kernel/sys.c

   1 /*
   2  *  linux/kernel/sys.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6
   7 #include <linux/module.h>
   8 #include <linux/mm.h>
   9 #include <linux/utsname.h>
  10 #include <linux/mman.h>
  11 #include <linux/notifier.h>
  12 #include <linux/reboot.h>
  13 #include <linux/prctl.h>
  14 #include <linux/highuid.h>
  15 #include <linux/fs.h>
  16 #include <linux/perf_event.h>
  17 #include <linux/resource.h>
  18 #include <linux/kernel.h>
  19 #include <linux/kexec.h>
  20 #include <linux/workqueue.h>
  21 #include <linux/capability.h>
  22 #include <linux/device.h>
  23 #include <linux/key.h>
  24 #include <linux/times.h>
  25 #include <linux/posix-timers.h>
  26 #include <linux/security.h>
  27 #include <linux/dcookies.h>
  28 #include <linux/suspend.h>
  29 #include <linux/tty.h>
  30 #include <linux/signal.h>
  31 #include <linux/cn_proc.h>
  32 #include <linux/getcpu.h>
  33 #include <linux/task_io_accounting_ops.h>
  34 #include <linux/seccomp.h>
  35 #include <linux/cpu.h>
  36 #include <linux/ptrace.h>
  37 #include <linux/fs_struct.h>
  38
  39 #include <linux/compat.h>
  40 #include <linux/syscalls.h>
  41 #include <linux/kprobes.h>
  42 #include <linux/user_namespace.h>
  43
  44 #include <asm/uaccess.h>
  45 #include <asm/io.h>
  46 #include <asm/unistd.h>
  47
  48 #ifndef SET_UNALIGN_CTL
  49 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  50 #endif
  51 #ifndef GET_UNALIGN_CTL
  52 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  53 #endif
  54 #ifndef SET_FPEMU_CTL
  55 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
  56 #endif
  57 #ifndef GET_FPEMU_CTL
  58 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
  59 #endif
  60 #ifndef SET_FPEXC_CTL
  61 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
  62 #endif
  63 #ifndef GET_FPEXC_CTL
  64 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
  65 #endif
  66 #ifndef GET_ENDIAN
  67 # define GET_ENDIAN(a,b)        (-EINVAL)
  68 #endif
  69 #ifndef SET_ENDIAN
  70 # define SET_ENDIAN(a,b)        (-EINVAL)
  71 #endif
  72 #ifndef GET_TSC_CTL
  73 # define GET_TSC_CTL(a)         (-EINVAL)
  74 #endif
  75 #ifndef SET_TSC_CTL
  76 # define SET_TSC_CTL(a)         (-EINVAL)
  77 #endif
  78
  79 /*
  80  * this is where the system-wide overflow UID and GID are defined, for
  81  * architectures that now have 32-bit UID/GID but didn't in the past
  82  */
  83
  84 int overflowuid = DEFAULT_OVERFLOWUID;
  85 int overflowgid = DEFAULT_OVERFLOWGID;
  86
  87 #ifdef CONFIG_UID16
  88 EXPORT_SYMBOL(overflowuid);
  89 EXPORT_SYMBOL(overflowgid);
  90 #endif
  91
  92 /*
  93  * the same as above, but for filesystems which can only store a 16-bit
  94  * UID and GID. as such, this is needed on all architectures
  95  */
  96
  97 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
  98 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
  99
 100 EXPORT_SYMBOL(fs_overflowuid);
 101 EXPORT_SYMBOL(fs_overflowgid);
 102
 103 /*
 104  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
 105  */
 106
 107 int C_A_D = 1;
 108 struct pid *cad_pid;
 109 EXPORT_SYMBOL(cad_pid);
 110
 111 /*
 112  * If set, this is used for preparing the system to power off.
 113  */
 114
 115 void (*pm_power_off_prepare)(void);
 116
 117 /*
 118  * set the priority of a task
 119  * - the caller must hold the RCU read lock
 120  */
 121 static int set_one_prio(struct task_struct *p, int niceval, int error)
 122 {
 123         const struct cred *cred = current_cred(), *pcred = __task_cred(p);
 124         int no_nice;
 125
 126         if (pcred->uid  != cred->euid &&
 127             pcred->euid != cred->euid && !capable(CAP_SYS_NICE)) {
 128                 error = -EPERM;
 129                 goto out;
 130         }
 131         if (niceval < task_nice(p) && !can_nice(p, niceval)) {
 132                 error = -EACCES;
 133                 goto out;
 134         }
 135         no_nice = security_task_setnice(p, niceval);
 136         if (no_nice) {
 137                 error = no_nice;
 138                 goto out;
 139         }
 140         if (error == -ESRCH)
 141                 error = 0;
 142         set_user_nice(p, niceval);
 143 out:
 144         return error;
 145 }
 146
 147 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
 148 {
 149         struct task_struct *g, *p;
 150         struct user_struct *user;
 151         const struct cred *cred = current_cred();
 152         int error = -EINVAL;
 153         struct pid *pgrp;
 154
 155         if (which > PRIO_USER || which < PRIO_PROCESS)
 156                 goto out;
 157
 158         /* normalize: avoid signed division (rounding problems) */
 159         error = -ESRCH;
 160         if (niceval < -20)
 161                 niceval = -20;
 162         if (niceval > 19)
 163                 niceval = 19;
 164
 165         rcu_read_lock();
 166         read_lock(&tasklist_lock);
 167         switch (which) {
 168                 case PRIO_PROCESS:
 169                         if (who)
 170                                 p = find_task_by_vpid(who);
 171                         else
 172                                 p = current;
 173                         if (p)
 174                                 error = set_one_prio(p, niceval, error);
 175                         break;
 176                 case PRIO_PGRP:
 177                         if (who)
 178                                 pgrp = find_vpid(who);
 179                         else
 180                                 pgrp = task_pgrp(current);
 181                         do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
 182                                 error = set_one_prio(p, niceval, error);
 183                         } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
 184                         break;
 185                 case PRIO_USER:
 186                         user = (struct user_struct *) cred->user;
 187                         if (!who)
 188                                 who = cred->uid;
 189                         else if ((who != cred->uid) &&
 190                                  !(user = find_user(who)))
 191                                 goto out_unlock;        /* No processes for this user */
 192
 193                         do_each_thread(g, p) {
 194                                 if (__task_cred(p)->uid == who)
 195                                         error = set_one_prio(p, niceval, error);
 196                         } while_each_thread(g, p);
 197                         if (who != cred->uid)
 198                                 free_uid(user);         /* For find_user() */
 199                         break;
 200         }
 201 out_unlock:
 202         read_unlock(&tasklist_lock);
 203         rcu_read_unlock();
 204 out:
 205         return error;
 206 }
 207
 208 /*
 209  * Ugh. To avoid negative return values, "getpriority()" will
 210  * not return the normal nice-value, but a negated value that
 211  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 212  * to stay compatible.
