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 SYSCALL_DEFINE3(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 SYSCALL_DEFINE2(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 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
 359                 void __user *, arg)
 360 {
 361         char buffer[256];
 362
 363         /* We only trust the superuser with rebooting the system. */
 364         if (!capable(CAP_SYS_BOOT))
 365                 return -EPERM;
 366
 367         /* For safety, we require "magic" arguments. */
 368         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 369             (magic2 != LINUX_REBOOT_MAGIC2 &&
 370                         magic2 != LINUX_REBOOT_MAGIC2A &&
 371                         magic2 != LINUX_REBOOT_MAGIC2B &&
 372                         magic2 != LINUX_REBOOT_MAGIC2C))
 373                 return -EINVAL;
 374
 375         /* Instead of trying to make the power_off code look like
 376          * halt when pm_power_off is not set do it the easy way.
 377          */
 378         if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
 379                 cmd = LINUX_REBOOT_CMD_HALT;
 380
 381         lock_kernel();
 382         switch (cmd) {
 383         case LINUX_REBOOT_CMD_RESTART:
 384                 kernel_restart(NULL);
 385                 break;
 386
 387         case LINUX_REBOOT_CMD_CAD_ON:
 388                 C_A_D = 1;
 389                 break;
 390
 391         case LINUX_REBOOT_CMD_CAD_OFF:
 392                 C_A_D = 0;
 393                 break;
 394
 395         case LINUX_REBOOT_CMD_HALT:
 396                 kernel_halt();
 397                 unlock_kernel();
 398                 do_exit(0);
 399                 break;
 400
 401         case LINUX_REBOOT_CMD_POWER_OFF:
 402                 kernel_power_off();
 403                 unlock_kernel();
 404                 do_exit(0);
 405                 break;
 406
 407         case LINUX_REBOOT_CMD_RESTART2:
 408                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 409                         unlock_kernel();
 410                         return -EFAULT;
 411                 }
 412                 buffer[sizeof(buffer) - 1] = '\0';
 413
 414                 kernel_restart(buffer);
 415                 break;
 416
 417 #ifdef CONFIG_KEXEC
 418         case LINUX_REBOOT_CMD_KEXEC:
 419                 {
 420                         int ret;
 421                         ret = kernel_kexec();
 422                         unlock_kernel();
 423                         return ret;
 424                 }
 425 #endif
 426
 427 #ifdef CONFIG_HIBERNATION
 428         case LINUX_REBOOT_CMD_SW_SUSPEND:
 429                 {
 430                         int ret = hibernate();
 431                         unlock_kernel();
 432                         return ret;
 433                 }
 434 #endif
 435
 436         default:
 437                 unlock_kernel();
 438                 return -EINVAL;
 439         }
 440         unlock_kernel();
 441         return 0;
 442 }
 443
 444 static void deferred_cad(struct work_struct *dummy)
 445 {
 446         kernel_restart(NULL);
 447 }
 448
 449 /*
 450  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 451  * As it's called within an interrupt, it may NOT sync: the only choice
 452  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 453  */
 454 void ctrl_alt_del(void)
 455 {
 456         static DECLARE_WORK(cad_work, deferred_cad);
 457
 458         if (C_A_D)
 459                 schedule_work(&cad_work);
 460         else
 461                 kill_cad_pid(SIGINT, 1);
 462 }
 463
 464 /*
 465  * Unprivileged users may change the real gid to the effective gid
 466  * or vice versa.  (BSD-style)
 467  *
 468  * If you set the real gid at all, or set the effective gid to a value not
 469  * equal to the real gid, then the saved gid is set to the new effective gid.
 470  *
 471  * This makes it possible for a setgid program to completely drop its
 472  * privileges, which is often a useful assertion to make when you are doing
 473  * a security audit over a program.
 474  *
 475  * The general idea is that a program which uses just setregid() will be
 476  * 100% compatible with BSD.  A program which uses just setgid() will be
 477  * 100% compatible with POSIX with saved IDs.
 478  *
 479  * SMP: There are not races, the GIDs are checked only by filesystem
 480  *      operations (as far as semantic preservation is concerned).
 481  */
 482 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
 483 {
 484         const struct cred *old;
 485         struct cred *new;
 486         int retval;
 487
 488         new = prepare_creds();
 489         if (!new)
 490                 return -ENOMEM;
 491         old = current_cred();
 492
 493         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 494         if (retval)
 495                 goto error;
 496
 497         retval = -EPERM;
 498         if (rgid != (gid_t) -1) {
 499                 if (old->gid == rgid ||
 500                     old->egid == rgid ||
 501                     capable(CAP_SETGID))
 502                         new->gid = rgid;
 503                 else
 504                         goto error;
 505         }
 506         if (egid != (gid_t) -1) {
 507                 if (old->gid == egid ||
 508                     old->egid == egid ||
 509                     old->sgid == egid ||
 510                     capable(CAP_SETGID))
 511                         new->egid = egid;
 512                 else
 513                         goto error;
 514         }
 515
 516         if (rgid != (gid_t) -1 ||
 517             (egid != (gid_t) -1 && egid != old->gid))
 518                 new->sgid = new->egid;
 519         new->fsgid = new->egid;
 520
 521         return commit_creds(new);
 522
 523 error:
 524         abort_creds(new);
 525         return retval;
 526 }
 527
 528 /*
 529  * setgid() is implemented like SysV w/ SAVED_IDS
 530  *
 531  * SMP: Same implicit races as above.
