */
1101         if (pid_task(sid, PIDTYPE_PGID))
1102                 goto out;
1103
1104         group_leader->signal->leader = 1;
1105         __set_special_pids(sid);
1106
1107         proc_clear_tty(group_leader);
1108
1109         err = session;
1110 out:
1111         write_unlock_irq(&tasklist_lock);
1112         if (err > 0)
1113                 proc_sid_connector(group_leader);
1114         return err;
1115 }
1116
1117 DECLARE_RWSEM(uts_sem);
1118
1119 #ifdef COMPAT_UTS_MACHINE
1120 #define override_architecture(name) \
1121         (personality(current->personality) == PER_LINUX32 && \
1122          copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1123                       sizeof(COMPAT_UTS_MACHINE)))
1124 #else
1125 #define override_architecture(name)     0
1126 #endif
1127
1128 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1129 {
1130         int errno = 0;
1131
1132         down_read(&uts_sem);
1133         if (copy_to_user(name, utsname(), sizeof *name))
1134                 errno = -EFAULT;
1135         up_read(&uts_sem);
1136
1137         if (!errno && override_architecture(name))
1138                 errno = -EFAULT;
1139         return errno;
1140 }
1141
1142 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1143 /*
1144  * Old cruft
1145  */
1146 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1147 {
1148         int error = 0;
1149
1150         if (!name)
1151                 return -EFAULT;
1152
1153         down_read(&uts_sem);
1154         if (copy_to_user(name, utsname(), sizeof(*name)))
1155                 error = -EFAULT;
1156         up_read(&uts_sem);
1157
1158         if (!error && override_architecture(name))
1159                 error = -EFAULT;
1160         return error;
1161 }
1162
1163 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1164 {
1165         int error;
1166
1167         if (!name)
1168                 return -EFAULT;
1169         if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1170                 return -EFAULT;
1171
1172         down_read(&uts_sem);
1173         error = __copy_to_user(&name->sysname, &utsname()->sysname,
1174                                __OLD_UTS_LEN);
1175         error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1176         error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1177                                 __OLD_UTS_LEN);
1178         error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1179         error |= __copy_to_user(&name->release, &utsname()->release,
1180                                 __OLD_UTS_LEN);
1181         error |= __put_user(0, name->release + __OLD_UTS_LEN);
1182         error |= __copy_to_user(&name->version, &utsname()->version,
1183                                 __OLD_UTS_LEN);
1184         error |= __put_user(0, name->version + __OLD_UTS_LEN);
1185         error |= __copy_to_user(&name->machine, &utsname()->machine,
1186                                 __OLD_UTS_LEN);
1187         error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1188         up_read(&uts_sem);
1189
1190         if (!error && override_architecture(name))
1191                 error = -EFAULT;
1192         return error ? -EFAULT : 0;
1193 }
1194 #endif
1195
1196 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1197 {
1198         int errno;
1199         char tmp[__NEW_UTS_LEN];
1200
1201         if (!capable(CAP_SYS_ADMIN))
1202                 return -EPERM;
1203         if (len < 0 || len > __NEW_UTS_LEN)
1204                 return -EINVAL;
1205         down_write(&uts_sem);
1206         errno = -EFAULT;
1207         if (!copy_from_user(tmp, name, len)) {
1208                 struct new_utsname *u = utsname();
1209
1210                 memcpy(u->nodename, tmp, len);
1211                 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1212                 errno = 0;
1213         }
1214         up_write(&uts_sem);
1215         return errno;
1216 }
1217
1218 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1219
1220 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1221 {
1222         int i, errno;
1223         struct new_utsname *u;
1224
1225         if (len < 0)
1226                 return -EINVAL;
1227         down_read(&uts_sem);
1228         u = utsname();
1229         i = 1 + strlen(u->nodename);
1230         if (i > len)
1231                 i = len;
1232         errno = 0;
1233         if (copy_to_user(name, u->nodename, i))
1234                 errno = -EFAULT;
1235         up_read(&uts_sem);
1236         return errno;
1237 }
1238
1239 #endif
1240
1241 /*
1242  * Only setdomainname; getdomainname can be implemented by calling
1243  * uname()
1244  */
1245 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1246 {
1247         int errno;
1248         char tmp[__NEW_UTS_LEN];
1249
1250         if (!capable(CAP_SYS_ADMIN))
1251                 return -EPERM;
1252         if (len < 0 || len > __NEW_UTS_LEN)
1253                 return -EINVAL;
1254
1255         down_write(&uts_sem);
1256         errno = -EFAULT;
1257         if (!copy_from_user(tmp, name, len)) {
1258                 struct new_utsname *u = utsname();
1259
1260                 memcpy(u->domainname, tmp, len);
1261                 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1262                 errno = 0;
1263         }
1264         up_write(&uts_sem);
1265         return errno;
1266 }
1267
1268 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1269 {
1270         if (resource >= RLIM_NLIMITS)
1271                 return -EINVAL;
1272         else {
1273                 struct rlimit value;
1274                 task_lock(current->group_leader);
1275                 value = current->signal->rlim[resource];
1276                 task_unlock(current->group_leader);
1277                 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1278         }
1279 }
1280
1281 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1282
1283 /*
1284  *      Back compatibility for getrlimit. Needed for some apps.
