kernel/sys.c

   1 /*
   2  *  linux/kernel/sys.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6
   7 #include <linux/config.h>
   8 #include <linux/module.h>
   9 #include <linux/mm.h>
  10 #include <linux/utsname.h>
  11 #include <linux/mman.h>
  12 #include <linux/smp_lock.h>
  13 #include <linux/notifier.h>
  14 #include <linux/kmod.h>
  15 #include <linux/reboot.h>
  16 #include <linux/prctl.h>
  17 #include <linux/init.h>
  18 #include <linux/highuid.h>
  19 #include <linux/fs.h>
  20 #include <linux/kernel.h>
  21 #include <linux/kexec.h>
  22 #include <linux/workqueue.h>
  23 #include <linux/device.h>
  24 #include <linux/key.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_events.h>
  30 #include <linux/tty.h>
  31 #include <linux/vs_cvirt.h>
  32 #include <linux/compat.h>
  33 #include <linux/syscalls.h>
  34
  35 #include <asm/uaccess.h>
  36 #include <asm/io.h>
  37 #include <asm/unistd.h>
  38
  39 #ifndef SET_UNALIGN_CTL
  40 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  41 #endif
  42 #ifndef GET_UNALIGN_CTL
  43 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  44 #endif
  45 #ifndef SET_FPEMU_CTL
  46 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
  47 #endif
  48 #ifndef GET_FPEMU_CTL
  49 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
  50 #endif
  51 #ifndef SET_FPEXC_CTL
  52 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
  53 #endif
  54 #ifndef GET_FPEXC_CTL
  55 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
  56 #endif
  57
  58 /*
  59  * this is where the system-wide overflow UID and GID are defined, for
  60  * architectures that now have 32-bit UID/GID but didn't in the past
  61  */
  62
  63 int overflowuid = DEFAULT_OVERFLOWUID;
  64 int overflowgid = DEFAULT_OVERFLOWGID;
  65
  66 #ifdef CONFIG_UID16
  67 EXPORT_SYMBOL(overflowuid);
  68 EXPORT_SYMBOL(overflowgid);
  69 #endif
  70
  71 /*
  72  * the same as above, but for filesystems which can only store a 16-bit
  73  * UID and GID. as such, this is needed on all architectures
  74  */
  75
  76 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
  77 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
  78
  79 EXPORT_SYMBOL(fs_overflowuid);
  80 EXPORT_SYMBOL(fs_overflowgid);
  81
  82 /*
  83  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
  84  */
  85
  86 int C_A_D = 1;
  87 int cad_pid = 1;
  88
  89 /*
  90  *      Notifier list for kernel code which wants to be called
  91  *      at shutdown. This is used to stop any idling DMA operations
  92  *      and the like.
  93  */
  94
  95 static struct notifier_block *reboot_notifier_list;
  96 rwlock_t notifier_lock = RW_LOCK_UNLOCKED;
  97
  98 /**
  99  *      notifier_chain_register - Add notifier to a notifier chain
 100  *      @list: Pointer to root list pointer
 101  *      @n: New entry in notifier chain
 102  *
 103  *      Adds a notifier to a notifier chain.
 104  *
 105  *      Currently always returns zero.
 106  */
 107
 108 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
 109 {
 110         write_lock(&notifier_lock);
 111         while(*list)
 112         {
 113                 if(n->priority > (*list)->priority)
 114                         break;
 115                 list= &((*list)->next);
 116         }
 117         n->next = *list;
 118         *list=n;
 119         write_unlock(&notifier_lock);
 120         return 0;
 121 }
 122
 123 EXPORT_SYMBOL(notifier_chain_register);
 124
 125 /**
 126  *      notifier_chain_unregister - Remove notifier from a notifier chain
 127  *      @nl: Pointer to root list pointer
 128  *      @n: New entry in notifier chain
 129  *
 130  *      Removes a notifier from a notifier chain.
 131  *
 132  *      Returns zero on success, or %-ENOENT on failure.
 133  */
 134
 135 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
 136 {
 137         write_lock(&notifier_lock);
 138         while((*nl)!=NULL)
 139         {
 140                 if((*nl)==n)
 141                 {
 142                         *nl=n->next;
 143                         write_unlock(&notifier_lock);
 144                         return 0;
 145                 }
 146                 nl=&((*nl)->next);
 147         }
 148         write_unlock(&notifier_lock);
 149         return -ENOENT;
 150 }
 151
 152 EXPORT_SYMBOL(notifier_chain_unregister);
 153
 154 /**
 155  *      notifier_call_chain - Call functions in a notifier chain
 156  *      @n: Pointer to root pointer of notifier chain
 157  *      @val: Value passed unmodified to notifier function
 158  *      @v: Pointer passed unmodified to notifier function
 159  *
 160  *      Calls each function in a notifier chain in turn.
 161  *
 162  *      If the return value of the notifier can be and'd
 163  *      with %NOTIFY_STOP_MASK, then notifier_call_chain
 164  *      will return immediately, with the return value of
 165  *      the notifier function which halted execution.
 166  *      Otherwise, the return value is the return value
 167  *      of the last notifier function called.
 168  */
 169
 170 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
 171 {
 172         int ret=NOTIFY_DONE;
 173         struct notifier_block *nb = *n;
 174
 175         while(nb)
 176         {
 177                 ret=nb->notifier_call(nb,val,v);
 178                 if(ret&NOTIFY_STOP_MASK)
 179                 {
 180                         return ret;
 181                 }
 182                 nb=nb->next;
 183         }
 184         return ret;
 185 }
 186
 187 EXPORT_SYMBOL(notifier_call_chain);
 188
 189 /**
 190  *      register_reboot_notifier - Register function to be called at reboot time
 191  *      @nb: Info about notifier function to be called
 192  *
 193  *      Registers a function with the list of functions
 194  *      to be called at reboot time.
 195  *
 196  *      Currently always returns zero, as notifier_chain_register
 197  *      always returns zero.
 198  */
 199
 200 int register_reboot_notifier(struct notifier_block * nb)
 201 {
 202         return notifier_chain_register(&reboot_notifier_list, nb);
 203 }
 204
 205 EXPORT_SYMBOL(register_reboot_notifier);
 206
 207 /**
 208  *      unregister_reboot_notifier - Unregister previously registered reboot notifier
 209  *      @nb: Hook to be unregistered
 210  *
 211  *      Unregisters a previously registered reboot
 212  *      notifier function.
 213  *
 214  *      Returns zero on success, or %-ENOENT on failure.