 532  */
 533 SYSCALL_DEFINE1(setgid, gid_t, gid)
 534 {
 535         const struct cred *old;
 536         struct cred *new;
 537         int retval;
 538
 539         new = prepare_creds();
 540         if (!new)
 541                 return -ENOMEM;
 542         old = current_cred();
 543
 544         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 545         if (retval)
 546                 goto error;
 547
 548         retval = -EPERM;
 549         if (capable(CAP_SETGID))
 550                 new->gid = new->egid = new->sgid = new->fsgid = gid;
 551         else if (gid == old->gid || gid == old->sgid)
 552                 new->egid = new->fsgid = gid;
 553         else
 554                 goto error;
 555
 556         return commit_creds(new);
 557
 558 error:
 559         abort_creds(new);
 560         return retval;
 561 }
 562
 563 /*
 564  * change the user struct in a credentials set to match the new UID
 565  */
 566 static int set_user(struct cred *new)
 567 {
 568         struct user_struct *new_user;
 569
 570         new_user = alloc_uid(current_user_ns(), new->uid);
 571         if (!new_user)
 572                 return -EAGAIN;
 573
 574         if (atomic_read(&new_user->processes) >=
 575                                 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
 576                         new_user != INIT_USER) {
 577                 free_uid(new_user);
 578                 return -EAGAIN;
 579         }
 580
 581         free_uid(new->user);
 582         new->user = new_user;
 583         return 0;
 584 }
 585
 586 /*
 587  * Unprivileged users may change the real uid to the effective uid
 588  * or vice versa.  (BSD-style)
 589  *
 590  * If you set the real uid at all, or set the effective uid to a value not
 591  * equal to the real uid, then the saved uid is set to the new effective uid.
 592  *
 593  * This makes it possible for a setuid program to completely drop its
 594  * privileges, which is often a useful assertion to make when you are doing
 595  * a security audit over a program.
 596  *
 597  * The general idea is that a program which uses just setreuid() will be
 598  * 100% compatible with BSD.  A program which uses just setuid() will be
 599  * 100% compatible with POSIX with saved IDs.
 600  */
 601 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
 602 {
 603         const struct cred *old;
 604         struct cred *new;
 605         int retval;
 606
 607         new = prepare_creds();
 608         if (!new)
 609                 return -ENOMEM;
 610         old = current_cred();
 611
 612         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 613         if (retval)
 614                 goto error;
 615
 616         retval = -EPERM;
 617         if (ruid != (uid_t) -1) {
 618                 new->uid = ruid;
 619                 if (old->uid != ruid &&
 620                     old->euid != ruid &&
 621                     !capable(CAP_SETUID))
 622                         goto error;
 623         }
 624
 625         if (euid != (uid_t) -1) {
 626                 new->euid = euid;
 627                 if (old->uid != euid &&
 628                     old->euid != euid &&
 629                     old->suid != euid &&
 630                     !capable(CAP_SETUID))
 631                         goto error;
 632         }
 633
 634         retval = -EAGAIN;
 635         if (new->uid != old->uid && set_user(new) < 0)
 636                 goto error;
 637
 638         if (ruid != (uid_t) -1 ||
 639             (euid != (uid_t) -1 && euid != old->uid))
 640                 new->suid = new->euid;
 641         new->fsuid = new->euid;
 642
 643         retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
 644         if (retval < 0)
 645                 goto error;
 646
 647         return commit_creds(new);
 648
 649 error:
 650         abort_creds(new);
 651         return retval;
 652 }
 653
 654 /*
 655  * setuid() is implemented like SysV with SAVED_IDS
 656  *
 657  * Note that SAVED_ID's is deficient in that a setuid root program
 658  * like sendmail, for example, cannot set its uid to be a normal
 659  * user and then switch back, because if you're root, setuid() sets
 660  * the saved uid too.  If you don't like this, blame the bright people
 661  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 662  * will allow a root program to temporarily drop privileges and be able to
 663  * regain them by swapping the real and effective uid.
 664  */
 665 SYSCALL_DEFINE1(setuid, uid_t, uid)
 666 {
 667         const struct cred *old;
 668         struct cred *new;
 669         int retval;
 670
 671         new = prepare_creds();
 672         if (!new)
 673                 return -ENOMEM;
 674         old = current_cred();
 675
 676         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 677         if (retval)
 678                 goto error;
 679
 680         retval = -EPERM;
 681         if (capable(CAP_SETUID)) {
 682                 new->suid = new->uid = uid;
 683                 if (uid != old->uid && set_user(new) < 0) {
 684                         retval = -EAGAIN;
 685                         goto error;
 686                 }
 687         } else if (uid != old->uid && uid != new->suid) {
 688                 goto error;
 689         }
 690
 691         new->fsuid = new->euid = uid;
 692
 693         retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
 694         if (retval < 0)
 695                 goto error;
 696
 697         return commit_creds(new);
 698
 699 error:
 700         abort_creds(new);
 701         return retval;
 702 }
 703
 704
 705 /*
 706  * This function implements a generic ability to update ruid, euid,
 707  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 708  */
 709 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
 710 {
 711         const struct cred *old;
 712         struct cred *new;
 713         int retval;
 714
 715         new = prepare_creds();
 716         if (!new)
 717                 return -ENOMEM;
 718
 719         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 720         if (retval)
 721                 goto error;
 722         old = current_cred();
 723
 724         retval = -EPERM;
 725         if (!capable(CAP_SETUID)) {
 726                 if (ruid != (uid_t) -1 && ruid != old->uid &&
 727                     ruid != old->euid  && ruid != old->suid)
 728                         goto error;
 729                 if (euid != (uid_t) -1 && euid != old->uid &&
 730                     euid != old->euid  && euid != old->suid)
 731                         goto error;
 732                 if (suid != (uid_t) -1 && suid != old->uid &&
 733                     suid != old->euid  && suid != old->suid)
 734                         goto error;
 735         }
 736
 737         retval = -EAGAIN;
 738         if (ruid != (uid_t) -1) {
 739                 new->uid = ruid;
 740                 if (ruid != old->uid && set_user(new) < 0)
 741                         goto error;
 742         }
 743         if (euid != (uid_t) -1)
 744                 new->euid = euid;
 745         if (suid != (uid_t) -1)
 746                 new->suid = suid;
 747         new->fsuid = new->euid;
 748
 749         retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
 750         if (retval < 0)
 751                 goto error;
 752
 753         return commit_creds(new);
 754
 755 error:
 756         abort_creds(new);
 757         return retval;
 758 }
 759
 760 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruid, uid_t __user *, euid, uid_t __user *, suid)
 761 {
 762         const struct cred *cred = current_cred();
 763         int retval;
 764
 765         if (!(retval   = put_user(cred->uid,  ruid)) &&
 766             !(retval   = put_user(cred->euid, euid)))
 767                 retval = put_user(cred->suid, suid);
 768
 769         return retval;
 770 }
 771
 772 /*
 773  * Same as above, but for rgid, egid, sgid.
