kernel/linux/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/compat.h>
   9 #include <linux/module.h>
  10 #include <linux/mm.h>
  11 #include <linux/utsname.h>
  12 #include <linux/mman.h>
  13 #include <linux/smp_lock.h>
  14 #include <linux/notifier.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/times.h>
  23 #include <linux/security.h>
  24 #include <linux/dcookies.h>
  25 #include <linux/suspend.h>
  26
  27 #include <asm/uaccess.h>
  28 #include <asm/io.h>
  29 #include <asm/unistd.h>
  30
  31 #ifndef SET_UNALIGN_CTL
  32 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  33 #endif
  34 #ifndef GET_UNALIGN_CTL
  35 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  36 #endif
  37 #ifndef SET_FPEMU_CTL
  38 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
  39 #endif
  40 #ifndef GET_FPEMU_CTL
  41 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
  42 #endif
  43 #ifndef SET_FPEXC_CTL
  44 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
  45 #endif
  46 #ifndef GET_FPEXC_CTL
  47 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
  48 #endif
  49
  50 /*
  51  * this is where the system-wide overflow UID and GID are defined, for
  52  * architectures that now have 32-bit UID/GID but didn't in the past
  53  */
  54
  55 int overflowuid = DEFAULT_OVERFLOWUID;
  56 int overflowgid = DEFAULT_OVERFLOWGID;
  57
  58 #ifdef CONFIG_UID16
  59 EXPORT_SYMBOL(overflowuid);
  60 EXPORT_SYMBOL(overflowgid);
  61 #endif
  62
  63 /*
  64  * the same as above, but for filesystems which can only store a 16-bit
  65  * UID and GID. as such, this is needed on all architectures
  66  */
  67
  68 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
  69 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
  70
  71 EXPORT_SYMBOL(fs_overflowuid);
  72 EXPORT_SYMBOL(fs_overflowgid);
  73
  74 /*
  75  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
  76  */
  77
  78 int C_A_D = 1;
  79 int cad_pid = 1;
  80
  81 /*
  82  *      Notifier list for kernel code which wants to be called
  83  *      at shutdown. This is used to stop any idling DMA operations
  84  *      and the like.
  85  */
  86
  87 static struct notifier_block *reboot_notifier_list;
  88 rwlock_t notifier_lock = RW_LOCK_UNLOCKED;
  89
  90 /**
  91  *      notifier_chain_register - Add notifier to a notifier chain
  92  *      @list: Pointer to root list pointer
  93  *      @n: New entry in notifier chain
  94  *
  95  *      Adds a notifier to a notifier chain.
  96  *
  97  *      Currently always returns zero.
  98  */
  99
 100 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
 101 {
 102         write_lock(&notifier_lock);
 103         while(*list)
 104         {
 105                 if(n->priority > (*list)->priority)
 106                         break;
 107                 list= &((*list)->next);
 108         }
 109         n->next = *list;
 110         *list=n;
 111         write_unlock(&notifier_lock);
 112         return 0;
 113 }
 114
 115 EXPORT_SYMBOL(notifier_chain_register);
 116
 117 /**
 118  *      notifier_chain_unregister - Remove notifier from a notifier chain
 119  *      @nl: Pointer to root list pointer
 120  *      @n: New entry in notifier chain
 121  *
 122  *      Removes a notifier from a notifier chain.
 123  *
 124  *      Returns zero on success, or %-ENOENT on failure.
 125  */
 126
 127 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
 128 {
 129         write_lock(&notifier_lock);
 130         while((*nl)!=NULL)
 131         {
 132                 if((*nl)==n)
 133                 {
 134                         *nl=n->next;
 135                         write_unlock(&notifier_lock);
 136                         return 0;
 137                 }
 138                 nl=&((*nl)->next);
 139         }
 140         write_unlock(&notifier_lock);
 141         return -ENOENT;
 142 }
 143
 144 EXPORT_SYMBOL(notifier_chain_unregister);
 145
 146 /**
 147  *      notifier_call_chain - Call functions in a notifier chain
 148  *      @n: Pointer to root pointer of notifier chain
 149  *      @val: Value passed unmodified to notifier function
 150  *      @v: Pointer passed unmodified to notifier function
 151  *
 152  *      Calls each function in a notifier chain in turn.
 153  *
 154  *      If the return value of the notifier can be and'd
 155  *      with %NOTIFY_STOP_MASK, then notifier_call_chain
 156  *      will return immediately, with the return value of
 157  *      the notifier function which halted execution.
 158  *      Otherwise, the return value is the return value
 159  *      of the last notifier function called.
 160  */
 161
 162 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
 163 {
 164         int ret=NOTIFY_DONE;
 165         struct notifier_block *nb = *n;
 166
 167         while(nb)
 168         {
 169                 ret=nb->notifier_call(nb,val,v);
 170                 if(ret&NOTIFY_STOP_MASK)
 171                 {
 172                         return ret;
 173                 }
 174                 nb=nb->next;
 175         }
 176         return ret;
 177 }
 178
 179 EXPORT_SYMBOL(notifier_call_chain);
 180
 181 /**
 182  *      register_reboot_notifier - Register function to be called at reboot time
 183  *      @nb: Info about notifier function to be called
 184  *
 185  *      Registers a function with the list of functions
 186  *      to be called at reboot time.
 187  *
 188  *      Currently always returns zero, as notifier_chain_register
 189  *      always returns zero.
 190  */
 191
 192 int register_reboot_notifier(struct notifier_block * nb)
 193 {
 194         return notifier_chain_register(&reboot_notifier_list, nb);
 195 }
 196
 197 EXPORT_SYMBOL(register_reboot_notifier);
 198
 199 /**
 200  *      unregister_reboot_notifier - Unregister previously registered reboot notifier
 201  *      @nb: Hook to be unregistered
 202  *
 203  *      Unregisters a previously registered reboot
 204  *      notifier function.
 205  *
 206  *      Returns zero on success, or %-ENOENT on failure.