 213  */
 214 SYSCALL_DEFINE2(getpriority, int, which, int, who)
 215 {
 216         struct task_struct *g, *p;
 217         struct user_struct *user;
 218         const struct cred *cred = current_cred();
 219         long niceval, retval = -ESRCH;
 220         struct pid *pgrp;
 221
 222         if (which > PRIO_USER || which < PRIO_PROCESS)
 223                 return -EINVAL;
 224
 225         rcu_read_lock();
 226         read_lock(&tasklist_lock);
 227         switch (which) {
 228                 case PRIO_PROCESS:
 229                         if (who)
 230                                 p = find_task_by_vpid(who);
 231                         else
 232                                 p = current;
 233                         if (p) {
 234                                 niceval = 20 - task_nice(p);
 235                                 if (niceval > retval)
 236                                         retval = niceval;
 237                         }
 238                         break;
 239                 case PRIO_PGRP:
 240                         if (who)
 241                                 pgrp = find_vpid(who);
 242                         else
 243                                 pgrp = task_pgrp(current);
 244                         do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
 245                                 niceval = 20 - task_nice(p);
 246                                 if (niceval > retval)
 247                                         retval = niceval;
 248                         } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
 249                         break;
 250                 case PRIO_USER:
 251                         user = (struct user_struct *) cred->user;
 252                         if (!who)
 253                                 who = cred->uid;
 254                         else if ((who != cred->uid) &&
 255                                  !(user = find_user(who)))
 256                                 goto out_unlock;        /* No processes for this user */
 257
 258                         do_each_thread(g, p) {
 259                                 if (__task_cred(p)->uid == who) {
 260                                         niceval = 20 - task_nice(p);
 261                                         if (niceval > retval)
 262                                                 retval = niceval;
 263                                 }
 264                         } while_each_thread(g, p);
 265                         if (who != cred->uid)
 266                                 free_uid(user);         /* for find_user() */
 267                         break;
 268         }
 269 out_unlock:
 270         read_unlock(&tasklist_lock);
 271         rcu_read_unlock();
 272
 273         return retval;
 274 }
 275
 276 /**
 277  *      emergency_restart - reboot the system
 278  *
 279  *      Without shutting down any hardware or taking any locks
 280  *      reboot the system.  This is called when we know we are in
 281  *      trouble so this is our best effort to reboot.  This is
 282  *      safe to call in interrupt context.
 283  */
 284 void emergency_restart(void)
 285 {
 286         machine_emergency_restart();
 287 }
 288 EXPORT_SYMBOL_GPL(emergency_restart);
 289
 290 void kernel_restart_prepare(char *cmd)
 291 {
 292         blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
 293         system_state = SYSTEM_RESTART;
 294         device_shutdown();
 295         sysdev_shutdown();
 296 }
 297
 298 /**
 299  *      kernel_restart - reboot the system
 300  *      @cmd: pointer to buffer containing command to execute for restart
 301  *              or %NULL
 302  *
 303  *      Shutdown everything and perform a clean reboot.
 304  *      This is not safe to call in interrupt context.
 305  */
 306 void kernel_restart(char *cmd)
 307 {
 308         kernel_restart_prepare(cmd);
 309         if (!cmd)
 310                 printk(KERN_EMERG "Restarting system.\n");
 311         else
 312                 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
 313         machine_restart(cmd);
 314 }
 315 EXPORT_SYMBOL_GPL(kernel_restart);
 316
 317 static void kernel_shutdown_prepare(enum system_states state)
 318 {
 319         blocking_notifier_call_chain(&reboot_notifier_list,
 320                 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
 321         system_state = state;
 322         device_shutdown();
 323 }
 324 /**
 325  *      kernel_halt - halt the system
 326  *
 327  *      Shutdown everything and perform a clean system halt.
 328  */
 329 void kernel_halt(void)
 330 {
 331         kernel_shutdown_prepare(SYSTEM_HALT);
 332         sysdev_shutdown();
 333         printk(KERN_EMERG "System halted.\n");
 334         machine_halt();
 335 }
 336
 337 EXPORT_SYMBOL_GPL(kernel_halt);
 338
 339 /**
 340  *      kernel_power_off - power_off the system
 341  *
 342  *      Shutdown everything and perform a clean system power_off.
 343  */
 344 void kernel_power_off(void)
 345 {
 346         kernel_shutdown_prepare(SYSTEM_POWER_OFF);
 347         if (pm_power_off_prepare)
 348                 pm_power_off_prepare();
 349         disable_nonboot_cpus();
 350         sysdev_shutdown();
 351         printk(KERN_EMERG "Power down.\n");
 352         machine_power_off();
 353 }
 354 EXPORT_SYMBOL_GPL(kernel_power_off);
 355
 356 static DEFINE_MUTEX(reboot_mutex);
 357
 358 /*
 359  * Reboot system call: for obvious reasons only root may call it,
 360  * and even root needs to set up some magic numbers in the registers
 361  * so that some mistake won't make this reboot the whole machine.
 362  * You can also set the meaning of the ctrl-alt-del-key here.
 363  *
 364  * reboot doesn't sync: do that yourself before calling this.
 365  */
 366 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
 367                 void __user *, arg)
 368 {
 369         char buffer[256];
 370         int ret = 0;
 371
 372         /* We only trust the superuser with rebooting the system. */
 373         if (!capable(CAP_SYS_BOOT))
 374                 return -EPERM;
 375
 376         /* For safety, we require "magic" arguments. */
 377         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 378             (magic2 != LINUX_REBOOT_MAGIC2 &&
 379                         magic2 != LINUX_REBOOT_MAGIC2A &&
 380                         magic2 != LINUX_REBOOT_MAGIC2B &&
 381                         magic2 != LINUX_REBOOT_MAGIC2C))
 382                 return -EINVAL;
 383
 384         /* Instead of trying to make the power_off code look like
 385          * halt when pm_power_off is not set do it the easy way.
