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/kmod.h>
  16 #include <linux/reboot.h>
  17 #include <linux/prctl.h>
  18 #include <linux/init.h>
  19 #include <linux/highuid.h>
  20 #include <linux/fs.h>
  21 #include <linux/kernel.h>
  22 #include <linux/kexec.h>
  23 #include <linux/workqueue.h>
  24 #include <linux/device.h>
  25 #include <linux/times.h>
  26 #include <linux/security.h>
  27 #include <linux/dcookies.h>
  28 #include <linux/suspend.h>
  29 #include <linux/ckrm.h>
  30 #include <linux/vs_base.h>
  31 #include <linux/vs_cvirt.h>
  32
  33 #include <asm/uaccess.h>
  34 #include <asm/io.h>
  35 #include <asm/unistd.h>
  36
  37 #ifndef SET_UNALIGN_CTL
  38 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  39 #endif
  40 #ifndef GET_UNALIGN_CTL
  41 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  42 #endif
  43 #ifndef SET_FPEMU_CTL
  44 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
  45 #endif
  46 #ifndef GET_FPEMU_CTL
  47 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
  48 #endif
  49 #ifndef SET_FPEXC_CTL
  50 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
  51 #endif
  52 #ifndef GET_FPEXC_CTL
  53 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
  54 #endif
  55
  56 /*
  57  * this is where the system-wide overflow UID and GID are defined, for
  58  * architectures that now have 32-bit UID/GID but didn't in the past
  59  */
  60
  61 int overflowuid = DEFAULT_OVERFLOWUID;
  62 int overflowgid = DEFAULT_OVERFLOWGID;
  63
  64 #ifdef CONFIG_UID16
  65 EXPORT_SYMBOL(overflowuid);
  66 EXPORT_SYMBOL(overflowgid);
  67 #endif
  68
  69 /*
  70  * the same as above, but for filesystems which can only store a 16-bit
  71  * UID and GID. as such, this is needed on all architectures
  72  */
  73
  74 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
  75 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
  76
  77 EXPORT_SYMBOL(fs_overflowuid);
  78 EXPORT_SYMBOL(fs_overflowgid);
  79
  80 /*
  81  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
  82  */
  83
  84 int C_A_D = 1;
  85 int cad_pid = 1;
  86
  87 /*
  88  *      Notifier list for kernel code which wants to be called
  89  *      at shutdown. This is used to stop any idling DMA operations
  90  *      and the like.
  91  */
  92
  93 static struct notifier_block *reboot_notifier_list;
  94 rwlock_t notifier_lock = RW_LOCK_UNLOCKED;
  95
  96 /**
  97  *      notifier_chain_register - Add notifier to a notifier chain
  98  *      @list: Pointer to root list pointer
  99  *      @n: New entry in notifier chain
 100  *
 101  *      Adds a notifier to a notifier chain.
 102  *
 103  *      Currently always returns zero.
 104  */
 105
 106 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
 107 {
 108         write_lock(&notifier_lock);
 109         while(*list)
 110         {
 111                 if(n->priority > (*list)->priority)
 112                         break;
 113                 list= &((*list)->next);
 114         }
 115         n->next = *list;
 116         *list=n;
 117         write_unlock(&notifier_lock);
 118         return 0;
 119 }
 120
 121 EXPORT_SYMBOL(notifier_chain_register);
 122
 123 /**
 124  *      notifier_chain_unregister - Remove notifier from a notifier chain
 125  *      @nl: Pointer to root list pointer
 126  *      @n: New entry in notifier chain
 127  *
 128  *      Removes a notifier from a notifier chain.
 129  *
 130  *      Returns zero on success, or %-ENOENT on failure.
 131  */
 132
 133 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
 134 {
 135         write_lock(&notifier_lock);
 136         while((*nl)!=NULL)
 137         {
 138                 if((*nl)==n)
 139                 {
 140                         *nl=n->next;
 141                         write_unlock(&notifier_lock);
 142                         return 0;
 143                 }
 144                 nl=&((*nl)->next);
 145         }
 146         write_unlock(&notifier_lock);
 147         return -ENOENT;
 148 }
 149
 150 EXPORT_SYMBOL(notifier_chain_unregister);
 151
 152 /**
 153  *      notifier_call_chain - Call functions in a notifier chain
 154  *      @n: Pointer to root pointer of notifier chain
 155  *      @val: Value passed unmodified to notifier function
 156  *      @v: Pointer passed unmodified to notifier function
 157  *
 158  *      Calls each function in a notifier chain in turn.
 159  *
 160  *      If the return value of the notifier can be and'd
 161  *      with %NOTIFY_STOP_MASK, then notifier_call_chain
 162  *      will return immediately, with the return value of
 163  *      the notifier function which halted execution.
 164  *      Otherwise, the return value is the return value
 165  *      of the last notifier function called.
 166  */
 167
 168 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
 169 {
 170         int ret=NOTIFY_DONE;
 171         struct notifier_block *nb = *n;
 172
 173         while(nb)
 174         {
 175                 ret=nb->notifier_call(nb,val,v);
 176                 if(ret&NOTIFY_STOP_MASK)
 177                 {
 178                         return ret;
 179                 }
 180                 nb=nb->next;
 181         }
 182         return ret;
 183 }
 184
 185 EXPORT_SYMBOL(notifier_call_chain);
 186
 187 /**
 188  *      register_reboot_notifier - Register function to be called at reboot time
 189  *      @nb: Info about notifier function to be called
 190  *
 191  *      Registers a function with the list of functions
 192  *      to be called at reboot time.
 193  *
 194  *      Currently always returns zero, as notifier_chain_register
 195  *      always returns zero.
 196  */
 197
 198 int register_reboot_notifier(struct notifier_block * nb)
 199 {
 200         return notifier_chain_register(&reboot_notifier_list, nb);
 201 }
 202
 203 EXPORT_SYMBOL(register_reboot_notifier);
 204
 205 /**
 206  *      unregister_reboot_notifier - Unregister previously registered reboot notifier
 207  *      @nb: Hook to be unregistered
 208  *
 209  *      Unregisters a previously registered reboot
 210  *      notifier function.
 211  *
 212  *      Returns zero on success, or %-ENOENT on failure.