 207  */
 208
 209 int unregister_reboot_notifier(struct notifier_block * nb)
 210 {
 211         return notifier_chain_unregister(&reboot_notifier_list, nb);
 212 }
 213
 214 EXPORT_SYMBOL(unregister_reboot_notifier);
 215
 216 asmlinkage long sys_ni_syscall(void)
 217 {
 218         return -ENOSYS;
 219 }
 220
 221 cond_syscall(sys_nfsservctl)
 222 cond_syscall(sys_quotactl)
 223 cond_syscall(sys_acct)
 224 cond_syscall(sys_lookup_dcookie)
 225 cond_syscall(sys_swapon)
 226 cond_syscall(sys_swapoff)
 227 cond_syscall(sys_init_module)
 228 cond_syscall(sys_delete_module)
 229 cond_syscall(sys_socketpair)
 230 cond_syscall(sys_bind)
 231 cond_syscall(sys_listen)
 232 cond_syscall(sys_accept)
 233 cond_syscall(sys_connect)
 234 cond_syscall(sys_getsockname)
 235 cond_syscall(sys_getpeername)
 236 cond_syscall(sys_sendto)
 237 cond_syscall(sys_send)
 238 cond_syscall(sys_recvfrom)
 239 cond_syscall(sys_recv)
 240 cond_syscall(sys_socket)
 241 cond_syscall(sys_setsockopt)
 242 cond_syscall(sys_getsockopt)
 243 cond_syscall(sys_shutdown)
 244 cond_syscall(sys_sendmsg)
 245 cond_syscall(sys_recvmsg)
 246 cond_syscall(sys_socketcall)
 247 cond_syscall(sys_futex)
 248 cond_syscall(compat_sys_futex)
 249 cond_syscall(sys_epoll_create)
 250 cond_syscall(sys_epoll_ctl)
 251 cond_syscall(sys_epoll_wait)
 252 cond_syscall(sys_semget)
 253 cond_syscall(sys_semop)
 254 cond_syscall(sys_semtimedop)
 255 cond_syscall(sys_semctl)
 256 cond_syscall(sys_msgget)
 257 cond_syscall(sys_msgsnd)
 258 cond_syscall(sys_msgrcv)
 259 cond_syscall(sys_msgctl)
 260 cond_syscall(sys_shmget)
 261 cond_syscall(sys_shmdt)
 262 cond_syscall(sys_shmctl)
 263 cond_syscall(sys_mq_open)
 264 cond_syscall(sys_mq_unlink)
 265 cond_syscall(sys_mq_timedsend)
 266 cond_syscall(sys_mq_timedreceive)
 267 cond_syscall(sys_mq_notify)
 268 cond_syscall(sys_mq_getsetattr)
 269 cond_syscall(compat_sys_mq_open)
 270 cond_syscall(compat_sys_mq_timedsend)
 271 cond_syscall(compat_sys_mq_timedreceive)
 272 cond_syscall(compat_sys_mq_notify)
 273 cond_syscall(compat_sys_mq_getsetattr)
 274 cond_syscall(sys_mbind)
 275 cond_syscall(sys_get_mempolicy)
 276 cond_syscall(sys_set_mempolicy)
 277 cond_syscall(compat_get_mempolicy)
 278
 279 /* arch-specific weak syscall entries */
 280 cond_syscall(sys_pciconfig_read)
 281 cond_syscall(sys_pciconfig_write)
 282 cond_syscall(sys_pciconfig_iobase)
 283
 284 static int set_one_prio(struct task_struct *p, int niceval, int error)
 285 {
 286         int no_nice;
 287
 288         if (p->uid != current->euid &&
 289                 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
 290                 error = -EPERM;
 291                 goto out;
 292         }
 293         if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
 294                 error = -EACCES;
 295                 goto out;
 296         }
 297         no_nice = security_task_setnice(p, niceval);
 298         if (no_nice) {
 299                 error = no_nice;
 300                 goto out;
 301         }
 302         if (error == -ESRCH)
 303                 error = 0;
 304         set_user_nice(p, niceval);
 305 out:
 306         return error;
 307 }
 308
 309 asmlinkage long sys_setpriority(int which, int who, int niceval)
 310 {
 311         struct task_struct *g, *p;
 312         struct user_struct *user;
 313         struct pid *pid;
 314         struct list_head *l;
 315         int error = -EINVAL;
 316
 317         if (which > 2 || which < 0)
 318                 goto out;
 319
 320         /* normalize: avoid signed division (rounding problems) */
 321         error = -ESRCH;
 322         if (niceval < -20)
 323                 niceval = -20;
 324         if (niceval > 19)
 325                 niceval = 19;
 326
 327         read_lock(&tasklist_lock);
 328         switch (which) {
 329                 case PRIO_PROCESS:
 330                         if (!who)
 331                                 who = current->pid;
 332                         p = find_task_by_pid(who);
 333                         if (p)
 334                                 error = set_one_prio(p, niceval, error);
 335                         break;
 336                 case PRIO_PGRP:
 337                         if (!who)
 338                                 who = process_group(current);
 339                         for_each_task_pid(who, PIDTYPE_PGID, p, l, pid)
 340                                 error = set_one_prio(p, niceval, error);
 341                         break;
 342                 case PRIO_USER:
 343                         if (!who)
 344                                 user = current->user;
 345                         else
 346                                 user = find_user(who);
 347
 348                         if (!user)
 349                                 goto out_unlock;
 350
 351                         do_each_thread(g, p)
 352                                 if (p->uid == who)
 353                                         error = set_one_prio(p, niceval, error);
 354                         while_each_thread(g, p);
 355                         if (who)
 356                                 free_uid(user);         /* For find_user() */
 357                         break;
 358         }
 359 out_unlock:
 360         read_unlock(&tasklist_lock);
 361 out:
 362         return error;
 363 }
 364
 365 /*
 366  * Ugh. To avoid negative return values, "getpriority()" will
 367  * not return the normal nice-value, but a negated value that
 368  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 369  * to stay compatible.
 370  */
 371 asmlinkage long sys_getpriority(int which, int who)
 372 {
 373         struct task_struct *g, *p;
 374         struct list_head *l;
 375         struct pid *pid;
 376         struct user_struct *user;
 377         long niceval, retval = -ESRCH;
 378
 379         if (which > 2 || which < 0)
 380                 return -EINVAL;
 381
 382         read_lock(&tasklist_lock);
 383         switch (which) {
 384                 case PRIO_PROCESS:
 385                         if (!who)
 386                                 who = current->pid;
 387                         p = find_task_by_pid(who);
 388                         if (p) {
 389                                 niceval = 20 - task_nice(p);
 390                                 if (niceval > retval)
 391                                         retval = niceval;
 392                         }
 393                         break;
 394                 case PRIO_PGRP:
 395                         if (!who)
 396                                 who = process_group(current);
 397                         for_each_task_pid(who, PIDTYPE_PGID, p, l, pid) {
 398                                 niceval = 20 - task_nice(p);
 399                                 if (niceval > retval)
 400                                         retval = niceval;
 401                         }
 402                         break;
 403                 case PRIO_USER:
 404                         if (!who)
 405                                 user = current->user;
 406                         else
 407                                 user = find_user(who);
 408
 409                         if (!user)
 410                                 goto out_unlock;
 411
 412                         do_each_thread(g, p)
 413                                 if (p->uid == who) {
 414                                         niceval = 20 - task_nice(p);
 415                                         if (niceval > retval)
 416                                                 retval = niceval;
 417                                 }
 418                         while_each_thread(g, p);
 419                         if (who)
 420                                 free_uid(user);         /* for find_user() */
 421                         break;
 422         }
 423 out_unlock:
 424         read_unlock(&tasklist_lock);
 425
 426         return retval;
 427 }
 428
 429
 430 /*
 431  * Reboot system call: for obvious reasons only root may call it,
 432  * and even root needs to set up some magic numbers in the registers
 433  * so that some mistake won't make this reboot the whole machine.
