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/reboot.h>
  15 #include <linux/prctl.h>
  16 #include <linux/init.h>
  17 #include <linux/highuid.h>
  18 #include <linux/fs.h>
  19 #include <linux/kernel.h>
  20 #include <linux/kexec.h>
  21 #include <linux/workqueue.h>
  22 #include <linux/device.h>
  23 #include <linux/key.h>
  24 #include <linux/times.h>
  25 #include <linux/posix-timers.h>
  26 #include <linux/security.h>
  27 #include <linux/dcookies.h>
  28 #include <linux/suspend.h>
  29 #include <linux/tty.h>
  30 #include <linux/signal.h>
  31
  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 static DEFINE_RWLOCK(notifier_lock);
  97
  98 /**
  99  *      notifier_chain_register - Add notifier to a notifier chain
 100  *      @list: Pointer to root list pointer
 101  *      @n: New entry in notifier chain
 102  *
 103  *      Adds a notifier to a notifier chain.
 104  *
 105  *      Currently always returns zero.
 106  */
 107
 108 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
 109 {
 110         write_lock(&notifier_lock);
 111         while(*list)
 112         {
 113                 if(n->priority > (*list)->priority)
 114                         break;
 115                 list= &((*list)->next);
 116         }
 117         n->next = *list;
 118         *list=n;
 119         write_unlock(&notifier_lock);
 120         return 0;
 121 }
 122
 123 EXPORT_SYMBOL(notifier_chain_register);
 124
 125 /**
 126  *      notifier_chain_unregister - Remove notifier from a notifier chain
 127  *      @nl: Pointer to root list pointer
 128  *      @n: New entry in notifier chain
 129  *
 130  *      Removes a notifier from a notifier chain.
 131  *
 132  *      Returns zero on success, or %-ENOENT on failure.
 133  */
 134
 135 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
 136 {
 137         write_lock(&notifier_lock);
 138         while((*nl)!=NULL)
 139         {
 140                 if((*nl)==n)
 141                 {
 142                         *nl=n->next;
 143                         write_unlock(&notifier_lock);
 144                         return 0;
 145                 }
 146                 nl=&((*nl)->next);
 147         }
 148         write_unlock(&notifier_lock);
 149         return -ENOENT;
 150 }
 151
 152 EXPORT_SYMBOL(notifier_chain_unregister);
 153
 154 /**
 155  *      notifier_call_chain - Call functions in a notifier chain
 156  *      @n: Pointer to root pointer of notifier chain
 157  *      @val: Value passed unmodified to notifier function
 158  *      @v: Pointer passed unmodified to notifier function
 159  *
 160  *      Calls each function in a notifier chain in turn.
 161  *
 162  *      If the return value of the notifier can be and'd
 163  *      with %NOTIFY_STOP_MASK, then notifier_call_chain
 164  *      will return immediately, with the return value of
 165  *      the notifier function which halted execution.
 166  *      Otherwise, the return value is the return value
 167  *      of the last notifier function called.
 168  */
 169
 170 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
 171 {
 172         int ret=NOTIFY_DONE;
 173         struct notifier_block *nb = *n;
 174
 175         while(nb)
 176         {
 177                 ret=nb->notifier_call(nb,val,v);
 178                 if(ret&NOTIFY_STOP_MASK)
 179                 {
 180                         return ret;
 181                 }
 182                 nb=nb->next;
 183         }
 184         return ret;
 185 }
 186
 187 EXPORT_SYMBOL(notifier_call_chain);
 188
 189 /**
 190  *      register_reboot_notifier - Register function to be called at reboot time
 191  *      @nb: Info about notifier function to be called
 192  *
 193  *      Registers a function with the list of functions
 194  *      to be called at reboot time.
 195  *
 196  *      Currently always returns zero, as notifier_chain_register
 197  *      always returns zero.