 215  */
 216
 217 int unregister_reboot_notifier(struct notifier_block * nb)
 218 {
 219         return notifier_chain_unregister(&reboot_notifier_list, nb);
 220 }
 221
 222 EXPORT_SYMBOL(unregister_reboot_notifier);
 223 static int set_one_prio(struct task_struct *p, int niceval, int error)
 224 {
 225         int no_nice;
 226
 227         if (p->uid != current->euid &&
 228                 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
 229                 error = -EPERM;
 230                 goto out;
 231         }
 232         if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
 233                 if (vx_flags(VXF_IGNEG_NICE, 0))
 234                         error = 0;
 235                 else
 236                         error = -EACCES;
 237                 goto out;
 238         }
 239         no_nice = security_task_setnice(p, niceval);
 240         if (no_nice) {
 241                 error = no_nice;
 242                 goto out;
 243         }
 244         if (error == -ESRCH)
 245                 error = 0;
 246         set_user_nice(p, niceval);
 247 out:
 248         return error;
 249 }
 250
 251 asmlinkage long sys_setpriority(int which, int who, int niceval)
 252 {
 253         struct task_struct *g, *p;
 254         struct user_struct *user;
 255         int error = -EINVAL;
 256
 257         if (which > 2 || which < 0)
 258                 goto out;
 259
 260         /* normalize: avoid signed division (rounding problems) */
 261         error = -ESRCH;
 262         if (niceval < -20)
 263                 niceval = -20;
 264         if (niceval > 19)
 265                 niceval = 19;
 266
 267         read_lock(&tasklist_lock);
 268         switch (which) {
 269                 case PRIO_PROCESS:
 270                         if (!who)
 271                                 who = current->pid;
 272                         p = find_task_by_pid(who);
 273                         if (p)
 274                                 error = set_one_prio(p, niceval, error);
 275                         break;
 276                 case PRIO_PGRP:
 277                         if (!who)
 278                                 who = process_group(current);
 279                         do_each_task_pid(who, PIDTYPE_PGID, p) {
 280                                 error = set_one_prio(p, niceval, error);
 281                         } while_each_task_pid(who, PIDTYPE_PGID, p);
 282                         break;
 283                 case PRIO_USER:
 284                         user = current->user;
 285                         if (!who)
 286                                 who = current->uid;
 287                         else
 288                                 if ((who != current->uid) &&
 289                                         !(user = find_user(vx_current_xid(), who)))
 290                                         goto out_unlock;        /* No processes for this user */
 291
 292                         do_each_thread(g, p)
 293                                 if (p->uid == who)
 294                                         error = set_one_prio(p, niceval, error);
 295                         while_each_thread(g, p);
 296                         if (who != current->uid)
 297                                 free_uid(user);         /* For find_user() */
 298                         break;
 299         }
 300 out_unlock:
 301         read_unlock(&tasklist_lock);
 302 out:
 303         return error;
 304 }
 305
 306 /*
 307  * Ugh. To avoid negative return values, "getpriority()" will
 308  * not return the normal nice-value, but a negated value that
 309  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 310  * to stay compatible.
 311  */
 312 asmlinkage long sys_getpriority(int which, int who)
 313 {
 314         struct task_struct *g, *p;
 315         struct user_struct *user;
 316         long niceval, retval = -ESRCH;
 317
 318         if (which > 2 || which < 0)
 319                 return -EINVAL;
 320
 321         read_lock(&tasklist_lock);
 322         switch (which) {
 323                 case PRIO_PROCESS:
 324                         if (!who)
 325                                 who = current->pid;
 326                         p = find_task_by_pid(who);
 327                         if (p) {
 328                                 niceval = 20 - task_nice(p);
 329                                 if (niceval > retval)
 330                                         retval = niceval;
 331                         }
 332                         break;
 333                 case PRIO_PGRP:
 334                         if (!who)
 335                                 who = process_group(current);
 336                         do_each_task_pid(who, PIDTYPE_PGID, p) {
 337                                 niceval = 20 - task_nice(p);
 338                                 if (niceval > retval)
 339                                         retval = niceval;
 340                         } while_each_task_pid(who, PIDTYPE_PGID, p);
 341                         break;
 342                 case PRIO_USER:
 343                         user = current->user;
 344                         if (!who)
 345                                 who = current->uid;
 346                         else
 347                                 if ((who != current->uid) &&
 348                                         !(user = find_user(vx_current_xid(), who)))
 349                                         goto out_unlock;        /* No processes for this user */
 350
 351                         do_each_thread(g, p)
 352                                 if (p->uid == who) {
 353                                         niceval = 20 - task_nice(p);
 354                                         if (niceval > retval)
 355                                                 retval = niceval;
 356                                 }
 357                         while_each_thread(g, p);
 358                         if (who != current->uid)
 359                                 free_uid(user);         /* for find_user() */
 360                         break;
 361         }
 362 out_unlock:
 363         read_unlock(&tasklist_lock);
 364
 365         key_fsgid_changed(current);
 366         return 0;
 367 }
 368
 369 long vs_reboot(unsigned int, void *);
 370
 371 /*
 372  * Reboot system call: for obvious reasons only root may call it,
 373  * and even root needs to set up some magic numbers in the registers
 374  * so that some mistake won't make this reboot the whole machine.
 375  * You can also set the meaning of the ctrl-alt-del-key here.
 376  *
 377  * reboot doesn't sync: do that yourself before calling this.