 386          */
 387         if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
 388                 cmd = LINUX_REBOOT_CMD_HALT;
 389
 390         mutex_lock(&reboot_mutex);
 391         switch (cmd) {
 392         case LINUX_REBOOT_CMD_RESTART:
 393                 kernel_restart(NULL);
 394                 break;
 395
 396         case LINUX_REBOOT_CMD_CAD_ON:
 397                 C_A_D = 1;
 398                 break;
 399
 400         case LINUX_REBOOT_CMD_CAD_OFF:
 401                 C_A_D = 0;
 402                 break;
 403
 404         case LINUX_REBOOT_CMD_HALT:
 405                 kernel_halt();
 406                 do_exit(0);
 407                 panic("cannot halt");
 408
 409         case LINUX_REBOOT_CMD_POWER_OFF:
 410                 kernel_power_off();
 411                 do_exit(0);
 412                 break;
 413
 414         case LINUX_REBOOT_CMD_RESTART2:
 415                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 416                         ret = -EFAULT;
 417                         break;
 418                 }
 419                 buffer[sizeof(buffer) - 1] = '\0';
 420
 421                 kernel_restart(buffer);
 422                 break;
 423
 424 #ifdef CONFIG_KEXEC
 425         case LINUX_REBOOT_CMD_KEXEC:
 426                 ret = kernel_kexec();
 427                 break;
 428 #endif
 429
 430 #ifdef CONFIG_HIBERNATION
 431         case LINUX_REBOOT_CMD_SW_SUSPEND:
 432                 ret = hibernate();
 433                 break;
 434 #endif
 435
 436         default:
 437                 ret = -EINVAL;
 438                 break;
 439         }
 440         mutex_unlock(&reboot_mutex);
 441         return ret;
 442 }
 443
 444 static void deferred_cad(struct work_struct *dummy)
 445 {
 446         kernel_restart(NULL);
 447 }
 448
 449 /*
 450  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 451  * As it's called within an interrupt, it may NOT sync: the only choice
 452  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 453  */
 454 void ctrl_alt_del(void)
 455 {
 456         static DECLARE_WORK(cad_work, deferred_cad);
 457
 458         if (C_A_D)
 459                 schedule_work(&cad_work);
 460         else
 461                 kill_cad_pid(SIGINT, 1);
 462 }
 463
 464 /*
 465  * Unprivileged users may change the real gid to the effective gid
 466  * or vice versa.  (BSD-style)
 467  *
 468  * If you set the real gid at all, or set the effective gid to a value not
 469  * equal to the real gid, then the saved gid is set to the new effective gid.
 470  *
 471  * This makes it possible for a setgid program to completely drop its
 472  * privileges, which is often a useful assertion to make when you are doing
 473  * a security audit over a program.
 474  *
 475  * The general idea is that a program which uses just setregid() will be
 476  * 100% compatible with BSD.  A program which uses just setgid() will be
 477  * 100% compatible with POSIX with saved IDs.
 478  *
 479  * SMP: There are not races, the GIDs are checked only by filesystem
 480  *      operations (as far as semantic preservation is concerned).
 481  */
 482 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
 483 {
 484         const struct cred *old;
 485         struct cred *new;
 486         int retval;
 487
 488         new = prepare_creds();
 489         if (!new)
 490                 return -ENOMEM;
 491         old = current_cred();
 492
 493         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 494         if (retval)
 495                 goto error;
 496
 497         retval = -EPERM;
 498         if (rgid != (gid_t) -1) {
 499                 if (old->gid == rgid ||
 500                     old->egid == rgid ||
 501                     capable(CAP_SETGID))
 502                         new->gid = rgid;
 503                 else
 504                         goto error;
 505         }
 506         if (egid != (gid_t) -1) {
 507                 if (old->gid == egid ||
 508                     old->egid == egid ||
 509                     old->sgid == egid ||
 510                     capable(CAP_SETGID))
 511                         new->egid = egid;
 512                 else
 513                         goto error;
 514         }
 515
 516         if (rgid != (gid_t) -1 ||
 517             (egid != (gid_t) -1 && egid != old->gid))
 518                 new->sgid = new->egid;
 519         new->fsgid = new->egid;
 520
 521         return commit_creds(new);
 522
 523 error:
 524         abort_creds(new);
 525         return retval;
 526 }
 527
 528 /*
 529  * setgid() is implemented like SysV w/ SAVED_IDS
 530  *
 531  * SMP: Same implicit races as above.
 532  */
 533 SYSCALL_DEFINE1(setgid, gid_t, gid)
 534 {
 535         const struct cred *old;
 536         struct cred *new;
 537         int retval;
 538
 539         new = prepare_creds();
 540         if (!new)
 541                 return -ENOMEM;
 542         old = current_cred();
 543
 544         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 545         if (retval)
 546                 goto error;
 547
 548         retval = -EPERM;
 549         if (capable(CAP_SETGID))
 550                 new->gid = new->egid = new->sgid = new->fsgid = gid;
 551         else if (gid == old->gid || gid == old->sgid)
 552                 new->egid = new->fsgid = gid;
 553         else
 554                 goto error;
 555
 556         return commit_creds(new);
 557
 558 error:
 559         abort_creds(new);
 560         return retval;
 561 }
 562
 563 /*
 564  * change the user struct in a credentials set to match the new UID
 565  */
 566 static int set_user(struct cred *new)
 567 {
 568         struct user_struct *new_user;
 569
 570         new_user = alloc_uid(current_user_ns(), new->uid);
 571         if (!new_user)
 572                 return -EAGAIN;
 573
 574         if (!task_can_switch_user(new_user, current)) {
 575                 free_uid(new_user);
 576                 return -EINVAL;
 577         }
 578
 579         if (atomic_read(&new_user->processes) >=
 580                                 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
 581                         new_user != INIT_USER) {
 582                 free_uid(new_user);
 583                 return -EAGAIN;
 584         }
 585
 586         free_uid(new->user);
 587         new->user = new_user;
 588         return 0;
 589 }
 590
 591 /*
 592  * Unprivileged users may change the real uid to the effective uid
 593  * or vice versa.  (BSD-style)
 594  *
 595  * If you set the real uid at all, or set the effective uid to a value not
 596  * equal to the real uid, then the saved uid is set to the new effective uid.
 597  *
 598  * This makes it possible for a setuid program to completely drop its
 599  * privileges, which is often a useful assertion to make when you are doing
 600  * a security audit over a program.
 601  *
 602  * The general idea is that a program which uses just setreuid() will be
 603  * 100% compatible with BSD.  A program which uses just setuid() will be
 604  * 100% compatible with POSIX with saved IDs.