 774  */
 775 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
 776 {
 777         const struct cred *old;
 778         struct cred *new;
 779         int retval;
 780
 781         new = prepare_creds();
 782         if (!new)
 783                 return -ENOMEM;
 784         old = current_cred();
 785
 786         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 787         if (retval)
 788                 goto error;
 789
 790         retval = -EPERM;
 791         if (!capable(CAP_SETGID)) {
 792                 if (rgid != (gid_t) -1 && rgid != old->gid &&
 793                     rgid != old->egid  && rgid != old->sgid)
 794                         goto error;
 795                 if (egid != (gid_t) -1 && egid != old->gid &&
 796                     egid != old->egid  && egid != old->sgid)
 797                         goto error;
 798                 if (sgid != (gid_t) -1 && sgid != old->gid &&
 799                     sgid != old->egid  && sgid != old->sgid)
 800                         goto error;
 801         }
 802
 803         if (rgid != (gid_t) -1)
 804                 new->gid = rgid;
 805         if (egid != (gid_t) -1)
 806                 new->egid = egid;
 807         if (sgid != (gid_t) -1)
 808                 new->sgid = sgid;
 809         new->fsgid = new->egid;
 810
 811         return commit_creds(new);
 812
 813 error:
 814         abort_creds(new);
 815         return retval;
 816 }
 817
 818 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgid, gid_t __user *, egid, gid_t __user *, sgid)
 819 {
 820         const struct cred *cred = current_cred();
 821         int retval;
 822
 823         if (!(retval   = put_user(cred->gid,  rgid)) &&
 824             !(retval   = put_user(cred->egid, egid)))
 825                 retval = put_user(cred->sgid, sgid);
 826
 827         return retval;
 828 }
 829
 830
 831 /*
 832  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 833  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 834  * whatever uid it wants to). It normally shadows "euid", except when
 835  * explicitly set by setfsuid() or for access..
 836  */
 837 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
 838 {
 839         const struct cred *old;
 840         struct cred *new;
 841         uid_t old_fsuid;
 842
 843         new = prepare_creds();
 844         if (!new)
 845                 return current_fsuid();
 846         old = current_cred();
 847         old_fsuid = old->fsuid;
 848
 849         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS) < 0)
 850                 goto error;
 851
 852         if (uid == old->uid  || uid == old->euid  ||
 853             uid == old->suid || uid == old->fsuid ||
 854             capable(CAP_SETUID)) {
 855                 if (uid != old_fsuid) {
 856                         new->fsuid = uid;
 857                         if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
 858                                 goto change_okay;
 859                 }
 860         }
 861
 862 error:
 863         abort_creds(new);
 864         return old_fsuid;
 865
 866 change_okay:
 867         commit_creds(new);
 868         return old_fsuid;
 869 }
 870
 871 /*
 872  * Samma på svenska..
 873  */
 874 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
 875 {
 876         const struct cred *old;
 877         struct cred *new;
 878         gid_t old_fsgid;
 879
 880         new = prepare_creds();
 881         if (!new)
 882                 return current_fsgid();
 883         old = current_cred();
 884         old_fsgid = old->fsgid;
 885
 886         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 887                 goto error;
 888
 889         if (gid == old->gid  || gid == old->egid  ||
 890             gid == old->sgid || gid == old->fsgid ||
 891             capable(CAP_SETGID)) {
 892                 if (gid != old_fsgid) {
 893                         new->fsgid = gid;
 894                         goto change_okay;
 895                 }
 896         }
 897
 898 error:
 899         abort_creds(new);
 900         return old_fsgid;
 901
 902 change_okay:
 903         commit_creds(new);
 904         return old_fsgid;
 905 }
 906
 907 void do_sys_times(struct tms *tms)
 908 {
 909         struct task_cputime cputime;
 910         cputime_t cutime, cstime;
 911
 912         thread_group_cputime(current, &cputime);
 913         spin_lock_irq(&current->sighand->siglock);
 914         cutime = current->signal->cutime;
 915         cstime = current->signal->cstime;
 916         spin_unlock_irq(&current->sighand->siglock);
 917         tms->tms_utime = cputime_to_clock_t(cputime.utime);
 918         tms->tms_stime = cputime_to_clock_t(cputime.stime);
 919         tms->tms_cutime = cputime_to_clock_t(cutime);
 920         tms->tms_cstime = cputime_to_clock_t(cstime);
 921 }
 922
 923 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
 924 {
 925         if (tbuf) {
 926                 struct tms tmp;
 927
 928                 do_sys_times(&tmp);
 929                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
 930                         return -EFAULT;
 931         }
 932         force_successful_syscall_return();
 933         return (long) jiffies_64_to_clock_t(get_jiffies_64());
 934 }
 935
 936 /*
 937  * This needs some heavy checking ...
 938  * I just haven't the stomach for it. I also don't fully
 939  * understand sessions/pgrp etc. Let somebody who does explain it.