 213  */
 214
 215 int unregister_reboot_notifier(struct notifier_block * nb)
 216 {
 217         return notifier_chain_unregister(&reboot_notifier_list, nb);
 218 }
 219
 220 EXPORT_SYMBOL(unregister_reboot_notifier);
 221
 222 asmlinkage long sys_ni_syscall(void)
 223 {
 224         return -ENOSYS;
 225 }
 226
 227 cond_syscall(sys_nfsservctl)
 228 cond_syscall(sys_quotactl)
 229 cond_syscall(sys_acct)
 230 cond_syscall(sys_lookup_dcookie)
 231 cond_syscall(sys_swapon)
 232 cond_syscall(sys_swapoff)
 233 cond_syscall(sys_init_module)
 234 cond_syscall(sys_delete_module)
 235 cond_syscall(sys_socketpair)
 236 cond_syscall(sys_bind)
 237 cond_syscall(sys_listen)
 238 cond_syscall(sys_accept)
 239 cond_syscall(sys_connect)
 240 cond_syscall(sys_getsockname)
 241 cond_syscall(sys_getpeername)
 242 cond_syscall(sys_sendto)
 243 cond_syscall(sys_send)
 244 cond_syscall(sys_recvfrom)
 245 cond_syscall(sys_recv)
 246 cond_syscall(sys_socket)
 247 cond_syscall(sys_setsockopt)
 248 cond_syscall(sys_getsockopt)
 249 cond_syscall(sys_shutdown)
 250 cond_syscall(sys_sendmsg)
 251 cond_syscall(sys_recvmsg)
 252 cond_syscall(sys_socketcall)
 253 cond_syscall(sys_futex)
 254 cond_syscall(compat_sys_futex)
 255 cond_syscall(sys_epoll_create)
 256 cond_syscall(sys_epoll_ctl)
 257 cond_syscall(sys_epoll_wait)
 258 cond_syscall(sys_semget)
 259 cond_syscall(sys_semop)
 260 cond_syscall(sys_semtimedop)
 261 cond_syscall(sys_semctl)
 262 cond_syscall(sys_msgget)
 263 cond_syscall(sys_msgsnd)
 264 cond_syscall(sys_msgrcv)
 265 cond_syscall(sys_msgctl)
 266 cond_syscall(sys_shmget)
 267 cond_syscall(sys_shmdt)
 268 cond_syscall(sys_shmctl)
 269 cond_syscall(sys_mq_open)
 270 cond_syscall(sys_mq_unlink)
 271 cond_syscall(sys_mq_timedsend)
 272 cond_syscall(sys_mq_timedreceive)
 273 cond_syscall(sys_mq_notify)
 274 cond_syscall(sys_mq_getsetattr)
 275 cond_syscall(compat_sys_mq_open)
 276 cond_syscall(compat_sys_mq_timedsend)
 277 cond_syscall(compat_sys_mq_timedreceive)
 278 cond_syscall(compat_sys_mq_notify)
 279 cond_syscall(compat_sys_mq_getsetattr)
 280 cond_syscall(sys_mbind)
 281 cond_syscall(sys_get_mempolicy)
 282 cond_syscall(sys_set_mempolicy)
 283 cond_syscall(compat_get_mempolicy)
 284
 285 /* arch-specific weak syscall entries */
 286 cond_syscall(sys_pciconfig_read)
 287 cond_syscall(sys_pciconfig_write)
 288 cond_syscall(sys_pciconfig_iobase)
 289
 290 static int set_one_prio(struct task_struct *p, int niceval, int error)
 291 {
 292         int no_nice;
 293
 294         if (p->uid != current->euid &&
 295                 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
 296                 error = -EPERM;
 297                 goto out;
 298         }
 299         if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
 300                 error = -EACCES;
 301                 goto out;
 302         }
 303         no_nice = security_task_setnice(p, niceval);
 304         if (no_nice) {
 305                 error = no_nice;
 306                 goto out;
 307         }
 308         if (error == -ESRCH)
 309                 error = 0;
 310         set_user_nice(p, niceval);
 311 out:
 312         return error;
 313 }
 314
 315 asmlinkage long sys_setpriority(int which, int who, int niceval)
 316 {
 317         struct task_struct *g, *p;
 318         struct user_struct *user;
 319         struct pid *pid;
 320         struct list_head *l;
 321         int error = -EINVAL;
 322
 323         if (which > 2 || which < 0)
 324                 goto out;
 325
 326         /* normalize: avoid signed division (rounding problems) */
 327         error = -ESRCH;
 328         if (niceval < -20)
 329                 niceval = -20;
 330         if (niceval > 19)
 331                 niceval = 19;
 332
 333         read_lock(&tasklist_lock);
 334         switch (which) {
 335                 case PRIO_PROCESS:
 336                         if (!who)
 337                                 who = current->pid;
 338                         p = find_task_by_pid(who);
 339                         if (p)
 340                                 error = set_one_prio(p, niceval, error);
 341                         break;
 342                 case PRIO_PGRP:
 343                         if (!who)
 344                                 who = process_group(current);
 345                         for_each_task_pid(who, PIDTYPE_PGID, p, l, pid)
 346                                 error = set_one_prio(p, niceval, error);
 347                         break;
 348                 case PRIO_USER:
 349                         if (!who)
 350                                 user = current->user;
 351                         else
 352                                 user = find_user(vx_current_xid(), who);
 353
 354                         if (!user)
 355                                 goto out_unlock;
 356
 357                         do_each_thread(g, p)
 358                                 if (p->uid == who)
 359                                         error = set_one_prio(p, niceval, error);
 360                         while_each_thread(g, p);
 361                         if (who)
 362                                 free_uid(user);         /* For find_user() */
 363                         break;
 364         }
 365 out_unlock:
 366         read_unlock(&tasklist_lock);
 367 out:
 368         return error;
 369 }
 370
 371 /*
 372  * Ugh. To avoid negative return values, "getpriority()" will
 373  * not return the normal nice-value, but a negated value that
 374  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 375  * to stay compatible.
 376  */
 377 asmlinkage long sys_getpriority(int which, int who)
 378 {
 379         struct task_struct *g, *p;
 380         struct list_head *l;
 381         struct pid *pid;
 382         struct user_struct *user;
 383         long niceval, retval = -ESRCH;
 384
 385         if (which > 2 || which < 0)
 386                 return -EINVAL;
 387
 388         read_lock(&tasklist_lock);
 389         switch (which) {
 390                 case PRIO_PROCESS:
 391                         if (!who)
 392                                 who = current->pid;
 393                         p = find_task_by_pid(who);
 394                         if (p) {
 395                                 niceval = 20 - task_nice(p);
 396                                 if (niceval > retval)
 397                                         retval = niceval;
 398                         }
 399                         break;
 400                 case PRIO_PGRP:
 401                         if (!who)
 402                                 who = process_group(current);
 403                         for_each_task_pid(who, PIDTYPE_PGID, p, l, pid) {
 404                                 niceval = 20 - task_nice(p);
 405                                 if (niceval > retval)
 406                                         retval = niceval;
 407                         }
 408                         break;
 409                 case PRIO_USER:
 410                         if (!who)
 411                                 user = current->user;
 412                         else
 413                                 user = find_user(vx_current_xid(), who);
 414
 415                         if (!user)
 416                                 goto out_unlock;
 417
 418                         do_each_thread(g, p)
 419                                 if (p->uid == who) {
 420                                         niceval = 20 - task_nice(p);
 421                                         if (niceval > retval)
 422                                                 retval = niceval;
 423                                 }
 424                         while_each_thread(g, p);
 425                         if (who)
 426                                 free_uid(user);         /* for find_user() */
 427                         break;
 428         }
 429 out_unlock:
 430         read_unlock(&tasklist_lock);
 431
 432         return retval;
 433 }
 434
 435 long vs_reboot(unsigned int, void *);
 436
 437 /*
 438  * Reboot system call: for obvious reasons only root may call it,
 439  * and even root needs to set up some magic numbers in the registers
 440  * so that some mistake won't make this reboot the whole machine.