 605  */
 606 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
 607 {
 608         const struct cred *old;
 609         struct cred *new;
 610         int retval;
 611
 612         new = prepare_creds();
 613         if (!new)
 614                 return -ENOMEM;
 615         old = current_cred();
 616
 617         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 618         if (retval)
 619                 goto error;
 620
 621         retval = -EPERM;
 622         if (ruid != (uid_t) -1) {
 623                 new->uid = ruid;
 624                 if (old->uid != ruid &&
 625                     old->euid != ruid &&
 626                     !capable(CAP_SETUID))
 627                         goto error;
 628         }
 629
 630         if (euid != (uid_t) -1) {
 631                 new->euid = euid;
 632                 if (old->uid != euid &&
 633                     old->euid != euid &&
 634                     old->suid != euid &&
 635                     !capable(CAP_SETUID))
 636                         goto error;
 637         }
 638
 639         if (new->uid != old->uid) {
 640                 retval = set_user(new);
 641                 if (retval < 0)
 642                         goto error;
 643         }
 644         if (ruid != (uid_t) -1 ||
 645             (euid != (uid_t) -1 && euid != old->uid))
 646                 new->suid = new->euid;
 647         new->fsuid = new->euid;
 648
 649         retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
 650         if (retval < 0)
 651                 goto error;
 652
 653         return commit_creds(new);
 654
 655 error:
 656         abort_creds(new);
 657         return retval;
 658 }
 659
 660 /*
 661  * setuid() is implemented like SysV with SAVED_IDS
 662  *
 663  * Note that SAVED_ID's is deficient in that a setuid root program
 664  * like sendmail, for example, cannot set its uid to be a normal
 665  * user and then switch back, because if you're root, setuid() sets
 666  * the saved uid too.  If you don't like this, blame the bright people
 667  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 668  * will allow a root program to temporarily drop privileges and be able to
 669  * regain them by swapping the real and effective uid.
 670  */
 671 SYSCALL_DEFINE1(setuid, uid_t, uid)
 672 {
 673         const struct cred *old;
 674         struct cred *new;
 675         int retval;
 676
 677         new = prepare_creds();
 678         if (!new)
 679                 return -ENOMEM;
 680         old = current_cred();
 681
 682         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 683         if (retval)
 684                 goto error;
 685
 686         retval = -EPERM;
 687         if (capable(CAP_SETUID)) {
 688                 new->suid = new->uid = uid;
 689                 if (uid != old->uid) {
 690                         retval = set_user(new);
 691                         if (retval < 0)
 692                                 goto error;
 693                 }
 694         } else if (uid != old->uid && uid != new->suid) {
 695                 goto error;
 696         }
 697
 698         new->fsuid = new->euid = uid;
 699
 700         retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
 701         if (retval < 0)
 702                 goto error;
 703
 704         return commit_creds(new);
 705
 706 error:
 707         abort_creds(new);
 708         return retval;
 709 }
 710
 711
 712 /*
 713  * This function implements a generic ability to update ruid, euid,
 714  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 715  */
 716 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
 717 {
 718         const struct cred *old;
 719         struct cred *new;
 720         int retval;
 721
 722         new = prepare_creds();
 723         if (!new)
 724                 return -ENOMEM;
 725
 726         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 727         if (retval)
 728                 goto error;
 729         old = current_cred();
 730
 731         retval = -EPERM;
 732         if (!capable(CAP_SETUID)) {
 733                 if (ruid != (uid_t) -1 && ruid != old->uid &&
 734                     ruid != old->euid  && ruid != old->suid)
 735                         goto error;
 736                 if (euid != (uid_t) -1 && euid != old->uid &&
 737                     euid != old->euid  && euid != old->suid)
 738                         goto error;
 739                 if (suid != (uid_t) -1 && suid != old->uid &&
 740                     suid != old->euid  && suid != old->suid)
 741                         goto error;
 742         }
 743
 744         if (ruid != (uid_t) -1) {
 745                 new->uid = ruid;
 746                 if (ruid != old->uid) {
 747                         retval = set_user(new);
 748                         if (retval < 0)
 749                                 goto error;
 750                 }
 751         }
 752         if (euid != (uid_t) -1)
 753                 new->euid = euid;
 754         if (suid != (uid_t) -1)
 755                 new->suid = suid;
 756         new->fsuid = new->euid;
 757
 758         retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
 759         if (retval < 0)
 760                 goto error;
 761
 762         return commit_creds(new);
 763
 764 error:
 765         abort_creds(new);
 766         return retval;
 767 }
 768
 769 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruid, uid_t __user *, euid, uid_t __user *, suid)
 770 {
 771         const struct cred *cred = current_cred();
 772         int retval;
 773
 774         if (!(retval   = put_user(cred->uid,  ruid)) &&
 775             !(retval   = put_user(cred->euid, euid)))
 776                 retval = put_user(cred->suid, suid);
 777
 778         return retval;
 779 }
 780
 781 /*
 782  * Same as above, but for rgid, egid, sgid.
 783  */
 784 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
 785 {
 786         const struct cred *old;
 787         struct cred *new;
 788         int retval;
 789
 790         new = prepare_creds();
 791         if (!new)
 792                 return -ENOMEM;
 793         old = current_cred();
 794
 795         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 796         if (retval)
 797                 goto error;
 798
 799         retval = -EPERM;
 800         if (!capable(CAP_SETGID)) {
 801                 if (rgid != (gid_t) -1 && rgid != old->gid &&
 802                     rgid != old->egid  && rgid != old->sgid)
 803                         goto error;
 804                 if (egid != (gid_t) -1 && egid != old->gid &&
 805                     egid != old->egid  && egid != old->sgid)
 806                         goto error;
 807                 if (sgid != (gid_t) -1 && sgid != old->gid &&
 808                     sgid != old->egid  && sgid != old->sgid)
 809                         goto error;
 810         }
 811
 812         if (rgid != (gid_t) -1)
 813                 new->gid = rgid;
 814         if (egid != (gid_t) -1)
 815                 new->egid = egid;
 816         if (sgid != (gid_t) -1)
 817                 new->sgid = sgid;
 818         new->fsgid = new->egid;
 819
 820         return commit_creds(new);
 821
 822 error:
 823         abort_creds(new);
 824         return retval;
 825 }
 826
 827 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgid, gid_t __user *, egid, gid_t __user *, sgid)
 828 {
 829         const struct cred *cred = current_cred();
 830         int retval;
 831
 832         if (!(retval   = put_user(cred->gid,  rgid)) &&
 833             !(retval   = put_user(cred->egid, egid)))
 834                 retval = put_user(cred->sgid, sgid);
 835
 836         return retval;
 837 }
 838
 839
 840 /*
 841  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 842  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 843  * whatever uid it wants to). It normally shadows "euid", except when
 844  * explicitly set by setfsuid() or for access..
