kernel/sys.c

   1 /*
   2  *  linux/kernel/sys.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6
   7 #include <linux/config.h>
   8 #include <linux/module.h>
   9 #include <linux/mm.h>
  10 #include <linux/utsname.h>
  11 #include <linux/mman.h>
  12 #include <linux/smp_lock.h>
  13 #include <linux/notifier.h>
  14 #include <linux/kmod.h>
  15 #include <linux/reboot.h>
  16 #include <linux/prctl.h>
  17 #include <linux/init.h>
  18 #include <linux/highuid.h>
  19 #include <linux/fs.h>
  20 #include <linux/workqueue.h>
  21 #include <linux/device.h>
  22 #include <linux/key.h>
  23 #include <linux/times.h>
  24 #include <linux/posix-timers.h>
  25 #include <linux/security.h>
  26 #include <linux/dcookies.h>
  27 #include <linux/suspend.h>
  28 #include <linux/tty.h>
  29 #include <linux/signal.h>
  30
  31 #include <linux/compat.h>
  32 #include <linux/syscalls.h>
  33 #include <linux/vs_cvirt.h>
  34
  35 #include <asm/uaccess.h>
  36 #include <asm/io.h>
  37 #include <asm/unistd.h>
  38
  39 #ifndef SET_UNALIGN_CTL
  40 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  41 #endif
  42 #ifndef GET_UNALIGN_CTL
  43 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  44 #endif
  45 #ifndef SET_FPEMU_CTL
  46 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
  47 #endif
  48 #ifndef GET_FPEMU_CTL
  49 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
  50 #endif
  51 #ifndef SET_FPEXC_CTL
  52 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
  53 #endif
  54 #ifndef GET_FPEXC_CTL
  55 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
  56 #endif
  57
  58 /*
  59  * this is where the system-wide overflow UID and GID are defined, for
  60  * architectures that now have 32-bit UID/GID but didn't in the past
  61  */
  62
  63 int overflowuid = DEFAULT_OVERFLOWUID;
  64 int overflowgid = DEFAULT_OVERFLOWGID;
  65
  66 #ifdef CONFIG_UID16
  67 EXPORT_SYMBOL(overflowuid);
  68 EXPORT_SYMBOL(overflowgid);
  69 #endif
  70
  71 /*
  72  * the same as above, but for filesystems which can only store a 16-bit
  73  * UID and GID. as such, this is needed on all architectures
  74  */
  75
  76 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
  77 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
  78
  79 EXPORT_SYMBOL(fs_overflowuid);
  80 EXPORT_SYMBOL(fs_overflowgid);
  81
  82 /*
  83  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
  84  */
  85
  86 int C_A_D = 1;
  87 int cad_pid = 1;
  88
  89 /*
  90  *      Notifier list for kernel code which wants to be called
  91  *      at shutdown. This is used to stop any idling DMA operations
  92  *      and the like.
  93  */
  94
  95 static struct notifier_block *reboot_notifier_list;
  96 static DEFINE_RWLOCK(notifier_lock);
  97
  98 /**
  99  *      notifier_chain_register - Add notifier to a notifier chain
 100  *      @list: Pointer to root list pointer
 101  *      @n: New entry in notifier chain
 102  *
 103  *      Adds a notifier to a notifier chain.
 104  *
 105  *      Currently always returns zero.
 106  */
 107
 108 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
 109 {
 110         write_lock(&notifier_lock);
 111         while(*list)
 112         {
 113                 if(n->priority > (*list)->priority)
 114                         break;
 115                 list= &((*list)->next);
 116         }
 117         n->next = *list;
 118         *list=n;
 119         write_unlock(&notifier_lock);
 120         return 0;
 121 }
 122
 123 EXPORT_SYMBOL(notifier_chain_register);
 124
 125 /**
 126  *      notifier_chain_unregister - Remove notifier from a notifier chain
 127  *      @nl: Pointer to root list pointer
 128  *      @n: New entry in notifier chain
 129  *
 130  *      Removes a notifier from a notifier chain.
 131  *
 132  *      Returns zero on success, or %-ENOENT on failure.
 133  */
 134
 135 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
 136 {
 137         write_lock(&notifier_lock);
 138         while((*nl)!=NULL)
 139         {
 140                 if((*nl)==n)
 141                 {
 142                         *nl=n->next;
 143                         write_unlock(&notifier_lock);
 144                         return 0;
 145                 }
 146                 nl=&((*nl)->next);
 147         }
 148         write_unlock(&notifier_lock);
 149         return -ENOENT;
 150 }
 151
 152 EXPORT_SYMBOL(notifier_chain_unregister);
 153
 154 /**
 155  *      notifier_call_chain - Call functions in a notifier chain
 156  *      @n: Pointer to root pointer of notifier chain
 157  *      @val: Value passed unmodified to notifier function
 158  *      @v: Pointer passed unmodified to notifier function
 159  *
 160  *      Calls each function in a notifier chain in turn.
 161  *
 162  *      If the return value of the notifier can be and'd
 163  *      with %NOTIFY_STOP_MASK, then notifier_call_chain
 164  *      will return immediately, with the return value of
 165  *      the notifier function which halted execution.
 166  *      Otherwise, the return value is the return value
 167  *      of the last notifier function called.
 168  */
 169
 170 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
 171 {
 172         int ret=NOTIFY_DONE;
 173         struct notifier_block *nb = *n;
 174
 175         while(nb)
 176         {
 177                 ret=nb->notifier_call(nb,val,v);
 178                 if(ret&NOTIFY_STOP_MASK)
 179                 {
 180                         return ret;
 181                 }
 182                 nb=nb->next;
 183         }
 184         return ret;
 185 }
 186
 187 EXPORT_SYMBOL(notifier_call_chain);
 188
 189 /**
 190  *      register_reboot_notifier - Register function to be called at reboot time
 191  *      @nb: Info about notifier function to be called
 192  *
 193  *      Registers a function with the list of functions
 194  *      to be called at reboot time.
 195  *
 196  *      Currently always returns zero, as notifier_chain_register
 197  *      always returns zero.
 198  */
 199
 200 int register_reboot_notifier(struct notifier_block * nb)
 201 {
 202         return notifier_chain_register(&reboot_notifier_list, nb);
 203 }
 204
 205 EXPORT_SYMBOL(register_reboot_notifier);
 206
 207 /**
 208  *      unregister_reboot_notifier - Unregister previously registered reboot notifier
 209  *      @nb: Hook to be unregistered
 210  *
 211  *      Unregisters a previously registered reboot
 212  *      notifier function.
 213  *
 214  *      Returns zero on success, or %-ENOENT on failure.