 940  *
 941  * OK, I think I have the protection semantics right.... this is really
 942  * only important on a multi-user system anyway, to make sure one user
 943  * can't send a signal to a process owned by another.  -TYT, 12/12/91
 944  *
 945  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
 946  * LBT 04.03.94
 947  */
 948 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
 949 {
 950         struct task_struct *p;
 951         struct task_struct *group_leader = current->group_leader;
 952         struct pid *pgrp;
 953         int err;
 954
 955         if (!pid)
 956                 pid = task_pid_vnr(group_leader);
 957         if (!pgid)
 958                 pgid = pid;
 959         if (pgid < 0)
 960                 return -EINVAL;
 961
 962         /* From this point forward we keep holding onto the tasklist lock
 963          * so that our parent does not change from under us. -DaveM
 964          */
 965         write_lock_irq(&tasklist_lock);
 966
 967         err = -ESRCH;
 968         p = find_task_by_vpid(pid);
 969         if (!p)
 970                 goto out;
 971
 972         err = -EINVAL;
 973         if (!thread_group_leader(p))
 974                 goto out;
 975
 976         if (same_thread_group(p->real_parent, group_leader)) {
 977                 err = -EPERM;
 978                 if (task_session(p) != task_session(group_leader))
 979                         goto out;
 980                 err = -EACCES;
 981                 if (p->did_exec)
 982                         goto out;
 983         } else {
 984                 err = -ESRCH;
 985                 if (p != group_leader)
 986                         goto out;
 987         }
 988
 989         err = -EPERM;
 990         if (p->signal->leader)
 991                 goto out;
 992
 993         pgrp = task_pid(p);
 994         if (pgid != pid) {
 995                 struct task_struct *g;
 996
 997                 pgrp = find_vpid(pgid);
 998                 g = pid_task(pgrp, PIDTYPE_PGID);
 999                 if (!g || task_session(g) != task_session(group_leader))
1000                         goto out;
1001         }
1002
1003         err = security_task_setpgid(p, pgid);
1004         if (err)
1005                 goto out;
1006
1007         if (task_pgrp(p) != pgrp) {
1008                 change_pid(p, PIDTYPE_PGID, pgrp);
1009                 set_task_pgrp(p, pid_nr(pgrp));
1010         }
1011
1012         err = 0;
1013 out:
1014         /* All paths lead to here, thus we are safe. -DaveM */
1015         write_unlock_irq(&tasklist_lock);
1016         return err;
1017 }
1018
1019 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1020 {
1021         struct task_struct *p;
1022         struct pid *grp;
1023         int retval;
1024
1025         rcu_read_lock();
1026         if (!pid)
1027                 grp = task_pgrp(current);
1028         else {
1029                 retval = -ESRCH;
1030                 p = find_task_by_vpid(pid);
1031                 if (!p)
1032                         goto out;
1033                 grp = task_pgrp(p);
1034                 if (!grp)
1035                         goto out;
1036
1037                 retval = security_task_getpgid(p);
1038                 if (retval)
1039                         goto out;
1040         }
1041         retval = pid_vnr(grp);
1042 out:
1043         rcu_read_unlock();
1044         return retval;
1045 }
1046
1047 #ifdef __ARCH_WANT_SYS_GETPGRP
1048
1049 SYSCALL_DEFINE0(getpgrp)
1050 {
1051         return sys_getpgid(0);
1052 }
1053
1054 #endif
1055
1056 SYSCALL_DEFINE1(getsid, pid_t, pid)
1057 {
1058         struct task_struct *p;
1059         struct pid *sid;
1060         int retval;
1061
1062         rcu_read_lock();
1063         if (!pid)
1064                 sid = task_session(current);
1065         else {
1066                 retval = -ESRCH;
1067                 p = find_task_by_vpid(pid);
1068                 if (!p)
1069                         goto out;
1070                 sid = task_session(p);
1071                 if (!sid)
1072                         goto out;
1073
1074                 retval = security_task_getsid(p);
1075                 if (retval)
1076                         goto out;
1077         }
1078         retval = pid_vnr(sid);
1079 out:
1080         rcu_read_unlock();
1081         return retval;
1082 }
1083
1084 SYSCALL_DEFINE0(setsid)
1085 {
1086         struct task_struct *group_leader = current->group_leader;
1087         struct pid *sid = task_pid(group_leader);
1088         pid_t session = pid_vnr(sid);
1089         int err = -EPERM;
1090
1091         write_lock_irq(&tasklist_lock);
1092         /* Fail if I am already a session leader */
1093         if (group_leader->signal->leader)
1094                 goto out;
1095
1096         /* Fail if a process group id already exists that equals the
1097          * proposed session id.