 845  */
 846 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
 847 {
 848         const struct cred *old;
 849         struct cred *new;
 850         uid_t old_fsuid;
 851
 852         new = prepare_creds();
 853         if (!new)
 854                 return current_fsuid();
 855         old = current_cred();
 856         old_fsuid = old->fsuid;
 857
 858         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS) < 0)
 859                 goto error;
 860
 861         if (uid == old->uid  || uid == old->euid  ||
 862             uid == old->suid || uid == old->fsuid ||
 863             capable(CAP_SETUID)) {
 864                 if (uid != old_fsuid) {
 865                         new->fsuid = uid;
 866                         if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
 867                                 goto change_okay;
 868                 }
 869         }
 870
 871 error:
 872         abort_creds(new);
 873         return old_fsuid;
 874
 875 change_okay:
 876         commit_creds(new);
 877         return old_fsuid;
 878 }
 879
 880 /*
 881  * Samma på svenska..
 882  */
 883 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
 884 {
 885         const struct cred *old;
 886         struct cred *new;
 887         gid_t old_fsgid;
 888
 889         new = prepare_creds();
 890         if (!new)
 891                 return current_fsgid();
 892         old = current_cred();
 893         old_fsgid = old->fsgid;
 894
 895         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 896                 goto error;
 897
 898         if (gid == old->gid  || gid == old->egid  ||
 899             gid == old->sgid || gid == old->fsgid ||
 900             capable(CAP_SETGID)) {
 901                 if (gid != old_fsgid) {
 902                         new->fsgid = gid;
 903                         goto change_okay;
 904                 }
 905         }
 906
 907 error:
 908         abort_creds(new);
 909         return old_fsgid;
 910
 911 change_okay:
 912         commit_creds(new);
 913         return old_fsgid;
 914 }
 915
 916 void do_sys_times(struct tms *tms)
 917 {
 918         cputime_t tgutime, tgstime, cutime, cstime;
 919
 920         spin_lock_irq(&current->sighand->siglock);
 921         thread_group_times(current, &tgutime, &tgstime);
 922         cutime = current->signal->cutime;
 923         cstime = current->signal->cstime;
 924         spin_unlock_irq(&current->sighand->siglock);
 925         tms->tms_utime = cputime_to_clock_t(tgutime);
 926         tms->tms_stime = cputime_to_clock_t(tgstime);
 927         tms->tms_cutime = cputime_to_clock_t(cutime);
 928         tms->tms_cstime = cputime_to_clock_t(cstime);
 929 }
 930
 931 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
 932 {
 933         if (tbuf) {
 934                 struct tms tmp;
 935
 936                 do_sys_times(&tmp);
 937                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
 938                         return -EFAULT;
 939         }
 940         force_successful_syscall_return();
 941         return (long) jiffies_64_to_clock_t(get_jiffies_64());
 942 }
 943
 944 /*
 945  * This needs some heavy checking ...
 946  * I just haven't the stomach for it. I also don't fully
 947  * understand sessions/pgrp etc. Let somebody who does explain it.
 948  *
 949  * OK, I think I have the protection semantics right.... this is really
 950  * only important on a multi-user system anyway, to make sure one user
 951  * can't send a signal to a process owned by another.  -TYT, 12/12/91
 952  *
 953  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
 954  * LBT 04.03.94
 955  */
 956 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
 957 {
 958         struct task_struct *p;
 959         struct task_struct *group_leader = current->group_leader;
 960         struct pid *pgrp;
 961         int err;
 962
 963         if (!pid)
 964                 pid = task_pid_vnr(group_leader);
 965         if (!pgid)
 966                 pgid = pid;
 967         if (pgid < 0)
 968                 return -EINVAL;
 969         rcu_read_lock();
 970
 971         /* From this point forward we keep holding onto the tasklist lock
 972          * so that our parent does not change from under us. -DaveM
 973          */
 974         write_lock_irq(&tasklist_lock);
 975
 976         err = -ESRCH;
 977         p = find_task_by_vpid(pid);
 978         if (!p)
 979                 goto out;
 980
 981         err = -EINVAL;
 982         if (!thread_group_leader(p))
 983                 goto out;
 984
 985         if (same_thread_group(p->real_parent, group_leader)) {
 986                 err = -EPERM;
 987                 if (task_session(p) != task_session(group_leader))
 988                         goto out;
 989                 err = -EACCES;
 990                 if (p->did_exec)
 991                         goto out;
 992         } else {
 993                 err = -ESRCH;
 994                 if (p != group_leader)
 995                         goto out;
 996         }
 997
 998         err = -EPERM;
 999         if (p->signal->leader)
1000                 goto out;
1001
1002         pgrp = task_pid(p);
1003         if (pgid != pid) {
1004                 struct task_struct *g;
1005
1006                 pgrp = find_vpid(pgid);
1007                 g = pid_task(pgrp, PIDTYPE_PGID);
1008                 if (!g || task_session(g) != task_session(group_leader))
1009                         goto out;
1010         }
1011
1012         err = security_task_setpgid(p, pgid);
1013         if (err)
1014                 goto out;
1015
1016         if (task_pgrp(p) != pgrp)
1017                 change_pid(p, PIDTYPE_PGID, pgrp);
1018
1019         err = 0;
1020 out:
1021         /* All paths lead to here, thus we are safe. -DaveM */
1022         write_unlock_irq(&tasklist_lock);
1023         rcu_read_unlock();
1024         return err;
1025 }
1026
1027 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1028 {
1029         struct task_struct *p;
1030         struct pid *grp;
1031         int retval;
1032
1033         rcu_read_lock();
1034         if (!pid)
1035                 grp = task_pgrp(current);
1036         else {
1037                 retval = -ESRCH;
1038                 p = find_task_by_vpid(pid);
1039                 if (!p)
1040                         goto out;
1041                 grp = task_pgrp(p);
1042                 if (!grp)
1043                         goto out;
1044
1045                 retval = security_task_getpgid(p);
1046                 if (retval)
1047                         goto out;
1048         }
1049         retval = pid_vnr(grp);
1050 out:
1051         rcu_read_unlock();
1052         return retval;
1053 }
1054
1055 #ifdef __ARCH_WANT_SYS_GETPGRP
1056
1057 SYSCALL_DEFINE0(getpgrp)
1058 {
1059         return sys_getpgid(0);
1060 }
1061
1062 #endif
1063
1064 SYSCALL_DEFINE1(getsid, pid_t, pid)
1065 {
1066         struct task_struct *p;
1067         struct pid *sid;
1068         int retval;
1069
1070         rcu_read_lock();
1071         if (!pid)
1072                 sid = task_session(current);
1073         else {
1074                 retval = -ESRCH;
1075                 p = find_task_by_vpid(pid);
1076                 if (!p)
1077                         goto out;
1078                 sid = task_session(p);
1079                 if (!sid)
1080                         goto out;
1081
1082                 retval = security_task_getsid(p);
1083                 if (retval)
1084                         goto out;
1085         }
1086         retval = pid_vnr(sid);
1087 out:
1088         rcu_read_unlock();
1089         return retval;
1090 }
1091
1092 SYSCALL_DEFINE0(setsid)
1093 {
1094         struct task_struct *group_leader = current->group_leader;
1095         struct pid *sid = task_pid(group_leader);
1096         pid_t session = pid_vnr(sid);
1097         int err = -EPERM;
1098
1099         write_lock_irq(&tasklist_lock);
1100         /* Fail if I am already a session leader */
1101         if (group_leader->signal->leader)
1102                 goto out;
1103
1104         /* Fail if a process group id already exists that equals the
1105          * proposed session id.