 441  * You can also set the meaning of the ctrl-alt-del-key here.
 442  *
 443  * reboot doesn't sync: do that yourself before calling this.
 444  */
 445 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
 446 {
 447         char buffer[256];
 448
 449         /* We only trust the superuser with rebooting the system. */
 450         if (!capable(CAP_SYS_BOOT))
 451                 return -EPERM;
 452
 453         /* For safety, we require "magic" arguments. */
 454         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 455             (magic2 != LINUX_REBOOT_MAGIC2 &&
 456                         magic2 != LINUX_REBOOT_MAGIC2A &&
 457                         magic2 != LINUX_REBOOT_MAGIC2B &&
 458                         magic2 != LINUX_REBOOT_MAGIC2C))
 459                 return -EINVAL;
 460
 461         if (!vx_check(0, VX_ADMIN|VX_WATCH))
 462                 return vs_reboot(cmd, arg);
 463
 464         lock_kernel();
 465         switch (cmd) {
 466         case LINUX_REBOOT_CMD_RESTART:
 467                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 468                 system_state = SYSTEM_RESTART;
 469                 device_shutdown();
 470                 printk(KERN_EMERG "Restarting system.\n");
 471                 machine_restart(NULL);
 472                 break;
 473
 474         case LINUX_REBOOT_CMD_CAD_ON:
 475                 C_A_D = 1;
 476                 break;
 477
 478         case LINUX_REBOOT_CMD_CAD_OFF:
 479                 C_A_D = 0;
 480                 break;
 481
 482         case LINUX_REBOOT_CMD_HALT:
 483                 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
 484                 system_state = SYSTEM_HALT;
 485                 device_shutdown();
 486                 printk(KERN_EMERG "System halted.\n");
 487                 machine_halt();
 488                 unlock_kernel();
 489                 do_exit(0);
 490                 break;
 491
 492         case LINUX_REBOOT_CMD_POWER_OFF:
 493                 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
 494                 system_state = SYSTEM_POWER_OFF;
 495                 device_shutdown();
 496                 printk(KERN_EMERG "Power down.\n");
 497                 machine_power_off();
 498                 unlock_kernel();
 499                 do_exit(0);
 500                 break;
 501
 502         case LINUX_REBOOT_CMD_RESTART2:
 503                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 504                         unlock_kernel();
 505                         return -EFAULT;
 506                 }
 507                 buffer[sizeof(buffer) - 1] = '\0';
 508
 509                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
 510                 system_state = SYSTEM_RESTART;
 511                 device_shutdown();
 512                 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
 513                 machine_restart(buffer);
 514                 break;
 515
 516 #ifdef CONFIG_KEXEC
 517         case LINUX_REBOOT_CMD_KEXEC:
 518         {
 519                 struct kimage *image;
 520                 image = xchg(&kexec_image, 0);
 521                 if (!image) {
 522                         unlock_kernel();
 523                         return -EINVAL;
 524                 }
 525                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 526                 system_state = SYSTEM_RESTART;
 527                 device_shutdown();
 528                 system_state = SYSTEM_BOOTING;
 529                 printk(KERN_EMERG "Starting new kernel\n");
 530                 machine_shutdown();
 531                 machine_kexec(image);
 532                 break;
 533         }
 534 #endif
 535 #ifdef CONFIG_SOFTWARE_SUSPEND
 536         case LINUX_REBOOT_CMD_SW_SUSPEND:
 537                 {
 538                         int ret = software_suspend();
 539                         unlock_kernel();
 540                         return ret;
 541                 }
 542 #endif
 543
 544         default:
 545                 unlock_kernel();
 546                 return -EINVAL;
 547         }
 548         unlock_kernel();
 549         return 0;
 550 }
 551
 552 static void deferred_cad(void *dummy)
 553 {
 554         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 555         machine_restart(NULL);
 556 }
 557
 558 /*
 559  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 560  * As it's called within an interrupt, it may NOT sync: the only choice
 561  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 562  */
 563 void ctrl_alt_del(void)
 564 {
 565         static DECLARE_WORK(cad_work, deferred_cad, NULL);
 566
 567         if (C_A_D)
 568                 schedule_work(&cad_work);
 569         else
 570                 kill_proc(cad_pid, SIGINT, 1);
 571 }
 572
 573
 574 /*
 575  * Unprivileged users may change the real gid to the effective gid
 576  * or vice versa.  (BSD-style)
 577  *
 578  * If you set the real gid at all, or set the effective gid to a value not
 579  * equal to the real gid, then the saved gid is set to the new effective gid.
 580  *
 581  * This makes it possible for a setgid program to completely drop its
 582  * privileges, which is often a useful assertion to make when you are doing
 583  * a security audit over a program.
 584  *
 585  * The general idea is that a program which uses just setregid() will be
 586  * 100% compatible with BSD.  A program which uses just setgid() will be
 587  * 100% compatible with POSIX with saved IDs.
 588  *
 589  * SMP: There are not races, the GIDs are checked only by filesystem
 590  *      operations (as far as semantic preservation is concerned).