 378  */
 379 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
 380 {
 381         char buffer[256];
 382
 383         /* We only trust the superuser with rebooting the system. */
 384         if (!capable(CAP_SYS_BOOT))
 385                 return -EPERM;
 386
 387         /* For safety, we require "magic" arguments. */
 388         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 389             (magic2 != LINUX_REBOOT_MAGIC2 &&
 390                         magic2 != LINUX_REBOOT_MAGIC2A &&
 391                         magic2 != LINUX_REBOOT_MAGIC2B &&
 392                         magic2 != LINUX_REBOOT_MAGIC2C))
 393                 return -EINVAL;
 394
 395         if (!vx_check(0, VX_ADMIN|VX_WATCH))
 396                 return vs_reboot(cmd, arg);
 397
 398         lock_kernel();
 399         switch (cmd) {
 400         case LINUX_REBOOT_CMD_RESTART:
 401                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 402                 system_state = SYSTEM_RESTART;
 403                 device_shutdown();
 404                 printk(KERN_EMERG "Restarting system.\n");
 405                 machine_restart(NULL);
 406                 break;
 407
 408         case LINUX_REBOOT_CMD_CAD_ON:
 409                 C_A_D = 1;
 410                 break;
 411
 412         case LINUX_REBOOT_CMD_CAD_OFF:
 413                 C_A_D = 0;
 414                 break;
 415
 416         case LINUX_REBOOT_CMD_HALT:
 417                 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
 418                 system_state = SYSTEM_HALT;
 419                 device_shutdown();
 420                 printk(KERN_EMERG "System halted.\n");
 421                 machine_halt();
 422                 unlock_kernel();
 423                 do_exit(0);
 424                 break;
 425
 426         case LINUX_REBOOT_CMD_POWER_OFF:
 427                 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
 428                 system_state = SYSTEM_POWER_OFF;
 429                 device_shutdown();
 430                 printk(KERN_EMERG "Power down.\n");
 431                 machine_power_off();
 432                 unlock_kernel();
 433                 do_exit(0);
 434                 break;
 435
 436         case LINUX_REBOOT_CMD_RESTART2:
 437                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 438                         unlock_kernel();
 439                         return -EFAULT;
 440                 }
 441                 buffer[sizeof(buffer) - 1] = '\0';
 442
 443                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
 444                 system_state = SYSTEM_RESTART;
 445                 device_shutdown();
 446                 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
 447                 machine_restart(buffer);
 448                 break;
 449
 450 #ifdef CONFIG_KEXEC
 451         case LINUX_REBOOT_CMD_KEXEC:
 452         {
 453                 struct kimage *image;
 454                 image = xchg(&kexec_image, 0);
 455                 if (!image) {
 456                         unlock_kernel();
 457                         return -EINVAL;
 458                 }
 459                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 460                 system_state = SYSTEM_RESTART;
 461                 device_shutdown();
 462                 system_state = SYSTEM_BOOTING;
 463                 printk(KERN_EMERG "Starting new kernel\n");
 464                 machine_shutdown();
 465                 machine_kexec(image);
 466                 break;
 467         }
 468 #endif
 469
 470 #ifdef CONFIG_SOFTWARE_SUSPEND
 471         case LINUX_REBOOT_CMD_SW_SUSPEND:
 472                 {
 473                         int ret = software_suspend();
 474                         unlock_kernel();
 475                         return ret;
 476                 }
 477 #endif
 478
 479         default:
 480                 unlock_kernel();
 481                 return -EINVAL;
 482         }
 483         unlock_kernel();
 484         return 0;
 485 }
 486
 487 static void deferred_cad(void *dummy)
 488 {
 489         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 490         machine_restart(NULL);
 491 }
 492
 493 /*
 494  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 495  * As it's called within an interrupt, it may NOT sync: the only choice
 496  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 497  */
 498 void ctrl_alt_del(void)
 499 {
 500         static DECLARE_WORK(cad_work, deferred_cad, NULL);
 501
 502         if (C_A_D)
 503                 schedule_work(&cad_work);
 504         else
 505                 kill_proc(cad_pid, SIGINT, 1);
 506 }
 507
 508
 509
 510 /*
 511  * Unprivileged users may change the real gid to the effective gid
 512  * or vice versa.  (BSD-style)
 513  *
 514  * If you set the real gid at all, or set the effective gid to a value not
 515  * equal to the real gid, then the saved gid is set to the new effective gid.
 516  *
 517  * This makes it possible for a setgid program to completely drop its
 518  * privileges, which is often a useful assertion to make when you are doing
 519  * a security audit over a program.
 520  *
 521  * The general idea is that a program which uses just setregid() will be
 522  * 100% compatible with BSD.  A program which uses just setgid() will be
 523  * 100% compatible with POSIX with saved IDs.
 524  *
 525  * SMP: There are not races, the GIDs are checked only by filesystem
 526  *      operations (as far as semantic preservation is concerned).
 527  */
 528 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 529 {
 530         int old_rgid = current->gid;
 531         int old_egid = current->egid;
 532         int new_rgid = old_rgid;
 533         int new_egid = old_egid;
 534         int retval;
 535
 536         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 537         if (retval)
 538                 return retval;
 539
 540         if (rgid != (gid_t) -1) {
 541                 if ((old_rgid == rgid) ||
 542                     (current->egid==rgid) ||
 543                     capable(CAP_SETGID))
 544                         new_rgid = rgid;
 545                 else
 546                         return -EPERM;
 547         }
 548         if (egid != (gid_t) -1) {
 549                 if ((old_rgid == egid) ||
 550                     (current->egid == egid) ||
 551                     (current->sgid == egid) ||
 552                     capable(CAP_SETGID))
 553                         new_egid = egid;
 554                 else {
 555                         return -EPERM;
 556                 }
 557         }
 558         if (new_egid != old_egid)
 559         {
 560                 current->mm->dumpable = suid_dumpable;
 561                 wmb();
 562         }
 563         if (rgid != (gid_t) -1 ||
 564             (egid != (gid_t) -1 && egid != old_rgid))
 565                 current->sgid = new_egid;
 566         current->fsgid = new_egid;
 567         current->egid = new_egid;
 568         current->gid = new_rgid;
 569
 570         key_fsgid_changed(current);
 571         ckrm_cb_gid();
 572         return 0;
 573 }
 574
 575 /*
 576  * setgid() is implemented like SysV w/ SAVED_IDS
 577  *
 578  * SMP: Same implicit races as above.
 579  */
 580 asmlinkage long sys_setgid(gid_t gid)
 581 {
 582         int old_egid = current->egid;
 583         int retval;
 584
 585         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 586         if (retval)
 587                 return retval;
 588
 589         if (capable(CAP_SETGID))
 590         {
 591                 if(old_egid != gid)
 592                 {
 593                         current->mm->dumpable = suid_dumpable;
 594                         wmb();
 595                 }
 596                 current->gid = current->egid = current->sgid = current->fsgid = gid;
 597         }
 598         else if ((gid == current->gid) || (gid == current->sgid))
 599         {
 600                 if(old_egid != gid)
 601                 {
 602                         current->mm->dumpable = suid_dumpable;
 603                         wmb();
 604                 }
 605                 current->egid = current->fsgid = gid;
 606         }
 607         else
 608                 return -EPERM;
 609
 610         key_fsgid_changed(current);
 611         ckrm_cb_gid();
 612
 613         return 0;
 614 }
 615
 616 static int set_user(uid_t new_ruid, int dumpclear)
 617 {
 618         struct user_struct *new_user;
 619
 620         new_user = alloc_uid(vx_current_xid(), new_ruid);
 621         if (!new_user)
 622                 return -EAGAIN;
 623
 624         if (atomic_read(&new_user->processes) >=
 625                                 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
 626                         new_user != &root_user) {
 627                 free_uid(new_user);
 628                 return -EAGAIN;
 629         }
 630
 631         switch_uid(new_user);
 632
 633         if(dumpclear)
 634         {
 635                 current->mm->dumpable = suid_dumpable;
 636                 wmb();
 637         }
 638         current->uid = new_ruid;
 639         return 0;
 640 }
 641
 642 /*
 643  * Unprivileged users may change the real uid to the effective uid
 644  * or vice versa.  (BSD-style)
 645  *
 646  * If you set the real uid at all, or set the effective uid to a value not
 647  * equal to the real uid, then the saved uid is set to the new effective uid.