1285  */
1286
1287 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1288                 struct rlimit __user *, rlim)
1289 {
1290         struct rlimit x;
1291         if (resource >= RLIM_NLIMITS)
1292                 return -EINVAL;
1293
1294         task_lock(current->group_leader);
1295         x = current->signal->rlim[resource];
1296         task_unlock(current->group_leader);
1297         if (x.rlim_cur > 0x7FFFFFFF)
1298                 x.rlim_cur = 0x7FFFFFFF;
1299         if (x.rlim_max > 0x7FFFFFFF)
1300                 x.rlim_max = 0x7FFFFFFF;
1301         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1302 }
1303
1304 #endif
1305
1306 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1307 {
1308         struct rlimit new_rlim, *old_rlim;
1309         int retval;
1310
1311         if (resource >= RLIM_NLIMITS)
1312                 return -EINVAL;
1313         if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1314                 return -EFAULT;
1315         if (new_rlim.rlim_cur > new_rlim.rlim_max)
1316                 return -EINVAL;
1317         old_rlim = current->signal->rlim + resource;
1318         if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1319             !capable(CAP_SYS_RESOURCE))
1320                 return -EPERM;
1321         if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > sysctl_nr_open)
1322                 return -EPERM;
1323
1324         retval = security_task_setrlimit(resource, &new_rlim);
1325         if (retval)
1326                 return retval;
1327
1328         if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1329                 /*
1330                  * The caller is asking for an immediate RLIMIT_CPU
1331                  * expiry.  But we use the zero value to mean "it was
1332                  * never set".  So let's cheat and make it one second
1333                  * instead
1334                  */
1335                 new_rlim.rlim_cur = 1;
1336         }
1337
1338         task_lock(current->group_leader);
1339         *old_rlim = new_rlim;
1340         task_unlock(current->group_leader);
1341
1342         if (resource != RLIMIT_CPU)
1343                 goto out;
1344
1345         /*
1346          * RLIMIT_CPU handling.   Note that the kernel fails to return an error
1347          * code if it rejected the user's attempt to set RLIMIT_CPU.  This is a
1348          * very long-standing error, and fixing it now risks breakage of
1349          * applications, so we live with it
1350          */
1351         if (new_rlim.rlim_cur == RLIM_INFINITY)
1352                 goto out;
1353
1354         update_rlimit_cpu(new_rlim.rlim_cur);
1355 out:
1356         return 0;
1357 }
1358
1359 /*
1360  * It would make sense to put struct rusage in the task_struct,
1361  * except that would make the task_struct be *really big*.  After
1362  * task_struct gets moved into malloc'ed memory, it would
1363  * make sense to do this.  It will make moving the rest of the information
1364  * a lot simpler!  (Which we're not doing right now because we're not
1365  * measuring them yet).
1366  *
1367  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1368  * races with threads incrementing their own counters.  But since word
1369  * reads are atomic, we either get new values or old values and we don't
1370  * care which for the sums.  We always take the siglock to protect reading
1371  * the c* fields from p->signal from races with exit.c updating those
1372  * fields when reaping, so a sample either gets all the additions of a
1373  * given child after it's reaped, or none so this sample is before reaping.