 591  */
 592 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 593 {
 594         int old_rgid = current->gid;
 595         int old_egid = current->egid;
 596         int new_rgid = old_rgid;
 597         int new_egid = old_egid;
 598         int retval;
 599
 600         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 601         if (retval)
 602                 return retval;
 603
 604         if (rgid != (gid_t) -1) {
 605                 if ((old_rgid == rgid) ||
 606                     (current->egid==rgid) ||
 607                     capable(CAP_SETGID))
 608                         new_rgid = rgid;
 609                 else
 610                         return -EPERM;
 611         }
 612         if (egid != (gid_t) -1) {
 613                 if ((old_rgid == egid) ||
 614                     (current->egid == egid) ||
 615                     (current->sgid == egid) ||
 616                     capable(CAP_SETGID))
 617                         new_egid = egid;
 618                 else {
 619                         return -EPERM;
 620                 }
 621         }
 622         if (new_egid != old_egid)
 623         {
 624                 current->mm->dumpable = 0;
 625                 wmb();
 626         }
 627         if (rgid != (gid_t) -1 ||
 628             (egid != (gid_t) -1 && egid != old_rgid))
 629                 current->sgid = new_egid;
 630         current->fsgid = new_egid;
 631         current->egid = new_egid;
 632         current->gid = new_rgid;
 633
 634         ckrm_cb_gid();
 635
 636         return 0;
 637 }
 638
 639 /*
 640  * setgid() is implemented like SysV w/ SAVED_IDS
 641  *
 642  * SMP: Same implicit races as above.
 643  */
 644 asmlinkage long sys_setgid(gid_t gid)
 645 {
 646         int old_egid = current->egid;
 647         int retval;
 648
 649         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 650         if (retval)
 651                 return retval;
 652
 653         if (capable(CAP_SETGID))
 654         {
 655                 if(old_egid != gid)
 656                 {
 657                         current->mm->dumpable=0;
 658                         wmb();
 659                 }
 660                 current->gid = current->egid = current->sgid = current->fsgid = gid;
 661         }
 662         else if ((gid == current->gid) || (gid == current->sgid))
 663         {
 664                 if(old_egid != gid)
 665                 {
 666                         current->mm->dumpable=0;
 667                         wmb();
 668                 }
 669                 current->egid = current->fsgid = gid;
 670         }
 671         else
 672                 return -EPERM;
 673
 674         ckrm_cb_gid();
 675
 676         return 0;
 677 }
 678
 679 static int set_user(uid_t new_ruid, int dumpclear)
 680 {
 681         struct user_struct *new_user;
 682
 683         new_user = alloc_uid(vx_current_xid(), new_ruid);
 684         if (!new_user)
 685                 return -EAGAIN;
 686
 687         if (atomic_read(&new_user->processes) >=
 688                                 current->rlim[RLIMIT_NPROC].rlim_cur &&
 689                         new_user != &root_user) {
 690                 free_uid(new_user);
 691                 return -EAGAIN;
 692         }
 693
 694         switch_uid(new_user);
 695
 696         if(dumpclear)
 697         {
 698                 current->mm->dumpable = 0;
 699                 wmb();
 700         }
 701         current->uid = new_ruid;
 702         return 0;
 703 }
 704
 705 /*
 706  * Unprivileged users may change the real uid to the effective uid
 707  * or vice versa.  (BSD-style)
 708  *
 709  * If you set the real uid at all, or set the effective uid to a value not
 710  * equal to the real uid, then the saved uid is set to the new effective uid.
 711  *
 712  * This makes it possible for a setuid program to completely drop its
 713  * privileges, which is often a useful assertion to make when you are doing
 714  * a security audit over a program.
 715  *
 716  * The general idea is that a program which uses just setreuid() will be
 717  * 100% compatible with BSD.  A program which uses just setuid() will be
 718  * 100% compatible with POSIX with saved IDs.
 719  */
 720 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 721 {
 722         int old_ruid, old_euid, old_suid, new_ruid, new_euid;
 723         int retval;
 724
 725         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 726         if (retval)
 727                 return retval;
 728
 729         new_ruid = old_ruid = current->uid;
 730         new_euid = old_euid = current->euid;
 731         old_suid = current->suid;
 732
 733         if (ruid != (uid_t) -1) {
 734                 new_ruid = ruid;
 735                 if ((old_ruid != ruid) &&
 736                     (current->euid != ruid) &&
 737                     !capable(CAP_SETUID))
 738                         return -EPERM;
 739         }
 740
 741         if (euid != (uid_t) -1) {
 742                 new_euid = euid;
 743                 if ((old_ruid != euid) &&
 744                     (current->euid != euid) &&
 745                     (current->suid != euid) &&
 746                     !capable(CAP_SETUID))
 747                         return -EPERM;
 748         }
 749
 750         if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
 751                 return -EAGAIN;
 752
 753         if (new_euid != old_euid)
 754         {
 755                 current->mm->dumpable=0;
 756                 wmb();
 757         }
 758         current->fsuid = current->euid = new_euid;
 759         if (ruid != (uid_t) -1 ||
 760             (euid != (uid_t) -1 && euid != old_ruid))
 761                 current->suid = current->euid;
 762         current->fsuid = current->euid;
 763
 764         ckrm_cb_uid();
 765
 766         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
 767 }
 768
 769
 770
 771 /*
 772  * setuid() is implemented like SysV with SAVED_IDS
 773  *
 774  * Note that SAVED_ID's is deficient in that a setuid root program
 775  * like sendmail, for example, cannot set its uid to be a normal
 776  * user and then switch back, because if you're root, setuid() sets
 777  * the saved uid too.  If you don't like this, blame the bright people
 778  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 779  * will allow a root program to temporarily drop privileges and be able to
 780  * regain them by swapping the real and effective uid.
 781  */
 782 asmlinkage long sys_setuid(uid_t uid)
 783 {
 784         int old_euid = current->euid;
 785         int old_ruid, old_suid, new_ruid, new_suid;
 786         int retval;
 787
 788         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 789         if (retval)
 790                 return retval;
 791
 792         old_ruid = new_ruid = current->uid;
 793         old_suid = current->suid;
 794         new_suid = old_suid;
 795
 796         if (capable(CAP_SETUID)) {
 797                 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
 798                         return -EAGAIN;
 799                 new_suid = uid;
 800         } else if ((uid != current->uid) && (uid != new_suid))
 801                 return -EPERM;
 802
 803         if (old_euid != uid)
 804         {
 805                 current->mm->dumpable = 0;
 806                 wmb();
 807         }
 808         current->fsuid = current->euid = uid;
 809         current->suid = new_suid;
 810
 811         ckrm_cb_uid();
 812
 813         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
 814 }
 815
 816
 817 /*
 818  * This function implements a generic ability to update ruid, euid,
 819  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 820  */
 821 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 822 {
 823         int old_ruid = current->uid;
 824         int old_euid = current->euid;
 825         int old_suid = current->suid;
 826         int retval;
 827
 828         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 829         if (retval)
 830                 return retval;
 831
 832         if (!capable(CAP_SETUID)) {
 833                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
 834                     (ruid != current->euid) && (ruid != current->suid))
 835                         return -EPERM;
 836                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
 837                     (euid != current->euid) && (euid != current->suid))
 838                         return -EPERM;
 839                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
 840                     (suid != current->euid) && (suid != current->suid))
 841                         return -EPERM;
 842         }
 843         if (ruid != (uid_t) -1) {
 844                 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
 845                         return -EAGAIN;
 846         }
 847         if (euid != (uid_t) -1) {
 848                 if (euid != current->euid)
 849                 {
 850                         current->mm->dumpable = 0;
 851                         wmb();
 852                 }
 853                 current->euid = euid;
 854         }
 855         current->fsuid = current->euid;
 856         if (suid != (uid_t) -1)
 857                 current->suid = suid;
 858
 859         ckrm_cb_uid();
 860
 861         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
 862 }
 863
 864 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
 865 {
 866         int retval;
 867
 868         if (!(retval = put_user(current->uid, ruid)) &&
 869             !(retval = put_user(current->euid, euid)))
 870                 retval = put_user(current->suid, suid);
 871
 872         return retval;
 873 }
 874
 875 /*
 876  * Same as above, but for rgid, egid, sgid.
