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/smp_lock.h>
  12 #include <linux/notifier.h>
  13 #include <linux/reboot.h>
  14 #include <linux/prctl.h>
  15 #include <linux/highuid.h>
  16 #include <linux/fs.h>
  17 #include <linux/resource.h>
  18 #include <linux/kernel.h>
  19 #include <linux/kexec.h>
  20 #include <linux/workqueue.h>
  21 #include <linux/capability.h>
  22 #include <linux/device.h>
  23 #include <linux/key.h>
  24 #include <linux/times.h>
  25 #include <linux/posix-timers.h>
  26 #include <linux/security.h>
  27 #include <linux/dcookies.h>
  28 #include <linux/suspend.h>
  29 #include <linux/tty.h>
  30 #include <linux/signal.h>
  31 #include <linux/cn_proc.h>
  32 #include <linux/getcpu.h>
  33 #include <linux/task_io_accounting_ops.h>
  34 #include <linux/seccomp.h>
  35 #include <linux/cpu.h>
  36 #include <linux/ptrace.h>
  37
  38 #include <linux/compat.h>
  39 #include <linux/syscalls.h>
  40 #include <linux/kprobes.h>
  41 #include <linux/user_namespace.h>
  42
  43 #include <asm/uaccess.h>
  44 #include <asm/io.h>
  45 #include <asm/unistd.h>
  46
  47 #ifndef SET_UNALIGN_CTL
  48 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  49 #endif
  50 #ifndef GET_UNALIGN_CTL
  51 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  52 #endif
  53 #ifndef SET_FPEMU_CTL
  54 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
  55 #endif
  56 #ifndef GET_FPEMU_CTL
  57 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
  58 #endif
  59 #ifndef SET_FPEXC_CTL
  60 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
  61 #endif
  62 #ifndef GET_FPEXC_CTL
  63 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
  64 #endif
  65 #ifndef GET_ENDIAN
  66 # define GET_ENDIAN(a,b)        (-EINVAL)
  67 #endif
  68 #ifndef SET_ENDIAN
  69 # define SET_ENDIAN(a,b)        (-EINVAL)
  70 #endif
  71 #ifndef GET_TSC_CTL
  72 # define GET_TSC_CTL(a)         (-EINVAL)
  73 #endif
  74 #ifndef SET_TSC_CTL
  75 # define SET_TSC_CTL(a)         (-EINVAL)
  76 #endif
  77
  78 /*
  79  * this is where the system-wide overflow UID and GID are defined, for
  80  * architectures that now have 32-bit UID/GID but didn't in the past
  81  */
  82
  83 int overflowuid = DEFAULT_OVERFLOWUID;
  84 int overflowgid = DEFAULT_OVERFLOWGID;
  85
  86 #ifdef CONFIG_UID16
  87 EXPORT_SYMBOL(overflowuid);
  88 EXPORT_SYMBOL(overflowgid);
  89 #endif
  90
  91 /*
  92  * the same as above, but for filesystems which can only store a 16-bit
  93  * UID and GID. as such, this is needed on all architectures
  94  */
  95
  96 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
  97 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
  98
  99 EXPORT_SYMBOL(fs_overflowuid);
 100 EXPORT_SYMBOL(fs_overflowgid);
 101
 102 /*
 103  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
 104  */
 105
 106 int C_A_D = 1;
 107 struct pid *cad_pid;
 108 EXPORT_SYMBOL(cad_pid);
 109
 110 /*
 111  * If set, this is used for preparing the system to power off.
 112  */
 113
 114 void (*pm_power_off_prepare)(void);
 115
 116 /*
 117  * set the priority of a task
 118  * - the caller must hold the RCU read lock
 119  */
 120 static int set_one_prio(struct task_struct *p, int niceval, int error)
 121 {
 122         const struct cred *cred = current_cred(), *pcred = __task_cred(p);
 123         int no_nice;
 124
 125         if (pcred->uid  != cred->euid &&
 126             pcred->euid != cred->euid && !capable(CAP_SYS_NICE)) {
 127                 error = -EPERM;
 128                 goto out;
 129         }
 130         if (niceval < task_nice(p) && !can_nice(p, niceval)) {
 131                 error = -EACCES;
 132                 goto out;
 133         }
 134         no_nice = security_task_setnice(p, niceval);
 135         if (no_nice) {
 136                 error = no_nice;
 137                 goto out;
 138         }
 139         if (error == -ESRCH)
 140                 error = 0;
 141         set_user_nice(p, niceval);
 142 out:
 143         return error;
 144 }
 145
 146 asmlinkage long sys_setpriority(int which, int who, int niceval)
 147 {
 148         struct task_struct *g, *p;
 149         struct user_struct *user;
 150         const struct cred *cred = current_cred();
 151         int error = -EINVAL;
 152         struct pid *pgrp;
 153
 154         if (which > PRIO_USER || which < PRIO_PROCESS)
 155                 goto out;
 156
 157         /* normalize: avoid signed division (rounding problems) */
 158         error = -ESRCH;
 159         if (niceval < -20)
 160                 niceval = -20;
 161         if (niceval > 19)
 162                 niceval = 19;
 163
 164         read_lock(&tasklist_lock);
 165         switch (which) {
 166                 case PRIO_PROCESS:
 167                         if (who)
 168                                 p = find_task_by_vpid(who);
 169                         else
 170                                 p = current;
 171                         if (p)
 172                                 error = set_one_prio(p, niceval, error);
 173                         break;
 174                 case PRIO_PGRP:
 175                         if (who)
 176                                 pgrp = find_vpid(who);
 177                         else
 178                                 pgrp = task_pgrp(current);
 179                         do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
 180                                 error = set_one_prio(p, niceval, error);
 181                         } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
 182                         break;
 183                 case PRIO_USER:
 184                         user = (struct user_struct *) cred->user;
 185                         if (!who)
 186                                 who = cred->uid;
 187                         else if ((who != cred->uid) &&
 188                                  !(user = find_user(who)))
 189                                 goto out_unlock;        /* No processes for this user */
 190
 191                         do_each_thread(g, p)
 192                                 if (__task_cred(p)->uid == who)
 193                                         error = set_one_prio(p, niceval, error);
 194                         while_each_thread(g, p);
 195                         if (who != cred->uid)
 196                                 free_uid(user);         /* For find_user() */
 197                         break;
 198         }
 199 out_unlock:
 200         read_unlock(&tasklist_lock);
 201 out:
 202         return error;
 203 }
 204
 205 /*
 206  * Ugh. To avoid negative return values, "getpriority()" will
 207  * not return the normal nice-value, but a negated value that
 208  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 209  * to stay compatible.