1098          */
1099         if (pid_task(sid, PIDTYPE_PGID))
1100                 goto out;
1101
1102         group_leader->signal->leader = 1;
1103         __set_special_pids(sid);
1104
1105         proc_clear_tty(group_leader);
1106
1107         err = session;
1108 out:
1109         write_unlock_irq(&tasklist_lock);
1110         return err;
1111 }
1112
1113 /*
1114  * Supplementary group IDs
1115  */
1116
1117 /* init to 2 - one for init_task, one to ensure it is never freed */
1118 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1119
1120 struct group_info *groups_alloc(int gidsetsize)
1121 {
1122         struct group_info *group_info;
1123         int nblocks;
1124         int i;
1125
1126         nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1127         /* Make sure we always allocate at least one indirect block pointer */
1128         nblocks = nblocks ? : 1;
1129         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1130         if (!group_info)
1131                 return NULL;
1132         group_info->ngroups = gidsetsize;
1133         group_info->nblocks = nblocks;
1134         atomic_set(&group_info->usage, 1);
1135
1136         if (gidsetsize <= NGROUPS_SMALL)
1137                 group_info->blocks[0] = group_info->small_block;
1138         else {
1139                 for (i = 0; i < nblocks; i++) {
1140                         gid_t *b;
1141                         b = (void *)__get_free_page(GFP_USER);
1142                         if (!b)
1143                                 goto out_undo_partial_alloc;
1144                         group_info->blocks[i] = b;
1145                 }
1146         }
1147         return group_info;
1148
1149 out_undo_partial_alloc:
1150         while (--i >= 0) {
1151                 free_page((unsigned long)group_info->blocks[i]);
1152         }
1153         kfree(group_info);
1154         return NULL;
1155 }
1156
1157 EXPORT_SYMBOL(groups_alloc);
1158
1159 void groups_free(struct group_info *group_info)
1160 {
1161         if (group_info->blocks[0] != group_info->small_block) {
1162                 int i;
1163                 for (i = 0; i < group_info->nblocks; i++)
1164                         free_page((unsigned long)group_info->blocks[i]);
1165         }
1166         kfree(group_info);
1167 }
1168
1169 EXPORT_SYMBOL(groups_free);
1170
1171 /* export the group_info to a user-space array */
1172 static int groups_to_user(gid_t __user *grouplist,
1173                           const struct group_info *group_info)
1174 {
1175         int i;
1176         unsigned int count = group_info->ngroups;
1177
1178         for (i = 0; i < group_info->nblocks; i++) {
1179                 unsigned int cp_count = min(NGROUPS_PER_BLOCK, count);
1180                 unsigned int len = cp_count * sizeof(*grouplist);
1181
1182                 if (copy_to_user(grouplist, group_info->blocks[i], len))
1183                         return -EFAULT;
1184
1185                 grouplist += NGROUPS_PER_BLOCK;
1186                 count -= cp_count;
1187         }
1188         return 0;
1189 }
1190
1191 /* fill a group_info from a user-space array - it must be allocated already */
1192 static int groups_from_user(struct group_info *group_info,
1193     gid_t __user *grouplist)
1194 {
1195         int i;
1196         unsigned int count = group_info->ngroups;
1197
1198         for (i = 0; i < group_info->nblocks; i++) {
1199                 unsigned int cp_count = min(NGROUPS_PER_BLOCK, count);
1200                 unsigned int len = cp_count * sizeof(*grouplist);
1201
1202                 if (copy_from_user(group_info->blocks[i], grouplist, len))
1203                         return -EFAULT;
1204
1205                 grouplist += NGROUPS_PER_BLOCK;
1206                 count -= cp_count;
1207         }
1208         return 0;
1209 }
1210
1211 /* a simple Shell sort */
1212 static void groups_sort(struct group_info *group_info)
1213 {
1214         int base, max, stride;
1215         int gidsetsize = group_info->ngroups;
1216
1217         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1218                 ; /* nothing */
1219         stride /= 3;
1220
1221         while (stride) {
1222                 max = gidsetsize - stride;
1223                 for (base = 0; base < max; base++) {
1224                         int left = base;
1225                         int right = left + stride;
1226                         gid_t tmp = GROUP_AT(group_info, right);
1227
1228                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1229                                 GROUP_AT(group_info, right) =
1230                                     GROUP_AT(group_info, left);
1231                                 right = left;
1232                                 left -= stride;
1233                         }
1234                         GROUP_AT(group_info, right) = tmp;
1235                 }
1236                 stride /= 3;
1237         }
1238 }
1239
1240 /* a simple bsearch */
1241 int groups_search(const struct group_info *group_info, gid_t grp)
1242 {
1243         unsigned int left, right;
1244
1245         if (!group_info)
1246                 return 0;
1247
1248         left = 0;
1249         right = group_info->ngroups;
1250         while (left < right) {
1251                 unsigned int mid = (left+right)/2;
1252                 int cmp = grp - GROUP_AT(group_info, mid);
1253                 if (cmp > 0)
1254                         left = mid + 1;
1255                 else if (cmp < 0)
1256                         right = mid;
1257                 else
1258                         return 1;
1259         }
1260         return 0;
1261 }
1262
1263 /**
1264  * set_groups - Change a group subscription in a set of credentials
1265  * @new: The newly prepared set of credentials to alter
1266  * @group_info: The group list to install
1267  *
1268  * Validate a group subscription and, if valid, insert it into a set
1269  * of credentials.
1270  */
1271 int set_groups(struct cred *new, struct group_info *group_info)
1272 {
1273         int retval;
1274
1275         retval = security_task_setgroups(group_info);
1276         if (retval)
1277                 return retval;
1278
1279         put_group_info(new->group_info);
1280         groups_sort(group_info);
1281         get_group_info(group_info);
1282         new->group_info = group_info;
1283         return 0;
1284 }
1285
1286 EXPORT_SYMBOL(set_groups);
1287
1288 /**
1289  * set_current_groups - Change current's group subscription
1290  * @group_info: The group list to impose
1291  *
1292  * Validate a group subscription and, if valid, impose it upon current's task
1293  * security record.
1294  */
1295 int set_current_groups(struct group_info *group_info)
1296 {
1297         struct cred *new;
1298         int ret;
1299
1300         new = prepare_creds();
1301         if (!new)
1302                 return -ENOMEM;
1303
1304         ret = set_groups(new, group_info);
1305         if (ret < 0) {
1306                 abort_creds(new);
1307                 return ret;
1308         }
1309
1310         return commit_creds(new);
1311 }
1312
1313 EXPORT_SYMBOL(set_current_groups);
1314
1315 SYSCALL_DEFINE2(getgroups, int, gidsetsize, gid_t __user *, grouplist)
1316 {
1317         const struct cred *cred = current_cred();
1318         int i;
1319
1320         if (gidsetsize < 0)
1321                 return -EINVAL;
1322
1323         /* no need to grab task_lock here; it cannot change */
1324         i = cred->group_info->ngroups;
1325         if (gidsetsize) {
1326                 if (i > gidsetsize) {
1327                         i = -EINVAL;
1328                         goto out;
1329                 }
1330                 if (groups_to_user(grouplist, cred->group_info)) {
1331                         i = -EFAULT;
1332                         goto out;
1333                 }
1334         }
1335 out:
1336         return i;
1337 }
1338
1339 /*
1340  *      SMP: Our groups are copy-on-write. We can set them safely
1341  *      without another task interfering.