1374  *
1375  * Locking:
1376  * We need to take the siglock for CHILDEREN, SELF and BOTH
1377  * for  the cases current multithreaded, non-current single threaded
1378  * non-current multithreaded.  Thread traversal is now safe with
1379  * the siglock held.
1380  * Strictly speaking, we donot need to take the siglock if we are current and
1381  * single threaded,  as no one else can take our signal_struct away, no one
1382  * else can  reap the  children to update signal->c* counters, and no one else
1383  * can race with the signal-> fields. If we do not take any lock, the
1384  * signal-> fields could be read out of order while another thread was just
1385  * exiting. So we should  place a read memory barrier when we avoid the lock.
1386  * On the writer side,  write memory barrier is implied in  __exit_signal
1387  * as __exit_signal releases  the siglock spinlock after updating the signal->
1388  * fields. But we don't do this yet to keep things simple.
1389  *
1390  */
1391
1392 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1393 {
1394         r->ru_nvcsw += t->nvcsw;
1395         r->ru_nivcsw += t->nivcsw;
1396         r->ru_minflt += t->min_flt;
1397         r->ru_majflt += t->maj_flt;
1398         r->ru_inblock += task_io_get_inblock(t);
1399         r->ru_oublock += task_io_get_oublock(t);
1400 }
1401
1402 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1403 {
1404         struct task_struct *t;
1405         unsigned long flags;
1406         cputime_t tgutime, tgstime, utime, stime;
1407         unsigned long maxrss = 0;
1408
1409         memset((char *) r, 0, sizeof *r);
1410         utime = stime = cputime_zero;
1411
1412         if (who == RUSAGE_THREAD) {
1413                 task_times(current, &utime, &stime);
1414                 accumulate_thread_rusage(p, r);
1415                 maxrss = p->signal->maxrss;
1416                 goto out;
1417         }
1418
1419         if (!lock_task_sighand(p, &flags))
1420                 return;
1421
1422         switch (who) {
1423                 case RUSAGE_BOTH:
1424                 case RUSAGE_CHILDREN:
1425                         utime = p->signal->cutime;
1426                         stime = p->signal->cstime;
1427                         r->ru_nvcsw = p->signal->cnvcsw;
1428                         r->ru_nivcsw = p->signal->cnivcsw;
1429                         r->ru_minflt = p->signal->cmin_flt;
1430                         r->ru_majflt = p->signal->cmaj_flt;
1431                         r->ru_inblock = p->signal->cinblock;
1432                         r->ru_oublock = p->signal->coublock;
1433                         maxrss = p->signal->cmaxrss;
1434
1435                         if (who == RUSAGE_CHILDREN)
1436                                 break;
1437
1438                 case RUSAGE_SELF:
1439                         thread_group_times(p, &tgutime, &tgstime);
1440                         utime = cputime_add(utime, tgutime);
1441                         stime = cputime_add(stime, tgstime);
1442                         r->ru_nvcsw += p->signal->nvcsw;
1443                         r->ru_nivcsw += p->signal->nivcsw;
1444                         r->ru_minflt += p->signal->min_flt;
1445                         r->ru_majflt += p->signal->maj_flt;
1446                         r->ru_inblock += p->signal->inblock;
1447                         r->ru_oublock += p->signal->oublock;
1448                         if (maxrss < p->signal->maxrss)
1449                                 maxrss = p->signal->maxrss;
1450                         t = p;
1451                         do {
1452                                 accumulate_thread_rusage(t, r);
1453                                 t = next_thread(t);
1454                         } while (t != p);
1455                         break;
1456
1457                 default:
1458                         BUG();
1459         }
1460         unlock_task_sighand(p, &flags);
1461
1462 out:
1463         cputime_to_timeval(utime, &r->ru_utime);
1464         cputime_to_timeval(stime, &r->ru_stime);
1465
1466         if (who != RUSAGE_CHILDREN) {
1467                 struct mm_struct *mm = get_task_mm(p);
1468                 if (mm) {
1469                         setmax_mm_hiwater_rss(&maxrss, mm);
1470                         mmput(mm);
1471                 }
1472         }
1473         r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1474 }
1475
1476 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1477 {
1478         struct rusage r;
1479         k_getrusage(p, who, &r);
1480         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1481 }
1482
1483 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1484 {
1485         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1486             who != RUSAGE_THREAD)
1487                 return -EINVAL;
1488         return getrusage(current, who, ru);
1489 }
1490