 434  * You can also set the meaning of the ctrl-alt-del-key here.
 435  *
 436  * reboot doesn't sync: do that yourself before calling this.
 437  */
 438 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
 439 {
 440         char buffer[256];
 441
 442         /* We only trust the superuser with rebooting the system. */
 443         if (!capable(CAP_SYS_BOOT))
 444                 return -EPERM;
 445
 446         /* For safety, we require "magic" arguments. */
 447         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 448             (magic2 != LINUX_REBOOT_MAGIC2 &&
 449                         magic2 != LINUX_REBOOT_MAGIC2A &&
 450                         magic2 != LINUX_REBOOT_MAGIC2B &&
 451                         magic2 != LINUX_REBOOT_MAGIC2C))
 452                 return -EINVAL;
 453
 454         lock_kernel();
 455         switch (cmd) {
 456         case LINUX_REBOOT_CMD_RESTART:
 457                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 458                 system_state = SYSTEM_RESTART;
 459                 device_shutdown();
 460                 printk(KERN_EMERG "Restarting system.\n");
 461                 machine_restart(NULL);
 462                 break;
 463
 464         case LINUX_REBOOT_CMD_CAD_ON:
 465                 C_A_D = 1;
 466                 break;
 467
 468         case LINUX_REBOOT_CMD_CAD_OFF:
 469                 C_A_D = 0;
 470                 break;
 471
 472         case LINUX_REBOOT_CMD_HALT:
 473                 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
 474                 system_state = SYSTEM_HALT;
 475                 device_shutdown();
 476                 printk(KERN_EMERG "System halted.\n");
 477                 machine_halt();
 478                 unlock_kernel();
 479                 do_exit(0);
 480                 break;
 481
 482         case LINUX_REBOOT_CMD_POWER_OFF:
 483                 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
 484                 system_state = SYSTEM_POWER_OFF;
 485                 device_shutdown();
 486                 printk(KERN_EMERG "Power down.\n");
 487                 machine_power_off();
 488                 unlock_kernel();
 489                 do_exit(0);
 490                 break;
 491
 492         case LINUX_REBOOT_CMD_RESTART2:
 493                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 494                         unlock_kernel();
 495                         return -EFAULT;
 496                 }
 497                 buffer[sizeof(buffer) - 1] = '\0';
 498
 499                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
 500                 system_state = SYSTEM_RESTART;
 501                 device_shutdown();
 502                 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
 503                 machine_restart(buffer);
 504                 break;
 505
 506 #ifdef CONFIG_SOFTWARE_SUSPEND
 507         case LINUX_REBOOT_CMD_SW_SUSPEND:
 508                 {
 509                         int ret = software_suspend();
 510                         unlock_kernel();
 511                         return ret;
 512                 }
 513 #endif
 514
 515         default:
 516                 unlock_kernel();
 517                 return -EINVAL;
 518         }
 519         unlock_kernel();
 520         return 0;
 521 }
 522
 523 static void deferred_cad(void *dummy)
 524 {
 525         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 526         machine_restart(NULL);
 527 }
 528
 529 /*
 530  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 531  * As it's called within an interrupt, it may NOT sync: the only choice
 532  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 533  */
 534 void ctrl_alt_del(void)
 535 {
 536         static DECLARE_WORK(cad_work, deferred_cad, NULL);
 537
 538         if (C_A_D)
 539                 schedule_work(&cad_work);
 540         else
 541                 kill_proc(cad_pid, SIGINT, 1);
 542 }
 543
 544
 545 /*
 546  * Unprivileged users may change the real gid to the effective gid
 547  * or vice versa.  (BSD-style)
 548  *
 549  * If you set the real gid at all, or set the effective gid to a value not
 550  * equal to the real gid, then the saved gid is set to the new effective gid.
 551  *
 552  * This makes it possible for a setgid program to completely drop its
 553  * privileges, which is often a useful assertion to make when you are doing
 554  * a security audit over a program.
 555  *
 556  * The general idea is that a program which uses just setregid() will be
 557  * 100% compatible with BSD.  A program which uses just setgid() will be
 558  * 100% compatible with POSIX with saved IDs.
 559  *
 560  * SMP: There are not races, the GIDs are checked only by filesystem
 561  *      operations (as far as semantic preservation is concerned).
 562  */
 563 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 564 {
 565         int old_rgid = current->gid;
 566         int old_egid = current->egid;
 567         int new_rgid = old_rgid;
 568         int new_egid = old_egid;
 569         int retval;
 570
 571         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 572         if (retval)
 573                 return retval;
 574
 575         if (rgid != (gid_t) -1) {
 576                 if ((old_rgid == rgid) ||
 577                     (current->egid==rgid) ||
 578                     capable(CAP_SETGID))
 579                         new_rgid = rgid;
 580                 else
 581                         return -EPERM;
 582         }
 583         if (egid != (gid_t) -1) {
 584                 if ((old_rgid == egid) ||
 585                     (current->egid == egid) ||
 586                     (current->sgid == egid) ||
 587                     capable(CAP_SETGID))
 588                         new_egid = egid;
 589                 else {
 590                         return -EPERM;
 591                 }
 592         }
 593         if (new_egid != old_egid)
 594         {
 595                 current->mm->dumpable = 0;
 596                 wmb();
 597         }
 598         if (rgid != (gid_t) -1 ||
 599             (egid != (gid_t) -1 && egid != old_rgid))
 600                 current->sgid = new_egid;
 601         current->fsgid = new_egid;
 602         current->egid = new_egid;
 603         current->gid = new_rgid;
 604         return 0;
 605 }
 606
 607 /*
 608  * setgid() is implemented like SysV w/ SAVED_IDS
 609  *
 610  * SMP: Same implicit races as above.