 210  */
 211 asmlinkage long sys_getpriority(int which, int who)
 212 {
 213         struct task_struct *g, *p;
 214         struct user_struct *user;
 215         const struct cred *cred = current_cred();
 216         long niceval, retval = -ESRCH;
 217         struct pid *pgrp;
 218
 219         if (which > PRIO_USER || which < PRIO_PROCESS)
 220                 return -EINVAL;
 221
 222         read_lock(&tasklist_lock);
 223         switch (which) {
 224                 case PRIO_PROCESS:
 225                         if (who)
 226                                 p = find_task_by_vpid(who);
 227                         else
 228                                 p = current;
 229                         if (p) {
 230                                 niceval = 20 - task_nice(p);
 231                                 if (niceval > retval)
 232                                         retval = niceval;
 233                         }
 234                         break;
 235                 case PRIO_PGRP:
 236                         if (who)
 237                                 pgrp = find_vpid(who);
 238                         else
 239                                 pgrp = task_pgrp(current);
 240                         do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
 241                                 niceval = 20 - task_nice(p);
 242                                 if (niceval > retval)
 243                                         retval = niceval;
 244                         } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
 245                         break;
 246                 case PRIO_USER:
 247                         user = (struct user_struct *) cred->user;
 248                         if (!who)
 249                                 who = cred->uid;
 250                         else if ((who != cred->uid) &&
 251                                  !(user = find_user(who)))
 252                                 goto out_unlock;        /* No processes for this user */
 253
 254                         do_each_thread(g, p)
 255                                 if (__task_cred(p)->uid == who) {
 256                                         niceval = 20 - task_nice(p);
 257                                         if (niceval > retval)
 258                                                 retval = niceval;
 259                                 }
 260                         while_each_thread(g, p);
 261                         if (who != cred->uid)
 262                                 free_uid(user);         /* for find_user() */
 263                         break;
 264         }
 265 out_unlock:
 266         read_unlock(&tasklist_lock);
 267
 268         return retval;
 269 }
 270
 271 /**
 272  *      emergency_restart - reboot the system
 273  *
 274  *      Without shutting down any hardware or taking any locks
 275  *      reboot the system.  This is called when we know we are in
 276  *      trouble so this is our best effort to reboot.  This is
 277  *      safe to call in interrupt context.
 278  */
 279 void emergency_restart(void)
 280 {
 281         machine_emergency_restart();
 282 }
 283 EXPORT_SYMBOL_GPL(emergency_restart);
 284
 285 void kernel_restart_prepare(char *cmd)
 286 {
 287         blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
 288         system_state = SYSTEM_RESTART;
 289         device_shutdown();
 290         sysdev_shutdown();
 291 }
 292
 293 /**
 294  *      kernel_restart - reboot the system
 295  *      @cmd: pointer to buffer containing command to execute for restart
 296  *              or %NULL
 297  *
 298  *      Shutdown everything and perform a clean reboot.
 299  *      This is not safe to call in interrupt context.
 300  */
 301 void kernel_restart(char *cmd)
 302 {
 303         kernel_restart_prepare(cmd);
 304         if (!cmd)
 305                 printk(KERN_EMERG "Restarting system.\n");
 306         else
 307                 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
 308         machine_restart(cmd);
 309 }
 310 EXPORT_SYMBOL_GPL(kernel_restart);
 311
 312 static void kernel_shutdown_prepare(enum system_states state)
 313 {
 314         blocking_notifier_call_chain(&reboot_notifier_list,
 315                 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
 316         system_state = state;
 317         device_shutdown();
 318 }
 319 /**
 320  *      kernel_halt - halt the system
 321  *
 322  *      Shutdown everything and perform a clean system halt.
 323  */
 324 void kernel_halt(void)
 325 {
 326         kernel_shutdown_prepare(SYSTEM_HALT);
 327         sysdev_shutdown();
 328         printk(KERN_EMERG "System halted.\n");
 329         machine_halt();
 330 }
 331
 332 EXPORT_SYMBOL_GPL(kernel_halt);
 333
 334 /**
 335  *      kernel_power_off - power_off the system
 336  *
 337  *      Shutdown everything and perform a clean system power_off.
 338  */
 339 void kernel_power_off(void)
 340 {
 341         kernel_shutdown_prepare(SYSTEM_POWER_OFF);
 342         if (pm_power_off_prepare)
 343                 pm_power_off_prepare();
 344         disable_nonboot_cpus();
 345         sysdev_shutdown();
 346         printk(KERN_EMERG "Power down.\n");
 347         machine_power_off();
 348 }
 349 EXPORT_SYMBOL_GPL(kernel_power_off);
 350 /*
 351  * Reboot system call: for obvious reasons only root may call it,
 352  * and even root needs to set up some magic numbers in the registers
 353  * so that some mistake won't make this reboot the whole machine.
 354  * You can also set the meaning of the ctrl-alt-del-key here.
 355  *
 356  * reboot doesn't sync: do that yourself before calling this.
 357  */
 358 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
 359 {
 360         char buffer[256];
 361
 362         /* We only trust the superuser with rebooting the system. */
 363         if (!capable(CAP_SYS_BOOT))
 364                 return -EPERM;
 365
 366         /* For safety, we require "magic" arguments. */
 367         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 368             (magic2 != LINUX_REBOOT_MAGIC2 &&
 369                         magic2 != LINUX_REBOOT_MAGIC2A &&
 370                         magic2 != LINUX_REBOOT_MAGIC2B &&
 371                         magic2 != LINUX_REBOOT_MAGIC2C))
 372                 return -EINVAL;
 373
 374         /* Instead of trying to make the power_off code look like
 375          * halt when pm_power_off is not set do it the easy way.
 376          */
 377         if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
 378                 cmd = LINUX_REBOOT_CMD_HALT;
 379
 380         lock_kernel();
 381         switch (cmd) {
 382         case LINUX_REBOOT_CMD_RESTART:
 383                 kernel_restart(NULL);
 384                 break;
 385
 386         case LINUX_REBOOT_CMD_CAD_ON:
 387                 C_A_D = 1;
 388                 break;
 389
 390         case LINUX_REBOOT_CMD_CAD_OFF:
 391                 C_A_D = 0;
 392                 break;
 393
 394         case LINUX_REBOOT_CMD_HALT:
 395                 kernel_halt();
 396                 unlock_kernel();
 397                 do_exit(0);
 398                 break;
 399
 400         case LINUX_REBOOT_CMD_POWER_OFF:
 401                 kernel_power_off();
 402                 unlock_kernel();
 403                 do_exit(0);
 404                 break;
 405
 406         case LINUX_REBOOT_CMD_RESTART2:
 407                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 408                         unlock_kernel();
 409                         return -EFAULT;
 410                 }
 411                 buffer[sizeof(buffer) - 1] = '\0';
 412
 413                 kernel_restart(buffer);
 414                 break;
 415
 416 #ifdef CONFIG_KEXEC
 417         case LINUX_REBOOT_CMD_KEXEC:
 418                 {
 419                         int ret;
 420                         ret = kernel_kexec();
 421                         unlock_kernel();
 422                         return ret;
 423                 }
 424 #endif
 425
 426 #ifdef CONFIG_HIBERNATION
 427         case LINUX_REBOOT_CMD_SW_SUSPEND:
 428                 {
 429                         int ret = hibernate();
 430                         unlock_kernel();
 431                         return ret;
 432                 }
 433 #endif
 434
 435         default:
 436                 unlock_kernel();
 437                 return -EINVAL;
 438         }
 439         unlock_kernel();
 440         return 0;
 441 }
 442
 443 static void deferred_cad(struct work_struct *dummy)
 444 {
 445         kernel_restart(NULL);
 446 }
 447
 448 /*
 449  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 450  * As it's called within an interrupt, it may NOT sync: the only choice
 451  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 452  */
 453 void ctrl_alt_del(void)
 454 {
 455         static DECLARE_WORK(cad_work, deferred_cad);
 456
 457         if (C_A_D)
 458                 schedule_work(&cad_work);
 459         else
 460                 kill_cad_pid(SIGINT, 1);
 461 }
 462
 463 /*
 464  * Unprivileged users may change the real gid to the effective gid
 465  * or vice versa.  (BSD-style)
 466  *
 467  * If you set the real gid at all, or set the effective gid to a value not
 468  * equal to the real gid, then the saved gid is set to the new effective gid.