 877  */
 878 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 879 {
 880         int retval;
 881
 882         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 883         if (retval)
 884                 return retval;
 885
 886         if (!capable(CAP_SETGID)) {
 887                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
 888                     (rgid != current->egid) && (rgid != current->sgid))
 889                         return -EPERM;
 890                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
 891                     (egid != current->egid) && (egid != current->sgid))
 892                         return -EPERM;
 893                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
 894                     (sgid != current->egid) && (sgid != current->sgid))
 895                         return -EPERM;
 896         }
 897         if (egid != (gid_t) -1) {
 898                 if (egid != current->egid)
 899                 {
 900                         current->mm->dumpable = 0;
 901                         wmb();
 902                 }
 903                 current->egid = egid;
 904         }
 905         current->fsgid = current->egid;
 906         if (rgid != (gid_t) -1)
 907                 current->gid = rgid;
 908         if (sgid != (gid_t) -1)
 909                 current->sgid = sgid;
 910
 911         ckrm_cb_gid();
 912
 913         return 0;
 914 }
 915
 916 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
 917 {
 918         int retval;
 919
 920         if (!(retval = put_user(current->gid, rgid)) &&
 921             !(retval = put_user(current->egid, egid)))
 922                 retval = put_user(current->sgid, sgid);
 923
 924         return retval;
 925 }
 926
 927
 928 /*
 929  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 930  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 931  * whatever uid it wants to). It normally shadows "euid", except when
 932  * explicitly set by setfsuid() or for access..
 933  */
 934 asmlinkage long sys_setfsuid(uid_t uid)
 935 {
 936         int old_fsuid;
 937
 938         old_fsuid = current->fsuid;
 939         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
 940                 return old_fsuid;
 941
 942         if (uid == current->uid || uid == current->euid ||
 943             uid == current->suid || uid == current->fsuid ||
 944             capable(CAP_SETUID))
 945         {
 946                 if (uid != old_fsuid)
 947                 {
 948                         current->mm->dumpable = 0;
 949                         wmb();
 950                 }
 951                 current->fsuid = uid;
 952         }
 953
 954         security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
 955
 956         return old_fsuid;
 957 }
 958
 959 /*
 960  * Samma på svenska..
 961  */
 962 asmlinkage long sys_setfsgid(gid_t gid)
 963 {
 964         int old_fsgid;
 965
 966         old_fsgid = current->fsgid;
 967         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 968                 return old_fsgid;
 969
 970         if (gid == current->gid || gid == current->egid ||
 971             gid == current->sgid || gid == current->fsgid ||
 972             capable(CAP_SETGID))
 973         {
 974                 if (gid != old_fsgid)
 975                 {
 976                         current->mm->dumpable = 0;
 977                         wmb();
 978                 }
 979                 current->fsgid = gid;
 980         }
 981         return old_fsgid;
 982 }
 983
 984 asmlinkage long sys_times(struct tms __user * tbuf)
 985 {
 986         /*
 987          *      In the SMP world we might just be unlucky and have one of
 988          *      the times increment as we use it. Since the value is an
 989          *      atomically safe type this is just fine. Conceptually its
 990          *      as if the syscall took an instant longer to occur.
 991          */
 992         if (tbuf) {
 993                 struct tms tmp;
 994                 tmp.tms_utime = jiffies_to_clock_t(current->utime);
 995                 tmp.tms_stime = jiffies_to_clock_t(current->stime);
 996                 tmp.tms_cutime = jiffies_to_clock_t(current->cutime);
 997                 tmp.tms_cstime = jiffies_to_clock_t(current->cstime);
 998                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
 999                         return -EFAULT;
1000         }
1001         return (long) jiffies_64_to_clock_t(get_jiffies_64());
1002 }
1003
1004 /*
1005  * This needs some heavy checking ...
1006  * I just haven't the stomach for it. I also don't fully
1007  * understand sessions/pgrp etc. Let somebody who does explain it.