 215  */
 216
 217 int unregister_reboot_notifier(struct notifier_block * nb)
 218 {
 219         return notifier_chain_unregister(&reboot_notifier_list, nb);
 220 }
 221
 222 EXPORT_SYMBOL(unregister_reboot_notifier);
 223
 224 static int set_one_prio(struct task_struct *p, int niceval, int error)
 225 {
 226         int no_nice;
 227
 228         if (p->uid != current->euid &&
 229                 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
 230                 error = -EPERM;
 231                 goto out;
 232         }
 233         if (niceval < task_nice(p) && !can_nice(p, niceval)) {
 234                 if (vx_flags(VXF_IGNEG_NICE, 0))
 235                         error = 0;
 236                 else
 237                         error = -EACCES;
 238                 goto out;
 239         }
 240         no_nice = security_task_setnice(p, niceval);
 241         if (no_nice) {
 242                 error = no_nice;
 243                 goto out;
 244         }
 245         if (error == -ESRCH)
 246                 error = 0;
 247         set_user_nice(p, niceval);
 248 out:
 249         return error;
 250 }
 251
 252 asmlinkage long sys_setpriority(int which, int who, int niceval)
 253 {
 254         struct task_struct *g, *p;
 255         struct user_struct *user;
 256         int error = -EINVAL;
 257
 258         if (which > 2 || which < 0)
 259                 goto out;
 260
 261         /* normalize: avoid signed division (rounding problems) */
 262         error = -ESRCH;
 263         if (niceval < -20)
 264                 niceval = -20;
 265         if (niceval > 19)
 266                 niceval = 19;
 267
 268         read_lock(&tasklist_lock);
 269         switch (which) {
 270                 case PRIO_PROCESS:
 271                         if (!who)
 272                                 who = current->pid;
 273                         p = find_task_by_pid(who);
 274                         if (p)
 275                                 error = set_one_prio(p, niceval, error);
 276                         break;
 277                 case PRIO_PGRP:
 278                         if (!who)
 279                                 who = process_group(current);
 280                         do_each_task_pid(who, PIDTYPE_PGID, p) {
 281                                 error = set_one_prio(p, niceval, error);
 282                         } while_each_task_pid(who, PIDTYPE_PGID, p);
 283                         break;
 284                 case PRIO_USER:
 285                         user = current->user;
 286                         if (!who)
 287                                 who = current->uid;
 288                         else
 289                                 if ((who != current->uid) &&
 290                                         !(user = find_user(vx_current_xid(), who)))
 291                                         goto out_unlock;        /* No processes for this user */
 292
 293                         do_each_thread(g, p)
 294                                 if (p->uid == who)
 295                                         error = set_one_prio(p, niceval, error);
 296                         while_each_thread(g, p);
 297                         if (who != current->uid)
 298                                 free_uid(user);         /* For find_user() */
 299                         break;
 300         }
 301 out_unlock:
 302         read_unlock(&tasklist_lock);
 303 out:
 304         return error;
 305 }
 306
 307 /*
 308  * Ugh. To avoid negative return values, "getpriority()" will
 309  * not return the normal nice-value, but a negated value that
 310  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 311  * to stay compatible.
 312  */
 313 asmlinkage long sys_getpriority(int which, int who)
 314 {
 315         struct task_struct *g, *p;
 316         struct user_struct *user;
 317         long niceval, retval = -ESRCH;
 318
 319         if (which > 2 || which < 0)
 320                 return -EINVAL;
 321
 322         read_lock(&tasklist_lock);
 323         switch (which) {
 324                 case PRIO_PROCESS:
 325                         if (!who)
 326                                 who = current->pid;
 327                         p = find_task_by_pid(who);
 328                         if (p) {
 329                                 niceval = 20 - task_nice(p);
 330                                 if (niceval > retval)
 331                                         retval = niceval;
 332                         }
 333                         break;
 334                 case PRIO_PGRP:
 335                         if (!who)
 336                                 who = process_group(current);
 337                         do_each_task_pid(who, PIDTYPE_PGID, p) {
 338                                 niceval = 20 - task_nice(p);
 339                                 if (niceval > retval)
 340                                         retval = niceval;
 341                         } while_each_task_pid(who, PIDTYPE_PGID, p);
 342                         break;
 343                 case PRIO_USER:
 344                         user = current->user;
 345                         if (!who)
 346                                 who = current->uid;
 347                         else
 348                                 if ((who != current->uid) &&
 349                                         !(user = find_user(vx_current_xid(), who)))
 350                                         goto out_unlock;        /* No processes for this user */
 351
 352                         do_each_thread(g, p)
 353                                 if (p->uid == who) {
 354                                         niceval = 20 - task_nice(p);
 355                                         if (niceval > retval)
 356                                                 retval = niceval;
 357                                 }
 358                         while_each_thread(g, p);
 359                         if (who != current->uid)
 360                                 free_uid(user);         /* for find_user() */
 361                         break;
 362         }
 363 out_unlock:
 364         read_unlock(&tasklist_lock);
 365
 366         return retval;
 367 }
 368
 369 long vs_reboot(unsigned int, void *);
 370
 371 /*
 372  * Reboot system call: for obvious reasons only root may call it,
 373  * and even root needs to set up some magic numbers in the registers
 374  * so that some mistake won't make this reboot the whole machine.
 375  * You can also set the meaning of the ctrl-alt-del-key here.
 376  *
 377  * reboot doesn't sync: do that yourself before calling this.