 469  *
 470  * This makes it possible for a setgid program to completely drop its
 471  * privileges, which is often a useful assertion to make when you are doing
 472  * a security audit over a program.
 473  *
 474  * The general idea is that a program which uses just setregid() will be
 475  * 100% compatible with BSD.  A program which uses just setgid() will be
 476  * 100% compatible with POSIX with saved IDs.
 477  *
 478  * SMP: There are not races, the GIDs are checked only by filesystem
 479  *      operations (as far as semantic preservation is concerned).
 480  */
 481 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 482 {
 483         const struct cred *old;
 484         struct cred *new;
 485         int retval;
 486
 487         new = prepare_creds();
 488         if (!new)
 489                 return -ENOMEM;
 490         old = current_cred();
 491
 492         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 493         if (retval)
 494                 goto error;
 495
 496         retval = -EPERM;
 497         if (rgid != (gid_t) -1) {
 498                 if (old->gid == rgid ||
 499                     old->egid == rgid ||
 500                     capable(CAP_SETGID))
 501                         new->gid = rgid;
 502                 else
 503                         goto error;
 504         }
 505         if (egid != (gid_t) -1) {
 506                 if (old->gid == egid ||
 507                     old->egid == egid ||
 508                     old->sgid == egid ||
 509                     capable(CAP_SETGID))
 510                         new->egid = egid;
 511                 else
 512                         goto error;
 513         }
 514
 515         if (rgid != (gid_t) -1 ||
 516             (egid != (gid_t) -1 && egid != old->gid))
 517                 new->sgid = new->egid;
 518         new->fsgid = new->egid;
 519
 520         return commit_creds(new);
 521
 522 error:
 523         abort_creds(new);
 524         return retval;
 525 }
 526
 527 /*
 528  * setgid() is implemented like SysV w/ SAVED_IDS
 529  *
 530  * SMP: Same implicit races as above.
 531  */
 532 asmlinkage long sys_setgid(gid_t gid)
 533 {
 534         const struct cred *old;
 535         struct cred *new;
 536         int retval;
 537
 538         new = prepare_creds();
 539         if (!new)
 540                 return -ENOMEM;
 541         old = current_cred();
 542
 543         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 544         if (retval)
 545                 goto error;
 546
 547         retval = -EPERM;
 548         if (capable(CAP_SETGID))
 549                 new->gid = new->egid = new->sgid = new->fsgid = gid;
 550         else if (gid == old->gid || gid == old->sgid)
 551                 new->egid = new->fsgid = gid;
 552         else
 553                 goto error;
 554
 555         return commit_creds(new);
 556
 557 error:
 558         abort_creds(new);
 559         return retval;
 560 }
 561
 562 /*
 563  * change the user struct in a credentials set to match the new UID
 564  */
 565 static int set_user(struct cred *new)
 566 {
 567         struct user_struct *new_user;
 568
 569         new_user = alloc_uid(current_user_ns(), new->uid);
 570         if (!new_user)
 571                 return -EAGAIN;
 572
 573         if (atomic_read(&new_user->processes) >=
 574                                 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
 575                         new_user != INIT_USER) {
 576                 free_uid(new_user);
 577                 return -EAGAIN;
 578         }
 579
 580         free_uid(new->user);
 581         new->user = new_user;
 582         return 0;
 583 }
 584
 585 /*
 586  * Unprivileged users may change the real uid to the effective uid
 587  * or vice versa.  (BSD-style)
 588  *
 589  * If you set the real uid at all, or set the effective uid to a value not
 590  * equal to the real uid, then the saved uid is set to the new effective uid.
 591  *
 592  * This makes it possible for a setuid program to completely drop its
 593  * privileges, which is often a useful assertion to make when you are doing
 594  * a security audit over a program.
 595  *
 596  * The general idea is that a program which uses just setreuid() will be
 597  * 100% compatible with BSD.  A program which uses just setuid() will be
 598  * 100% compatible with POSIX with saved IDs.
 599  */
 600 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 601 {
 602         const struct cred *old;
 603         struct cred *new;
 604         int retval;
 605
 606         new = prepare_creds();
 607         if (!new)
 608                 return -ENOMEM;
 609         old = current_cred();
 610
 611         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 612         if (retval)
 613                 goto error;
 614
 615         retval = -EPERM;
 616         if (ruid != (uid_t) -1) {
 617                 new->uid = ruid;
 618                 if (old->uid != ruid &&
 619                     old->euid != ruid &&
 620                     !capable(CAP_SETUID))
 621                         goto error;
 622         }
 623
 624         if (euid != (uid_t) -1) {
 625                 new->euid = euid;
 626                 if (old->uid != euid &&
 627                     old->euid != euid &&
 628                     old->suid != euid &&
 629                     !capable(CAP_SETUID))
 630                         goto error;
 631         }
 632
 633         retval = -EAGAIN;
 634         if (new->uid != old->uid && set_user(new) < 0)
 635                 goto error;
 636
 637         if (ruid != (uid_t) -1 ||
 638             (euid != (uid_t) -1 && euid != old->uid))
 639                 new->suid = new->euid;
 640         new->fsuid = new->euid;
 641
 642         retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
 643         if (retval < 0)
 644                 goto error;
 645
 646         return commit_creds(new);
 647
 648 error:
 649         abort_creds(new);
 650         return retval;
 651 }
 652
 653 /*
 654  * setuid() is implemented like SysV with SAVED_IDS
 655  *
 656  * Note that SAVED_ID's is deficient in that a setuid root program
 657  * like sendmail, for example, cannot set its uid to be a normal
 658  * user and then switch back, because if you're root, setuid() sets
 659  * the saved uid too.  If you don't like this, blame the bright people
 660  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 661  * will allow a root program to temporarily drop privileges and be able to
 662  * regain them by swapping the real and effective uid.