1342  */
1343
1344 SYSCALL_DEFINE2(setgroups, int, gidsetsize, gid_t __user *, grouplist)
1345 {
1346         struct group_info *group_info;
1347         int retval;
1348
1349         if (!capable(CAP_SETGID))
1350                 return -EPERM;
1351         if ((unsigned)gidsetsize > NGROUPS_MAX)
1352                 return -EINVAL;
1353
1354         group_info = groups_alloc(gidsetsize);
1355         if (!group_info)
1356                 return -ENOMEM;
1357         retval = groups_from_user(group_info, grouplist);
1358         if (retval) {
1359                 put_group_info(group_info);
1360                 return retval;
1361         }
1362
1363         retval = set_current_groups(group_info);
1364         put_group_info(group_info);
1365
1366         return retval;
1367 }
1368
1369 /*
1370  * Check whether we're fsgid/egid or in the supplemental group..
1371  */
1372 int in_group_p(gid_t grp)
1373 {
1374         const struct cred *cred = current_cred();
1375         int retval = 1;
1376
1377         if (grp != cred->fsgid)
1378                 retval = groups_search(cred->group_info, grp);
1379         return retval;
1380 }
1381
1382 EXPORT_SYMBOL(in_group_p);
1383
1384 int in_egroup_p(gid_t grp)
1385 {
1386         const struct cred *cred = current_cred();
1387         int retval = 1;
1388
1389         if (grp != cred->egid)
1390                 retval = groups_search(cred->group_info, grp);
1391         return retval;
1392 }
1393
1394 EXPORT_SYMBOL(in_egroup_p);
1395
1396 DECLARE_RWSEM(uts_sem);
1397
1398 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1399 {
1400         int errno = 0;
1401
1402         down_read(&uts_sem);
1403         if (copy_to_user(name, utsname(), sizeof *name))
1404                 errno = -EFAULT;
1405         up_read(&uts_sem);
1406         return errno;
1407 }
1408
1409 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1410 {
1411         int errno;
1412         char tmp[__NEW_UTS_LEN];
1413
1414         if (!capable(CAP_SYS_ADMIN))
1415                 return -EPERM;
1416         if (len < 0 || len > __NEW_UTS_LEN)
1417                 return -EINVAL;
1418         down_write(&uts_sem);
1419         errno = -EFAULT;
1420         if (!copy_from_user(tmp, name, len)) {
1421                 struct new_utsname *u = utsname();
1422
1423                 memcpy(u->nodename, tmp, len);
1424                 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1425                 errno = 0;
1426         }
1427         up_write(&uts_sem);
1428         return errno;
1429 }
1430
1431 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1432
1433 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1434 {
1435         int i, errno;
1436         struct new_utsname *u;
1437
1438         if (len < 0)
1439                 return -EINVAL;
1440         down_read(&uts_sem);
1441         u = utsname();
1442         i = 1 + strlen(u->nodename);
1443         if (i > len)
1444                 i = len;
1445         errno = 0;
1446         if (copy_to_user(name, u->nodename, i))
1447                 errno = -EFAULT;
1448         up_read(&uts_sem);
1449         return errno;
1450 }
1451
1452 #endif
1453
1454 /*
1455  * Only setdomainname; getdomainname can be implemented by calling
1456  * uname()
1457  */
1458 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1459 {
1460         int errno;
1461         char tmp[__NEW_UTS_LEN];
1462
1463         if (!capable(CAP_SYS_ADMIN))
1464                 return -EPERM;
1465         if (len < 0 || len > __NEW_UTS_LEN)
1466                 return -EINVAL;
1467
1468         down_write(&uts_sem);
1469         errno = -EFAULT;
1470         if (!copy_from_user(tmp, name, len)) {
1471                 struct new_utsname *u = utsname();
1472
1473                 memcpy(u->domainname, tmp, len);
1474                 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1475                 errno = 0;
1476         }
1477         up_write(&uts_sem);
1478         return errno;
1479 }
1480
1481 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1482 {
1483         if (resource >= RLIM_NLIMITS)
1484                 return -EINVAL;
1485         else {
1486                 struct rlimit value;
1487                 task_lock(current->group_leader);
1488                 value = current->signal->rlim[resource];
1489                 task_unlock(current->group_leader);
1490                 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1491         }
1492 }
1493
1494 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1495
1496 /*
1497  *      Back compatibility for getrlimit. Needed for some apps.
1498  */
1499
1500 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1501                 struct rlimit __user *, rlim)
1502 {
1503         struct rlimit x;
1504         if (resource >= RLIM_NLIMITS)
1505                 return -EINVAL;
1506
1507         task_lock(current->group_leader);
1508         x = current->signal->rlim[resource];
1509         task_unlock(current->group_leader);
1510         if (x.rlim_cur > 0x7FFFFFFF)
1511                 x.rlim_cur = 0x7FFFFFFF;
1512         if (x.rlim_max > 0x7FFFFFFF)
1513                 x.rlim_max = 0x7FFFFFFF;
1514         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1515 }
1516
1517 #endif
1518
1519 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1520 {
1521         struct rlimit new_rlim, *old_rlim;
1522         int retval;
1523
1524         if (resource >= RLIM_NLIMITS)
1525                 return -EINVAL;
1526         if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1527                 return -EFAULT;
1528         old_rlim = current->signal->rlim + resource;
1529         if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1530             !capable(CAP_SYS_RESOURCE))
1531                 return -EPERM;
1532
1533         if (resource == RLIMIT_NOFILE) {
1534                 if (new_rlim.rlim_max == RLIM_INFINITY)
1535                         new_rlim.rlim_max = sysctl_nr_open;
1536                 if (new_rlim.rlim_cur == RLIM_INFINITY)
1537                         new_rlim.rlim_cur = sysctl_nr_open;
1538                 if (new_rlim.rlim_max > sysctl_nr_open)
1539                         return -EPERM;
1540         }
1541
1542         if (new_rlim.rlim_cur > new_rlim.rlim_max)
1543                 return -EINVAL;
1544
1545         retval = security_task_setrlimit(resource, &new_rlim);
1546         if (retval)
1547                 return retval;
1548
1549         if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1550                 /*
1551                  * The caller is asking for an immediate RLIMIT_CPU
1552                  * expiry.  But we use the zero value to mean "it was
1553                  * never set".  So let's cheat and make it one second
1554                  * instead
1555                  */
1556                 new_rlim.rlim_cur = 1;
1557         }
1558
1559         task_lock(current->group_leader);
1560         *old_rlim = new_rlim;
1561         task_unlock(current->group_leader);
1562
1563         if (resource != RLIMIT_CPU)
1564                 goto out;
1565
1566         /*
1567          * RLIMIT_CPU handling.   Note that the kernel fails to return an error
1568          * code if it rejected the user's attempt to set RLIMIT_CPU.  This is a
1569          * very long-standing error, and fixing it now risks breakage of
1570          * applications, so we live with it
1571          */
1572         if (new_rlim.rlim_cur == RLIM_INFINITY)
1573                 goto out;
1574
1575         update_rlimit_cpu(new_rlim.rlim_cur);
1576 out:
1577         return 0;
1578 }
1579
1580 /*
1581  * It would make sense to put struct rusage in the task_struct,
1582  * except that would make the task_struct be *really big*.  After
1583  * task_struct gets moved into malloc'ed memory, it would
1584  * make sense to do this.  It will make moving the rest of the information
1585  * a lot simpler!  (Which we're not doing right now because we're not
1586  * measuring them yet).