1106          */
1107         if (pid_task(sid, PIDTYPE_PGID))
1108                 goto out;
1109
1110         group_leader->signal->leader = 1;
1111         __set_special_pids(sid);
1112
1113         proc_clear_tty(group_leader);
1114
1115         err = session;
1116 out:
1117         write_unlock_irq(&tasklist_lock);
1118         if (err > 0)
1119                 proc_sid_connector(group_leader);
1120         return err;
1121 }
1122
1123 DECLARE_RWSEM(uts_sem);
1124
1125 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1126 {
1127         int errno = 0;
1128
1129         down_read(&uts_sem);
1130         if (copy_to_user(name, utsname(), sizeof *name))
1131                 errno = -EFAULT;
1132         up_read(&uts_sem);
1133         return errno;
1134 }
1135
1136 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1137 {
1138         int errno;
1139         char tmp[__NEW_UTS_LEN];
1140
1141         if (!capable(CAP_SYS_ADMIN))
1142                 return -EPERM;
1143         if (len < 0 || len > __NEW_UTS_LEN)
1144                 return -EINVAL;
1145         down_write(&uts_sem);
1146         errno = -EFAULT;
1147         if (!copy_from_user(tmp, name, len)) {
1148                 struct new_utsname *u = utsname();
1149
1150                 memcpy(u->nodename, tmp, len);
1151                 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1152                 errno = 0;
1153         }
1154         up_write(&uts_sem);
1155         return errno;
1156 }
1157
1158 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1159
1160 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1161 {
1162         int i, errno;
1163         struct new_utsname *u;
1164
1165         if (len < 0)
1166                 return -EINVAL;
1167         down_read(&uts_sem);
1168         u = utsname();
1169         i = 1 + strlen(u->nodename);
1170         if (i > len)
1171                 i = len;
1172         errno = 0;
1173         if (copy_to_user(name, u->nodename, i))
1174                 errno = -EFAULT;
1175         up_read(&uts_sem);
1176         return errno;
1177 }
1178
1179 #endif
1180
1181 /*
1182  * Only setdomainname; getdomainname can be implemented by calling
1183  * uname()
1184  */
1185 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1186 {
1187         int errno;
1188         char tmp[__NEW_UTS_LEN];
1189
1190         if (!capable(CAP_SYS_ADMIN))
1191                 return -EPERM;
1192         if (len < 0 || len > __NEW_UTS_LEN)
1193                 return -EINVAL;
1194
1195         down_write(&uts_sem);
1196         errno = -EFAULT;
1197         if (!copy_from_user(tmp, name, len)) {
1198                 struct new_utsname *u = utsname();
1199
1200                 memcpy(u->domainname, tmp, len);
1201                 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1202                 errno = 0;
1203         }
1204         up_write(&uts_sem);
1205         return errno;
1206 }
1207
1208 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1209 {
1210         if (resource >= RLIM_NLIMITS)
1211                 return -EINVAL;
1212         else {
1213                 struct rlimit value;
1214                 task_lock(current->group_leader);
1215                 value = current->signal->rlim[resource];
1216                 task_unlock(current->group_leader);
1217                 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1218         }
1219 }
1220
1221 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1222
1223 /*
1224  *      Back compatibility for getrlimit. Needed for some apps.
1225  */
1226
1227 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1228                 struct rlimit __user *, rlim)
1229 {
1230         struct rlimit x;
1231         if (resource >= RLIM_NLIMITS)
1232                 return -EINVAL;
1233
1234         task_lock(current->group_leader);
1235         x = current->signal->rlim[resource];
1236         task_unlock(current->group_leader);
1237         if (x.rlim_cur > 0x7FFFFFFF)
1238                 x.rlim_cur = 0x7FFFFFFF;
1239         if (x.rlim_max > 0x7FFFFFFF)
1240                 x.rlim_max = 0x7FFFFFFF;
1241         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1242 }
1243
1244 #endif
1245
1246 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1247 {
1248         struct rlimit new_rlim, *old_rlim;
1249         int retval;
1250
1251         if (resource >= RLIM_NLIMITS)
1252                 return -EINVAL;
1253         if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1254                 return -EFAULT;
1255         if (new_rlim.rlim_cur > new_rlim.rlim_max)
1256                 return -EINVAL;
1257         old_rlim = current->signal->rlim + resource;
1258         if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1259             !capable(CAP_SYS_RESOURCE))
1260                 return -EPERM;
1261         if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > sysctl_nr_open)
1262                 return -EPERM;
1263
1264         retval = security_task_setrlimit(resource, &new_rlim);
1265         if (retval)
1266                 return retval;
1267
1268         if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1269                 /*
1270                  * The caller is asking for an immediate RLIMIT_CPU
1271                  * expiry.  But we use the zero value to mean "it was
1272                  * never set".  So let's cheat and make it one second
1273                  * instead
1274                  */
1275                 new_rlim.rlim_cur = 1;
1276         }
1277
1278         task_lock(current->group_leader);
1279         *old_rlim = new_rlim;
1280         task_unlock(current->group_leader);
1281
1282         if (resource != RLIMIT_CPU)
1283                 goto out;
1284
1285         /*
1286          * RLIMIT_CPU handling.   Note that the kernel fails to return an error
1287          * code if it rejected the user's attempt to set RLIMIT_CPU.  This is a
1288          * very long-standing error, and fixing it now risks breakage of
1289          * applications, so we live with it
1290          */
1291         if (new_rlim.rlim_cur == RLIM_INFINITY)
1292                 goto out;
1293
1294         update_rlimit_cpu(new_rlim.rlim_cur);
1295 out:
1296         return 0;
1297 }
1298
1299 /*
1300  * It would make sense to put struct rusage in the task_struct,
1301  * except that would make the task_struct be *really big*.  After
1302  * task_struct gets moved into malloc'ed memory, it would
1303  * make sense to do this.  It will make moving the rest of the information
1304  * a lot simpler!  (Which we're not doing right now because we're not
1305  * measuring them yet).