 663  */
 664 asmlinkage long sys_setuid(uid_t uid)
 665 {
 666         const struct cred *old;
 667         struct cred *new;
 668         int retval;
 669
 670         new = prepare_creds();
 671         if (!new)
 672                 return -ENOMEM;
 673         old = current_cred();
 674
 675         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 676         if (retval)
 677                 goto error;
 678
 679         retval = -EPERM;
 680         if (capable(CAP_SETUID)) {
 681                 new->suid = new->uid = uid;
 682                 if (uid != old->uid && set_user(new) < 0) {
 683                         retval = -EAGAIN;
 684                         goto error;
 685                 }
 686         } else if (uid != old->uid && uid != new->suid) {
 687                 goto error;
 688         }
 689
 690         new->fsuid = new->euid = uid;
 691
 692         retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
 693         if (retval < 0)
 694                 goto error;
 695
 696         return commit_creds(new);
 697
 698 error:
 699         abort_creds(new);
 700         return retval;
 701 }
 702
 703
 704 /*
 705  * This function implements a generic ability to update ruid, euid,
 706  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 707  */
 708 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 709 {
 710         const struct cred *old;
 711         struct cred *new;
 712         int retval;
 713
 714         new = prepare_creds();
 715         if (!new)
 716                 return -ENOMEM;
 717
 718         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 719         if (retval)
 720                 goto error;
 721         old = current_cred();
 722
 723         retval = -EPERM;
 724         if (!capable(CAP_SETUID)) {
 725                 if (ruid != (uid_t) -1 && ruid != old->uid &&
 726                     ruid != old->euid  && ruid != old->suid)
 727                         goto error;
 728                 if (euid != (uid_t) -1 && euid != old->uid &&
 729                     euid != old->euid  && euid != old->suid)
 730                         goto error;
 731                 if (suid != (uid_t) -1 && suid != old->uid &&
 732                     suid != old->euid  && suid != old->suid)
 733                         goto error;
 734         }
 735
 736         retval = -EAGAIN;
 737         if (ruid != (uid_t) -1) {
 738                 new->uid = ruid;
 739                 if (ruid != old->uid && set_user(new) < 0)
 740                         goto error;
 741         }
 742         if (euid != (uid_t) -1)
 743                 new->euid = euid;
 744         if (suid != (uid_t) -1)
 745                 new->suid = suid;
 746         new->fsuid = new->euid;
 747
 748         retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
 749         if (retval < 0)
 750                 goto error;
 751
 752         return commit_creds(new);
 753
 754 error:
 755         abort_creds(new);
 756         return retval;
 757 }
 758
 759 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
 760 {
 761         const struct cred *cred = current_cred();
 762         int retval;
 763
 764         if (!(retval   = put_user(cred->uid,  ruid)) &&
 765             !(retval   = put_user(cred->euid, euid)))
 766                 retval = put_user(cred->suid, suid);
 767
 768         return retval;
 769 }
 770
 771 /*
 772  * Same as above, but for rgid, egid, sgid.
 773  */
 774 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 775 {
 776         const struct cred *old;
 777         struct cred *new;
 778         int retval;
 779
 780         new = prepare_creds();
 781         if (!new)
 782                 return -ENOMEM;
 783         old = current_cred();
 784
 785         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 786         if (retval)
 787                 goto error;
 788
 789         retval = -EPERM;
 790         if (!capable(CAP_SETGID)) {
 791                 if (rgid != (gid_t) -1 && rgid != old->gid &&
 792                     rgid != old->egid  && rgid != old->sgid)
 793                         goto error;
 794                 if (egid != (gid_t) -1 && egid != old->gid &&
 795                     egid != old->egid  && egid != old->sgid)
 796                         goto error;
 797                 if (sgid != (gid_t) -1 && sgid != old->gid &&
 798                     sgid != old->egid  && sgid != old->sgid)
 799                         goto error;
 800         }
 801
 802         if (rgid != (gid_t) -1)
 803                 new->gid = rgid;
 804         if (egid != (gid_t) -1)
 805                 new->egid = egid;
 806         if (sgid != (gid_t) -1)
 807                 new->sgid = sgid;
 808         new->fsgid = new->egid;
 809
 810         return commit_creds(new);
 811
 812 error:
 813         abort_creds(new);
 814         return retval;
 815 }
 816
 817 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
 818 {
 819         const struct cred *cred = current_cred();
 820         int retval;
 821
 822         if (!(retval   = put_user(cred->gid,  rgid)) &&
 823             !(retval   = put_user(cred->egid, egid)))
 824                 retval = put_user(cred->sgid, sgid);
 825
 826         return retval;
 827 }
 828
 829
 830 /*
 831  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 832  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 833  * whatever uid it wants to). It normally shadows "euid", except when
 834  * explicitly set by setfsuid() or for access..
 835  */
 836 asmlinkage long sys_setfsuid(uid_t uid)
 837 {
 838         const struct cred *old;
 839         struct cred *new;
 840         uid_t old_fsuid;
 841
 842         new = prepare_creds();
 843         if (!new)
 844                 return current_fsuid();
 845         old = current_cred();
 846         old_fsuid = old->fsuid;
 847
 848         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS) < 0)
 849                 goto error;
 850
 851         if (uid == old->uid  || uid == old->euid  ||
 852             uid == old->suid || uid == old->fsuid ||
 853             capable(CAP_SETUID)) {
 854                 if (uid != old_fsuid) {
 855                         new->fsuid = uid;
 856                         if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
 857                                 goto change_okay;
 858                 }
 859         }
 860
 861 error:
 862         abort_creds(new);
 863         return old_fsuid;
 864
 865 change_okay:
 866         commit_creds(new);
 867         return old_fsuid;
 868 }
 869
 870 /*
 871  * Samma på svenska..
 872  */
 873 asmlinkage long sys_setfsgid(gid_t gid)
 874 {
 875         const struct cred *old;
 876         struct cred *new;
 877         gid_t old_fsgid;
 878
 879         new = prepare_creds();
 880         if (!new)
 881                 return current_fsgid();
 882         old = current_cred();
 883         old_fsgid = old->fsgid;
 884
 885         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 886                 goto error;
 887
 888         if (gid == old->gid  || gid == old->egid  ||
 889             gid == old->sgid || gid == old->fsgid ||
 890             capable(CAP_SETGID)) {
 891                 if (gid != old_fsgid) {
 892                         new->fsgid = gid;
 893                         goto change_okay;
 894                 }
 895         }
 896
 897 error:
 898         abort_creds(new);
 899         return old_fsgid;
 900
 901 change_okay:
 902         commit_creds(new);
 903         return old_fsgid;
 904 }
 905
 906 void do_sys_times(struct tms *tms)
 907 {
 908         struct task_cputime cputime;
 909         cputime_t cutime, cstime;
 910
 911         thread_group_cputime(current, &cputime);
 912         spin_lock_irq(&current->sighand->siglock);
 913         cutime = current->signal->cutime;
 914         cstime = current->signal->cstime;
 915         spin_unlock_irq(&current->sighand->siglock);
 916         tms->tms_utime = cputime_to_clock_t(cputime.utime);
 917         tms->tms_stime = cputime_to_clock_t(cputime.stime);
 918         tms->tms_cutime = cputime_to_clock_t(cutime);
 919         tms->tms_cstime = cputime_to_clock_t(cstime);
 920 }
 921
 922 asmlinkage long sys_times(struct tms __user * tbuf)
 923 {
 924         if (tbuf) {
 925                 struct tms tmp;
 926
 927                 do_sys_times(&tmp);
 928                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
 929                         return -EFAULT;
 930         }
 931         force_successful_syscall_return();
 932         return (long) jiffies_64_to_clock_t(get_jiffies_64());
 933 }
 934
 935 /*
 936  * This needs some heavy checking ...
 937  * I just haven't the stomach for it. I also don't fully
 938  * understand sessions/pgrp etc. Let somebody who does explain it.