 198  */
 199
 200 int register_reboot_notifier(struct notifier_block * nb)
 201 {
 202         return notifier_chain_register(&reboot_notifier_list, nb);
 203 }
 204
 205 EXPORT_SYMBOL(register_reboot_notifier);
 206
 207 /**
 208  *      unregister_reboot_notifier - Unregister previously registered reboot notifier
 209  *      @nb: Hook to be unregistered
 210  *
 211  *      Unregisters a previously registered reboot
 212  *      notifier function.
 213  *
 214  *      Returns zero on success, or %-ENOENT on failure.
 215  */
 216
 217 int unregister_reboot_notifier(struct notifier_block * nb)
 218 {
 219         return notifier_chain_unregister(&reboot_notifier_list, nb);
 220 }
 221
 222 EXPORT_SYMBOL(unregister_reboot_notifier);
 223
 224 static int set_one_prio(struct task_struct *p, int niceval, int error)
 225 {
 226         int no_nice;
 227
 228         if (p->uid != current->euid &&
 229                 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
 230                 error = -EPERM;
 231                 goto out;
 232         }
 233         if (niceval < task_nice(p) && !can_nice(p, niceval)) {
 234                 error = -EACCES;
 235                 goto out;
 236         }
 237         no_nice = security_task_setnice(p, niceval);
 238         if (no_nice) {
 239                 error = no_nice;
 240                 goto out;
 241         }
 242         if (error == -ESRCH)
 243                 error = 0;
 244         set_user_nice(p, niceval);
 245 out:
 246         return error;
 247 }
 248
 249 asmlinkage long sys_setpriority(int which, int who, int niceval)
 250 {
 251         struct task_struct *g, *p;
 252         struct user_struct *user;
 253         int error = -EINVAL;
 254
 255         if (which > 2 || which < 0)
 256                 goto out;
 257
 258         /* normalize: avoid signed division (rounding problems) */
 259         error = -ESRCH;
 260         if (niceval < -20)
 261                 niceval = -20;
 262         if (niceval > 19)
 263                 niceval = 19;
 264
 265         read_lock(&tasklist_lock);
 266         switch (which) {
 267                 case PRIO_PROCESS:
 268                         if (!who)
 269                                 who = current->pid;
 270                         p = find_task_by_pid(who);
 271                         if (p)
 272                                 error = set_one_prio(p, niceval, error);
 273                         break;
 274                 case PRIO_PGRP:
 275                         if (!who)
 276                                 who = process_group(current);
 277                         do_each_task_pid(who, PIDTYPE_PGID, p) {
 278                                 error = set_one_prio(p, niceval, error);
 279                         } while_each_task_pid(who, PIDTYPE_PGID, p);
 280                         break;
 281                 case PRIO_USER:
 282                         user = current->user;
 283                         if (!who)
 284                                 who = current->uid;
 285                         else
 286                                 if ((who != current->uid) && !(user = find_user(who)))
 287                                         goto out_unlock;        /* No processes for this user */
 288
 289                         do_each_thread(g, p)
 290                                 if (p->uid == who)
 291                                         error = set_one_prio(p, niceval, error);
 292                         while_each_thread(g, p);
 293                         if (who != current->uid)
 294                                 free_uid(user);         /* For find_user() */
 295                         break;
 296         }
 297 out_unlock:
 298         read_unlock(&tasklist_lock);
 299 out:
 300         return error;
 301 }
 302
 303 /*
 304  * Ugh. To avoid negative return values, "getpriority()" will
 305  * not return the normal nice-value, but a negated value that
 306  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 307  * to stay compatible.