1491 SYSCALL_DEFINE1(umask, int, mask)
1492 {
1493         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1494         return mask;
1495 }
1496
1497 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1498                 unsigned long, arg4, unsigned long, arg5)
1499 {
1500         struct task_struct *me = current;
1501         unsigned char comm[sizeof(me->comm)];
1502         long error;
1503
1504         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1505         if (error != -ENOSYS)
1506                 return error;
1507
1508         error = 0;
1509         switch (option) {
1510                 case PR_SET_PDEATHSIG:
1511                         if (!valid_signal(arg2)) {
1512                                 error = -EINVAL;
1513                                 break;
1514                         }
1515                         me->pdeath_signal = arg2;
1516                         error = 0;
1517                         break;
1518                 case PR_GET_PDEATHSIG:
1519                         error = put_user(me->pdeath_signal, (int __user *)arg2);
1520                         break;
1521                 case PR_GET_DUMPABLE:
1522                         error = get_dumpable(me->mm);
1523                         break;
1524                 case PR_SET_DUMPABLE:
1525                         if (arg2 < 0 || arg2 > 1) {
1526                                 error = -EINVAL;
1527                                 break;
1528                         }
1529                         set_dumpable(me->mm, arg2);
1530                         error = 0;
1531                         break;
1532
1533                 case PR_SET_UNALIGN:
1534                         error = SET_UNALIGN_CTL(me, arg2);
1535                         break;
1536                 case PR_GET_UNALIGN:
1537                         error = GET_UNALIGN_CTL(me, arg2);
1538                         break;
1539                 case PR_SET_FPEMU:
1540                         error = SET_FPEMU_CTL(me, arg2);
1541                         break;
1542                 case PR_GET_FPEMU:
1543                         error = GET_FPEMU_CTL(me, arg2);
1544                         break;
1545                 case PR_SET_FPEXC:
1546                         error = SET_FPEXC_CTL(me, arg2);
1547                         break;
1548                 case PR_GET_FPEXC:
1549                         error = GET_FPEXC_CTL(me, arg2);
1550                         break;
1551                 case PR_GET_TIMING:
1552                         error = PR_TIMING_STATISTICAL;
1553                         break;
1554                 case PR_SET_TIMING:
1555                         if (arg2 != PR_TIMING_STATISTICAL)
1556                                 error = -EINVAL;
1557                         else
1558                                 error = 0;
1559                         break;
1560
1561                 case PR_SET_NAME:
1562                         comm[sizeof(me->comm)-1] = 0;
1563                         if (strncpy_from_user(comm, (char __user *)arg2,
1564                                               sizeof(me->comm) - 1) < 0)
1565                                 return -EFAULT;
1566                         set_task_comm(me, comm);
1567                         return 0;
1568                 case PR_GET_NAME:
1569                         get_task_comm(comm, me);
1570                         if (copy_to_user((char __user *)arg2, comm,
1571                                          sizeof(comm)))
1572                                 return -EFAULT;
1573                         return 0;
1574                 case PR_GET_ENDIAN:
1575                         error = GET_ENDIAN(me, arg2);
1576                         break;
1577                 case PR_SET_ENDIAN:
1578                         error = SET_ENDIAN(me, arg2);
1579                         break;
1580
1581                 case PR_GET_SECCOMP:
1582                         error = prctl_get_seccomp();
1583                         break;
1584                 case PR_SET_SECCOMP:
1585                         error = prctl_set_seccomp(arg2);
1586                         break;
1587                 case PR_GET_TSC:
1588                         error = GET_TSC_CTL(arg2);
1589                         break;
1590                 case PR_SET_TSC:
1591                         error = SET_TSC_CTL(arg2);
1592                         break;
1593                 case PR_TASK_PERF_EVENTS_DISABLE:
1594                         error = perf_event_task_disable();
1595                         break;
1596                 case PR_TASK_PERF_EVENTS_ENABLE:
1597                         error = perf_event_task_enable();
1598                         break;
1599                 case PR_GET_TIMERSLACK:
1600                         error = current->timer_slack_ns;
1601                         break;
1602                 case PR_SET_TIMERSLACK:
1603                         if (arg2 <= 0)
1604                                 current->timer_slack_ns =
1605                                         current->default_timer_slack_ns;
1606                         else
1607                                 current->timer_slack_ns = arg2;
1608                         error = 0;
1609                         break;
1610                 case PR_MCE_KILL:
1611                         if (arg4 | arg5)
1612                                 return -EINVAL;
1613                         switch (arg2) {
1614                         case PR_MCE_KILL_CLEAR:
1615                                 if (arg3 != 0)
1616                                         return -EINVAL;
1617                                 current->flags &= ~PF_MCE_PROCESS;
1618                                 break;
1619                         case PR_MCE_KILL_SET:
1620                                 current->flags |= PF_MCE_PROCESS;
1621                                 if (arg3 == PR_MCE_KILL_EARLY)
1622                                         current->flags |= PF_MCE_EARLY;
1623                                 else if (arg3 == PR_MCE_KILL_LATE)
1624                                         current->flags &= ~PF_MCE_EARLY;
1625                                 else if (arg3 == PR_MCE_KILL_DEFAULT)
1626                                         current->flags &=
1627                                                 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
1628                                 else
1629                                         return -EINVAL;
1630                                 break;
1631                         default:
1632                                 return -EINVAL;
1633                         }
1634                         error = 0;
1635                         break;
1636                 case PR_MCE_KILL_GET:
1637                         if (arg2 | arg3 | arg4 | arg5)
1638                                 return -EINVAL;
1639                         if (current->flags & PF_MCE_PROCESS)
1640                                 error = (current->flags & PF_MCE_EARLY) ?
1641                                         PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
1642                         else
1643                                 error = PR_MCE_KILL_DEFAULT;
1644                         break;
1645                 default:
1646                         error = -EINVAL;
1647                         break;
1648         }
1649         return error;
1650 }
1651
1652 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1653                 struct getcpu_cache __user *, unused)
1654 {
1655         int err = 0;
1656         int cpu = raw_smp_processor_id();
1657         if (cpup)
1658                 err |= put_user(cpu, cpup);
1659         if (nodep)
1660                 err |= put_user(cpu_to_node(cpu), nodep);
1661         return err ? -EFAULT : 0;
1662 }
1663
1664 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1665
1666 static void argv_cleanup(char **argv, char **envp)
1667 {
1668         argv_free(argv);
1669 }
1670
1671 /**
1672  * orderly_poweroff - Trigger an orderly system poweroff
1673  * @force: force poweroff if command execution fails
1674  *
1675  * This may be called from any context to trigger a system shutdown.
1676  * If the orderly shutdown fails, it will force an immediate shutdown.
1677  */
1678 int orderly_poweroff(bool force)
1679 {
1680         int argc;
1681         char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1682         static char *envp[] = {
1683                 "HOME=/",
1684                 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1685                 NULL
1686         };
1687         int ret = -ENOMEM;
1688         struct subprocess_info *info;
1689
1690         if (argv == NULL) {
1691                 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1692                        __func__, poweroff_cmd);
1693                 goto out;
1694         }
1695
1696         info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1697         if (info == NULL) {
1698                 argv_free(argv);
1699                 goto out;
1700         }
1701
1702         call_usermodehelper_setcleanup(info, argv_cleanup);
1703
1704         ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1705
1706   out:
1707         if (ret && force) {
1708                 printk(KERN_WARNING "Failed to start orderly shutdown: "
1709                        "forcing the issue\n");
1710
1711                 /* I guess this should try to kick off some daemon to
1712                    sync and poweroff asap.  Or not even bother syncing
1713                    if we're doing an emergency shutdown? */
1714                 emergency_sync();
1715                 kernel_power_off();
1716         }
1717
1718         return ret;
1719 }
1720 EXPORT_SYMBOL_GPL(orderly_poweroff);