 939  *
 940  * OK, I think I have the protection semantics right.... this is really
 941  * only important on a multi-user system anyway, to make sure one user
 942  * can't send a signal to a process owned by another.  -TYT, 12/12/91
 943  *
 944  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
 945  * LBT 04.03.94
 946  */
 947 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
 948 {
 949         struct task_struct *p;
 950         struct task_struct *group_leader = current->group_leader;
 951         struct pid *pgrp;
 952         int err;
 953
 954         if (!pid)
 955                 pid = task_pid_vnr(group_leader);
 956         if (!pgid)
 957                 pgid = pid;
 958         if (pgid < 0)
 959                 return -EINVAL;
 960
 961         /* From this point forward we keep holding onto the tasklist lock
 962          * so that our parent does not change from under us. -DaveM
 963          */
 964         write_lock_irq(&tasklist_lock);
 965
 966         err = -ESRCH;
 967         p = find_task_by_vpid(pid);
 968         if (!p)
 969                 goto out;
 970
 971         err = -EINVAL;
 972         if (!thread_group_leader(p))
 973                 goto out;
 974
 975         if (same_thread_group(p->real_parent, group_leader)) {
 976                 err = -EPERM;
 977                 if (task_session(p) != task_session(group_leader))
 978                         goto out;
 979                 err = -EACCES;
 980                 if (p->did_exec)
 981                         goto out;
 982         } else {
 983                 err = -ESRCH;
 984                 if (p != group_leader)
 985                         goto out;
 986         }
 987
 988         err = -EPERM;
 989         if (p->signal->leader)
 990                 goto out;
 991
 992         pgrp = task_pid(p);
 993         if (pgid != pid) {
 994                 struct task_struct *g;
 995
 996                 pgrp = find_vpid(pgid);
 997                 g = pid_task(pgrp, PIDTYPE_PGID);
 998                 if (!g || task_session(g) != task_session(group_leader))
 999                         goto out;
1000         }
1001
1002         err = security_task_setpgid(p, pgid);
1003         if (err)
1004                 goto out;
1005
1006         if (task_pgrp(p) != pgrp) {
1007                 change_pid(p, PIDTYPE_PGID, pgrp);
1008                 set_task_pgrp(p, pid_nr(pgrp));
1009         }
1010
1011         err = 0;
1012 out:
1013         /* All paths lead to here, thus we are safe. -DaveM */
1014         write_unlock_irq(&tasklist_lock);
1015         return err;
1016 }
1017
1018 asmlinkage long sys_getpgid(pid_t pid)
1019 {
1020         struct task_struct *p;
1021         struct pid *grp;
1022         int retval;
1023
1024         rcu_read_lock();
1025         if (!pid)
1026                 grp = task_pgrp(current);
1027         else {
1028                 retval = -ESRCH;
1029                 p = find_task_by_vpid(pid);
1030                 if (!p)
1031                         goto out;
1032                 grp = task_pgrp(p);
1033                 if (!grp)
1034                         goto out;
1035
1036                 retval = security_task_getpgid(p);
1037                 if (retval)
1038                         goto out;
1039         }
1040         retval = pid_vnr(grp);
1041 out:
1042         rcu_read_unlock();
1043         return retval;
1044 }
1045
1046 #ifdef __ARCH_WANT_SYS_GETPGRP
1047
1048 asmlinkage long sys_getpgrp(void)
1049 {
1050         return sys_getpgid(0);
1051 }
1052
1053 #endif
1054
1055 asmlinkage long sys_getsid(pid_t pid)
1056 {
1057         struct task_struct *p;
1058         struct pid *sid;
1059         int retval;
1060
1061         rcu_read_lock();
1062         if (!pid)
1063                 sid = task_session(current);
1064         else {
1065                 retval = -ESRCH;
1066                 p = find_task_by_vpid(pid);
1067                 if (!p)
1068                         goto out;
1069                 sid = task_session(p);
1070                 if (!sid)
1071                         goto out;
1072
1073                 retval = security_task_getsid(p);
1074                 if (retval)
1075                         goto out;
1076         }
1077         retval = pid_vnr(sid);
1078 out:
1079         rcu_read_unlock();
1080         return retval;
1081 }
1082
1083 asmlinkage long sys_setsid(void)
1084 {
1085         struct task_struct *group_leader = current->group_leader;
1086         struct pid *sid = task_pid(group_leader);
1087         pid_t session = pid_vnr(sid);
1088         int err = -EPERM;
1089
1090         write_lock_irq(&tasklist_lock);
1091         /* Fail if I am already a session leader */
1092         if (group_leader->signal->leader)
1093                 goto out;
1094
1095         /* Fail if a process group id already exists that equals the
1096          * proposed session id.
1097          */
1098         if (pid_task(sid, PIDTYPE_PGID))
1099                 goto out;
1100
1101         group_leader->signal->leader = 1;
1102         __set_special_pids(sid);
1103
1104         proc_clear_tty(group_leader);
1105
1106         err = session;
1107 out:
1108         write_unlock_irq(&tasklist_lock);
1109         return err;
1110 }
1111
1112 /*
1113  * Supplementary group IDs
1114  */
1115
1116 /* init to 2 - one for init_task, one to ensure it is never freed */
1117 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1118
1119 struct group_info *groups_alloc(int gidsetsize)
1120 {
1121         struct group_info *group_info;
1122         int nblocks;
1123         int i;
1124
1125         nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1126         /* Make sure we always allocate at least one indirect block pointer */
1127         nblocks = nblocks ? : 1;
1128         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1129         if (!group_info)
1130                 return NULL;
1131         group_info->ngroups = gidsetsize;
1132         group_info->nblocks = nblocks;
1133         atomic_set(&group_info->usage, 1);
1134
1135         if (gidsetsize <= NGROUPS_SMALL)
1136                 group_info->blocks[0] = group_info->small_block;
1137         else {
1138                 for (i = 0; i < nblocks; i++) {
1139                         gid_t *b;
1140                         b = (void *)__get_free_page(GFP_USER);
1141                         if (!b)
1142                                 goto out_undo_partial_alloc;
1143                         group_info->blocks[i] = b;
1144                 }
1145         }
1146         return group_info;
1147
1148 out_undo_partial_alloc:
1149         while (--i >= 0) {
1150                 free_page((unsigned long)group_info->blocks[i]);
1151         }
1152         kfree(group_info);
1153         return NULL;
1154 }
1155
1156 EXPORT_SYMBOL(groups_alloc);
1157
1158 void groups_free(struct group_info *group_info)
1159 {
1160         if (group_info->blocks[0] != group_info->small_block) {
1161                 int i;
1162                 for (i = 0; i < group_info->nblocks; i++)
1163                         free_page((unsigned long)group_info->blocks[i]);
1164         }
1165         kfree(group_info);
1166 }
1167
1168 EXPORT_SYMBOL(groups_free);
1169
1170 /* export the group_info to a user-space array */
1171 static int groups_to_user(gid_t __user *grouplist,
1172                           const struct group_info *group_info)
1173 {
1174         int i;
1175         unsigned int count = group_info->ngroups;
1176
1177         for (i = 0; i < group_info->nblocks; i++) {
1178                 unsigned int cp_count = min(NGROUPS_PER_BLOCK, count);
1179                 unsigned int len = cp_count * sizeof(*grouplist);
1180
1181                 if (copy_to_user(grouplist, group_info->blocks[i], len))
1182                         return -EFAULT;
1183
1184                 grouplist += NGROUPS_PER_BLOCK;
1185                 count -= cp_count;
1186         }
1187         return 0;
1188 }
1189
1190 /* fill a group_info from a user-space array - it must be allocated already */
1191 static int groups_from_user(struct group_info *group_info,
1192     gid_t __user *grouplist)
1193 {
1194         int i;
1195         unsigned int count = group_info->ngroups;
1196
1197         for (i = 0; i < group_info->nblocks; i++) {
1198                 unsigned int cp_count = min(NGROUPS_PER_BLOCK, count);
1199                 unsigned int len = cp_count * sizeof(*grouplist);
1200
1201                 if (copy_from_user(group_info->blocks[i], grouplist, len))
1202                         return -EFAULT;
1203
1204                 grouplist += NGROUPS_PER_BLOCK;
1205                 count -= cp_count;
1206         }
1207         return 0;
1208 }
1209
1210 /* a simple Shell sort */
1211 static void groups_sort(struct group_info *group_info)
1212 {
1213         int base, max, stride;
1214         int gidsetsize = group_info->ngroups;
1215
1216         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1217                 ; /* nothing */
1218         stride /= 3;
1219
1220         while (stride) {
1221                 max = gidsetsize - stride;
1222                 for (base = 0; base < max; base++) {
1223                         int left = base;
1224                         int right = left + stride;
1225                         gid_t tmp = GROUP_AT(group_info, right);
1226
1227                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1228                                 GROUP_AT(group_info, right) =
1229                                     GROUP_AT(group_info, left);
1230                                 right = left;
1231                                 left -= stride;
1232                         }
1233                         GROUP_AT(group_info, right) = tmp;
1234                 }
1235                 stride /= 3;
1236         }
1237 }
1238
1239 /* a simple bsearch */
1240 int groups_search(const struct group_info *group_info, gid_t grp)
1241 {
1242         unsigned int left, right;
1243
1244         if (!group_info)
1245                 return 0;
1246
1247         left = 0;
1248         right = group_info->ngroups;
1249         while (left < right) {
1250                 unsigned int mid = (left+right)/2;
1251                 int cmp = grp - GROUP_AT(group_info, mid);
1252                 if (cmp > 0)
1253                         left = mid + 1;
1254                 else if (cmp < 0)
1255                         right = mid;
1256                 else
1257                         return 1;
1258         }
1259         return 0;
1260 }
1261
1262 /**
1263  * set_groups - Change a group subscription in a set of credentials
1264  * @new: The newly prepared set of credentials to alter
1265  * @group_info: The group list to install
1266  *
1267  * Validate a group subscription and, if valid, insert it into a set
1268  * of credentials.