 378  */
 379 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
 380 {
 381         char buffer[256];
 382
 383         /* We only trust the superuser with rebooting the system. */
 384         if (!capable(CAP_SYS_BOOT))
 385                 return -EPERM;
 386
 387         /* For safety, we require "magic" arguments. */
 388         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 389             (magic2 != LINUX_REBOOT_MAGIC2 &&
 390                         magic2 != LINUX_REBOOT_MAGIC2A &&
 391                         magic2 != LINUX_REBOOT_MAGIC2B &&
 392                         magic2 != LINUX_REBOOT_MAGIC2C))
 393                 return -EINVAL;
 394
 395         if (!vx_check(0, VX_ADMIN|VX_WATCH))
 396                 return vs_reboot(cmd, arg);
 397
 398         lock_kernel();
 399         switch (cmd) {
 400         case LINUX_REBOOT_CMD_RESTART:
 401                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 402                 system_state = SYSTEM_RESTART;
 403                 device_shutdown();
 404                 printk(KERN_EMERG "Restarting system.\n");
 405                 machine_restart(NULL);
 406                 break;
 407
 408         case LINUX_REBOOT_CMD_CAD_ON:
 409                 C_A_D = 1;
 410                 break;
 411
 412         case LINUX_REBOOT_CMD_CAD_OFF:
 413                 C_A_D = 0;
 414                 break;
 415
 416         case LINUX_REBOOT_CMD_HALT:
 417                 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
 418                 system_state = SYSTEM_HALT;
 419                 device_shutdown();
 420                 printk(KERN_EMERG "System halted.\n");
 421                 machine_halt();
 422                 unlock_kernel();
 423                 do_exit(0);
 424                 break;
 425
 426         case LINUX_REBOOT_CMD_POWER_OFF:
 427                 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
 428                 system_state = SYSTEM_POWER_OFF;
 429                 device_shutdown();
 430                 printk(KERN_EMERG "Power down.\n");
 431                 machine_power_off();
 432                 unlock_kernel();
 433                 do_exit(0);
 434                 break;
 435
 436         case LINUX_REBOOT_CMD_RESTART2:
 437                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 438                         unlock_kernel();
 439                         return -EFAULT;
 440                 }
 441                 buffer[sizeof(buffer) - 1] = '\0';
 442
 443                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
 444                 system_state = SYSTEM_RESTART;
 445                 device_shutdown();
 446                 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
 447                 machine_restart(buffer);
 448                 break;
 449
 450 #ifdef CONFIG_SOFTWARE_SUSPEND
 451         case LINUX_REBOOT_CMD_SW_SUSPEND:
 452                 {
 453                         int ret = software_suspend();
 454                         unlock_kernel();
 455                         return ret;
 456                 }
 457 #endif
 458
 459         default:
 460                 unlock_kernel();
 461                 return -EINVAL;
 462         }
 463         unlock_kernel();
 464         return 0;
 465 }
 466
 467 static void deferred_cad(void *dummy)
 468 {
 469         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 470         machine_restart(NULL);
 471 }
 472
 473 /*
 474  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 475  * As it's called within an interrupt, it may NOT sync: the only choice
 476  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 477  */
 478 void ctrl_alt_del(void)
 479 {
 480         static DECLARE_WORK(cad_work, deferred_cad, NULL);
 481
 482         if (C_A_D)
 483                 schedule_work(&cad_work);
 484         else
 485                 kill_proc(cad_pid, SIGINT, 1);
 486 }
 487
 488
 489 /*
 490  * Unprivileged users may change the real gid to the effective gid
 491  * or vice versa.  (BSD-style)
 492  *
 493  * If you set the real gid at all, or set the effective gid to a value not
 494  * equal to the real gid, then the saved gid is set to the new effective gid.
 495  *
 496  * This makes it possible for a setgid program to completely drop its
 497  * privileges, which is often a useful assertion to make when you are doing
 498  * a security audit over a program.
 499  *
 500  * The general idea is that a program which uses just setregid() will be
 501  * 100% compatible with BSD.  A program which uses just setgid() will be
 502  * 100% compatible with POSIX with saved IDs.
 503  *
 504  * SMP: There are not races, the GIDs are checked only by filesystem
 505  *      operations (as far as semantic preservation is concerned).
 506  */
 507 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 508 {
 509         int old_rgid = current->gid;
 510         int old_egid = current->egid;
 511         int new_rgid = old_rgid;
 512         int new_egid = old_egid;
 513         int retval;
 514
 515         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 516         if (retval)
 517                 return retval;
 518
 519         if (rgid != (gid_t) -1) {
 520                 if ((old_rgid == rgid) ||
 521                     (current->egid==rgid) ||
 522                     capable(CAP_SETGID))
 523                         new_rgid = rgid;
 524                 else
 525                         return -EPERM;
 526         }
 527         if (egid != (gid_t) -1) {
 528                 if ((old_rgid == egid) ||
 529                     (current->egid == egid) ||
 530                     (current->sgid == egid) ||
 531                     capable(CAP_SETGID))
 532                         new_egid = egid;
 533                 else {
 534                         return -EPERM;
 535                 }
 536         }
 537         if (new_egid != old_egid)
 538         {
 539                 current->mm->dumpable = 0;
 540                 smp_wmb();
 541         }
 542         if (rgid != (gid_t) -1 ||
 543             (egid != (gid_t) -1 && egid != old_rgid))
 544                 current->sgid = new_egid;
 545         current->fsgid = new_egid;
 546         current->egid = new_egid;
 547         current->gid = new_rgid;
 548         key_fsgid_changed(current);
 549         return 0;
 550 }
 551
 552 /*
 553  * setgid() is implemented like SysV w/ SAVED_IDS
 554  *
 555  * SMP: Same implicit races as above.
 556  */
 557 asmlinkage long sys_setgid(gid_t gid)
 558 {
 559         int old_egid = current->egid;
 560         int retval;
 561
 562         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 563         if (retval)
 564                 return retval;
 565
 566         if (capable(CAP_SETGID))
 567         {
 568                 if(old_egid != gid)
 569                 {
 570                         current->mm->dumpable=0;
 571                         smp_wmb();
 572                 }
 573                 current->gid = current->egid = current->sgid = current->fsgid = gid;
 574         }
 575         else if ((gid == current->gid) || (gid == current->sgid))
 576         {
 577                 if(old_egid != gid)
 578                 {
 579                         current->mm->dumpable=0;
 580                         smp_wmb();
 581                 }
 582                 current->egid = current->fsgid = gid;
 583         }
 584         else
 585                 return -EPERM;
 586
 587         key_fsgid_changed(current);
 588         return 0;
 589 }
 590
 591 static int set_user(uid_t new_ruid, int dumpclear)
 592 {
 593         struct user_struct *new_user;
 594
 595         new_user = alloc_uid(vx_current_xid(), new_ruid);
 596         if (!new_user)
 597                 return -EAGAIN;
 598
 599         if (atomic_read(&new_user->processes) >=
 600                                 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
 601                         new_user != &root_user) {
 602                 free_uid(new_user);
 603                 return -EAGAIN;
 604         }
 605
 606         switch_uid(new_user);
 607
 608         if(dumpclear)
 609         {
 610                 current->mm->dumpable = 0;
 611                 smp_wmb();
 612         }
 613         current->uid = new_ruid;
 614         return 0;
 615 }
 616
 617 /*
 618  * Unprivileged users may change the real uid to the effective uid
 619  * or vice versa.  (BSD-style)
 620  *
 621  * If you set the real uid at all, or set the effective uid to a value not
 622  * equal to the real uid, then the saved uid is set to the new effective uid.
 623  *
 624  * This makes it possible for a setuid program to completely drop its
 625  * privileges, which is often a useful assertion to make when you are doing
 626  * a security audit over a program.
 627  *
 628  * The general idea is that a program which uses just setreuid() will be
 629  * 100% compatible with BSD.  A program which uses just setuid() will be
 630  * 100% compatible with POSIX with saved IDs.