 611  */
 612 asmlinkage long sys_setgid(gid_t gid)
 613 {
 614         int old_egid = current->egid;
 615         int retval;
 616
 617         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 618         if (retval)
 619                 return retval;
 620
 621         if (capable(CAP_SETGID))
 622         {
 623                 if(old_egid != gid)
 624                 {
 625                         current->mm->dumpable=0;
 626                         wmb();
 627                 }
 628                 current->gid = current->egid = current->sgid = current->fsgid = gid;
 629         }
 630         else if ((gid == current->gid) || (gid == current->sgid))
 631         {
 632                 if(old_egid != gid)
 633                 {
 634                         current->mm->dumpable=0;
 635                         wmb();
 636                 }
 637                 current->egid = current->fsgid = gid;
 638         }
 639         else
 640                 return -EPERM;
 641         return 0;
 642 }
 643
 644 static int set_user(uid_t new_ruid, int dumpclear)
 645 {
 646         struct user_struct *new_user;
 647
 648         new_user = alloc_uid(new_ruid);
 649         if (!new_user)
 650                 return -EAGAIN;
 651
 652         if (atomic_read(&new_user->processes) >=
 653                                 current->rlim[RLIMIT_NPROC].rlim_cur &&
 654                         new_user != &root_user) {
 655                 free_uid(new_user);
 656                 return -EAGAIN;
 657         }
 658
 659         switch_uid(new_user);
 660
 661         if(dumpclear)
 662         {
 663                 current->mm->dumpable = 0;
 664                 wmb();
 665         }
 666         current->uid = new_ruid;
 667         return 0;
 668 }
 669
 670 /*
 671  * Unprivileged users may change the real uid to the effective uid
 672  * or vice versa.  (BSD-style)
 673  *
 674  * If you set the real uid at all, or set the effective uid to a value not
 675  * equal to the real uid, then the saved uid is set to the new effective uid.
 676  *
 677  * This makes it possible for a setuid program to completely drop its
 678  * privileges, which is often a useful assertion to make when you are doing
 679  * a security audit over a program.
 680  *
 681  * The general idea is that a program which uses just setreuid() will be
 682  * 100% compatible with BSD.  A program which uses just setuid() will be
 683  * 100% compatible with POSIX with saved IDs.
 684  */
 685 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 686 {
 687         int old_ruid, old_euid, old_suid, new_ruid, new_euid;
 688         int retval;
 689
 690         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 691         if (retval)
 692                 return retval;
 693
 694         new_ruid = old_ruid = current->uid;
 695         new_euid = old_euid = current->euid;
 696         old_suid = current->suid;
 697
 698         if (ruid != (uid_t) -1) {
 699                 new_ruid = ruid;
 700                 if ((old_ruid != ruid) &&
 701                     (current->euid != ruid) &&
 702                     !capable(CAP_SETUID))
 703                         return -EPERM;
 704         }
 705
 706         if (euid != (uid_t) -1) {
 707                 new_euid = euid;
 708                 if ((old_ruid != euid) &&
 709                     (current->euid != euid) &&
 710                     (current->suid != euid) &&
 711                     !capable(CAP_SETUID))
 712                         return -EPERM;
 713         }
 714
 715         if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
 716                 return -EAGAIN;
 717
 718         if (new_euid != old_euid)
 719         {
 720                 current->mm->dumpable=0;
 721                 wmb();
 722         }
 723         current->fsuid = current->euid = new_euid;
 724         if (ruid != (uid_t) -1 ||
 725             (euid != (uid_t) -1 && euid != old_ruid))
 726                 current->suid = current->euid;
 727         current->fsuid = current->euid;
 728
 729         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
 730 }
 731
 732
 733
 734 /*
 735  * setuid() is implemented like SysV with SAVED_IDS
 736  *
 737  * Note that SAVED_ID's is deficient in that a setuid root program
 738  * like sendmail, for example, cannot set its uid to be a normal
 739  * user and then switch back, because if you're root, setuid() sets
 740  * the saved uid too.  If you don't like this, blame the bright people
 741  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 742  * will allow a root program to temporarily drop privileges and be able to
 743  * regain them by swapping the real and effective uid.
 744  */
 745 asmlinkage long sys_setuid(uid_t uid)
 746 {
 747         int old_euid = current->euid;
 748         int old_ruid, old_suid, new_ruid, new_suid;
 749         int retval;
 750
 751         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 752         if (retval)
 753                 return retval;
 754
 755         old_ruid = new_ruid = current->uid;
 756         old_suid = current->suid;
 757         new_suid = old_suid;
 758
 759         if (capable(CAP_SETUID)) {
 760                 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
 761                         return -EAGAIN;
 762                 new_suid = uid;
 763         } else if ((uid != current->uid) && (uid != new_suid))
 764                 return -EPERM;
 765
 766         if (old_euid != uid)
 767         {
 768                 current->mm->dumpable = 0;
 769                 wmb();
 770         }
 771         current->fsuid = current->euid = uid;
 772         current->suid = new_suid;
 773
 774         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
 775 }
 776
 777
 778 /*
 779  * This function implements a generic ability to update ruid, euid,
 780  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 781  */
 782 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 783 {
 784         int old_ruid = current->uid;
 785         int old_euid = current->euid;
 786         int old_suid = current->suid;
 787         int retval;
 788
 789         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 790         if (retval)
 791                 return retval;
 792
 793         if (!capable(CAP_SETUID)) {
 794                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
 795                     (ruid != current->euid) && (ruid != current->suid))
 796                         return -EPERM;
 797                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
 798                     (euid != current->euid) && (euid != current->suid))
 799                         return -EPERM;
 800                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
 801                     (suid != current->euid) && (suid != current->suid))
 802                         return -EPERM;
 803         }
 804         if (ruid != (uid_t) -1) {
 805                 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
 806                         return -EAGAIN;
 807         }
 808         if (euid != (uid_t) -1) {
 809                 if (euid != current->euid)
 810                 {
 811                         current->mm->dumpable = 0;
 812                         wmb();
 813                 }
 814                 current->euid = euid;
 815         }
 816         current->fsuid = current->euid;
 817         if (suid != (uid_t) -1)
 818                 current->suid = suid;
 819
 820         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
 821 }
 822
 823 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
 824 {
 825         int retval;
 826
 827         if (!(retval = put_user(current->uid, ruid)) &&
 828             !(retval = put_user(current->euid, euid)))
 829                 retval = put_user(current->suid, suid);
 830
 831         return retval;
 832 }
 833
 834 /*
 835  * Same as above, but for rgid, egid, sgid.