1269  */
1270 int set_groups(struct cred *new, struct group_info *group_info)
1271 {
1272         int retval;
1273
1274         retval = security_task_setgroups(group_info);
1275         if (retval)
1276                 return retval;
1277
1278         put_group_info(new->group_info);
1279         groups_sort(group_info);
1280         get_group_info(group_info);
1281         new->group_info = group_info;
1282         return 0;
1283 }
1284
1285 EXPORT_SYMBOL(set_groups);
1286
1287 /**
1288  * set_current_groups - Change current's group subscription
1289  * @group_info: The group list to impose
1290  *
1291  * Validate a group subscription and, if valid, impose it upon current's task
1292  * security record.
1293  */
1294 int set_current_groups(struct group_info *group_info)
1295 {
1296         struct cred *new;
1297         int ret;
1298
1299         new = prepare_creds();
1300         if (!new)
1301                 return -ENOMEM;
1302
1303         ret = set_groups(new, group_info);
1304         if (ret < 0) {
1305                 abort_creds(new);
1306                 return ret;
1307         }
1308
1309         return commit_creds(new);
1310 }
1311
1312 EXPORT_SYMBOL(set_current_groups);
1313
1314 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1315 {
1316         const struct cred *cred = current_cred();
1317         int i;
1318
1319         if (gidsetsize < 0)
1320                 return -EINVAL;
1321
1322         /* no need to grab task_lock here; it cannot change */
1323         i = cred->group_info->ngroups;
1324         if (gidsetsize) {
1325                 if (i > gidsetsize) {
1326                         i = -EINVAL;
1327                         goto out;
1328                 }
1329                 if (groups_to_user(grouplist, cred->group_info)) {
1330                         i = -EFAULT;
1331                         goto out;
1332                 }
1333         }
1334 out:
1335         return i;
1336 }
1337
1338 /*
1339  *      SMP: Our groups are copy-on-write. We can set them safely
1340  *      without another task interfering.
1341  */
1342
1343 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1344 {
1345         struct group_info *group_info;
1346         int retval;
1347
1348         if (!capable(CAP_SETGID))
1349                 return -EPERM;
1350         if ((unsigned)gidsetsize > NGROUPS_MAX)
1351                 return -EINVAL;
1352
1353         group_info = groups_alloc(gidsetsize);
1354         if (!group_info)
1355                 return -ENOMEM;
1356         retval = groups_from_user(group_info, grouplist);
1357         if (retval) {
1358                 put_group_info(group_info);
1359                 return retval;
1360         }
1361
1362         retval = set_current_groups(group_info);
1363         put_group_info(group_info);
1364
1365         return retval;
1366 }
1367
1368 /*
1369  * Check whether we're fsgid/egid or in the supplemental group..
1370  */
1371 int in_group_p(gid_t grp)
1372 {
1373         const struct cred *cred = current_cred();
1374         int retval = 1;
1375
1376         if (grp != cred->fsgid)
1377                 retval = groups_search(cred->group_info, grp);
1378         return retval;
1379 }
1380
1381 EXPORT_SYMBOL(in_group_p);
1382
1383 int in_egroup_p(gid_t grp)
1384 {
1385         const struct cred *cred = current_cred();
1386         int retval = 1;
1387
1388         if (grp != cred->egid)
1389                 retval = groups_search(cred->group_info, grp);
1390         return retval;
1391 }
1392
1393 EXPORT_SYMBOL(in_egroup_p);
1394
1395 DECLARE_RWSEM(uts_sem);
1396
1397 asmlinkage long sys_newuname(struct new_utsname __user * name)
1398 {
1399         int errno = 0;
1400
1401         down_read(&uts_sem);
1402         if (copy_to_user(name, utsname(), sizeof *name))
1403                 errno = -EFAULT;
1404         up_read(&uts_sem);
1405         return errno;
1406 }
1407
1408 asmlinkage long sys_sethostname(char __user *name, int len)
1409 {
1410         int errno;
1411         char tmp[__NEW_UTS_LEN];
1412
1413         if (!capable(CAP_SYS_ADMIN))
1414                 return -EPERM;
1415         if (len < 0 || len > __NEW_UTS_LEN)
1416                 return -EINVAL;
1417         down_write(&uts_sem);
1418         errno = -EFAULT;
1419         if (!copy_from_user(tmp, name, len)) {
1420                 struct new_utsname *u = utsname();
1421
1422                 memcpy(u->nodename, tmp, len);
1423                 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1424                 errno = 0;
1425         }
1426         up_write(&uts_sem);
1427         return errno;
1428 }
1429
1430 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1431
1432 asmlinkage long sys_gethostname(char __user *name, int len)
1433 {
1434         int i, errno;
1435         struct new_utsname *u;
1436
1437         if (len < 0)
1438                 return -EINVAL;
1439         down_read(&uts_sem);
1440         u = utsname();
1441         i = 1 + strlen(u->nodename);
1442         if (i > len)
1443                 i = len;
1444         errno = 0;
1445         if (copy_to_user(name, u->nodename, i))
1446                 errno = -EFAULT;
1447         up_read(&uts_sem);
1448         return errno;
1449 }
1450
1451 #endif
1452
1453 /*
1454  * Only setdomainname; getdomainname can be implemented by calling
1455  * uname()
1456  */
1457 asmlinkage long sys_setdomainname(char __user *name, int len)
1458 {
1459         int errno;
1460         char tmp[__NEW_UTS_LEN];
1461
1462         if (!capable(CAP_SYS_ADMIN))
1463                 return -EPERM;
1464         if (len < 0 || len > __NEW_UTS_LEN)
1465                 return -EINVAL;
1466
1467         down_write(&uts_sem);
1468         errno = -EFAULT;
1469         if (!copy_from_user(tmp, name, len)) {
1470                 struct new_utsname *u = utsname();
1471
1472                 memcpy(u->domainname, tmp, len);
1473                 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1474                 errno = 0;
1475         }
1476         up_write(&uts_sem);
1477         return errno;
1478 }
1479
1480 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1481 {
1482         if (resource >= RLIM_NLIMITS)
1483                 return -EINVAL;
1484         else {
1485                 struct rlimit value;
1486                 task_lock(current->group_leader);
1487                 value = current->signal->rlim[resource];
1488                 task_unlock(current->group_leader);
1489                 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1490         }
1491 }
1492
1493 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1494
1495 /*
1496  *      Back compatibility for getrlimit. Needed for some apps.