1008  *
1009  * OK, I think I have the protection semantics right.... this is really
1010  * only important on a multi-user system anyway, to make sure one user
1011  * can't send a signal to a process owned by another.  -TYT, 12/12/91
1012  *
1013  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1014  * LBT 04.03.94
1015  */
1016
1017 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1018 {
1019         struct task_struct *p;
1020         int err = -EINVAL;
1021
1022         if (!pid)
1023                 pid = current->pid;
1024         if (!pgid)
1025                 pgid = pid;
1026         if (pgid < 0)
1027                 return -EINVAL;
1028
1029         /* From this point forward we keep holding onto the tasklist lock
1030          * so that our parent does not change from under us. -DaveM
1031          */
1032         write_lock_irq(&tasklist_lock);
1033
1034         err = -ESRCH;
1035         p = find_task_by_pid(pid);
1036         if (!p)
1037                 goto out;
1038
1039         err = -EINVAL;
1040         if (!thread_group_leader(p))
1041                 goto out;
1042
1043         if (p->parent == current || p->real_parent == current) {
1044                 err = -EPERM;
1045                 if (p->signal->session != current->signal->session)
1046                         goto out;
1047                 err = -EACCES;
1048                 if (p->did_exec)
1049                         goto out;
1050         } else {
1051                 err = -ESRCH;
1052                 if (p != current)
1053                         goto out;
1054         }
1055
1056         err = -EPERM;
1057         if (p->signal->leader)
1058                 goto out;
1059
1060         if (pgid != pid) {
1061                 struct task_struct *p;
1062                 struct pid *pid;
1063                 struct list_head *l;
1064
1065                 for_each_task_pid(pgid, PIDTYPE_PGID, p, l, pid)
1066                         if (p->signal->session == current->signal->session)
1067                                 goto ok_pgid;
1068                 goto out;
1069         }
1070
1071 ok_pgid:
1072         err = security_task_setpgid(p, pgid);
1073         if (err)
1074                 goto out;
1075
1076         if (process_group(p) != pgid) {
1077                 detach_pid(p, PIDTYPE_PGID);
1078                 p->signal->pgrp = pgid;
1079                 attach_pid(p, PIDTYPE_PGID, pgid);
1080         }
1081
1082         err = 0;
1083 out:
1084         /* All paths lead to here, thus we are safe. -DaveM */
1085         write_unlock_irq(&tasklist_lock);
1086         return err;
1087 }
1088
1089 asmlinkage long sys_getpgid(pid_t pid)
1090 {
1091         if (!pid) {
1092                 return process_group(current);
1093         } else {
1094                 int retval;
1095                 struct task_struct *p;
1096
1097                 read_lock(&tasklist_lock);
1098                 p = find_task_by_pid(pid);
1099
1100                 retval = -ESRCH;
1101                 if (p) {
1102                         retval = security_task_getpgid(p);
1103                         if (!retval)
1104                                 retval = process_group(p);
1105                 }
1106                 read_unlock(&tasklist_lock);
1107                 return retval;
1108         }
1109 }
1110
1111 #ifdef __ARCH_WANT_SYS_GETPGRP
1112
1113 asmlinkage long sys_getpgrp(void)
1114 {
1115         /* SMP - assuming writes are word atomic this is fine */
1116         return process_group(current);
1117 }
1118
1119 #endif
1120
1121 asmlinkage long sys_getsid(pid_t pid)
1122 {
1123         if (!pid) {
1124                 return current->signal->session;
1125         } else {
1126                 int retval;
1127                 struct task_struct *p;
1128
1129                 read_lock(&tasklist_lock);
1130                 p = find_task_by_pid(pid);
1131
1132                 retval = -ESRCH;
1133                 if(p) {
1134                         retval = security_task_getsid(p);
1135                         if (!retval)
1136                                 retval = p->signal->session;
1137                 }
1138                 read_unlock(&tasklist_lock);
1139                 return retval;
1140         }
1141 }
1142
1143 asmlinkage long sys_setsid(void)
1144 {
1145         struct pid *pid;
1146         int err = -EPERM;
1147
1148         if (!thread_group_leader(current))
1149                 return -EINVAL;
1150
1151         write_lock_irq(&tasklist_lock);
1152
1153         pid = find_pid(PIDTYPE_PGID, current->pid);
1154         if (pid)
1155                 goto out;
1156
1157         current->signal->leader = 1;
1158         __set_special_pids(current->pid, current->pid);
1159         current->signal->tty = NULL;
1160         current->signal->tty_old_pgrp = 0;
1161         err = process_group(current);
1162 out:
1163         write_unlock_irq(&tasklist_lock);
1164         return err;
1165 }
1166
1167 /*
1168  * Supplementary group IDs
1169  */
1170
1171 /* init to 2 - one for init_task, one to ensure it is never freed */
1172 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1173
1174 struct group_info *groups_alloc(int gidsetsize)
1175 {
1176         struct group_info *group_info;
1177         int nblocks;
1178         int i;
1179
1180         nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1181         /* Make sure we always allocate at least one indirect block pointer */
1182         nblocks = nblocks ? : 1;
1183         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1184         if (!group_info)
1185                 return NULL;
1186         group_info->ngroups = gidsetsize;
1187         group_info->nblocks = nblocks;
1188         atomic_set(&group_info->usage, 1);
1189
1190         if (gidsetsize <= NGROUPS_SMALL) {
1191                 group_info->blocks[0] = group_info->small_block;
1192         } else {
1193                 for (i = 0; i < nblocks; i++) {
1194                         gid_t *b;
1195                         b = (void *)__get_free_page(GFP_USER);
1196                         if (!b)
1197                                 goto out_undo_partial_alloc;
1198                         group_info->blocks[i] = b;
1199                 }
1200         }
1201         return group_info;
1202
1203 out_undo_partial_alloc:
1204         while (--i >= 0) {
1205                 free_page((unsigned long)group_info->blocks[i]);
1206         }
1207         kfree(group_info);
1208         return NULL;
1209 }
1210
1211 EXPORT_SYMBOL(groups_alloc);
1212
1213 void groups_free(struct group_info *group_info)
1214 {
1215         if (group_info->blocks[0] != group_info->small_block) {
1216                 int i;
1217                 for (i = 0; i < group_info->nblocks; i++)
1218                         free_page((unsigned long)group_info->blocks[i]);
1219         }
1220         kfree(group_info);
1221 }
1222
1223 EXPORT_SYMBOL(groups_free);
1224
1225 /* export the group_info to a user-space array */
1226 static int groups_to_user(gid_t __user *grouplist,
1227     struct group_info *group_info)
1228 {
1229         int i;
1230         int count = group_info->ngroups;
1231
1232         for (i = 0; i < group_info->nblocks; i++) {
1233                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1234                 int off = i * NGROUPS_PER_BLOCK;
1235                 int len = cp_count * sizeof(*grouplist);
1236
1237                 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1238                         return -EFAULT;
1239
1240                 count -= cp_count;
1241         }
1242         return 0;
1243 }
1244
1245 /* fill a group_info from a user-space array - it must be allocated already */
1246 static int groups_from_user(struct group_info *group_info,
1247     gid_t __user *grouplist)
1248  {
1249         int i;
1250         int count = group_info->ngroups;
1251
1252         for (i = 0; i < group_info->nblocks; i++) {
1253                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1254                 int off = i * NGROUPS_PER_BLOCK;
1255                 int len = cp_count * sizeof(*grouplist);
1256
1257                 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1258                         return -EFAULT;
1259
1260                 count -= cp_count;
1261         }
1262         return 0;
1263 }
1264
1265 /* a simple shell-metzner sort */
1266 static void groups_sort(struct group_info *group_info)
1267 {
1268         int base, max, stride;
1269         int gidsetsize = group_info->ngroups;
1270
1271         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1272                 ; /* nothing */
1273         stride /= 3;
1274
1275         while (stride) {
1276                 max = gidsetsize - stride;
1277                 for (base = 0; base < max; base++) {
1278                         int left = base;
1279                         int right = left + stride;
1280                         gid_t tmp = GROUP_AT(group_info, right);
1281
1282                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1283                                 GROUP_AT(group_info, right) =
1284                                     GROUP_AT(group_info, left);
1285                                 right = left;
1286                                 left -= stride;
1287                         }
1288                         GROUP_AT(group_info, right) = tmp;
1289                 }
1290                 stride /= 3;
1291         }
1292 }
1293
1294 /* a simple bsearch */
1295 static int groups_search(struct group_info *group_info, gid_t grp)
1296 {
1297         int left, right;
1298
1299         if (!group_info)
1300                 return 0;
1301
1302         left = 0;
1303         right = group_info->ngroups;
1304         while (left < right) {
1305                 int mid = (left+right)/2;
1306                 int cmp = grp - GROUP_AT(group_info, mid);
1307                 if (cmp > 0)
1308                         left = mid + 1;
1309                 else if (cmp < 0)
1310                         right = mid;
1311                 else
1312                         return 1;
1313         }
1314         return 0;
1315 }
1316
1317 /* validate and set current->group_info */
1318 int set_current_groups(struct group_info *group_info)
1319 {
1320         int retval;
1321         struct group_info *old_info;
1322
1323         retval = security_task_setgroups(group_info);
1324         if (retval)
1325                 return retval;
1326
1327         groups_sort(group_info);
1328         get_group_info(group_info);
1329
1330         task_lock(current);
1331         old_info = current->group_info;
1332         current->group_info = group_info;
1333         task_unlock(current);
1334
1335         put_group_info(old_info);
1336
1337         return 0;
1338 }
1339
1340 EXPORT_SYMBOL(set_current_groups);
1341
1342 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1343 {
1344         int i = 0;
1345
1346         /*
1347          *      SMP: Nobody else can change our grouplist. Thus we are
1348          *      safe.