 836  */
 837 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 838 {
 839         int retval;
 840
 841         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 842         if (retval)
 843                 return retval;
 844
 845         if (!capable(CAP_SETGID)) {
 846                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
 847                     (rgid != current->egid) && (rgid != current->sgid))
 848                         return -EPERM;
 849                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
 850                     (egid != current->egid) && (egid != current->sgid))
 851                         return -EPERM;
 852                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
 853                     (sgid != current->egid) && (sgid != current->sgid))
 854                         return -EPERM;
 855         }
 856         if (egid != (gid_t) -1) {
 857                 if (egid != current->egid)
 858                 {
 859                         current->mm->dumpable = 0;
 860                         wmb();
 861                 }
 862                 current->egid = egid;
 863         }
 864         current->fsgid = current->egid;
 865         if (rgid != (gid_t) -1)
 866                 current->gid = rgid;
 867         if (sgid != (gid_t) -1)
 868                 current->sgid = sgid;
 869         return 0;
 870 }
 871
 872 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
 873 {
 874         int retval;
 875
 876         if (!(retval = put_user(current->gid, rgid)) &&
 877             !(retval = put_user(current->egid, egid)))
 878                 retval = put_user(current->sgid, sgid);
 879
 880         return retval;
 881 }
 882
 883
 884 /*
 885  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 886  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 887  * whatever uid it wants to). It normally shadows "euid", except when
 888  * explicitly set by setfsuid() or for access..
 889  */
 890 asmlinkage long sys_setfsuid(uid_t uid)
 891 {
 892         int old_fsuid;
 893
 894         old_fsuid = current->fsuid;
 895         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
 896                 return old_fsuid;
 897
 898         if (uid == current->uid || uid == current->euid ||
 899             uid == current->suid || uid == current->fsuid ||
 900             capable(CAP_SETUID))
 901         {
 902                 if (uid != old_fsuid)
 903                 {
 904                         current->mm->dumpable = 0;
 905                         wmb();
 906                 }
 907                 current->fsuid = uid;
 908         }
 909
 910         security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
 911
 912         return old_fsuid;
 913 }
 914
 915 /*
 916  * Samma på svenska..
 917  */
 918 asmlinkage long sys_setfsgid(gid_t gid)
 919 {
 920         int old_fsgid;
 921
 922         old_fsgid = current->fsgid;
 923         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 924                 return old_fsgid;
 925
 926         if (gid == current->gid || gid == current->egid ||
 927             gid == current->sgid || gid == current->fsgid ||
 928             capable(CAP_SETGID))
 929         {
 930                 if (gid != old_fsgid)
 931                 {
 932                         current->mm->dumpable = 0;
 933                         wmb();
 934                 }
 935                 current->fsgid = gid;
 936         }
 937         return old_fsgid;
 938 }
 939
 940 asmlinkage long sys_times(struct tms __user * tbuf)
 941 {
 942         /*
 943          *      In the SMP world we might just be unlucky and have one of
 944          *      the times increment as we use it. Since the value is an
 945          *      atomically safe type this is just fine. Conceptually its
 946          *      as if the syscall took an instant longer to occur.
 947          */
 948         if (tbuf) {
 949                 struct tms tmp;
 950                 tmp.tms_utime = jiffies_to_clock_t(current->utime);
 951                 tmp.tms_stime = jiffies_to_clock_t(current->stime);
 952                 tmp.tms_cutime = jiffies_to_clock_t(current->cutime);
 953                 tmp.tms_cstime = jiffies_to_clock_t(current->cstime);
 954                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
 955                         return -EFAULT;
 956         }
 957         return (long) jiffies_64_to_clock_t(get_jiffies_64());
 958 }
 959
 960 /*
 961  * This needs some heavy checking ...
 962  * I just haven't the stomach for it. I also don't fully
 963  * understand sessions/pgrp etc. Let somebody who does explain it.
 964  *
 965  * OK, I think I have the protection semantics right.... this is really
 966  * only important on a multi-user system anyway, to make sure one user
 967  * can't send a signal to a process owned by another.  -TYT, 12/12/91
 968  *
 969  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
 970  * LBT 04.03.94
 971  */
 972
 973 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
 974 {
 975         struct task_struct *p;
 976         int err = -EINVAL;
 977
 978         if (!pid)
 979                 pid = current->pid;
 980         if (!pgid)
 981                 pgid = pid;
 982         if (pgid < 0)
 983                 return -EINVAL;
 984
 985         /* From this point forward we keep holding onto the tasklist lock
 986          * so that our parent does not change from under us. -DaveM
 987          */
 988         write_lock_irq(&tasklist_lock);
 989
 990         err = -ESRCH;
 991         p = find_task_by_pid(pid);
 992         if (!p)
 993                 goto out;
 994
 995         err = -EINVAL;
 996         if (!thread_group_leader(p))
 997                 goto out;
 998
 999         if (p->parent == current || p->real_parent == current) {
1000                 err = -EPERM;
1001                 if (p->signal->session != current->signal->session)
1002                         goto out;
1003                 err = -EACCES;
1004                 if (p->did_exec)
1005                         goto out;
1006         } else {
1007                 err = -ESRCH;
1008                 if (p != current)
1009                         goto out;
1010         }
1011
1012         err = -EPERM;
1013         if (p->signal->leader)
1014                 goto out;
1015
1016         if (pgid != pid) {
1017                 struct task_struct *p;
1018                 struct pid *pid;
1019                 struct list_head *l;
1020
1021                 for_each_task_pid(pgid, PIDTYPE_PGID, p, l, pid)
1022                         if (p->signal->session == current->signal->session)
1023                                 goto ok_pgid;
1024                 goto out;
1025         }
1026
1027 ok_pgid:
1028         err = security_task_setpgid(p, pgid);
1029         if (err)
1030                 goto out;
1031
1032         if (process_group(p) != pgid) {
1033                 detach_pid(p, PIDTYPE_PGID);
1034                 p->signal->pgrp = pgid;
1035                 attach_pid(p, PIDTYPE_PGID, pgid);
1036         }
1037
1038         err = 0;
1039 out:
1040         /* All paths lead to here, thus we are safe. -DaveM */
1041         write_unlock_irq(&tasklist_lock);
1042         return err;
1043 }
1044
1045 asmlinkage long sys_getpgid(pid_t pid)
1046 {
1047         if (!pid) {
1048                 return process_group(current);
1049         } else {
1050                 int retval;
1051                 struct task_struct *p;
1052
1053                 read_lock(&tasklist_lock);
1054                 p = find_task_by_pid(pid);
1055
1056                 retval = -ESRCH;
1057                 if (p) {
1058                         retval = security_task_getpgid(p);
1059                         if (!retval)
1060                                 retval = process_group(p);