1306  *
1307  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1308  * races with threads incrementing their own counters.  But since word
1309  * reads are atomic, we either get new values or old values and we don't
1310  * care which for the sums.  We always take the siglock to protect reading
1311  * the c* fields from p->signal from races with exit.c updating those
1312  * fields when reaping, so a sample either gets all the additions of a
1313  * given child after it's reaped, or none so this sample is before reaping.
1314  *
1315  * Locking:
1316  * We need to take the siglock for CHILDEREN, SELF and BOTH
1317  * for  the cases current multithreaded, non-current single threaded
1318  * non-current multithreaded.  Thread traversal is now safe with
1319  * the siglock held.
1320  * Strictly speaking, we donot need to take the siglock if we are current and
1321  * single threaded,  as no one else can take our signal_struct away, no one
1322  * else can  reap the  children to update signal->c* counters, and no one else
1323  * can race with the signal-> fields. If we do not take any lock, the
1324  * signal-> fields could be read out of order while another thread was just
1325  * exiting. So we should  place a read memory barrier when we avoid the lock.
1326  * On the writer side,  write memory barrier is implied in  __exit_signal
1327  * as __exit_signal releases  the siglock spinlock after updating the signal->
1328  * fields. But we don't do this yet to keep things simple.
1329  *
1330  */
1331
1332 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1333 {
1334         r->ru_nvcsw += t->nvcsw;
1335         r->ru_nivcsw += t->nivcsw;
1336         r->ru_minflt += t->min_flt;
1337         r->ru_majflt += t->maj_flt;
1338         r->ru_inblock += task_io_get_inblock(t);
1339         r->ru_oublock += task_io_get_oublock(t);
1340 }
1341
1342 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1343 {
1344         struct task_struct *t;
1345         unsigned long flags;
1346         cputime_t tgutime, tgstime, utime, stime;
1347         unsigned long maxrss = 0;
1348
1349         memset((char *) r, 0, sizeof *r);
1350         utime = stime = cputime_zero;
1351
1352         if (who == RUSAGE_THREAD) {
1353                 task_times(current, &utime, &stime);
1354                 accumulate_thread_rusage(p, r);
1355                 maxrss = p->signal->maxrss;
1356                 goto out;
1357         }
1358
1359         if (!lock_task_sighand(p, &flags))
1360                 return;
1361
1362         switch (who) {
1363                 case RUSAGE_BOTH:
1364                 case RUSAGE_CHILDREN:
1365                         utime = p->signal->cutime;
1366                         stime = p->signal->cstime;
1367                         r->ru_nvcsw = p->signal->cnvcsw;
1368                         r->ru_nivcsw = p->signal->cnivcsw;
1369                         r->ru_minflt = p->signal->cmin_flt;
1370                         r->ru_majflt = p->signal->cmaj_flt;
1371                         r->ru_inblock = p->signal->cinblock;
1372                         r->ru_oublock = p->signal->coublock;
1373                         maxrss = p->signal->cmaxrss;
1374
1375                         if (who == RUSAGE_CHILDREN)
1376                                 break;
1377
1378                 case RUSAGE_SELF:
1379                         thread_group_times(p, &tgutime, &tgstime);
1380                         utime = cputime_add(utime, tgutime);
1381                         stime = cputime_add(stime, tgstime);
1382                         r->ru_nvcsw += p->signal->nvcsw;
1383                         r->ru_nivcsw += p->signal->nivcsw;
1384                         r->ru_minflt += p->signal->min_flt;
1385                         r->ru_majflt += p->signal->maj_flt;
1386                         r->ru_inblock += p->signal->inblock;
1387                         r->ru_oublock += p->signal->oublock;
1388                         if (maxrss < p->signal->maxrss)
1389                                 maxrss = p->signal->maxrss;
1390                         t = p;
1391                         do {
1392                                 accumulate_thread_rusage(t, r);
1393                                 t = next_thread(t);
1394                         } while (t != p);
1395                         break;
1396
1397                 default:
1398                         BUG();
1399         }
1400         unlock_task_sighand(p, &flags);
1401
1402 out:
1403         cputime_to_timeval(utime, &r->ru_utime);
1404         cputime_to_timeval(stime, &r->ru_stime);
1405
1406         if (who != RUSAGE_CHILDREN) {
1407                 struct mm_struct *mm = get_task_mm(p);
1408                 if (mm) {
1409                         setmax_mm_hiwater_rss(&maxrss, mm);
1410                         mmput(mm);
1411                 }
1412         }
1413         r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1414 }
1415
1416 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1417 {
1418         struct rusage r;
1419         k_getrusage(p, who, &r);
1420         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1421 }
1422
1423 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1424 {
1425         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1426             who != RUSAGE_THREAD)
1427                 return -EINVAL;
1428         return getrusage(current, who, ru);
1429 }
1430
1431 SYSCALL_DEFINE1(umask, int, mask)
1432 {
1433         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1434         return mask;
1435 }
1436
1437 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1438                 unsigned long, arg4, unsigned long, arg5)
1439 {
1440         struct task_struct *me = current;
1441         unsigned char comm[sizeof(me->comm)];
1442         long error;
1443
1444         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1445         if (error != -ENOSYS)
1446                 return error;
1447
1448         error = 0;
1449         switch (option) {
1450                 case PR_SET_PDEATHSIG:
1451                         if (!valid_signal(arg2)) {
1452                                 error = -EINVAL;
1453                                 break;
1454                         }
1455                         me->pdeath_signal = arg2;
1456                         error = 0;
1457                         break;
1458                 case PR_GET_PDEATHSIG:
1459                         error = put_user(me->pdeath_signal, (int __user *)arg2);
1460                         break;
1461                 case PR_GET_DUMPABLE:
1462                         error = get_dumpable(me->mm);
1463                         break;
1464                 case PR_SET_DUMPABLE:
1465                         if (arg2 < 0 || arg2 > 1) {
1466                                 error = -EINVAL;
1467                                 break;
1468                         }
1469                         set_dumpable(me->mm, arg2);
1470                         error = 0;
1471                         break;
1472
1473                 case PR_SET_UNALIGN:
1474                         error = SET_UNALIGN_CTL(me, arg2);
1475                         break;
1476                 case PR_GET_UNALIGN:
1477                         error = GET_UNALIGN_CTL(me, arg2);
1478                         break;
1479                 case PR_SET_FPEMU:
1480                         error = SET_FPEMU_CTL(me, arg2);
1481                         break;
1482                 case PR_GET_FPEMU:
1483                         error = GET_FPEMU_CTL(me, arg2);
1484                         break;