 648  *
 649  * This makes it possible for a setuid program to completely drop its
 650  * privileges, which is often a useful assertion to make when you are doing
 651  * a security audit over a program.
 652  *
 653  * The general idea is that a program which uses just setreuid() will be
 654  * 100% compatible with BSD.  A program which uses just setuid() will be
 655  * 100% compatible with POSIX with saved IDs.
 656  */
 657 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 658 {
 659         int old_ruid, old_euid, old_suid, new_ruid, new_euid;
 660         int retval;
 661
 662         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 663         if (retval)
 664                 return retval;
 665
 666         new_ruid = old_ruid = current->uid;
 667         new_euid = old_euid = current->euid;
 668         old_suid = current->suid;
 669
 670         if (ruid != (uid_t) -1) {
 671                 new_ruid = ruid;
 672                 if ((old_ruid != ruid) &&
 673                     (current->euid != ruid) &&
 674                     !capable(CAP_SETUID))
 675                         return -EPERM;
 676         }
 677
 678         if (euid != (uid_t) -1) {
 679                 new_euid = euid;
 680                 if ((old_ruid != euid) &&
 681                     (current->euid != euid) &&
 682                     (current->suid != euid) &&
 683                     !capable(CAP_SETUID))
 684                         return -EPERM;
 685         }
 686
 687         if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
 688                 return -EAGAIN;
 689
 690         if (new_euid != old_euid)
 691         {
 692                 current->mm->dumpable = suid_dumpable;
 693                 wmb();
 694         }
 695         current->fsuid = current->euid = new_euid;
 696         if (ruid != (uid_t) -1 ||
 697             (euid != (uid_t) -1 && euid != old_ruid))
 698                 current->suid = current->euid;
 699         current->fsuid = current->euid;
 700
 701         key_fsuid_changed(current);
 702         ckrm_cb_uid();
 703
 704         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
 705 }
 706
 707
 708
 709 /*
 710  * setuid() is implemented like SysV with SAVED_IDS
 711  *
 712  * Note that SAVED_ID's is deficient in that a setuid root program
 713  * like sendmail, for example, cannot set its uid to be a normal
 714  * user and then switch back, because if you're root, setuid() sets
 715  * the saved uid too.  If you don't like this, blame the bright people
 716  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 717  * will allow a root program to temporarily drop privileges and be able to
 718  * regain them by swapping the real and effective uid.
 719  */
 720 asmlinkage long sys_setuid(uid_t uid)
 721 {
 722         int old_euid = current->euid;
 723         int old_ruid, old_suid, new_ruid, new_suid;
 724         int retval;
 725
 726         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 727         if (retval)
 728                 return retval;
 729
 730         old_ruid = new_ruid = current->uid;
 731         old_suid = current->suid;
 732         new_suid = old_suid;
 733
 734         if (capable(CAP_SETUID)) {
 735                 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
 736                         return -EAGAIN;
 737                 new_suid = uid;
 738         } else if ((uid != current->uid) && (uid != new_suid))
 739                 return -EPERM;
 740
 741         if (old_euid != uid)
 742         {
 743                 current->mm->dumpable = suid_dumpable;
 744                 wmb();
 745         }
 746         current->fsuid = current->euid = uid;
 747         current->suid = new_suid;
 748
 749         key_fsuid_changed(current);
 750         ckrm_cb_uid();
 751
 752         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
 753 }
 754
 755
 756 /*
 757  * This function implements a generic ability to update ruid, euid,
 758  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 759  */
 760 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 761 {
 762         int old_ruid = current->uid;
 763         int old_euid = current->euid;
 764         int old_suid = current->suid;
 765         int retval;
 766
 767         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 768         if (retval)
 769                 return retval;
 770
 771         if (!capable(CAP_SETUID)) {
 772                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
 773                     (ruid != current->euid) && (ruid != current->suid))
 774                         return -EPERM;
 775                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
 776                     (euid != current->euid) && (euid != current->suid))
 777                         return -EPERM;
 778                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
 779                     (suid != current->euid) && (suid != current->suid))
 780                         return -EPERM;
 781         }
 782         if (ruid != (uid_t) -1) {
 783                 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
 784                         return -EAGAIN;
 785         }
 786         if (euid != (uid_t) -1) {
 787                 if (euid != current->euid)
 788                 {
 789                         current->mm->dumpable = suid_dumpable;
 790                         wmb();
 791                 }
 792                 current->euid = euid;
 793         }
 794         current->fsuid = current->euid;
 795         if (suid != (uid_t) -1)
 796                 current->suid = suid;
 797
 798         key_fsuid_changed(current);
 799         ckrm_cb_uid();
 800
 801         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
 802 }
 803
 804 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
 805 {
 806         int retval;
 807
 808         if (!(retval = put_user(current->uid, ruid)) &&
 809             !(retval = put_user(current->euid, euid)))
 810                 retval = put_user(current->suid, suid);
 811
 812         return retval;
 813 }
 814
 815 /*
 816  * Same as above, but for rgid, egid, sgid.
 817  */
 818 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 819 {
 820         int retval;
 821
 822         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 823         if (retval)
 824                 return retval;
 825
 826         if (!capable(CAP_SETGID)) {
 827                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
 828                     (rgid != current->egid) && (rgid != current->sgid))
 829                         return -EPERM;
 830                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
 831                     (egid != current->egid) && (egid != current->sgid))
 832                         return -EPERM;
 833                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
 834                     (sgid != current->egid) && (sgid != current->sgid))
 835                         return -EPERM;
 836         }
 837         if (egid != (gid_t) -1) {
 838                 if (egid != current->egid)
 839                 {
 840                         current->mm->dumpable = suid_dumpable;
 841                         wmb();
 842                 }
 843                 current->egid = egid;
 844         }
 845         current->fsgid = current->egid;
 846         if (rgid != (gid_t) -1)
 847                 current->gid = rgid;
 848         if (sgid != (gid_t) -1)
 849                 current->sgid = sgid;
 850
 851         key_fsgid_changed(current);
 852         ckrm_cb_gid();
 853         return 0;
 854 }
 855
 856 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
 857 {
 858         int retval;
 859
 860         if (!(retval = put_user(current->gid, rgid)) &&
 861             !(retval = put_user(current->egid, egid)))
 862                 retval = put_user(current->sgid, sgid);
 863
 864         return retval;
 865 }
 866
 867
 868 /*
 869  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 870  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 871  * whatever uid it wants to). It normally shadows "euid", except when
 872  * explicitly set by setfsuid() or for access..