1349          */
1350
1351         if (gidsetsize < 0)
1352                 return -EINVAL;
1353
1354         /* no need to grab task_lock here; it cannot change */
1355         get_group_info(current->group_info);
1356         i = current->group_info->ngroups;
1357         if (gidsetsize) {
1358                 if (i > gidsetsize) {
1359                         i = -EINVAL;
1360                         goto out;
1361                 }
1362                 if (groups_to_user(grouplist, current->group_info)) {
1363                         i = -EFAULT;
1364                         goto out;
1365                 }
1366         }
1367 out:
1368         put_group_info(current->group_info);
1369         return i;
1370 }
1371
1372 /*
1373  *      SMP: Our groups are copy-on-write. We can set them safely
1374  *      without another task interfering.
1375  */
1376
1377 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1378 {
1379         struct group_info *group_info;
1380         int retval;
1381
1382         if (!capable(CAP_SETGID))
1383                 return -EPERM;
1384         if ((unsigned)gidsetsize > NGROUPS_MAX)
1385                 return -EINVAL;
1386
1387         group_info = groups_alloc(gidsetsize);
1388         if (!group_info)
1389                 return -ENOMEM;
1390         retval = groups_from_user(group_info, grouplist);
1391         if (retval) {
1392                 put_group_info(group_info);
1393                 return retval;
1394         }
1395
1396         retval = set_current_groups(group_info);
1397         put_group_info(group_info);
1398
1399         return retval;
1400 }
1401
1402 /*
1403  * Check whether we're fsgid/egid or in the supplemental group..
1404  */
1405 int in_group_p(gid_t grp)
1406 {
1407         int retval = 1;
1408         if (grp != current->fsgid) {
1409                 get_group_info(current->group_info);
1410                 retval = groups_search(current->group_info, grp);
1411                 put_group_info(current->group_info);
1412         }
1413         return retval;
1414 }
1415
1416 EXPORT_SYMBOL(in_group_p);
1417
1418 int in_egroup_p(gid_t grp)
1419 {
1420         int retval = 1;
1421         if (grp != current->egid) {
1422                 get_group_info(current->group_info);
1423                 retval = groups_search(current->group_info, grp);
1424                 put_group_info(current->group_info);
1425         }
1426         return retval;
1427 }
1428
1429 EXPORT_SYMBOL(in_egroup_p);
1430
1431 DECLARE_RWSEM(uts_sem);
1432
1433 EXPORT_SYMBOL(uts_sem);
1434
1435 asmlinkage long sys_newuname(struct new_utsname __user * name)
1436 {
1437         int errno = 0;
1438
1439         down_read(&uts_sem);
1440         if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1441                 errno = -EFAULT;
1442         up_read(&uts_sem);
1443         return errno;
1444 }
1445
1446 asmlinkage long sys_sethostname(char __user *name, int len)
1447 {
1448         int errno;
1449         char tmp[__NEW_UTS_LEN];
1450
1451         if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1452                 return -EPERM;
1453         if (len < 0 || len > __NEW_UTS_LEN)
1454                 return -EINVAL;
1455         down_write(&uts_sem);
1456         errno = -EFAULT;
1457         if (!copy_from_user(tmp, name, len)) {
1458                 char *ptr = vx_new_uts(nodename);
1459
1460                 memcpy(ptr, tmp, len);
1461                 ptr[len] = 0;
1462                 errno = 0;
1463         }
1464         up_write(&uts_sem);
1465         return errno;
1466 }
1467
1468 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1469
1470 asmlinkage long sys_gethostname(char __user *name, int len)
1471 {
1472         int i, errno;
1473         char *ptr;
1474
1475         if (len < 0)
1476                 return -EINVAL;
1477         down_read(&uts_sem);
1478         ptr = vx_new_uts(nodename);
1479         i = 1 + strlen(ptr);
1480         if (i > len)
1481                 i = len;
1482         errno = 0;
1483         if (copy_to_user(name, ptr, i))
1484                 errno = -EFAULT;
1485         up_read(&uts_sem);
1486         return errno;
1487 }
1488
1489 #endif
1490
1491 /*
1492  * Only setdomainname; getdomainname can be implemented by calling
1493  * uname()
1494  */
1495 asmlinkage long sys_setdomainname(char __user *name, int len)
1496 {
1497         int errno;
1498         char tmp[__NEW_UTS_LEN];
1499
1500         if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1501                 return -EPERM;
1502         if (len < 0 || len > __NEW_UTS_LEN)
1503                 return -EINVAL;
1504
1505         down_write(&uts_sem);
1506         errno = -EFAULT;
1507         if (!copy_from_user(tmp, name, len)) {
1508                 char *ptr = vx_new_uts(domainname);
1509
1510                 memcpy(ptr, tmp, len);
1511                 ptr[len] = 0;
1512                 errno = 0;
1513         }
1514         up_write(&uts_sem);
1515         return errno;
1516 }
1517
1518 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1519 {
1520         if (resource >= RLIM_NLIMITS)
1521                 return -EINVAL;
1522         else
1523                 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
1524                         ? -EFAULT : 0;
1525 }
1526
1527 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1528
1529 /*
1530  *      Back compatibility for getrlimit. Needed for some apps.