1497  */
1498
1499 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1500 {
1501         struct rlimit x;
1502         if (resource >= RLIM_NLIMITS)
1503                 return -EINVAL;
1504
1505         task_lock(current->group_leader);
1506         x = current->signal->rlim[resource];
1507         task_unlock(current->group_leader);
1508         if (x.rlim_cur > 0x7FFFFFFF)
1509                 x.rlim_cur = 0x7FFFFFFF;
1510         if (x.rlim_max > 0x7FFFFFFF)
1511                 x.rlim_max = 0x7FFFFFFF;
1512         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1513 }
1514
1515 #endif
1516
1517 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1518 {
1519         struct rlimit new_rlim, *old_rlim;
1520         int retval;
1521
1522         if (resource >= RLIM_NLIMITS)
1523                 return -EINVAL;
1524         if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1525                 return -EFAULT;
1526         old_rlim = current->signal->rlim + resource;
1527         if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1528             !capable(CAP_SYS_RESOURCE))
1529                 return -EPERM;
1530
1531         if (resource == RLIMIT_NOFILE) {
1532                 if (new_rlim.rlim_max == RLIM_INFINITY)
1533                         new_rlim.rlim_max = sysctl_nr_open;
1534                 if (new_rlim.rlim_cur == RLIM_INFINITY)
1535                         new_rlim.rlim_cur = sysctl_nr_open;
1536                 if (new_rlim.rlim_max > sysctl_nr_open)
1537                         return -EPERM;
1538         }
1539
1540         if (new_rlim.rlim_cur > new_rlim.rlim_max)
1541                 return -EINVAL;
1542
1543         retval = security_task_setrlimit(resource, &new_rlim);
1544         if (retval)
1545                 return retval;
1546
1547         if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1548                 /*
1549                  * The caller is asking for an immediate RLIMIT_CPU
1550                  * expiry.  But we use the zero value to mean "it was
1551                  * never set".  So let's cheat and make it one second
1552                  * instead
1553                  */
1554                 new_rlim.rlim_cur = 1;
1555         }
1556
1557         task_lock(current->group_leader);
1558         *old_rlim = new_rlim;
1559         task_unlock(current->group_leader);
1560
1561         if (resource != RLIMIT_CPU)
1562                 goto out;
1563
1564         /*
1565          * RLIMIT_CPU handling.   Note that the kernel fails to return an error
1566          * code if it rejected the user's attempt to set RLIMIT_CPU.  This is a
1567          * very long-standing error, and fixing it now risks breakage of
1568          * applications, so we live with it
1569          */
1570         if (new_rlim.rlim_cur == RLIM_INFINITY)
1571                 goto out;
1572
1573         update_rlimit_cpu(new_rlim.rlim_cur);
1574 out:
1575         return 0;
1576 }
1577
1578 /*
1579  * It would make sense to put struct rusage in the task_struct,
1580  * except that would make the task_struct be *really big*.  After
1581  * task_struct gets moved into malloc'ed memory, it would
1582  * make sense to do this.  It will make moving the rest of the information
1583  * a lot simpler!  (Which we're not doing right now because we're not
1584  * measuring them yet).
1585  *
1586  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1587  * races with threads incrementing their own counters.  But since word
1588  * reads are atomic, we either get new values or old values and we don't
1589  * care which for the sums.  We always take the siglock to protect reading
1590  * the c* fields from p->signal from races with exit.c updating those
1591  * fields when reaping, so a sample either gets all the additions of a
1592  * given child after it's reaped, or none so this sample is before reaping.
1593  *
1594  * Locking:
1595  * We need to take the siglock for CHILDEREN, SELF and BOTH
1596  * for  the cases current multithreaded, non-current single threaded
1597  * non-current multithreaded.  Thread traversal is now safe with
1598  * the siglock held.
1599  * Strictly speaking, we donot need to take the siglock if we are current and
1600  * single threaded,  as no one else can take our signal_struct away, no one
1601  * else can  reap the  children to update signal->c* counters, and no one else
1602  * can race with the signal-> fields. If we do not take any lock, the
1603  * signal-> fields could be read out of order while another thread was just
1604  * exiting. So we should  place a read memory barrier when we avoid the lock.
1605  * On the writer side,  write memory barrier is implied in  __exit_signal
1606  * as __exit_signal releases  the siglock spinlock after updating the signal->
1607  * fields. But we don't do this yet to keep things simple.