 631  */
 632 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 633 {
 634         int old_ruid, old_euid, old_suid, new_ruid, new_euid;
 635         int retval;
 636
 637         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 638         if (retval)
 639                 return retval;
 640
 641         new_ruid = old_ruid = current->uid;
 642         new_euid = old_euid = current->euid;
 643         old_suid = current->suid;
 644
 645         if (ruid != (uid_t) -1) {
 646                 new_ruid = ruid;
 647                 if ((old_ruid != ruid) &&
 648                     (current->euid != ruid) &&
 649                     !capable(CAP_SETUID))
 650                         return -EPERM;
 651         }
 652
 653         if (euid != (uid_t) -1) {
 654                 new_euid = euid;
 655                 if ((old_ruid != euid) &&
 656                     (current->euid != euid) &&
 657                     (current->suid != euid) &&
 658                     !capable(CAP_SETUID))
 659                         return -EPERM;
 660         }
 661
 662         if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
 663                 return -EAGAIN;
 664
 665         if (new_euid != old_euid)
 666         {
 667                 current->mm->dumpable=0;
 668                 smp_wmb();
 669         }
 670         current->fsuid = current->euid = new_euid;
 671         if (ruid != (uid_t) -1 ||
 672             (euid != (uid_t) -1 && euid != old_ruid))
 673                 current->suid = current->euid;
 674         current->fsuid = current->euid;
 675
 676         key_fsuid_changed(current);
 677
 678         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
 679 }
 680
 681
 682
 683 /*
 684  * setuid() is implemented like SysV with SAVED_IDS
 685  *
 686  * Note that SAVED_ID's is deficient in that a setuid root program
 687  * like sendmail, for example, cannot set its uid to be a normal
 688  * user and then switch back, because if you're root, setuid() sets
 689  * the saved uid too.  If you don't like this, blame the bright people
 690  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 691  * will allow a root program to temporarily drop privileges and be able to
 692  * regain them by swapping the real and effective uid.
 693  */
 694 asmlinkage long sys_setuid(uid_t uid)
 695 {
 696         int old_euid = current->euid;
 697         int old_ruid, old_suid, new_ruid, new_suid;
 698         int retval;
 699
 700         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 701         if (retval)
 702                 return retval;
 703
 704         old_ruid = new_ruid = current->uid;
 705         old_suid = current->suid;
 706         new_suid = old_suid;
 707
 708         if (capable(CAP_SETUID)) {
 709                 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
 710                         return -EAGAIN;
 711                 new_suid = uid;
 712         } else if ((uid != current->uid) && (uid != new_suid))
 713                 return -EPERM;
 714
 715         if (old_euid != uid)
 716         {
 717                 current->mm->dumpable = 0;
 718                 smp_wmb();
 719         }
 720         current->fsuid = current->euid = uid;
 721         current->suid = new_suid;
 722
 723         key_fsuid_changed(current);
 724
 725         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
 726 }
 727
 728
 729 /*
 730  * This function implements a generic ability to update ruid, euid,
 731  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 732  */
 733 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 734 {
 735         int old_ruid = current->uid;
 736         int old_euid = current->euid;
 737         int old_suid = current->suid;
 738         int retval;
 739
 740         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 741         if (retval)
 742                 return retval;
 743
 744         if (!capable(CAP_SETUID)) {
 745                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
 746                     (ruid != current->euid) && (ruid != current->suid))
 747                         return -EPERM;
 748                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
 749                     (euid != current->euid) && (euid != current->suid))
 750                         return -EPERM;
 751                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
 752                     (suid != current->euid) && (suid != current->suid))
 753                         return -EPERM;
 754         }
 755         if (ruid != (uid_t) -1) {
 756                 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
 757                         return -EAGAIN;
 758         }
 759         if (euid != (uid_t) -1) {
 760                 if (euid != current->euid)
 761                 {
 762                         current->mm->dumpable = 0;
 763                         smp_wmb();
 764                 }
 765                 current->euid = euid;
 766         }
 767         current->fsuid = current->euid;
 768         if (suid != (uid_t) -1)
 769                 current->suid = suid;
 770
 771         key_fsuid_changed(current);
 772
 773         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
 774 }
 775
 776 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
 777 {
 778         int retval;
 779
 780         if (!(retval = put_user(current->uid, ruid)) &&
 781             !(retval = put_user(current->euid, euid)))
 782                 retval = put_user(current->suid, suid);
 783
 784         return retval;
 785 }
 786
 787 /*
 788  * Same as above, but for rgid, egid, sgid.
 789  */
 790 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 791 {
 792         int retval;
 793
 794         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 795         if (retval)
 796                 return retval;
 797
 798         if (!capable(CAP_SETGID)) {
 799                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
 800                     (rgid != current->egid) && (rgid != current->sgid))
 801                         return -EPERM;
 802                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
 803                     (egid != current->egid) && (egid != current->sgid))
 804                         return -EPERM;
 805                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
 806                     (sgid != current->egid) && (sgid != current->sgid))
 807                         return -EPERM;
 808         }
 809         if (egid != (gid_t) -1) {
 810                 if (egid != current->egid)
 811                 {
 812                         current->mm->dumpable = 0;
 813                         smp_wmb();
 814                 }
 815                 current->egid = egid;
 816         }
 817         current->fsgid = current->egid;
 818         if (rgid != (gid_t) -1)
 819                 current->gid = rgid;
 820         if (sgid != (gid_t) -1)
 821                 current->sgid = sgid;
 822
 823         key_fsgid_changed(current);
 824         return 0;
 825 }
 826
 827 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
 828 {
 829         int retval;
 830
 831         if (!(retval = put_user(current->gid, rgid)) &&
 832             !(retval = put_user(current->egid, egid)))
 833                 retval = put_user(current->sgid, sgid);
 834
 835         return retval;
 836 }
 837
 838
 839 /*
 840  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 841  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 842  * whatever uid it wants to). It normally shadows "euid", except when
 843  * explicitly set by setfsuid() or for access..
 844  */
 845 asmlinkage long sys_setfsuid(uid_t uid)
 846 {
 847         int old_fsuid;
 848
 849         old_fsuid = current->fsuid;
 850         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
 851                 return old_fsuid;
 852
 853         if (uid == current->uid || uid == current->euid ||
 854             uid == current->suid || uid == current->fsuid ||
 855             capable(CAP_SETUID))
 856         {
 857                 if (uid != old_fsuid)
 858                 {
 859                         current->mm->dumpable = 0;
 860                         smp_wmb();
 861                 }
 862                 current->fsuid = uid;
 863         }
 864
 865         key_fsuid_changed(current);
 866
 867         security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
 868
 869         return old_fsuid;
 870 }
 871
 872 /*
 873  * Samma på svenska..
 874  */
 875 asmlinkage long sys_setfsgid(gid_t gid)
 876 {
 877         int old_fsgid;
 878
 879         old_fsgid = current->fsgid;
 880         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 881                 return old_fsgid;
 882
 883         if (gid == current->gid || gid == current->egid ||
 884             gid == current->sgid || gid == current->fsgid ||
 885             capable(CAP_SETGID))
 886         {
 887                 if (gid != old_fsgid)
 888                 {
 889                         current->mm->dumpable = 0;
 890                         smp_wmb();
 891                 }
 892                 current->fsgid = gid;
 893                 key_fsgid_changed(current);
 894         }
 895         return old_fsgid;
 896 }
 897
 898 asmlinkage long sys_times(struct tms __user * tbuf)
 899 {
 900         /*
 901          *      In the SMP world we might just be unlucky and have one of
 902          *      the times increment as we use it. Since the value is an
 903          *      atomically safe type this is just fine. Conceptually its
 904          *      as if the syscall took an instant longer to occur.