 873  */
 874 asmlinkage long sys_setfsuid(uid_t uid)
 875 {
 876         int old_fsuid;
 877
 878         old_fsuid = current->fsuid;
 879         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
 880                 return old_fsuid;
 881
 882         if (uid == current->uid || uid == current->euid ||
 883             uid == current->suid || uid == current->fsuid ||
 884             capable(CAP_SETUID))
 885         {
 886                 if (uid != old_fsuid)
 887                 {
 888                         current->mm->dumpable = suid_dumpable;
 889                         wmb();
 890                 }
 891                 current->fsuid = uid;
 892         }
 893
 894         key_fsuid_changed(current);
 895
 896         security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
 897
 898         return old_fsuid;
 899 }
 900
 901 /*
 902  * Samma på svenska..
 903  */
 904 asmlinkage long sys_setfsgid(gid_t gid)
 905 {
 906         int old_fsgid;
 907
 908         old_fsgid = current->fsgid;
 909         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 910                 return old_fsgid;
 911
 912         if (gid == current->gid || gid == current->egid ||
 913             gid == current->sgid || gid == current->fsgid ||
 914             capable(CAP_SETGID))
 915         {
 916                 if (gid != old_fsgid)
 917                 {
 918                         current->mm->dumpable = suid_dumpable;
 919                         wmb();
 920                 }
 921                 current->fsgid = gid;
 922                 key_fsgid_changed(current);
 923         }
 924         return old_fsgid;
 925 }
 926
 927 asmlinkage long sys_times(struct tms __user * tbuf)
 928 {
 929         /*
 930          *      In the SMP world we might just be unlucky and have one of
 931          *      the times increment as we use it. Since the value is an
 932          *      atomically safe type this is just fine. Conceptually its
 933          *      as if the syscall took an instant longer to occur.
 934          */
 935         if (tbuf) {
 936                 struct tms tmp;
 937                 struct task_struct *tsk = current;
 938                 struct task_struct *t;
 939                 unsigned long utime, stime, cutime, cstime;
 940
 941                 read_lock(&tasklist_lock);
 942                 utime = tsk->signal->utime;
 943                 stime = tsk->signal->stime;
 944                 t = tsk;
 945                 do {
 946                         utime += t->utime;
 947                         stime += t->stime;
 948                         t = next_thread(t);
 949                 } while (t != tsk);
 950
 951                 /*
 952                  * While we have tasklist_lock read-locked, no dying thread
 953                  * can be updating current->signal->[us]time.  Instead,
 954                  * we got their counts included in the live thread loop.
 955                  * However, another thread can come in right now and
 956                  * do a wait call that updates current->signal->c[us]time.
 957                  * To make sure we always see that pair updated atomically,
 958                  * we take the siglock around fetching them.
 959                  */
 960                 spin_lock_irq(&tsk->sighand->siglock);
 961                 cutime = tsk->signal->cutime;
 962                 cstime = tsk->signal->cstime;
 963                 spin_unlock_irq(&tsk->sighand->siglock);
 964                 read_unlock(&tasklist_lock);
 965
 966                 tmp.tms_utime = jiffies_to_clock_t(utime);
 967                 tmp.tms_stime = jiffies_to_clock_t(stime);
 968                 tmp.tms_cutime = jiffies_to_clock_t(cutime);
 969                 tmp.tms_cstime = jiffies_to_clock_t(cstime);
 970                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
 971                         return -EFAULT;
 972         }
 973         return (long) jiffies_64_to_clock_t(get_jiffies_64());
 974 }
 975
 976 /*
 977  * This needs some heavy checking ...
 978  * I just haven't the stomach for it. I also don't fully
 979  * understand sessions/pgrp etc. Let somebody who does explain it.
 980  *
 981  * OK, I think I have the protection semantics right.... this is really
 982  * only important on a multi-user system anyway, to make sure one user
 983  * can't send a signal to a process owned by another.  -TYT, 12/12/91
 984  *
 985  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
 986  * LBT 04.03.94
 987  */
 988
 989 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
 990 {
 991         struct task_struct *p;
 992         int err = -EINVAL;
 993         pid_t rpgid;
 994
 995         if (!pid)
 996                 pid = vx_map_pid(current->pid);
 997         if (!pgid)
 998                 pgid = pid;
 999         if (pgid < 0)
1000                 return -EINVAL;
1001
1002         rpgid = vx_rmap_pid(pgid);
1003
1004         /* From this point forward we keep holding onto the tasklist lock
1005          * so that our parent does not change from under us. -DaveM
1006          */
1007         write_lock_irq(&tasklist_lock);
1008
1009         err = -ESRCH;
1010         p = find_task_by_pid(pid);
1011         if (!p)
1012                 goto out;
1013
1014         err = -EINVAL;
1015         if (!thread_group_leader(p))
1016                 goto out;
1017
1018         if (p->parent == current || p->real_parent == current) {
1019                 err = -EPERM;
1020                 if (p->signal->session != current->signal->session)
1021                         goto out;
1022                 err = -EACCES;
1023                 if (p->did_exec)
1024                         goto out;
1025         } else {
1026                 err = -ESRCH;
1027                 if (p != current)
1028                         goto out;
1029         }
1030
1031         err = -EPERM;
1032         if (p->signal->leader)
1033                 goto out;
1034
1035         if (pgid != pid) {
1036                 struct task_struct *p;
1037
1038                 do_each_task_pid(rpgid, PIDTYPE_PGID, p) {
1039                         if (p->signal->session == current->signal->session)
1040                                 goto ok_pgid;
1041                 } while_each_task_pid(rpgid, PIDTYPE_PGID, p);
1042                 goto out;
1043         }
1044
1045 ok_pgid:
1046         err = security_task_setpgid(p, rpgid);
1047         if (err)
1048                 goto out;
1049
1050         if (process_group(p) != rpgid) {
1051                 detach_pid(p, PIDTYPE_PGID);
1052                 p->signal->pgrp = rpgid;
1053                 attach_pid(p, PIDTYPE_PGID, rpgid);
1054         }
1055
1056         err = 0;
1057 out:
1058         /* All paths lead to here, thus we are safe. -DaveM */
1059         write_unlock_irq(&tasklist_lock);
1060         return err;
1061 }
1062
1063 asmlinkage long sys_getpgid(pid_t pid)
1064 {
1065         if (!pid) {
1066                 return vx_rmap_pid(process_group(current));
1067         } else {
1068                 int retval;
1069                 struct task_struct *p;
1070
1071                 read_lock(&tasklist_lock);
1072                 p = find_task_by_pid(pid);
1073
1074                 retval = -ESRCH;
1075                 if (p) {
1076                         retval = security_task_getpgid(p);
1077                         if (!retval)
1078                                 retval = vx_rmap_pid(process_group(p));
1079                 }
1080                 read_unlock(&tasklist_lock);