 308  */
 309 asmlinkage long sys_getpriority(int which, int who)
 310 {
 311         struct task_struct *g, *p;
 312         struct user_struct *user;
 313         long niceval, retval = -ESRCH;
 314
 315         if (which > 2 || which < 0)
 316                 return -EINVAL;
 317
 318         read_lock(&tasklist_lock);
 319         switch (which) {
 320                 case PRIO_PROCESS:
 321                         if (!who)
 322                                 who = current->pid;
 323                         p = find_task_by_pid(who);
 324                         if (p) {
 325                                 niceval = 20 - task_nice(p);
 326                                 if (niceval > retval)
 327                                         retval = niceval;
 328                         }
 329                         break;
 330                 case PRIO_PGRP:
 331                         if (!who)
 332                                 who = process_group(current);
 333                         do_each_task_pid(who, PIDTYPE_PGID, p) {
 334                                 niceval = 20 - task_nice(p);
 335                                 if (niceval > retval)
 336                                         retval = niceval;
 337                         } while_each_task_pid(who, PIDTYPE_PGID, p);
 338                         break;
 339                 case PRIO_USER:
 340                         user = current->user;
 341                         if (!who)
 342                                 who = current->uid;
 343                         else
 344                                 if ((who != current->uid) && !(user = find_user(who)))
 345                                         goto out_unlock;        /* No processes for this user */
 346
 347                         do_each_thread(g, p)
 348                                 if (p->uid == who) {
 349                                         niceval = 20 - task_nice(p);
 350                                         if (niceval > retval)
 351                                                 retval = niceval;
 352                                 }
 353                         while_each_thread(g, p);
 354                         if (who != current->uid)
 355                                 free_uid(user);         /* for find_user() */
 356                         break;
 357         }
 358 out_unlock:
 359         read_unlock(&tasklist_lock);
 360
 361         return retval;
 362 }
 363
 364 void kernel_restart(char *cmd)
 365 {
 366         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
 367         system_state = SYSTEM_RESTART;
 368         device_suspend(PMSG_FREEZE);
 369         device_shutdown();
 370         if (!cmd) {
 371                 printk(KERN_EMERG "Restarting system.\n");
 372         } else {
 373                 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
 374         }
 375         printk(".\n");
 376         machine_restart(cmd);
 377 }
 378 EXPORT_SYMBOL_GPL(kernel_restart);
 379
 380 void kernel_kexec(void)
 381 {
 382 #ifdef CONFIG_KEXEC
 383         struct kimage *image;
 384         image = xchg(&kexec_image, 0);
 385         if (!image) {
 386                 return;
 387         }
 388         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 389         system_state = SYSTEM_RESTART;
 390         device_suspend(PMSG_FREEZE);
 391         device_shutdown();
 392         printk(KERN_EMERG "Starting new kernel\n");
 393         machine_shutdown();
 394         machine_kexec(image);
 395 #endif
 396 }
 397 EXPORT_SYMBOL_GPL(kernel_kexec);
 398
 399 void kernel_halt(void)
 400 {
 401         notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
 402         system_state = SYSTEM_HALT;
 403         device_suspend(PMSG_SUSPEND);
 404         device_shutdown();
 405         printk(KERN_EMERG "System halted.\n");
 406         machine_halt();
 407 }
 408 EXPORT_SYMBOL_GPL(kernel_halt);
 409
 410 void kernel_power_off(void)
 411 {
 412         notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
 413         system_state = SYSTEM_POWER_OFF;
 414         device_suspend(PMSG_SUSPEND);
 415         device_shutdown();
 416         printk(KERN_EMERG "Power down.\n");
 417         machine_power_off();
 418 }
 419 EXPORT_SYMBOL_GPL(kernel_power_off);
 420
 421 /*
 422  * Reboot system call: for obvious reasons only root may call it,
 423  * and even root needs to set up some magic numbers in the registers
 424  * so that some mistake won't make this reboot the whole machine.
 425  * You can also set the meaning of the ctrl-alt-del-key here.
 426  *
 427  * reboot doesn't sync: do that yourself before calling this.