1061                 }
1062                 read_unlock(&tasklist_lock);
1063                 return retval;
1064         }
1065 }
1066
1067 #ifdef __ARCH_WANT_SYS_GETPGRP
1068
1069 asmlinkage long sys_getpgrp(void)
1070 {
1071         /* SMP - assuming writes are word atomic this is fine */
1072         return process_group(current);
1073 }
1074
1075 #endif
1076
1077 asmlinkage long sys_getsid(pid_t pid)
1078 {
1079         if (!pid) {
1080                 return current->signal->session;
1081         } else {
1082                 int retval;
1083                 struct task_struct *p;
1084
1085                 read_lock(&tasklist_lock);
1086                 p = find_task_by_pid(pid);
1087
1088                 retval = -ESRCH;
1089                 if(p) {
1090                         retval = security_task_getsid(p);
1091                         if (!retval)
1092                                 retval = p->signal->session;
1093                 }
1094                 read_unlock(&tasklist_lock);
1095                 return retval;
1096         }
1097 }
1098
1099 asmlinkage long sys_setsid(void)
1100 {
1101         struct pid *pid;
1102         int err = -EPERM;
1103
1104         if (!thread_group_leader(current))
1105                 return -EINVAL;
1106
1107         write_lock_irq(&tasklist_lock);
1108
1109         pid = find_pid(PIDTYPE_PGID, current->pid);
1110         if (pid)
1111                 goto out;
1112
1113         current->signal->leader = 1;
1114         __set_special_pids(current->pid, current->pid);
1115         current->signal->tty = NULL;
1116         current->signal->tty_old_pgrp = 0;
1117         err = process_group(current);
1118 out:
1119         write_unlock_irq(&tasklist_lock);
1120         return err;
1121 }
1122
1123 /*
1124  * Supplementary group IDs
1125  */
1126
1127 /* init to 2 - one for init_task, one to ensure it is never freed */
1128 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1129
1130 struct group_info *groups_alloc(int gidsetsize)
1131 {
1132         struct group_info *group_info;
1133         int nblocks;
1134         int i;
1135
1136         nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1137         /* Make sure we always allocate at least one indirect block pointer */
1138         nblocks = nblocks ? : 1;
1139         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1140         if (!group_info)
1141                 return NULL;
1142         group_info->ngroups = gidsetsize;
1143         group_info->nblocks = nblocks;
1144         atomic_set(&group_info->usage, 1);
1145
1146         if (gidsetsize <= NGROUPS_SMALL) {
1147                 group_info->blocks[0] = group_info->small_block;
1148         } else {
1149                 for (i = 0; i < nblocks; i++) {
1150                         gid_t *b;
1151                         b = (void *)__get_free_page(GFP_USER);
1152                         if (!b)
1153                                 goto out_undo_partial_alloc;
1154                         group_info->blocks[i] = b;
1155                 }
1156         }
1157         return group_info;
1158
1159 out_undo_partial_alloc:
1160         while (--i >= 0) {
1161                 free_page((unsigned long)group_info->blocks[i]);
1162         }
1163         kfree(group_info);
1164         return NULL;
1165 }
1166
1167 EXPORT_SYMBOL(groups_alloc);
1168
1169 void groups_free(struct group_info *group_info)
1170 {
1171         if (group_info->blocks[0] != group_info->small_block) {
1172                 int i;
1173                 for (i = 0; i < group_info->nblocks; i++)
1174                         free_page((unsigned long)group_info->blocks[i]);
1175         }
1176         kfree(group_info);
1177 }
1178
1179 EXPORT_SYMBOL(groups_free);
1180
1181 /* export the group_info to a user-space array */
1182 static int groups_to_user(gid_t __user *grouplist,
1183     struct group_info *group_info)
1184 {
1185         int i;
1186         int count = group_info->ngroups;
1187
1188         for (i = 0; i < group_info->nblocks; i++) {
1189                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1190                 int off = i * NGROUPS_PER_BLOCK;
1191                 int len = cp_count * sizeof(*grouplist);
1192
1193                 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1194                         return -EFAULT;
1195
1196                 count -= cp_count;
1197         }
1198         return 0;
1199 }
1200
1201 /* fill a group_info from a user-space array - it must be allocated already */
1202 static int groups_from_user(struct group_info *group_info,
1203     gid_t __user *grouplist)
1204  {
1205         int i;
1206         int count = group_info->ngroups;
1207
1208         for (i = 0; i < group_info->nblocks; i++) {
1209                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1210                 int off = i * NGROUPS_PER_BLOCK;
1211                 int len = cp_count * sizeof(*grouplist);
1212
1213                 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1214                         return -EFAULT;
1215
1216                 count -= cp_count;
1217         }
1218         return 0;
1219 }
1220
1221 /* a simple shell-metzner sort */
1222 static void groups_sort(struct group_info *group_info)
1223 {
1224         int base, max, stride;
1225         int gidsetsize = group_info->ngroups;
1226
1227         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1228                 ; /* nothing */
1229         stride /= 3;
1230
1231         while (stride) {
1232                 max = gidsetsize - stride;
1233                 for (base = 0; base < max; base++) {
1234                         int left = base;
1235                         int right = left + stride;
1236                         gid_t tmp = GROUP_AT(group_info, right);
1237
1238                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1239                                 GROUP_AT(group_info, right) =
1240                                     GROUP_AT(group_info, left);
1241                                 right = left;
1242                                 left -= stride;
1243                         }
1244                         GROUP_AT(group_info, right) = tmp;
1245                 }
1246                 stride /= 3;
1247         }
1248 }
1249
1250 /* a simple bsearch */
1251 static int groups_search(struct group_info *group_info, gid_t grp)
1252 {
1253         int left, right;
1254
1255         if (!group_info)
1256                 return 0;
1257
1258         left = 0;
1259         right = group_info->ngroups;
1260         while (left < right) {
1261                 int mid = (left+right)/2;
1262                 int cmp = grp - GROUP_AT(group_info, mid);
1263                 if (cmp > 0)
1264                         left = mid + 1;
1265                 else if (cmp < 0)
1266                         right = mid;
1267                 else
1268                         return 1;
1269         }
1270         return 0;
1271 }
1272
1273 /* validate and set current->group_info */
1274 int set_current_groups(struct group_info *group_info)
1275 {
1276         int retval;
1277         struct group_info *old_info;
1278
1279         retval = security_task_setgroups(group_info);
1280         if (retval)
1281                 return retval;
1282
1283         groups_sort(group_info);
1284         get_group_info(group_info);
1285
1286         task_lock(current);
1287         old_info = current->group_info;
1288         current->group_info = group_info;
1289         task_unlock(current);
1290
1291         put_group_info(old_info);
1292
1293         return 0;
1294 }
1295
1296 EXPORT_SYMBOL(set_current_groups);
1297
1298 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1299 {
1300         int i = 0;
1301
1302         /*
1303          *      SMP: Nobody else can change our grouplist. Thus we are
1304          *      safe.