1081                 return retval;
1082         }
1083 }
1084
1085 #ifdef __ARCH_WANT_SYS_GETPGRP
1086
1087 asmlinkage long sys_getpgrp(void)
1088 {
1089         /* SMP - assuming writes are word atomic this is fine */
1090         return process_group(current);
1091 }
1092
1093 #endif
1094
1095 asmlinkage long sys_getsid(pid_t pid)
1096 {
1097         if (!pid) {
1098                 return current->signal->session;
1099         } else {
1100                 int retval;
1101                 struct task_struct *p;
1102
1103                 read_lock(&tasklist_lock);
1104                 p = find_task_by_pid(pid);
1105
1106                 retval = -ESRCH;
1107                 if(p) {
1108                         retval = security_task_getsid(p);
1109                         if (!retval)
1110                                 retval = p->signal->session;
1111                 }
1112                 read_unlock(&tasklist_lock);
1113                 return retval;
1114         }
1115 }
1116
1117 asmlinkage long sys_setsid(void)
1118 {
1119         struct pid *pid;
1120         int err = -EPERM;
1121
1122         if (!thread_group_leader(current))
1123                 return -EINVAL;
1124
1125         down(&tty_sem);
1126         write_lock_irq(&tasklist_lock);
1127
1128         pid = find_pid(PIDTYPE_PGID, current->pid);
1129         if (pid)
1130                 goto out;
1131
1132         current->signal->leader = 1;
1133         __set_special_pids(current->pid, current->pid);
1134         current->signal->tty = NULL;
1135         current->signal->tty_old_pgrp = 0;
1136         err = process_group(current);
1137 out:
1138         write_unlock_irq(&tasklist_lock);
1139         up(&tty_sem);
1140         return err;
1141 }
1142
1143 /*
1144  * Supplementary group IDs
1145  */
1146
1147 /* init to 2 - one for init_task, one to ensure it is never freed */
1148 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1149
1150 struct group_info *groups_alloc(int gidsetsize)
1151 {
1152         struct group_info *group_info;
1153         int nblocks;
1154         int i;
1155
1156         nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1157         /* Make sure we always allocate at least one indirect block pointer */
1158         nblocks = nblocks ? : 1;
1159         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1160         if (!group_info)
1161                 return NULL;
1162         group_info->ngroups = gidsetsize;
1163         group_info->nblocks = nblocks;
1164         atomic_set(&group_info->usage, 1);
1165
1166         if (gidsetsize <= NGROUPS_SMALL) {
1167                 group_info->blocks[0] = group_info->small_block;
1168         } else {
1169                 for (i = 0; i < nblocks; i++) {
1170                         gid_t *b;
1171                         b = (void *)__get_free_page(GFP_USER);
1172                         if (!b)
1173                                 goto out_undo_partial_alloc;
1174                         group_info->blocks[i] = b;
1175                 }
1176         }
1177         return group_info;
1178
1179 out_undo_partial_alloc:
1180         while (--i >= 0) {
1181                 free_page((unsigned long)group_info->blocks[i]);
1182         }
1183         kfree(group_info);
1184         return NULL;
1185 }
1186
1187 EXPORT_SYMBOL(groups_alloc);
1188
1189 void groups_free(struct group_info *group_info)
1190 {
1191         if (group_info->blocks[0] != group_info->small_block) {
1192                 int i;
1193                 for (i = 0; i < group_info->nblocks; i++)
1194                         free_page((unsigned long)group_info->blocks[i]);
1195         }
1196         kfree(group_info);
1197 }
1198
1199 EXPORT_SYMBOL(groups_free);
1200
1201 /* export the group_info to a user-space array */
1202 static int groups_to_user(gid_t __user *grouplist,
1203     struct group_info *group_info)
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_to_user(grouplist+off, group_info->blocks[i], len))
1214                         return -EFAULT;
1215
1216                 count -= cp_count;
1217         }
1218         return 0;
1219 }
1220
1221 /* fill a group_info from a user-space array - it must be allocated already */
1222 static int groups_from_user(struct group_info *group_info,
1223     gid_t __user *grouplist)
1224  {
1225         int i;
1226         int count = group_info->ngroups;
1227
1228         for (i = 0; i < group_info->nblocks; i++) {
1229                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1230                 int off = i * NGROUPS_PER_BLOCK;
1231                 int len = cp_count * sizeof(*grouplist);
1232
1233                 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1234                         return -EFAULT;
1235
1236                 count -= cp_count;
1237         }
1238         return 0;
1239 }
1240
1241 /* a simple shell-metzner sort */
1242 static void groups_sort(struct group_info *group_info)
1243 {
1244         int base, max, stride;
1245         int gidsetsize = group_info->ngroups;
1246
1247         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1248                 ; /* nothing */
1249         stride /= 3;
1250
1251         while (stride) {
1252                 max = gidsetsize - stride;
1253                 for (base = 0; base < max; base++) {
1254                         int left = base;
1255                         int right = left + stride;
1256                         gid_t tmp = GROUP_AT(group_info, right);
1257
1258                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1259                                 GROUP_AT(group_info, right) =
1260                                     GROUP_AT(group_info, left);
1261                                 right = left;
1262                                 left -= stride;
1263                         }
1264                         GROUP_AT(group_info, right) = tmp;
1265                 }
1266                 stride /= 3;
1267         }
1268 }
1269
1270 /* a simple bsearch */
1271 static int groups_search(struct group_info *group_info, gid_t grp)
1272 {
1273         int left, right;
1274
1275         if (!group_info)
1276                 return 0;
1277
1278         left = 0;
1279         right = group_info->ngroups;
1280         while (left < right) {
1281                 int mid = (left+right)/2;
1282                 int cmp = grp - GROUP_AT(group_info, mid);
1283                 if (cmp > 0)
1284                         left = mid + 1;
1285                 else if (cmp < 0)
1286                         right = mid;
1287                 else
1288                         return 1;
1289         }
1290         return 0;
1291 }
1292
1293 /* validate and set current->group_info */
1294 int set_current_groups(struct group_info *group_info)
1295 {
1296         int retval;
1297         struct group_info *old_info;
1298
1299         retval = security_task_setgroups(group_info);
1300         if (retval)
1301                 return retval;
1302
1303         groups_sort(group_info);
1304         get_group_info(group_info);
1305
1306         task_lock(current);
1307         old_info = current->group_info;
1308         current->group_info = group_info;
1309         task_unlock(current);
1310
1311         put_group_info(old_info);
1312
1313         return 0;
1314 }
1315
1316 EXPORT_SYMBOL(set_current_groups);
1317
1318 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1319 {
1320         int i = 0;
1321
1322         /*
1323          *      SMP: Nobody else can change our grouplist. Thus we are
1324          *      safe.