1587  *
1588  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1589  * races with threads incrementing their own counters.  But since word
1590  * reads are atomic, we either get new values or old values and we don't
1591  * care which for the sums.  We always take the siglock to protect reading
1592  * the c* fields from p->signal from races with exit.c updating those
1593  * fields when reaping, so a sample either gets all the additions of a
1594  * given child after it's reaped, or none so this sample is before reaping.
1595  *
1596  * Locking:
1597  * We need to take the siglock for CHILDEREN, SELF and BOTH
1598  * for  the cases current multithreaded, non-current single threaded
1599  * non-current multithreaded.  Thread traversal is now safe with
1600  * the siglock held.
1601  * Strictly speaking, we donot need to take the siglock if we are current and
1602  * single threaded,  as no one else can take our signal_struct away, no one
1603  * else can  reap the  children to update signal->c* counters, and no one else
1604  * can race with the signal-> fields. If we do not take any lock, the
1605  * signal-> fields could be read out of order while another thread was just
1606  * exiting. So we should  place a read memory barrier when we avoid the lock.
1607  * On the writer side,  write memory barrier is implied in  __exit_signal
1608  * as __exit_signal releases  the siglock spinlock after updating the signal->
1609  * fields. But we don't do this yet to keep things simple.
1610  *
1611  */
1612
1613 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1614 {
1615         r->ru_nvcsw += t->nvcsw;
1616         r->ru_nivcsw += t->nivcsw;
1617         r->ru_minflt += t->min_flt;
1618         r->ru_majflt += t->maj_flt;
1619         r->ru_inblock += task_io_get_inblock(t);
1620         r->ru_oublock += task_io_get_oublock(t);
1621 }
1622
1623 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1624 {
1625         struct task_struct *t;
1626         unsigned long flags;
1627         cputime_t utime, stime;
1628         struct task_cputime cputime;
1629
1630         memset((char *) r, 0, sizeof *r);
1631         utime = stime = cputime_zero;
1632
1633         if (who == RUSAGE_THREAD) {
1634                 utime = task_utime(current);
1635                 stime = task_stime(current);
1636                 accumulate_thread_rusage(p, r);
1637                 goto out;
1638         }
1639
1640         if (!lock_task_sighand(p, &flags))
1641                 return;
1642
1643         switch (who) {
1644                 case RUSAGE_BOTH:
1645                 case RUSAGE_CHILDREN:
1646                         utime = p->signal->cutime;
1647                         stime = p->signal->cstime;
1648                         r->ru_nvcsw = p->signal->cnvcsw;
1649                         r->ru_nivcsw = p->signal->cnivcsw;
1650                         r->ru_minflt = p->signal->cmin_flt;
1651                         r->ru_majflt = p->signal->cmaj_flt;
1652                         r->ru_inblock = p->signal->cinblock;
1653                         r->ru_oublock = p->signal->coublock;
1654
1655                         if (who == RUSAGE_CHILDREN)
1656                                 break;
1657
1658                 case RUSAGE_SELF:
1659                         thread_group_cputime(p, &cputime);
1660                         utime = cputime_add(utime, cputime.utime);
1661                         stime = cputime_add(stime, cputime.stime);
1662                         r->ru_nvcsw += p->signal->nvcsw;
1663                         r->ru_nivcsw += p->signal->nivcsw;
1664                         r->ru_minflt += p->signal->min_flt;
1665                         r->ru_majflt += p->signal->maj_flt;
1666                         r->ru_inblock += p->signal->inblock;
1667                         r->ru_oublock += p->signal->oublock;
1668                         t = p;
1669                         do {
1670                                 accumulate_thread_rusage(t, r);
1671                                 t = next_thread(t);
1672                         } while (t != p);
1673                         break;
1674
1675                 default:
1676                         BUG();
1677         }
1678         unlock_task_sighand(p, &flags);
1679
1680 out:
1681         cputime_to_timeval(utime, &r->ru_utime);
1682         cputime_to_timeval(stime, &r->ru_stime);
1683 }
1684
1685 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1686 {
1687         struct rusage r;
1688         k_getrusage(p, who, &r);
1689         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1690 }
1691
1692 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1693 {
1694         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1695             who != RUSAGE_THREAD)
1696                 return -EINVAL;
1697         return getrusage(current, who, ru);
1698 }
1699
1700 SYSCALL_DEFINE1(umask, int, mask)
1701 {
1702         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1703         return mask;
1704 }
1705
1706 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1707                 unsigned long, arg4, unsigned long, arg5)
1708 {
1709         struct task_struct *me = current;
1710         unsigned char comm[sizeof(me->comm)];
1711         long error;
1712
1713         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1714         if (error != -ENOSYS)
1715                 return error;
1716
1717         error = 0;
1718         switch (option) {
1719                 case PR_SET_PDEATHSIG:
1720                         if (!valid_signal(arg2)) {
1721                                 error = -EINVAL;
1722                                 break;
1723                         }
1724                         me->pdeath_signal = arg2;
1725                         error = 0;
1726                         break;
1727                 case PR_GET_PDEATHSIG:
1728                         error = put_user(me->pdeath_signal, (int __user *)arg2);
1729                         break;
1730                 case PR_GET_DUMPABLE:
1731                         error = get_dumpable(me->mm);
1732                         break;
1733                 case PR_SET_DUMPABLE:
1734                         if (arg2 < 0 || arg2 > 1) {
1735                                 error = -EINVAL;
1736                                 break;
1737                         }
1738                         set_dumpable(me->mm, arg2);
1739                         error = 0;
1740                         break;
1741
1742                 case PR_SET_UNALIGN:
1743                         error = SET_UNALIGN_CTL(me, arg2);
1744                         break;
1745                 case PR_GET_UNALIGN:
1746                         error = GET_UNALIGN_CTL(me, arg2);
1747                         break;
1748                 case PR_SET_FPEMU:
1749                         error = SET_FPEMU_CTL(me, arg2);
1750                         break;
1751                 case PR_GET_FPEMU:
1752                         error = GET_FPEMU_CTL(me, arg2);
1753                         break;
1754                 case PR_SET_FPEXC:
1755                         error = SET_FPEXC_CTL(me, arg2);
1756                         break;
1757                 case PR_GET_FPEXC:
1758                         error = GET_FPEXC_CTL(me, arg2);
1759                         break;
1760                 case PR_GET_TIMING:
1761                         error = PR_TIMING_STATISTICAL;
1762                         break;
1763                 case PR_SET_TIMING:
1764                         if (arg2 != PR_TIMING_STATISTICAL)
1765                                 error = -EINVAL;
1766                         else
1767                                 error = 0;
1768                         break;
1769
1770                 case PR_SET_NAME:
1771                         comm[sizeof(me->comm)-1] = 0;
1772                         if (strncpy_from_user(comm, (char __user *)arg2,
1773                                               sizeof(me->comm) - 1) < 0)
1774                                 return -EFAULT;
1775                         set_task_comm(me, comm);
1776                         return 0;
1777                 case PR_GET_NAME:
1778                         get_task_comm(comm, me);
1779                         if (copy_to_user((char __user *)arg2, comm,
1780                                          sizeof(comm)))
1781                                 return -EFAULT;
1782                         return 0;
1783                 case PR_GET_ENDIAN:
1784                         error = GET_ENDIAN(me, arg2);
1785                         break;
1786                 case PR_SET_ENDIAN:
1787                         error = SET_ENDIAN(me, arg2);
1788                         break;
1789
1790                 case PR_GET_SECCOMP:
1791                         error = prctl_get_seccomp();
1792                         break;
1793                 case PR_SET_SECCOMP:
1794                         error = prctl_set_seccomp(arg2);
1795                         break;
1796                 case PR_GET_TSC:
1797                         error = GET_TSC_CTL(arg2);
1798                         break;
1799                 case PR_SET_TSC:
1800                         error = SET_TSC_CTL(arg2);
1801                         break;
1802                 case PR_GET_TIMERSLACK:
1803                         error = current->timer_slack_ns;
1804                         break;
1805                 case PR_SET_TIMERSLACK:
1806                         if (arg2 <= 0)
1807                                 current->timer_slack_ns =
1808                                         current->default_timer_slack_ns;
1809                         else
1810                                 current->timer_slack_ns = arg2;
1811                         error = 0;
1812                         break;
1813                 default:
1814                         error = -EINVAL;
1815                         break;
1816         }
1817         return error;
1818 }
1819
1820 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1821                 struct getcpu_cache __user *, unused)
1822 {
1823         int err = 0;
1824         int cpu = raw_smp_processor_id();
1825         if (cpup)
1826                 err |= put_user(cpu, cpup);
1827         if (nodep)
1828                 err |= put_user(cpu_to_node(cpu), nodep);
1829         return err ? -EFAULT : 0;
1830 }
1831
1832 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1833
1834 static void argv_cleanup(char **argv, char **envp)
1835 {
1836         argv_free(argv);
1837 }
1838
1839 /**
1840  * orderly_poweroff - Trigger an orderly system poweroff
1841  * @force: force poweroff if command execution fails
1842  *
1843  * This may be called from any context to trigger a system shutdown.
1844  * If the orderly shutdown fails, it will force an immediate shutdown.
1845  */
1846 int orderly_poweroff(bool force)
1847 {
1848         int argc;
1849         char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1850         static char *envp[] = {
1851                 "HOME=/",
1852                 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1853                 NULL
1854         };
1855         int ret = -ENOMEM;
1856         struct subprocess_info *info;
1857
1858         if (argv == NULL) {
1859                 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1860                        __func__, poweroff_cmd);
1861                 goto out;
1862         }
1863
1864         info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1865         if (info == NULL) {
1866                 argv_free(argv);
1867                 goto out;
1868         }
1869
1870         call_usermodehelper_setcleanup(info, argv_cleanup);
1871
1872         ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1873
1874   out:
1875         if (ret && force) {
1876                 printk(KERN_WARNING "Failed to start orderly shutdown: "
1877                        "forcing the issue\n");
1878
1879                 /* I guess this should try to kick off some daemon to
1880                    sync and poweroff asap.  Or not even bother syncing
1881                    if we're doing an emergency shutdown? */
1882                 emergency_sync();
1883                 kernel_power_off();
1884         }
1885
1886         return ret;
1887 }
1888 EXPORT_SYMBOL_GPL(orderly_poweroff);