1485                 case PR_SET_FPEXC:
1486                         error = SET_FPEXC_CTL(me, arg2);
1487                         break;
1488                 case PR_GET_FPEXC:
1489                         error = GET_FPEXC_CTL(me, arg2);
1490                         break;
1491                 case PR_GET_TIMING:
1492                         error = PR_TIMING_STATISTICAL;
1493                         break;
1494                 case PR_SET_TIMING:
1495                         if (arg2 != PR_TIMING_STATISTICAL)
1496                                 error = -EINVAL;
1497                         else
1498                                 error = 0;
1499                         break;
1500
1501                 case PR_SET_NAME:
1502                         comm[sizeof(me->comm)-1] = 0;
1503                         if (strncpy_from_user(comm, (char __user *)arg2,
1504                                               sizeof(me->comm) - 1) < 0)
1505                                 return -EFAULT;
1506                         set_task_comm(me, comm);
1507                         return 0;
1508                 case PR_GET_NAME:
1509                         get_task_comm(comm, me);
1510                         if (copy_to_user((char __user *)arg2, comm,
1511                                          sizeof(comm)))
1512                                 return -EFAULT;
1513                         return 0;
1514                 case PR_GET_ENDIAN:
1515                         error = GET_ENDIAN(me, arg2);
1516                         break;
1517                 case PR_SET_ENDIAN:
1518                         error = SET_ENDIAN(me, arg2);
1519                         break;
1520
1521                 case PR_GET_SECCOMP:
1522                         error = prctl_get_seccomp();
1523                         break;
1524                 case PR_SET_SECCOMP:
1525                         error = prctl_set_seccomp(arg2);
1526                         break;
1527                 case PR_GET_TSC:
1528                         error = GET_TSC_CTL(arg2);
1529                         break;
1530                 case PR_SET_TSC:
1531                         error = SET_TSC_CTL(arg2);
1532                         break;
1533                 case PR_TASK_PERF_EVENTS_DISABLE:
1534                         error = perf_event_task_disable();
1535                         break;
1536                 case PR_TASK_PERF_EVENTS_ENABLE:
1537                         error = perf_event_task_enable();
1538                         break;
1539                 case PR_GET_TIMERSLACK:
1540                         error = current->timer_slack_ns;
1541                         break;
1542                 case PR_SET_TIMERSLACK:
1543                         if (arg2 <= 0)
1544                                 current->timer_slack_ns =
1545                                         current->default_timer_slack_ns;
1546                         else
1547                                 current->timer_slack_ns = arg2;
1548                         error = 0;
1549                         break;
1550                 case PR_MCE_KILL:
1551                         if (arg4 | arg5)
1552                                 return -EINVAL;
1553                         switch (arg2) {
1554                         case PR_MCE_KILL_CLEAR:
1555                                 if (arg3 != 0)
1556                                         return -EINVAL;
1557                                 current->flags &= ~PF_MCE_PROCESS;
1558                                 break;
1559                         case PR_MCE_KILL_SET:
1560                                 current->flags |= PF_MCE_PROCESS;
1561                                 if (arg3 == PR_MCE_KILL_EARLY)
1562                                         current->flags |= PF_MCE_EARLY;
1563                                 else if (arg3 == PR_MCE_KILL_LATE)
1564                                         current->flags &= ~PF_MCE_EARLY;
1565                                 else if (arg3 == PR_MCE_KILL_DEFAULT)
1566                                         current->flags &=
1567                                                 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
1568                                 else
1569                                         return -EINVAL;
1570                                 break;
1571                         default:
1572                                 return -EINVAL;
1573                         }
1574                         error = 0;
1575                         break;
1576                 case PR_MCE_KILL_GET:
1577                         if (arg2 | arg3 | arg4 | arg5)
1578                                 return -EINVAL;
1579                         if (current->flags & PF_MCE_PROCESS)
1580                                 error = (current->flags & PF_MCE_EARLY) ?
1581                                         PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
1582                         else
1583                                 error = PR_MCE_KILL_DEFAULT;
1584                         break;
1585                 default:
1586                         error = -EINVAL;
1587                         break;
1588         }
1589         return error;
1590 }
1591
1592 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1593                 struct getcpu_cache __user *, unused)
1594 {
1595         int err = 0;
1596         int cpu = raw_smp_processor_id();
1597         if (cpup)
1598                 err |= put_user(cpu, cpup);
1599         if (nodep)
1600                 err |= put_user(cpu_to_node(cpu), nodep);
1601         return err ? -EFAULT : 0;
1602 }
1603
1604 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1605
1606 static void argv_cleanup(char **argv, char **envp)
1607 {
1608         argv_free(argv);
1609 }
1610
1611 /**
1612  * orderly_poweroff - Trigger an orderly system poweroff
1613  * @force: force poweroff if command execution fails
1614  *
1615  * This may be called from any context to trigger a system shutdown.
1616  * If the orderly shutdown fails, it will force an immediate shutdown.
1617  */
1618 int orderly_poweroff(bool force)
1619 {
1620         int argc;
1621         char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1622         static char *envp[] = {
1623                 "HOME=/",
1624                 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1625                 NULL
1626         };
1627         int ret = -ENOMEM;
1628         struct subprocess_info *info;
1629
1630         if (argv == NULL) {
1631                 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1632                        __func__, poweroff_cmd);
1633                 goto out;
1634         }
1635
1636         info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1637         if (info == NULL) {
1638                 argv_free(argv);
1639                 goto out;
1640         }
1641
1642         call_usermodehelper_setcleanup(info, argv_cleanup);
1643
1644         ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1645
1646   out:
1647         if (ret && force) {
1648                 printk(KERN_WARNING "Failed to start orderly shutdown: "
1649                        "forcing the issue\n");
1650
1651                 /* I guess this should try to kick off some daemon to
1652                    sync and poweroff asap.  Or not even bother syncing
1653                    if we're doing an emergency shutdown? */
1654                 emergency_sync();
1655                 kernel_power_off();
1656         }
1657
1658         return ret;
1659 }
1660 EXPORT_SYMBOL_GPL(orderly_poweroff);