1325          */
1326
1327         if (gidsetsize < 0)
1328                 return -EINVAL;
1329
1330         /* no need to grab task_lock here; it cannot change */
1331         get_group_info(current->group_info);
1332         i = current->group_info->ngroups;
1333         if (gidsetsize) {
1334                 if (i > gidsetsize) {
1335                         i = -EINVAL;
1336                         goto out;
1337                 }
1338                 if (groups_to_user(grouplist, current->group_info)) {
1339                         i = -EFAULT;
1340                         goto out;
1341                 }
1342         }
1343 out:
1344         put_group_info(current->group_info);
1345         return i;
1346 }
1347
1348 /*
1349  *      SMP: Our groups are copy-on-write. We can set them safely
1350  *      without another task interfering.
1351  */
1352
1353 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1354 {
1355         struct group_info *group_info;
1356         int retval;
1357
1358         if (!capable(CAP_SETGID))
1359                 return -EPERM;
1360         if ((unsigned)gidsetsize > NGROUPS_MAX)
1361                 return -EINVAL;
1362
1363         group_info = groups_alloc(gidsetsize);
1364         if (!group_info)
1365                 return -ENOMEM;
1366         retval = groups_from_user(group_info, grouplist);
1367         if (retval) {
1368                 put_group_info(group_info);
1369                 return retval;
1370         }
1371
1372         retval = set_current_groups(group_info);
1373         put_group_info(group_info);
1374
1375         return retval;
1376 }
1377
1378 /*
1379  * Check whether we're fsgid/egid or in the supplemental group..
1380  */
1381 int in_group_p(gid_t grp)
1382 {
1383         int retval = 1;
1384         if (grp != current->fsgid) {
1385                 get_group_info(current->group_info);
1386                 retval = groups_search(current->group_info, grp);
1387                 put_group_info(current->group_info);
1388         }
1389         return retval;
1390 }
1391
1392 EXPORT_SYMBOL(in_group_p);
1393
1394 int in_egroup_p(gid_t grp)
1395 {
1396         int retval = 1;
1397         if (grp != current->egid) {
1398                 get_group_info(current->group_info);
1399                 retval = groups_search(current->group_info, grp);
1400                 put_group_info(current->group_info);
1401         }
1402         return retval;
1403 }
1404
1405 EXPORT_SYMBOL(in_egroup_p);
1406
1407 DECLARE_RWSEM(uts_sem);
1408
1409 EXPORT_SYMBOL(uts_sem);
1410
1411 asmlinkage long sys_newuname(struct new_utsname __user * name)
1412 {
1413         int errno = 0;
1414
1415         down_read(&uts_sem);
1416         if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1417                 errno = -EFAULT;
1418         up_read(&uts_sem);
1419         return errno;
1420 }
1421
1422 asmlinkage long sys_sethostname(char __user *name, int len)
1423 {
1424         int errno;
1425         char tmp[__NEW_UTS_LEN];
1426
1427         if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1428                 return -EPERM;
1429         if (len < 0 || len > __NEW_UTS_LEN)
1430                 return -EINVAL;
1431         down_write(&uts_sem);
1432         errno = -EFAULT;
1433         if (!copy_from_user(tmp, name, len)) {
1434                 char *ptr = vx_new_uts(nodename);
1435
1436                 memcpy(ptr, tmp, len);
1437                 ptr[len] = 0;
1438                 errno = 0;
1439         }
1440         up_write(&uts_sem);
1441         return errno;
1442 }
1443
1444 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1445
1446 asmlinkage long sys_gethostname(char __user *name, int len)
1447 {
1448         int i, errno;
1449         char *ptr;
1450
1451         if (len < 0)
1452                 return -EINVAL;
1453         down_read(&uts_sem);
1454         ptr = vx_new_uts(nodename);
1455         i = 1 + strlen(ptr);
1456         if (i > len)
1457                 i = len;
1458         errno = 0;
1459         if (copy_to_user(name, ptr, i))
1460                 errno = -EFAULT;
1461         up_read(&uts_sem);
1462         return errno;
1463 }
1464
1465 #endif
1466
1467 /*
1468  * Only setdomainname; getdomainname can be implemented by calling
1469  * uname()
1470  */
1471 asmlinkage long sys_setdomainname(char __user *name, int len)
1472 {
1473         int errno;
1474         char tmp[__NEW_UTS_LEN];
1475
1476         if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1477                 return -EPERM;
1478         if (len < 0 || len > __NEW_UTS_LEN)
1479                 return -EINVAL;
1480
1481         down_write(&uts_sem);
1482         errno = -EFAULT;
1483         if (!copy_from_user(tmp, name, len)) {
1484                 char *ptr = vx_new_uts(domainname);
1485
1486                 memcpy(ptr, tmp, len);
1487                 ptr[len] = 0;
1488                 errno = 0;
1489         }
1490         up_write(&uts_sem);
1491         return errno;
1492 }
1493
1494 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1495 {
1496         if (resource >= RLIM_NLIMITS)
1497                 return -EINVAL;
1498         else {
1499                 struct rlimit value;
1500                 task_lock(current->group_leader);
1501                 value = current->signal->rlim[resource];
1502                 task_unlock(current->group_leader);
1503                 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1504         }
1505 }
1506
1507 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1508
1509 /*
1510  *      Back compatibility for getrlimit. Needed for some apps.
1511  */
1512
1513 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1514 {
1515         struct rlimit x;
1516         if (resource >= RLIM_NLIMITS)
1517                 return -EINVAL;
1518
1519         task_lock(current->group_leader);
1520         x = current->signal->rlim[resource];
1521         task_unlock(current->group_leader);
1522         if(x.rlim_cur > 0x7FFFFFFF)
1523                 x.rlim_cur = 0x7FFFFFFF;
1524         if(x.rlim_max > 0x7FFFFFFF)
1525                 x.rlim_max = 0x7FFFFFFF;
1526         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1527 }
1528
1529 #endif
1530
1531 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1532 {
1533         struct rlimit new_rlim, *old_rlim;
1534         int retval;
1535
1536         if (resource >= RLIM_NLIMITS)
1537                 return -EINVAL;
1538         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1539                 return -EFAULT;
1540        if (new_rlim.rlim_cur > new_rlim.rlim_max)
1541                return -EINVAL;
1542         old_rlim = current->signal->rlim + resource;
1543         if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1544             !capable(CAP_SYS_RESOURCE) && !vx_ccaps(VXC_SET_RLIMIT))
1545                 return -EPERM;
1546         if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1547                         return -EPERM;
1548
1549         retval = security_task_setrlimit(resource, &new_rlim);
1550         if (retval)
1551                 return retval;
1552
1553         task_lock(current->group_leader);
1554         *old_rlim = new_rlim;
1555         task_unlock(current->group_leader);
1556         return 0;
1557 }
1558
1559 /*
1560  * It would make sense to put struct rusage in the task_struct,
1561  * except that would make the task_struct be *really big*.  After
1562  * task_struct gets moved into malloc'ed memory, it would
1563  * make sense to do this.  It will make moving the rest of the information
1564  * a lot simpler!  (Which we're not doing right now because we're not
1565  * measuring them yet).