 428  */
 429 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
 430 {
 431         char buffer[256];
 432
 433         /* We only trust the superuser with rebooting the system. */
 434         if (!capable(CAP_SYS_BOOT))
 435                 return -EPERM;
 436
 437         /* For safety, we require "magic" arguments. */
 438         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 439             (magic2 != LINUX_REBOOT_MAGIC2 &&
 440                         magic2 != LINUX_REBOOT_MAGIC2A &&
 441                         magic2 != LINUX_REBOOT_MAGIC2B &&
 442                         magic2 != LINUX_REBOOT_MAGIC2C))
 443                 return -EINVAL;
 444
 445         lock_kernel();
 446         switch (cmd) {
 447         case LINUX_REBOOT_CMD_RESTART:
 448                 kernel_restart(NULL);
 449                 break;
 450
 451         case LINUX_REBOOT_CMD_CAD_ON:
 452                 C_A_D = 1;
 453                 break;
 454
 455         case LINUX_REBOOT_CMD_CAD_OFF:
 456                 C_A_D = 0;
 457                 break;
 458
 459         case LINUX_REBOOT_CMD_HALT:
 460                 kernel_halt();
 461                 unlock_kernel();
 462                 do_exit(0);
 463                 break;
 464
 465         case LINUX_REBOOT_CMD_POWER_OFF:
 466                 kernel_power_off();
 467                 unlock_kernel();
 468                 do_exit(0);
 469                 break;
 470
 471         case LINUX_REBOOT_CMD_RESTART2:
 472                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 473                         unlock_kernel();
 474                         return -EFAULT;
 475                 }
 476                 buffer[sizeof(buffer) - 1] = '\0';
 477
 478                 kernel_restart(buffer);
 479                 break;
 480
 481         case LINUX_REBOOT_CMD_KEXEC:
 482                 kernel_kexec();
 483                 unlock_kernel();
 484                 return -EINVAL;
 485
 486 #ifdef CONFIG_SOFTWARE_SUSPEND
 487         case LINUX_REBOOT_CMD_SW_SUSPEND:
 488                 {
 489                         int ret = software_suspend();
 490                         unlock_kernel();
 491                         return ret;
 492                 }
 493 #endif
 494
 495         default:
 496                 unlock_kernel();
 497                 return -EINVAL;
 498         }
 499         unlock_kernel();
 500         return 0;
 501 }
 502
 503 static void deferred_cad(void *dummy)
 504 {
 505         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 506         machine_restart(NULL);
 507 }
 508
 509 /*
 510  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 511  * As it's called within an interrupt, it may NOT sync: the only choice
 512  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 513  */
 514 void ctrl_alt_del(void)
 515 {
 516         static DECLARE_WORK(cad_work, deferred_cad, NULL);
 517
 518         if (C_A_D)
 519                 schedule_work(&cad_work);
 520         else
 521                 kill_proc(cad_pid, SIGINT, 1);
 522 }
 523
 524
 525 /*
 526  * Unprivileged users may change the real gid to the effective gid
 527  * or vice versa.  (BSD-style)
 528  *
 529  * If you set the real gid at all, or set the effective gid to a value not
 530  * equal to the real gid, then the saved gid is set to the new effective gid.
 531  *
 532  * This makes it possible for a setgid program to completely drop its
 533  * privileges, which is often a useful assertion to make when you are doing
 534  * a security audit over a program.
 535  *
 536  * The general idea is that a program which uses just setregid() will be
 537  * 100% compatible with BSD.  A program which uses just setgid() will be
 538  * 100% compatible with POSIX with saved IDs.
 539  *
 540  * SMP: There are not races, the GIDs are checked only by filesystem
 541  *      operations (as far as semantic preservation is concerned).
 542  */
 543 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 544 {
 545         int old_rgid = current->gid;
 546         int old_egid = current->egid;
 547         int new_rgid = old_rgid;
 548         int new_egid = old_egid;
 549         int retval;
 550
 551         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 552         if (retval)
 553                 return retval;
 554
 555         if (rgid != (gid_t) -1) {
 556                 if ((old_rgid == rgid) ||
 557                     (current->egid==rgid) ||
 558                     capable(CAP_SETGID))
 559                         new_rgid = rgid;
 560                 else
 561                         return -EPERM;
 562         }
 563         if (egid != (gid_t) -1) {
 564                 if ((old_rgid == egid) ||
 565                     (current->egid == egid) ||
 566                     (current->sgid == egid) ||
 567                     capable(CAP_SETGID))
 568                         new_egid = egid;
 569                 else {
 570                         return -EPERM;
 571                 }
 572         }
 573         if (new_egid != old_egid)
 574         {
 575                 current->mm->dumpable = suid_dumpable;
 576                 smp_wmb();
 577         }
 578         if (rgid != (gid_t) -1 ||
 579             (egid != (gid_t) -1 && egid != old_rgid))
 580                 current->sgid = new_egid;
 581         current->fsgid = new_egid;
 582         current->egid = new_egid;
 583         current->gid = new_rgid;
 584         key_fsgid_changed(current);
 585         return 0;
 586 }
 587
 588 /*
 589  * setgid() is implemented like SysV w/ SAVED_IDS
 590  *
 591  * SMP: Same implicit races as above.