1531  */
1532
1533 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1534 {
1535         struct rlimit x;
1536         if (resource >= RLIM_NLIMITS)
1537                 return -EINVAL;
1538
1539         memcpy(&x, current->rlim + resource, sizeof(*rlim));
1540         if(x.rlim_cur > 0x7FFFFFFF)
1541                 x.rlim_cur = 0x7FFFFFFF;
1542         if(x.rlim_max > 0x7FFFFFFF)
1543                 x.rlim_max = 0x7FFFFFFF;
1544         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1545 }
1546
1547 #endif
1548
1549 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1550 {
1551         struct rlimit new_rlim, *old_rlim;
1552         int retval;
1553
1554         if (resource >= RLIM_NLIMITS)
1555                 return -EINVAL;
1556         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1557                 return -EFAULT;
1558        if (new_rlim.rlim_cur > new_rlim.rlim_max)
1559                return -EINVAL;
1560         old_rlim = current->rlim + resource;
1561         if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
1562              (new_rlim.rlim_max > old_rlim->rlim_max)) &&
1563             !capable(CAP_SYS_RESOURCE) && !vx_ccaps(VXC_SET_RLIMIT))
1564                 return -EPERM;
1565         if (resource == RLIMIT_NOFILE) {
1566                 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
1567                         return -EPERM;
1568         }
1569
1570         retval = security_task_setrlimit(resource, &new_rlim);
1571         if (retval)
1572                 return retval;
1573
1574         *old_rlim = new_rlim;
1575         return 0;
1576 }
1577
1578 /*
1579  * It would make sense to put struct rusage in the task_struct,
1580  * except that would make the task_struct be *really big*.  After
1581  * task_struct gets moved into malloc'ed memory, it would
1582  * make sense to do this.  It will make moving the rest of the information
1583  * a lot simpler!  (Which we're not doing right now because we're not
1584  * measuring them yet).
1585  *
1586  * This is SMP safe.  Either we are called from sys_getrusage on ourselves
1587  * below (we know we aren't going to exit/disappear and only we change our
1588  * rusage counters), or we are called from wait4() on a process which is
1589  * either stopped or zombied.  In the zombied case the task won't get
1590  * reaped till shortly after the call to getrusage(), in both cases the
1591  * task being examined is in a frozen state so the counters won't change.
1592  */
1593 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1594 {
1595         struct rusage r;
1596
1597         memset((char *) &r, 0, sizeof(r));
1598         switch (who) {
1599                 case RUSAGE_SELF:
1600                         jiffies_to_timeval(p->utime, &r.ru_utime);
1601                         jiffies_to_timeval(p->stime, &r.ru_stime);
1602                         r.ru_nvcsw = p->nvcsw;
1603                         r.ru_nivcsw = p->nivcsw;
1604                         r.ru_minflt = p->min_flt;
1605                         r.ru_majflt = p->maj_flt;
1606                         break;
1607                 case RUSAGE_CHILDREN:
1608                         jiffies_to_timeval(p->cutime, &r.ru_utime);
1609                         jiffies_to_timeval(p->cstime, &r.ru_stime);
1610                         r.ru_nvcsw = p->cnvcsw;
1611                         r.ru_nivcsw = p->cnivcsw;
1612                         r.ru_minflt = p->cmin_flt;
1613                         r.ru_majflt = p->cmaj_flt;
1614                         break;
1615                 default:
1616                         jiffies_to_timeval(p->utime + p->cutime, &r.ru_utime);
1617                         jiffies_to_timeval(p->stime + p->cstime, &r.ru_stime);
1618                         r.ru_nvcsw = p->nvcsw + p->cnvcsw;
1619                         r.ru_nivcsw = p->nivcsw + p->cnivcsw;
1620                         r.ru_minflt = p->min_flt + p->cmin_flt;
1621                         r.ru_majflt = p->maj_flt + p->cmaj_flt;
1622                         break;
1623         }
1624         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1625 }
1626
1627 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1628 {
1629         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1630                 return -EINVAL;
1631         return getrusage(current, who, ru);
1632 }
1633
1634 asmlinkage long sys_umask(int mask)
1635 {
1636         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1637         return mask;
1638 }
1639
1640 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1641                           unsigned long arg4, unsigned long arg5)
1642 {
1643         int error;
1644         int sig;
1645
1646         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1647         if (error)
1648                 return error;
1649
1650         switch (option) {
1651                 case PR_SET_PDEATHSIG:
1652                         sig = arg2;
1653                         if (sig < 0 || sig > _NSIG) {
1654                                 error = -EINVAL;
1655                                 break;
1656                         }
1657                         current->pdeath_signal = sig;
1658                         break;
1659                 case PR_GET_PDEATHSIG:
1660                         error = put_user(current->pdeath_signal, (int __user *)arg2);
1661                         break;
1662                 case PR_GET_DUMPABLE:
1663                         if (current->mm->dumpable)
1664                                 error = 1;
1665                         break;
1666                 case PR_SET_DUMPABLE:
1667                         if (arg2 != 0 && arg2 != 1) {
1668                                 error = -EINVAL;
1669                                 break;
1670                         }
1671                         current->mm->dumpable = arg2;
1672                         break;
1673
1674                 case PR_SET_UNALIGN:
1675                         error = SET_UNALIGN_CTL(current, arg2);
1676                         break;
1677                 case PR_GET_UNALIGN:
1678                         error = GET_UNALIGN_CTL(current, arg2);
1679                         break;
1680                 case PR_SET_FPEMU:
1681                         error = SET_FPEMU_CTL(current, arg2);
1682                         break;
1683                 case PR_GET_FPEMU:
1684                         error = GET_FPEMU_CTL(current, arg2);
1685                         break;
1686                 case PR_SET_FPEXC:
1687                         error = SET_FPEXC_CTL(current, arg2);
1688                         break;
1689                 case PR_GET_FPEXC:
1690                         error = GET_FPEXC_CTL(current, arg2);
1691                         break;
1692                 case PR_GET_TIMING:
1693                         error = PR_TIMING_STATISTICAL;
1694                         break;
1695                 case PR_SET_TIMING:
1696                         if (arg2 == PR_TIMING_STATISTICAL)
1697                                 error = 0;
1698                         else
1699                                 error = -EINVAL;
1700                         break;
1701
1702                 case PR_GET_KEEPCAPS:
1703                         if (current->keep_capabilities)
1704                                 error = 1;
1705                         break;
1706                 case PR_SET_KEEPCAPS:
1707                         if (arg2 != 0 && arg2 != 1) {
1708                                 error = -EINVAL;
1709                                 break;
1710                         }
1711                         current->keep_capabilities = arg2;
1712                         break;
1713                 default:
1714                         error = -EINVAL;
1715                         break;
1716         }
1717         return error;
1718 }