 905          */
 906         if (tbuf) {
 907                 struct tms tmp;
 908                 struct task_struct *tsk = current;
 909                 struct task_struct *t;
 910                 cputime_t utime, stime, cutime, cstime;
 911
 912                 read_lock(&tasklist_lock);
 913                 utime = tsk->signal->utime;
 914                 stime = tsk->signal->stime;
 915                 t = tsk;
 916                 do {
 917                         utime = cputime_add(utime, t->utime);
 918                         stime = cputime_add(stime, t->stime);
 919                         t = next_thread(t);
 920                 } while (t != tsk);
 921
 922                 /*
 923                  * While we have tasklist_lock read-locked, no dying thread
 924                  * can be updating current->signal->[us]time.  Instead,
 925                  * we got their counts included in the live thread loop.
 926                  * However, another thread can come in right now and
 927                  * do a wait call that updates current->signal->c[us]time.
 928                  * To make sure we always see that pair updated atomically,
 929                  * we take the siglock around fetching them.
 930                  */
 931                 spin_lock_irq(&tsk->sighand->siglock);
 932                 cutime = tsk->signal->cutime;
 933                 cstime = tsk->signal->cstime;
 934                 spin_unlock_irq(&tsk->sighand->siglock);
 935                 read_unlock(&tasklist_lock);
 936
 937                 tmp.tms_utime = cputime_to_clock_t(utime);
 938                 tmp.tms_stime = cputime_to_clock_t(stime);
 939                 tmp.tms_cutime = cputime_to_clock_t(cutime);
 940                 tmp.tms_cstime = cputime_to_clock_t(cstime);
 941                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
 942                         return -EFAULT;
 943         }
 944         return (long) jiffies_64_to_clock_t(get_jiffies_64());
 945 }
 946
 947 /*
 948  * This needs some heavy checking ...
 949  * I just haven't the stomach for it. I also don't fully
 950  * understand sessions/pgrp etc. Let somebody who does explain it.
 951  *
 952  * OK, I think I have the protection semantics right.... this is really
 953  * only important on a multi-user system anyway, to make sure one user
 954  * can't send a signal to a process owned by another.  -TYT, 12/12/91
 955  *
 956  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
 957  * LBT 04.03.94
 958  */
 959
 960 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
 961 {
 962         struct task_struct *p;
 963         int err = -EINVAL;
 964         pid_t rpgid;
 965
 966         if (!pid)
 967                 pid = vx_map_pid(current->pid);
 968         if (!pgid)
 969                 pgid = pid;
 970         if (pgid < 0)
 971                 return -EINVAL;
 972
 973         rpgid = vx_rmap_pid(pgid);
 974
 975         /* From this point forward we keep holding onto the tasklist lock
 976          * so that our parent does not change from under us. -DaveM
 977          */
 978         write_lock_irq(&tasklist_lock);
 979
 980         err = -ESRCH;
 981         p = find_task_by_pid(pid);
 982         if (!p)
 983                 goto out;
 984
 985         err = -EINVAL;
 986         if (!thread_group_leader(p))
 987                 goto out;
 988
 989         if (p->parent == current || p->real_parent == current) {
 990                 err = -EPERM;
 991                 if (p->signal->session != current->signal->session)
 992                         goto out;
 993                 err = -EACCES;
 994                 if (p->did_exec)
 995                         goto out;
 996         } else {
 997                 err = -ESRCH;
 998                 if (p != current)
 999                         goto out;
1000         }
1001
1002         err = -EPERM;
1003         if (p->signal->leader)
1004                 goto out;
1005
1006         if (pgid != pid) {
1007                 struct task_struct *p;
1008
1009                 do_each_task_pid(rpgid, PIDTYPE_PGID, p) {
1010                         if (p->signal->session == current->signal->session)
1011                                 goto ok_pgid;
1012                 } while_each_task_pid(rpgid, PIDTYPE_PGID, p);
1013                 goto out;
1014         }
1015
1016 ok_pgid:
1017         err = security_task_setpgid(p, rpgid);
1018         if (err)
1019                 goto out;
1020
1021         if (process_group(p) != rpgid) {
1022                 detach_pid(p, PIDTYPE_PGID);
1023                 p->signal->pgrp = rpgid;
1024                 attach_pid(p, PIDTYPE_PGID, rpgid);
1025         }
1026
1027         err = 0;
1028 out:
1029         /* All paths lead to here, thus we are safe. -DaveM */
1030         write_unlock_irq(&tasklist_lock);
1031         return err;
1032 }
1033
1034 asmlinkage long sys_getpgid(pid_t pid)
1035 {
1036         if (!pid) {
1037                 return vx_rmap_pid(process_group(current));
1038         } else {
1039                 int retval;
1040                 struct task_struct *p;
1041
1042                 read_lock(&tasklist_lock);
1043                 p = find_task_by_pid(pid);
1044
1045                 retval = -ESRCH;
1046                 if (p) {
1047                         retval = security_task_getpgid(p);
1048                         if (!retval)
1049                                 retval = vx_rmap_pid(process_group(p));
1050                 }
1051                 read_unlock(&tasklist_lock);
1052                 return retval;
1053         }
1054 }
1055
1056 #ifdef __ARCH_WANT_SYS_GETPGRP
1057
1058 asmlinkage long sys_getpgrp(void)
1059 {
1060         /* SMP - assuming writes are word atomic this is fine */
1061         return process_group(current);
1062 }
1063
1064 #endif
1065
1066 asmlinkage long sys_getsid(pid_t pid)
1067 {
1068         if (!pid) {
1069                 return current->signal->session;
1070         } else {
1071                 int retval;
1072                 struct task_struct *p;
1073
1074                 read_lock(&tasklist_lock);
1075                 p = find_task_by_pid(pid);
1076
1077                 retval = -ESRCH;
1078                 if(p) {
1079                         retval = security_task_getsid(p);
1080                         if (!retval)
1081                                 retval = p->signal->session;
1082                 }
1083                 read_unlock(&tasklist_lock);
1084                 return retval;
1085         }
1086 }
1087
1088 asmlinkage long sys_setsid(void)
1089 {
1090         struct pid *pid;
1091         int err = -EPERM;
1092
1093         if (!thread_group_leader(current))
1094                 return -EINVAL;
1095
1096         down(&tty_sem);
1097         write_lock_irq(&tasklist_lock);
1098
1099         pid = find_pid(PIDTYPE_PGID, current->pid);
1100         if (pid)
1101                 goto out;
1102
1103         current->signal->leader = 1;
1104         __set_special_pids(current->pid, current->pid);
1105         current->signal->tty = NULL;
1106         current->signal->tty_old_pgrp = 0;
1107         err = process_group(current);
1108 out:
1109         write_unlock_irq(&tasklist_lock);
1110         up(&tty_sem);
1111         return err;
1112 }
1113
1114 /*
1115  * Supplementary group IDs
1116  */
1117
1118 /* init to 2 - one for init_task, one to ensure it is never freed */
1119 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1120
1121 struct group_info *groups_alloc(int gidsetsize)
1122 {