1566  *
1567  * This expects to be called with tasklist_lock read-locked or better,
1568  * and the siglock not locked.  It may momentarily take the siglock.
1569  *
1570  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1571  * races with threads incrementing their own counters.  But since word
1572  * reads are atomic, we either get new values or old values and we don't
1573  * care which for the sums.  We always take the siglock to protect reading
1574  * the c* fields from p->signal from races with exit.c updating those
1575  * fields when reaping, so a sample either gets all the additions of a
1576  * given child after it's reaped, or none so this sample is before reaping.
1577  */
1578
1579 void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1580 {
1581         struct task_struct *t;
1582         unsigned long flags;
1583         unsigned long utime, stime;
1584
1585         memset((char *) r, 0, sizeof *r);
1586
1587         if (unlikely(!p->signal))
1588                 return;
1589
1590         switch (who) {
1591                 case RUSAGE_CHILDREN:
1592                         spin_lock_irqsave(&p->sighand->siglock, flags);
1593                         utime = p->signal->cutime;
1594                         stime = p->signal->cstime;
1595                         r->ru_nvcsw = p->signal->cnvcsw;
1596                         r->ru_nivcsw = p->signal->cnivcsw;
1597                         r->ru_minflt = p->signal->cmin_flt;
1598                         r->ru_majflt = p->signal->cmaj_flt;
1599                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1600                         jiffies_to_timeval(utime, &r->ru_utime);
1601                         jiffies_to_timeval(stime, &r->ru_stime);
1602                         break;
1603                 case RUSAGE_SELF:
1604                         spin_lock_irqsave(&p->sighand->siglock, flags);
1605                         utime = stime = 0;
1606                         goto sum_group;
1607                 case RUSAGE_BOTH:
1608                         spin_lock_irqsave(&p->sighand->siglock, flags);
1609                         utime = p->signal->cutime;
1610                         stime = p->signal->cstime;
1611                         r->ru_nvcsw = p->signal->cnvcsw;
1612                         r->ru_nivcsw = p->signal->cnivcsw;
1613                         r->ru_minflt = p->signal->cmin_flt;
1614                         r->ru_majflt = p->signal->cmaj_flt;
1615                 sum_group:
1616                         utime += p->signal->utime;
1617                         stime += p->signal->stime;
1618                         r->ru_nvcsw += p->signal->nvcsw;
1619                         r->ru_nivcsw += p->signal->nivcsw;
1620                         r->ru_minflt += p->signal->min_flt;
1621                         r->ru_majflt += p->signal->maj_flt;
1622                         t = p;
1623                         do {
1624                                 utime += t->utime;
1625                                 stime += t->stime;
1626                                 r->ru_nvcsw += t->nvcsw;
1627                                 r->ru_nivcsw += t->nivcsw;
1628                                 r->ru_minflt += t->min_flt;
1629                                 r->ru_majflt += t->maj_flt;
1630                                 t = next_thread(t);
1631                         } while (t != p);
1632                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1633                         jiffies_to_timeval(utime, &r->ru_utime);
1634                         jiffies_to_timeval(stime, &r->ru_stime);
1635                         break;
1636                 default:
1637                         BUG();
1638         }
1639 }
1640
1641 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1642 {
1643         struct rusage r;
1644         read_lock(&tasklist_lock);
1645         k_getrusage(p, who, &r);
1646         read_unlock(&tasklist_lock);
1647         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1648 }
1649
1650 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1651 {
1652         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1653                 return -EINVAL;
1654         return getrusage(current, who, ru);
1655 }
1656
1657 asmlinkage long sys_umask(int mask)
1658 {
1659         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1660         return mask;
1661 }
1662
1663 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1664                           unsigned long arg4, unsigned long arg5)
1665 {
1666         long error;
1667         int sig;
1668
1669         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1670         if (error)
1671                 return error;
1672
1673         switch (option) {
1674                 case PR_SET_PDEATHSIG:
1675                         sig = arg2;
1676                         if (sig < 0 || sig > _NSIG) {
1677                                 error = -EINVAL;
1678                                 break;
1679                         }
1680                         current->pdeath_signal = sig;
1681                         break;
1682                 case PR_GET_PDEATHSIG:
1683                         error = put_user(current->pdeath_signal, (int __user *)arg2);
1684                         break;
1685                 case PR_GET_DUMPABLE:
1686                         if (current->mm->dumpable)
1687                                 error = 1;
1688                         break;
1689                 case PR_SET_DUMPABLE:
1690                         if (arg2 < 0 && arg2 > 2) {
1691                                 error = -EINVAL;
1692                                 break;
1693                         }
1694                         current->mm->dumpable = arg2;
1695                         break;
1696
1697                 case PR_SET_UNALIGN:
1698                         error = SET_UNALIGN_CTL(current, arg2);
1699                         break;
1700                 case PR_GET_UNALIGN:
1701                         error = GET_UNALIGN_CTL(current, arg2);
1702                         break;
1703                 case PR_SET_FPEMU:
1704                         error = SET_FPEMU_CTL(current, arg2);
1705                         break;
1706                 case PR_GET_FPEMU:
1707                         error = GET_FPEMU_CTL(current, arg2);
1708                         break;
1709                 case PR_SET_FPEXC:
1710                         error = SET_FPEXC_CTL(current, arg2);
1711                         break;
1712                 case PR_GET_FPEXC:
1713                         error = GET_FPEXC_CTL(current, arg2);
1714                         break;
1715                 case PR_GET_TIMING:
1716                         error = PR_TIMING_STATISTICAL;
1717                         break;
1718                 case PR_SET_TIMING:
1719                         if (arg2 == PR_TIMING_STATISTICAL)
1720                                 error = 0;
1721                         else
1722                                 error = -EINVAL;
1723                         break;
1724
1725                 case PR_GET_KEEPCAPS:
1726                         if (current->keep_capabilities)
1727                                 error = 1;
1728                         break;
1729                 case PR_SET_KEEPCAPS:
1730                         if (arg2 != 0 && arg2 != 1) {
1731                                 error = -EINVAL;
1732                                 break;
1733                         }
1734                         current->keep_capabilities = arg2;
1735                         break;
1736                 case PR_SET_NAME: {
1737                         struct task_struct *me = current;
1738                         unsigned char ncomm[sizeof(me->comm)];
1739
1740                         ncomm[sizeof(me->comm)-1] = 0;
1741                         if (strncpy_from_user(ncomm, (char __user *)arg2,
1742                                                 sizeof(me->comm)-1) < 0)
1743                                 return -EFAULT;
1744                         set_task_comm(me, ncomm);
1745                         return 0;
1746                 }
1747                 default:
1748                         error = -EINVAL;
1749                         break;
1750         }
1751         return error;
1752 }