1305          */
1306
1307         if (gidsetsize < 0)
1308                 return -EINVAL;
1309
1310         /* no need to grab task_lock here; it cannot change */
1311         get_group_info(current->group_info);
1312         i = current->group_info->ngroups;
1313         if (gidsetsize) {
1314                 if (i > gidsetsize) {
1315                         i = -EINVAL;
1316                         goto out;
1317                 }
1318                 if (groups_to_user(grouplist, current->group_info)) {
1319                         i = -EFAULT;
1320                         goto out;
1321                 }
1322         }
1323 out:
1324         put_group_info(current->group_info);
1325         return i;
1326 }
1327
1328 /*
1329  *      SMP: Our groups are copy-on-write. We can set them safely
1330  *      without another task interfering.
1331  */
1332
1333 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1334 {
1335         struct group_info *group_info;
1336         int retval;
1337
1338         if (!capable(CAP_SETGID))
1339                 return -EPERM;
1340         if ((unsigned)gidsetsize > NGROUPS_MAX)
1341                 return -EINVAL;
1342
1343         group_info = groups_alloc(gidsetsize);
1344         if (!group_info)
1345                 return -ENOMEM;
1346         retval = groups_from_user(group_info, grouplist);
1347         if (retval) {
1348                 put_group_info(group_info);
1349                 return retval;
1350         }
1351
1352         retval = set_current_groups(group_info);
1353         put_group_info(group_info);
1354
1355         return retval;
1356 }
1357
1358 /*
1359  * Check whether we're fsgid/egid or in the supplemental group..
1360  */
1361 int in_group_p(gid_t grp)
1362 {
1363         int retval = 1;
1364         if (grp != current->fsgid) {
1365                 get_group_info(current->group_info);
1366                 retval = groups_search(current->group_info, grp);
1367                 put_group_info(current->group_info);
1368         }
1369         return retval;
1370 }
1371
1372 EXPORT_SYMBOL(in_group_p);
1373
1374 int in_egroup_p(gid_t grp)
1375 {
1376         int retval = 1;
1377         if (grp != current->egid) {
1378                 get_group_info(current->group_info);
1379                 retval = groups_search(current->group_info, grp);
1380                 put_group_info(current->group_info);
1381         }
1382         return retval;
1383 }
1384
1385 EXPORT_SYMBOL(in_egroup_p);
1386
1387 DECLARE_RWSEM(uts_sem);
1388
1389 EXPORT_SYMBOL(uts_sem);
1390
1391 asmlinkage long sys_newuname(struct new_utsname __user * name)
1392 {
1393         int errno = 0;
1394
1395         down_read(&uts_sem);
1396         if (copy_to_user(name,&system_utsname,sizeof *name))
1397                 errno = -EFAULT;
1398         up_read(&uts_sem);
1399         return errno;
1400 }
1401
1402 asmlinkage long sys_sethostname(char __user *name, int len)
1403 {
1404         int errno;
1405         char tmp[__NEW_UTS_LEN];
1406
1407         if (!capable(CAP_SYS_ADMIN))
1408                 return -EPERM;
1409         if (len < 0 || len > __NEW_UTS_LEN)
1410                 return -EINVAL;
1411         down_write(&uts_sem);
1412         errno = -EFAULT;
1413         if (!copy_from_user(tmp, name, len)) {
1414                 memcpy(system_utsname.nodename, tmp, len);
1415                 system_utsname.nodename[len] = 0;
1416                 errno = 0;
1417         }
1418         up_write(&uts_sem);
1419         return errno;
1420 }
1421
1422 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1423
1424 asmlinkage long sys_gethostname(char __user *name, int len)
1425 {
1426         int i, errno;
1427
1428         if (len < 0)
1429                 return -EINVAL;
1430         down_read(&uts_sem);
1431         i = 1 + strlen(system_utsname.nodename);
1432         if (i > len)
1433                 i = len;
1434         errno = 0;
1435         if (copy_to_user(name, system_utsname.nodename, i))
1436                 errno = -EFAULT;
1437         up_read(&uts_sem);
1438         return errno;
1439 }
1440
1441 #endif
1442
1443 /*
1444  * Only setdomainname; getdomainname can be implemented by calling
1445  * uname()
1446  */
1447 asmlinkage long sys_setdomainname(char __user *name, int len)
1448 {
1449         int errno;
1450         char tmp[__NEW_UTS_LEN];
1451
1452         if (!capable(CAP_SYS_ADMIN))
1453                 return -EPERM;
1454         if (len < 0 || len > __NEW_UTS_LEN)
1455                 return -EINVAL;
1456
1457         down_write(&uts_sem);
1458         errno = -EFAULT;
1459         if (!copy_from_user(tmp, name, len)) {
1460                 memcpy(system_utsname.domainname, tmp, len);
1461                 system_utsname.domainname[len] = 0;
1462                 errno = 0;
1463         }
1464         up_write(&uts_sem);
1465         return errno;
1466 }
1467
1468 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1469 {
1470         if (resource >= RLIM_NLIMITS)
1471                 return -EINVAL;
1472         else
1473                 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
1474                         ? -EFAULT : 0;
1475 }
1476
1477 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1478
1479 /*
1480  *      Back compatibility for getrlimit. Needed for some apps.