1123         struct group_info *group_info;
1124         int nblocks;
1125         int i;
1126
1127         nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1128         /* Make sure we always allocate at least one indirect block pointer */
1129         nblocks = nblocks ? : 1;
1130         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1131         if (!group_info)
1132                 return NULL;
1133         group_info->ngroups = gidsetsize;
1134         group_info->nblocks = nblocks;
1135         atomic_set(&group_info->usage, 1);
1136
1137         if (gidsetsize <= NGROUPS_SMALL) {
1138                 group_info->blocks[0] = group_info->small_block;
1139         } else {
1140                 for (i = 0; i < nblocks; i++) {
1141                         gid_t *b;
1142                         b = (void *)__get_free_page(GFP_USER);
1143                         if (!b)
1144                                 goto out_undo_partial_alloc;
1145                         group_info->blocks[i] = b;
1146                 }
1147         }
1148         return group_info;
1149
1150 out_undo_partial_alloc:
1151         while (--i >= 0) {
1152                 free_page((unsigned long)group_info->blocks[i]);
1153         }
1154         kfree(group_info);
1155         return NULL;
1156 }
1157
1158 EXPORT_SYMBOL(groups_alloc);
1159
1160 void groups_free(struct group_info *group_info)
1161 {
1162         if (group_info->blocks[0] != group_info->small_block) {
1163                 int i;
1164                 for (i = 0; i < group_info->nblocks; i++)
1165                         free_page((unsigned long)group_info->blocks[i]);
1166         }
1167         kfree(group_info);
1168 }
1169
1170 EXPORT_SYMBOL(groups_free);
1171
1172 /* export the group_info to a user-space array */
1173 static int groups_to_user(gid_t __user *grouplist,
1174     struct group_info *group_info)
1175 {
1176         int i;
1177         int count = group_info->ngroups;
1178
1179         for (i = 0; i < group_info->nblocks; i++) {
1180                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1181                 int off = i * NGROUPS_PER_BLOCK;
1182                 int len = cp_count * sizeof(*grouplist);
1183
1184                 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1185                         return -EFAULT;
1186
1187                 count -= cp_count;
1188         }
1189         return 0;
1190 }
1191
1192 /* fill a group_info from a user-space array - it must be allocated already */
1193 static int groups_from_user(struct group_info *group_info,
1194     gid_t __user *grouplist)
1195  {
1196         int i;
1197         int count = group_info->ngroups;
1198
1199         for (i = 0; i < group_info->nblocks; i++) {
1200                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1201                 int off = i * NGROUPS_PER_BLOCK;
1202                 int len = cp_count * sizeof(*grouplist);
1203
1204                 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1205                         return -EFAULT;
1206
1207                 count -= cp_count;
1208         }
1209         return 0;
1210 }
1211
1212 /* a simple Shell sort */
1213 static void groups_sort(struct group_info *group_info)
1214 {
1215         int base, max, stride;
1216         int gidsetsize = group_info->ngroups;
1217
1218         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1219                 ; /* nothing */
1220         stride /= 3;
1221
1222         while (stride) {
1223                 max = gidsetsize - stride;
1224                 for (base = 0; base < max; base++) {
1225                         int left = base;
1226                         int right = left + stride;
1227                         gid_t tmp = GROUP_AT(group_info, right);
1228
1229                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1230                                 GROUP_AT(group_info, right) =
1231                                     GROUP_AT(group_info, left);
1232                                 right = left;
1233                                 left -= stride;
1234                         }
1235                         GROUP_AT(group_info, right) = tmp;
1236                 }
1237                 stride /= 3;
1238         }
1239 }
1240
1241 /* a simple bsearch */
1242 static int groups_search(struct group_info *group_info, gid_t grp)
1243 {
1244         int left, right;
1245
1246         if (!group_info)
1247                 return 0;
1248
1249         left = 0;
1250         right = group_info->ngroups;
1251         while (left < right) {
1252                 int mid = (left+right)/2;
1253                 int cmp = grp - GROUP_AT(group_info, mid);
1254                 if (cmp > 0)
1255                         left = mid + 1;
1256                 else if (cmp < 0)
1257                         right = mid;
1258                 else
1259                         return 1;
1260         }
1261         return 0;
1262 }
1263
1264 /* validate and set current->group_info */
1265 int set_current_groups(struct group_info *group_info)
1266 {
1267         int retval;
1268         struct group_info *old_info;
1269
1270         retval = security_task_setgroups(group_info);
1271         if (retval)
1272                 return retval;
1273
1274         groups_sort(group_info);
1275         get_group_info(group_info);
1276
1277         task_lock(current);
1278         old_info = current->group_info;
1279         current->group_info = group_info;
1280         task_unlock(current);
1281
1282         put_group_info(old_info);
1283
1284         return 0;
1285 }
1286
1287 EXPORT_SYMBOL(set_current_groups);
1288
1289 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1290 {
1291         int i = 0;
1292
1293         /*
1294          *      SMP: Nobody else can change our grouplist. Thus we are
1295          *      safe.
1296          */
1297
1298         if (gidsetsize < 0)
1299                 return -EINVAL;
1300
1301         /* no need to grab task_lock here; it cannot change */
1302         get_group_info(current->group_info);
1303         i = current->group_info->ngroups;
1304         if (gidsetsize) {
1305                 if (i > gidsetsize) {
1306                         i = -EINVAL;
1307                         goto out;
1308                 }
1309                 if (groups_to_user(grouplist, current->group_info)) {
1310                         i = -EFAULT;
1311                         goto out;
1312                 }
1313         }
1314 out:
1315         put_group_info(current->group_info);
1316         return i;
1317 }
1318
1319 /*
1320  *      SMP: Our groups are copy-on-write. We can set them safely
1321  *      without another task interfering.
1322  */
1323
1324 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1325 {
1326         struct group_info *group_info;
1327         int retval;
1328
1329         if (!capable(CAP_SETGID))
1330                 return -EPERM;
1331         if ((unsigned)gidsetsize > NGROUPS_MAX)
1332                 return -EINVAL;
1333
1334         group_info = groups_alloc(gidsetsize);
1335         if (!group_info)
1336                 return -ENOMEM;
1337         retval = groups_from_user(group_info, grouplist);
1338         if (retval) {
1339                 put_group_info(group_info);
1340                 return retval;
1341         }
1342
1343         retval = set_current_groups(group_info);
1344         put_group_info(group_info);
1345
1346         return retval;
1347 }
1348
1349 /*
1350  * Check whether we're fsgid/egid or in the supplemental group..