1608  *
1609  */
1610
1611 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1612 {
1613         r->ru_nvcsw += t->nvcsw;
1614         r->ru_nivcsw += t->nivcsw;
1615         r->ru_minflt += t->min_flt;
1616         r->ru_majflt += t->maj_flt;
1617         r->ru_inblock += task_io_get_inblock(t);
1618         r->ru_oublock += task_io_get_oublock(t);
1619 }
1620
1621 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1622 {
1623         struct task_struct *t;
1624         unsigned long flags;
1625         cputime_t utime, stime;
1626         struct task_cputime cputime;
1627
1628         memset((char *) r, 0, sizeof *r);
1629         utime = stime = cputime_zero;
1630
1631         if (who == RUSAGE_THREAD) {
1632                 utime = task_utime(current);
1633                 stime = task_stime(current);
1634                 accumulate_thread_rusage(p, r);
1635                 goto out;
1636         }
1637
1638         if (!lock_task_sighand(p, &flags))
1639                 return;
1640
1641         switch (who) {
1642                 case RUSAGE_BOTH:
1643                 case RUSAGE_CHILDREN:
1644                         utime = p->signal->cutime;
1645                         stime = p->signal->cstime;
1646                         r->ru_nvcsw = p->signal->cnvcsw;
1647                         r->ru_nivcsw = p->signal->cnivcsw;
1648                         r->ru_minflt = p->signal->cmin_flt;
1649                         r->ru_majflt = p->signal->cmaj_flt;
1650                         r->ru_inblock = p->signal->cinblock;
1651                         r->ru_oublock = p->signal->coublock;
1652
1653                         if (who == RUSAGE_CHILDREN)
1654                                 break;
1655
1656                 case RUSAGE_SELF:
1657                         thread_group_cputime(p, &cputime);
1658                         utime = cputime_add(utime, cputime.utime);
1659                         stime = cputime_add(stime, cputime.stime);
1660                         r->ru_nvcsw += p->signal->nvcsw;
1661                         r->ru_nivcsw += p->signal->nivcsw;
1662                         r->ru_minflt += p->signal->min_flt;
1663                         r->ru_majflt += p->signal->maj_flt;
1664                         r->ru_inblock += p->signal->inblock;
1665                         r->ru_oublock += p->signal->oublock;
1666                         t = p;
1667                         do {
1668                                 accumulate_thread_rusage(t, r);
1669                                 t = next_thread(t);
1670                         } while (t != p);
1671                         break;
1672
1673                 default:
1674                         BUG();
1675         }
1676         unlock_task_sighand(p, &flags);
1677
1678 out:
1679         cputime_to_timeval(utime, &r->ru_utime);
1680         cputime_to_timeval(stime, &r->ru_stime);
1681 }
1682
1683 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1684 {
1685         struct rusage r;
1686         k_getrusage(p, who, &r);
1687         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1688 }
1689
1690 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1691 {
1692         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1693             who != RUSAGE_THREAD)
1694                 return -EINVAL;
1695         return getrusage(current, who, ru);
1696 }
1697
1698 asmlinkage long sys_umask(int mask)
1699 {
1700         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1701         return mask;
1702 }
1703
1704 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1705                           unsigned long arg4, unsigned long arg5)
1706 {
1707         struct task_struct *me = current;
1708         unsigned char comm[sizeof(me->comm)];
1709         long error;
1710
1711         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1712         if (error != -ENOSYS)
1713                 return error;
1714
1715         error = 0;
1716         switch (option) {
1717                 case PR_SET_PDEATHSIG:
1718                         if (!valid_signal(arg2)) {
1719                                 error = -EINVAL;
1720                                 break;
1721                         }
1722                         me->pdeath_signal = arg2;
1723                         error = 0;
1724                         break;
1725                 case PR_GET_PDEATHSIG:
1726                         error = put_user(me->pdeath_signal, (int __user *)arg2);
1727                         break;
1728                 case PR_GET_DUMPABLE:
1729                         error = get_dumpable(me->mm);
1730                         break;
1731                 case PR_SET_DUMPABLE:
1732                         if (arg2 < 0 || arg2 > 1) {
1733                                 error = -EINVAL;
1734                                 break;
1735                         }
1736                         set_dumpable(me->mm, arg2);
1737                         error = 0;
1738                         break;
1739
1740                 case PR_SET_UNALIGN:
1741                         error = SET_UNALIGN_CTL(me, arg2);
1742                         break;
1743                 case PR_GET_UNALIGN:
1744                         error = GET_UNALIGN_CTL(me, arg2);
1745                         break;
1746                 case PR_SET_FPEMU:
1747                         error = SET_FPEMU_CTL(me, arg2);
1748                         break;
1749                 case PR_GET_FPEMU:
1750                         error = GET_FPEMU_CTL(me, arg2);
1751                         break;
1752                 case PR_SET_FPEXC:
1753                         error = SET_FPEXC_CTL(me, arg2);
1754                         break;
1755                 case PR_GET_FPEXC:
1756                         error = GET_FPEXC_CTL(me, arg2);
1757                         break;
1758                 case PR_GET_TIMING:
1759                         error = PR_TIMING_STATISTICAL;
1760                         break;
1761                 case PR_SET_TIMING:
1762                         if (arg2 != PR_TIMING_STATISTICAL)
1763                                 error = -EINVAL;
1764                         else
1765                                 error = 0;
1766                         break;
1767
1768                 case PR_SET_NAME:
1769                         comm[sizeof(me->comm)-1] = 0;
1770                         if (strncpy_from_user(comm, (char __user *)arg2,
1771                                               sizeof(me->comm) - 1) < 0)
1772                                 return -EFAULT;
1773                         set_task_comm(me, comm);
1774                         return 0;
1775                 case PR_GET_NAME:
1776                         get_task_comm(comm, me);
1777                         if (copy_to_user((char __user *)arg2, comm,
1778                                          sizeof(comm)))
1779                                 return -EFAULT;
1780                         return 0;
1781                 case PR_GET_ENDIAN:
1782                         error = GET_ENDIAN(me, arg2);
1783                         break;
1784                 case PR_SET_ENDIAN:
1785                         error = SET_ENDIAN(me, arg2);
1786                         break;
1787
1788                 case PR_GET_SECCOMP:
1789                         error = prctl_get_seccomp();
1790                         break;
1791                 case PR_SET_SECCOMP:
1792                         error = prctl_set_seccomp(arg2);
1793                         break;
1794                 case PR_GET_TSC:
1795                         error = GET_TSC_CTL(arg2);
1796                         break;
1797                 case PR_SET_TSC:
1798                         error = SET_TSC_CTL(arg2);
1799                         break;
1800                 case PR_GET_TIMERSLACK:
1801                         error = current->timer_slack_ns;
1802                         break;
1803                 case PR_SET_TIMERSLACK:
1804                         if (arg2 <= 0)
1805                                 current->timer_slack_ns =
1806                                         current->default_timer_slack_ns;
1807                         else
1808                                 current->timer_slack_ns = arg2;
1809                         error = 0;
1810                         break;
1811                 default:
1812                         error = -EINVAL;
1813                         break;
1814         }
1815         return error;
1816 }
1817
1818 asmlinkage long sys_getcpu(unsigned __user *cpup, unsigned __user *nodep,
1819                            struct getcpu_cache __user *unused)
1820 {
1821         int err = 0;
1822         int cpu = raw_smp_processor_id();
1823         if (cpup)
1824                 err |= put_user(cpu, cpup);
1825         if (nodep)
1826                 err |= put_user(cpu_to_node(cpu), nodep);
1827         return err ? -EFAULT : 0;
1828 }
1829
1830 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1831
1832 static void argv_cleanup(char **argv, char **envp)
1833 {
1834         argv_free(argv);
1835 }
1836
1837 /**
1838  * orderly_poweroff - Trigger an orderly system poweroff
1839  * @force: force poweroff if command execution fails
1840  *
1841  * This may be called from any context to trigger a system shutdown.
1842  * If the orderly shutdown fails, it will force an immediate shutdown.
1843  */
1844 int orderly_poweroff(bool force)
1845 {
1846         int argc;
1847         char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1848         static char *envp[] = {
1849                 "HOME=/",
1850                 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1851                 NULL
1852         };
1853         int ret = -ENOMEM;
1854         struct subprocess_info *info;
1855
1856         if (argv == NULL) {
1857                 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1858                        __func__, poweroff_cmd);
1859                 goto out;
1860         }
1861
1862         info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1863         if (info == NULL) {
1864                 argv_free(argv);
1865                 goto out;
1866         }
1867
1868         call_usermodehelper_setcleanup(info, argv_cleanup);
1869
1870         ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1871
1872   out:
1873         if (ret && force) {
1874                 printk(KERN_WARNING "Failed to start orderly shutdown: "
1875                        "forcing the issue\n");
1876
1877                 /* I guess this should try to kick off some daemon to
1878                    sync and poweroff asap.  Or not even bother syncing
1879                    if we're doing an emergency shutdown? */
1880                 emergency_sync();
1881                 kernel_power_off();
1882         }
1883
1884         return ret;
1885 }
1886 EXPORT_SYMBOL_GPL(orderly_poweroff);