 592  */
 593 asmlinkage long sys_setgid(gid_t gid)
 594 {
 595         int old_egid = current->egid;
 596         int retval;
 597
 598         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 599         if (retval)
 600                 return retval;
 601
 602         if (capable(CAP_SETGID))
 603         {
 604                 if(old_egid != gid)
 605                 {
 606                         current->mm->dumpable = suid_dumpable;
 607                         smp_wmb();
 608                 }
 609                 current->gid = current->egid = current->sgid = current->fsgid = gid;
 610         }
 611         else if ((gid == current->gid) || (gid == current->sgid))
 612         {
 613                 if(old_egid != gid)
 614                 {
 615                         current->mm->dumpable = suid_dumpable;
 616                         smp_wmb();
 617                 }
 618                 current->egid = current->fsgid = gid;
 619         }
 620         else
 621                 return -EPERM;
 622
 623         key_fsgid_changed(current);
 624         return 0;
 625 }
 626
 627 static int set_user(uid_t new_ruid, int dumpclear)
 628 {
 629         struct user_struct *new_user;
 630
 631         new_user = alloc_uid(new_ruid);
 632         if (!new_user)
 633                 return -EAGAIN;
 634
 635         if (atomic_read(&new_user->processes) >=
 636                                 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
 637                         new_user != &root_user) {
 638                 free_uid(new_user);
 639                 return -EAGAIN;
 640         }
 641
 642         switch_uid(new_user);
 643
 644         if(dumpclear)
 645         {
 646                 current->mm->dumpable = suid_dumpable;
 647                 smp_wmb();
 648         }
 649         current->uid = new_ruid;
 650         return 0;
 651 }
 652
 653 /*
 654  * Unprivileged users may change the real uid to the effective uid
 655  * or vice versa.  (BSD-style)
 656  *
 657  * If you set the real uid at all, or set the effective uid to a value not
 658  * equal to the real uid, then the saved uid is set to the new effective uid.
 659  *
 660  * This makes it possible for a setuid program to completely drop its
 661  * privileges, which is often a useful assertion to make when you are doing
 662  * a security audit over a program.
 663  *
 664  * The general idea is that a program which uses just setreuid() will be
 665  * 100% compatible with BSD.  A program which uses just setuid() will be
 666  * 100% compatible with POSIX with saved IDs.
 667  */
 668 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 669 {
 670         int old_ruid, old_euid, old_suid, new_ruid, new_euid;
 671         int retval;
 672
 673         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 674         if (retval)
 675                 return retval;
 676
 677         new_ruid = old_ruid = current->uid;
 678         new_euid = old_euid = current->euid;
 679         old_suid = current->suid;
 680
 681         if (ruid != (uid_t) -1) {
 682                 new_ruid = ruid;
 683                 if ((old_ruid != ruid) &&
 684                     (current->euid != ruid) &&
 685                     !capable(CAP_SETUID))
 686                         return -EPERM;
 687         }
 688
 689         if (euid != (uid_t) -1) {
 690                 new_euid = euid;
 691                 if ((old_ruid != euid) &&
 692                     (current->euid != euid) &&
 693                     (current->suid != euid) &&
 694                     !capable(CAP_SETUID))
 695                         return -EPERM;
 696         }
 697
 698         if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
 699                 return -EAGAIN;
 700
 701         if (new_euid != old_euid)
 702         {
 703                 current->mm->dumpable = suid_dumpable;
 704                 smp_wmb();
 705         }
 706         current->fsuid = current->euid = new_euid;
 707         if (ruid != (uid_t) -1 ||
 708             (euid != (uid_t) -1 && euid != old_ruid))
 709                 current->suid = current->euid;
 710         current->fsuid = current->euid;
 711
 712         key_fsuid_changed(current);
 713
 714         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
 715 }
 716
 717
 718
 719 /*
 720  * setuid() is implemented like SysV with SAVED_IDS
 721  *
 722  * Note that SAVED_ID's is deficient in that a setuid root program
 723  * like sendmail, for example, cannot set its uid to be a normal
 724  * user and then switch back, because if you're root, setuid() sets
 725  * the saved uid too.  If you don't like this, blame the bright people
 726  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 727  * will allow a root program to temporarily drop privileges and be able to
 728  * regain them by swapping the real and effective uid.