1351  */
1352 int in_group_p(gid_t grp)
1353 {
1354         int retval = 1;
1355         if (grp != current->fsgid) {
1356                 get_group_info(current->group_info);
1357                 retval = groups_search(current->group_info, grp);
1358                 put_group_info(current->group_info);
1359         }
1360         return retval;
1361 }
1362
1363 EXPORT_SYMBOL(in_group_p);
1364
1365 int in_egroup_p(gid_t grp)
1366 {
1367         int retval = 1;
1368         if (grp != current->egid) {
1369                 get_group_info(current->group_info);
1370                 retval = groups_search(current->group_info, grp);
1371                 put_group_info(current->group_info);
1372         }
1373         return retval;
1374 }
1375
1376 EXPORT_SYMBOL(in_egroup_p);
1377
1378 DECLARE_RWSEM(uts_sem);
1379
1380 EXPORT_SYMBOL(uts_sem);
1381
1382 asmlinkage long sys_newuname(struct new_utsname __user * name)
1383 {
1384         int errno = 0;
1385
1386         down_read(&uts_sem);
1387         if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1388                 errno = -EFAULT;
1389         up_read(&uts_sem);
1390         return errno;
1391 }
1392
1393 asmlinkage long sys_sethostname(char __user *name, int len)
1394 {
1395         int errno;
1396         char tmp[__NEW_UTS_LEN];
1397
1398         if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1399                 return -EPERM;
1400         if (len < 0 || len > __NEW_UTS_LEN)
1401                 return -EINVAL;
1402         down_write(&uts_sem);
1403         errno = -EFAULT;
1404         if (!copy_from_user(tmp, name, len)) {
1405                 char *ptr = vx_new_uts(nodename);
1406
1407                 memcpy(ptr, tmp, len);
1408                 ptr[len] = 0;
1409                 errno = 0;
1410         }
1411         up_write(&uts_sem);
1412         return errno;
1413 }
1414
1415 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1416
1417 asmlinkage long sys_gethostname(char __user *name, int len)
1418 {
1419         int i, errno;
1420         char *ptr;
1421
1422         if (len < 0)
1423                 return -EINVAL;
1424         down_read(&uts_sem);
1425         ptr = vx_new_uts(nodename);
1426         i = 1 + strlen(ptr);
1427         if (i > len)
1428                 i = len;
1429         errno = 0;
1430         if (copy_to_user(name, ptr, i))
1431                 errno = -EFAULT;
1432         up_read(&uts_sem);
1433         return errno;
1434 }
1435
1436 #endif
1437
1438 /*
1439  * Only setdomainname; getdomainname can be implemented by calling
1440  * uname()
1441  */
1442 asmlinkage long sys_setdomainname(char __user *name, int len)
1443 {
1444         int errno;
1445         char tmp[__NEW_UTS_LEN];
1446
1447         if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1448                 return -EPERM;
1449         if (len < 0 || len > __NEW_UTS_LEN)
1450                 return -EINVAL;
1451
1452         down_write(&uts_sem);
1453         errno = -EFAULT;
1454         if (!copy_from_user(tmp, name, len)) {
1455                 char *ptr = vx_new_uts(domainname);
1456
1457                 memcpy(ptr, tmp, len);
1458                 ptr[len] = 0;
1459                 errno = 0;
1460         }
1461         up_write(&uts_sem);
1462         return errno;
1463 }
1464
1465 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1466 {
1467         if (resource >= RLIM_NLIMITS)
1468                 return -EINVAL;
1469         else {
1470                 struct rlimit value;
1471                 task_lock(current->group_leader);
1472                 value = current->signal->rlim[resource];
1473                 task_unlock(current->group_leader);
1474                 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1475         }
1476 }
1477
1478 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1479
1480 /*
1481  *      Back compatibility for getrlimit. Needed for some apps.
1482  */
1483
1484 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1485 {
1486         struct rlimit x;
1487         if (resource >= RLIM_NLIMITS)
1488                 return -EINVAL;
1489
1490         task_lock(current->group_leader);
1491         x = current->signal->rlim[resource];
1492         task_unlock(current->group_leader);
1493         if(x.rlim_cur > 0x7FFFFFFF)
1494                 x.rlim_cur = 0x7FFFFFFF;
1495         if(x.rlim_max > 0x7FFFFFFF)
1496                 x.rlim_max = 0x7FFFFFFF;
1497         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1498 }
1499
1500 #endif
1501
1502 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1503 {
1504         struct rlimit new_rlim, *old_rlim;
1505         int retval;
1506
1507         if (resource >= RLIM_NLIMITS)
1508                 return -EINVAL;
1509         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1510                 return -EFAULT;
1511        if (new_rlim.rlim_cur > new_rlim.rlim_max)
1512                return -EINVAL;
1513         old_rlim = current->signal->rlim + resource;
1514         if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1515             !capable(CAP_SYS_RESOURCE) && !vx_ccaps(VXC_SET_RLIMIT))
1516                 return -EPERM;
1517         if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1518                         return -EPERM;
1519
1520         retval = security_task_setrlimit(resource, &new_rlim);
1521         if (retval)
1522                 return retval;
1523
1524         task_lock(current->group_leader);
1525         *old_rlim = new_rlim;
1526         task_unlock(current->group_leader);
1527
1528         if (resource == RLIMIT_CPU && new_rlim.rlim_cur != RLIM_INFINITY &&
1529             (cputime_eq(current->signal->it_prof_expires, cputime_zero) ||
1530              new_rlim.rlim_cur <= cputime_to_secs(
1531                      current->signal->it_prof_expires))) {
1532                 cputime_t cputime = secs_to_cputime(new_rlim.rlim_cur);
1533                 read_lock(&tasklist_lock);
1534                 spin_lock_irq(&current->sighand->siglock);
1535                 set_process_cpu_timer(current, CPUCLOCK_PROF,
1536                                       &cputime, NULL);
1537                 spin_unlock_irq(&current->sighand->siglock);
1538                 read_unlock(&tasklist_lock);
1539         }
1540
1541         return 0;
1542 }
1543
1544 /*
1545  * It would make sense to put struct rusage in the task_struct,
1546  * except that would make the task_struct be *really big*.  After
1547  * task_struct gets moved into malloc'ed memory, it would
1548  * make sense to do this.  It will make moving the rest of the information
1549  * a lot simpler!  (Which we're not doing right now because we're not
1550  * measuring them yet).
1551  *
1552  * This expects to be called with tasklist_lock read-locked or better,
1553  * and the siglock not locked.  It may momentarily take the siglock.