1481  */
1482
1483 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1484 {
1485         struct rlimit x;
1486         if (resource >= RLIM_NLIMITS)
1487                 return -EINVAL;
1488
1489         memcpy(&x, current->rlim + resource, sizeof(*rlim));
1490         if(x.rlim_cur > 0x7FFFFFFF)
1491                 x.rlim_cur = 0x7FFFFFFF;
1492         if(x.rlim_max > 0x7FFFFFFF)
1493                 x.rlim_max = 0x7FFFFFFF;
1494         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1495 }
1496
1497 #endif
1498
1499 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1500 {
1501         struct rlimit new_rlim, *old_rlim;
1502         int retval;
1503
1504         if (resource >= RLIM_NLIMITS)
1505                 return -EINVAL;
1506         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1507                 return -EFAULT;
1508        if (new_rlim.rlim_cur > new_rlim.rlim_max)
1509                return -EINVAL;
1510         old_rlim = current->rlim + resource;
1511         if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
1512              (new_rlim.rlim_max > old_rlim->rlim_max)) &&
1513             !capable(CAP_SYS_RESOURCE))
1514                 return -EPERM;
1515         if (resource == RLIMIT_NOFILE) {
1516                 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
1517                         return -EPERM;
1518         }
1519
1520         retval = security_task_setrlimit(resource, &new_rlim);
1521         if (retval)
1522                 return retval;
1523
1524         *old_rlim = new_rlim;
1525         return 0;
1526 }
1527
1528 /*
1529  * It would make sense to put struct rusage in the task_struct,
1530  * except that would make the task_struct be *really big*.  After
1531  * task_struct gets moved into malloc'ed memory, it would
1532  * make sense to do this.  It will make moving the rest of the information
1533  * a lot simpler!  (Which we're not doing right now because we're not
1534  * measuring them yet).
1535  *
1536  * This is SMP safe.  Either we are called from sys_getrusage on ourselves
1537  * below (we know we aren't going to exit/disappear and only we change our
1538  * rusage counters), or we are called from wait4() on a process which is
1539  * either stopped or zombied.  In the zombied case the task won't get
1540  * reaped till shortly after the call to getrusage(), in both cases the
1541  * task being examined is in a frozen state so the counters won't change.
1542  */
1543 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1544 {
1545         struct rusage r;
1546
1547         memset((char *) &r, 0, sizeof(r));
1548         switch (who) {
1549                 case RUSAGE_SELF:
1550                         jiffies_to_timeval(p->utime, &r.ru_utime);
1551                         jiffies_to_timeval(p->stime, &r.ru_stime);
1552                         r.ru_nvcsw = p->nvcsw;
1553                         r.ru_nivcsw = p->nivcsw;
1554                         r.ru_minflt = p->min_flt;
1555                         r.ru_majflt = p->maj_flt;
1556                         break;
1557                 case RUSAGE_CHILDREN:
1558                         jiffies_to_timeval(p->cutime, &r.ru_utime);
1559                         jiffies_to_timeval(p->cstime, &r.ru_stime);
1560                         r.ru_nvcsw = p->cnvcsw;
1561                         r.ru_nivcsw = p->cnivcsw;
1562                         r.ru_minflt = p->cmin_flt;
1563                         r.ru_majflt = p->cmaj_flt;
1564                         break;
1565                 default:
1566                         jiffies_to_timeval(p->utime + p->cutime, &r.ru_utime);
1567                         jiffies_to_timeval(p->stime + p->cstime, &r.ru_stime);
1568                         r.ru_nvcsw = p->nvcsw + p->cnvcsw;
1569                         r.ru_nivcsw = p->nivcsw + p->cnivcsw;
1570                         r.ru_minflt = p->min_flt + p->cmin_flt;
1571                         r.ru_majflt = p->maj_flt + p->cmaj_flt;
1572                         break;
1573         }
1574         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1575 }
1576
1577 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1578 {
1579         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1580                 return -EINVAL;
1581         return getrusage(current, who, ru);
1582 }
1583
1584 asmlinkage long sys_umask(int mask)
1585 {
1586         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1587         return mask;
1588 }
1589
1590 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1591                           unsigned long arg4, unsigned long arg5)
1592 {
1593         int error;
1594         int sig;
1595
1596         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1597         if (error)
1598                 return error;
1599
1600         switch (option) {
1601                 case PR_SET_PDEATHSIG:
1602                         sig = arg2;
1603                         if (sig < 0 || sig > _NSIG) {
1604                                 error = -EINVAL;
1605                                 break;
1606                         }
1607                         current->pdeath_signal = sig;
1608                         break;
1609                 case PR_GET_PDEATHSIG:
1610                         error = put_user(current->pdeath_signal, (int __user *)arg2);
1611                         break;
1612                 case PR_GET_DUMPABLE:
1613                         if (current->mm->dumpable)
1614                                 error = 1;
1615                         break;
1616                 case PR_SET_DUMPABLE:
1617                         if (arg2 != 0 && arg2 != 1) {
1618                                 error = -EINVAL;
1619                                 break;
1620                         }
1621                         current->mm->dumpable = arg2;
1622                         break;
1623
1624                 case PR_SET_UNALIGN:
1625                         error = SET_UNALIGN_CTL(current, arg2);
1626                         break;
1627                 case PR_GET_UNALIGN:
1628                         error = GET_UNALIGN_CTL(current, arg2);
1629                         break;
1630                 case PR_SET_FPEMU:
1631                         error = SET_FPEMU_CTL(current, arg2);
1632                         break;
1633                 case PR_GET_FPEMU:
1634                         error = GET_FPEMU_CTL(current, arg2);
1635                         break;
1636                 case PR_SET_FPEXC:
1637                         error = SET_FPEXC_CTL(current, arg2);
1638                         break;
1639                 case PR_GET_FPEXC:
1640                         error = GET_FPEXC_CTL(current, arg2);
1641                         break;
1642                 case PR_GET_TIMING:
1643                         error = PR_TIMING_STATISTICAL;
1644                         break;
1645                 case PR_SET_TIMING:
1646                         if (arg2 == PR_TIMING_STATISTICAL)
1647                                 error = 0;
1648                         else
1649                                 error = -EINVAL;
1650                         break;
1651
1652                 case PR_GET_KEEPCAPS:
1653                         if (current->keep_capabilities)
1654                                 error = 1;
1655                         break;
1656                 case PR_SET_KEEPCAPS:
1657                         if (arg2 != 0 && arg2 != 1) {
1658                                 error = -EINVAL;
1659                                 break;
1660                         }
1661                         current->keep_capabilities = arg2;
1662                         break;
1663                 default:
1664                         error = -EINVAL;
1665                         break;
1666         }
1667         return error;
1668 }