1554  *
1555  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1556  * races with threads incrementing their own counters.  But since word
1557  * reads are atomic, we either get new values or old values and we don't
1558  * care which for the sums.  We always take the siglock to protect reading
1559  * the c* fields from p->signal from races with exit.c updating those
1560  * fields when reaping, so a sample either gets all the additions of a
1561  * given child after it's reaped, or none so this sample is before reaping.
1562  */
1563
1564 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1565 {
1566         struct task_struct *t;
1567         unsigned long flags;
1568         cputime_t utime, stime;
1569
1570         memset((char *) r, 0, sizeof *r);
1571
1572         if (unlikely(!p->signal))
1573                 return;
1574
1575         switch (who) {
1576                 case RUSAGE_CHILDREN:
1577                         spin_lock_irqsave(&p->sighand->siglock, flags);
1578                         utime = p->signal->cutime;
1579                         stime = p->signal->cstime;
1580                         r->ru_nvcsw = p->signal->cnvcsw;
1581                         r->ru_nivcsw = p->signal->cnivcsw;
1582                         r->ru_minflt = p->signal->cmin_flt;
1583                         r->ru_majflt = p->signal->cmaj_flt;
1584                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1585                         cputime_to_timeval(utime, &r->ru_utime);
1586                         cputime_to_timeval(stime, &r->ru_stime);
1587                         break;
1588                 case RUSAGE_SELF:
1589                         spin_lock_irqsave(&p->sighand->siglock, flags);
1590                         utime = stime = cputime_zero;
1591                         goto sum_group;
1592                 case RUSAGE_BOTH:
1593                         spin_lock_irqsave(&p->sighand->siglock, flags);
1594                         utime = p->signal->cutime;
1595                         stime = p->signal->cstime;
1596                         r->ru_nvcsw = p->signal->cnvcsw;
1597                         r->ru_nivcsw = p->signal->cnivcsw;
1598                         r->ru_minflt = p->signal->cmin_flt;
1599                         r->ru_majflt = p->signal->cmaj_flt;
1600                 sum_group:
1601                         utime = cputime_add(utime, p->signal->utime);
1602                         stime = cputime_add(stime, p->signal->stime);
1603                         r->ru_nvcsw += p->signal->nvcsw;
1604                         r->ru_nivcsw += p->signal->nivcsw;
1605                         r->ru_minflt += p->signal->min_flt;
1606                         r->ru_majflt += p->signal->maj_flt;
1607                         t = p;
1608                         do {
1609                                 utime = cputime_add(utime, t->utime);
1610                                 stime = cputime_add(stime, t->stime);
1611                                 r->ru_nvcsw += t->nvcsw;
1612                                 r->ru_nivcsw += t->nivcsw;
1613                                 r->ru_minflt += t->min_flt;
1614                                 r->ru_majflt += t->maj_flt;
1615                                 t = next_thread(t);
1616                         } while (t != p);
1617                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1618                         cputime_to_timeval(utime, &r->ru_utime);
1619                         cputime_to_timeval(stime, &r->ru_stime);
1620                         break;
1621                 default:
1622                         BUG();
1623         }
1624 }
1625
1626 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1627 {
1628         struct rusage r;
1629         read_lock(&tasklist_lock);
1630         k_getrusage(p, who, &r);
1631         read_unlock(&tasklist_lock);
1632         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1633 }
1634
1635 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1636 {
1637         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1638                 return -EINVAL;
1639         return getrusage(current, who, ru);
1640 }
1641
1642 asmlinkage long sys_umask(int mask)
1643 {
1644         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1645         return mask;
1646 }
1647
1648 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1649                           unsigned long arg4, unsigned long arg5)
1650 {
1651         long error;
1652         int sig;
1653
1654         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1655         if (error)
1656                 return error;
1657
1658         switch (option) {
1659                 case PR_SET_PDEATHSIG:
1660                         sig = arg2;
1661                         if (!valid_signal(sig)) {
1662                                 error = -EINVAL;
1663                                 break;
1664                         }
1665                         current->pdeath_signal = sig;
1666                         break;
1667                 case PR_GET_PDEATHSIG:
1668                         error = put_user(current->pdeath_signal, (int __user *)arg2);
1669                         break;
1670                 case PR_GET_DUMPABLE:
1671                         if (current->mm->dumpable)
1672                                 error = 1;
1673                         break;
1674                 case PR_SET_DUMPABLE:
1675                         if (arg2 != 0 && arg2 != 1) {
1676                                 error = -EINVAL;
1677                                 break;
1678                         }
1679                         current->mm->dumpable = arg2;
1680                         break;
1681
1682                 case PR_SET_UNALIGN:
1683                         error = SET_UNALIGN_CTL(current, arg2);
1684                         break;
1685                 case PR_GET_UNALIGN:
1686                         error = GET_UNALIGN_CTL(current, arg2);
1687                         break;
1688                 case PR_SET_FPEMU:
1689                         error = SET_FPEMU_CTL(current, arg2);
1690                         break;
1691                 case PR_GET_FPEMU:
1692                         error = GET_FPEMU_CTL(current, arg2);
1693                         break;
1694                 case PR_SET_FPEXC:
1695                         error = SET_FPEXC_CTL(current, arg2);
1696                         break;
1697                 case PR_GET_FPEXC:
1698                         error = GET_FPEXC_CTL(current, arg2);
1699                         break;
1700                 case PR_GET_TIMING:
1701                         error = PR_TIMING_STATISTICAL;
1702                         break;
1703                 case PR_SET_TIMING:
1704                         if (arg2 == PR_TIMING_STATISTICAL)
1705                                 error = 0;
1706                         else
1707                                 error = -EINVAL;
1708                         break;
1709
1710                 case PR_GET_KEEPCAPS:
1711                         if (current->keep_capabilities)
1712                                 error = 1;
1713                         break;
1714                 case PR_SET_KEEPCAPS:
1715                         if (arg2 != 0 && arg2 != 1) {
1716                                 error = -EINVAL;
1717                                 break;
1718                         }
1719                         current->keep_capabilities = arg2;
1720                         break;
1721                 case PR_SET_NAME: {
1722                         struct task_struct *me = current;
1723                         unsigned char ncomm[sizeof(me->comm)];
1724
1725                         ncomm[sizeof(me->comm)-1] = 0;
1726                         if (strncpy_from_user(ncomm, (char __user *)arg2,
1727                                                 sizeof(me->comm)-1) < 0)
1728                                 return -EFAULT;
1729                         set_task_comm(me, ncomm);
1730                         return 0;
1731                 }
1732                 case PR_GET_NAME: {
1733                         struct task_struct *me = current;
1734                         unsigned char tcomm[sizeof(me->comm)];
1735
1736                         get_task_comm(tcomm, me);
1737                         if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
1738                                 return -EFAULT;
1739                         return 0;
1740                 }
1741                 default:
1742                         error = -EINVAL;
1743                         break;
1744         }
1745         return error;
1746 }