 729  */
 730 asmlinkage long sys_setuid(uid_t uid)
 731 {
 732         int old_euid = current->euid;
 733         int old_ruid, old_suid, new_ruid, new_suid;
 734         int retval;
 735
 736         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 737         if (retval)
 738                 return retval;
 739
 740         old_ruid = new_ruid = current->uid;
 741         old_suid = current->suid;
 742         new_suid = old_suid;
 743
 744         if (capable(CAP_SETUID)) {
 745                 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
 746                         return -EAGAIN;
 747                 new_suid = uid;
 748         } else if ((uid != current->uid) && (uid != new_suid))
 749                 return -EPERM;
 750
 751         if (old_euid != uid)
 752         {
 753                 current->mm->dumpable = suid_dumpable;
 754                 smp_wmb();
 755         }
 756         current->fsuid = current->euid = uid;
 757         current->suid = new_suid;
 758
 759         key_fsuid_changed(current);
 760
 761         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
 762 }
 763
 764
 765 /*
 766  * This function implements a generic ability to update ruid, euid,
 767  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 768  */
 769 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 770 {
 771         int old_ruid = current->uid;
 772         int old_euid = current->euid;
 773         int old_suid = current->suid;
 774         int retval;
 775
 776         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 777         if (retval)
 778                 return retval;
 779
 780         if (!capable(CAP_SETUID)) {
 781                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
 782                     (ruid != current->euid) && (ruid != current->suid))
 783                         return -EPERM;
 784                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
 785                     (euid != current->euid) && (euid != current->suid))
 786                         return -EPERM;
 787                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
 788                     (suid != current->euid) && (suid != current->suid))
 789                         return -EPERM;
 790         }
 791         if (ruid != (uid_t) -1) {
 792                 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
 793                         return -EAGAIN;
 794         }
 795         if (euid != (uid_t) -1) {
 796                 if (euid != current->euid)
 797                 {
 798                         current->mm->dumpable = suid_dumpable;
 799                         smp_wmb();
 800                 }
 801                 current->euid = euid;
 802         }
 803         current->fsuid = current->euid;
 804         if (suid != (uid_t) -1)
 805                 current->suid = suid;
 806
 807         key_fsuid_changed(current);
 808
 809         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
 810 }
 811
 812 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
 813 {
 814         int retval;
 815
 816         if (!(retval = put_user(current->uid, ruid)) &&
 817             !(retval = put_user(current->euid, euid)))
 818                 retval = put_user(current->suid, suid);
 819
 820         return retval;
 821 }
 822
 823 /*
 824  * Same as above, but for rgid, egid, sgid.
 825  */
 826 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 827 {
 828         int retval;
 829
 830         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 831         if (retval)
 832                 return retval;
 833
 834         if (!capable(CAP_SETGID)) {
 835                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
 836                     (rgid != current->egid) && (rgid != current->sgid))
 837                         return -EPERM;
 838                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
 839                     (egid != current->egid) && (egid != current->sgid))
 840                         return -EPERM;
 841                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
 842                     (sgid != current->egid) && (sgid != current->sgid))
 843                         return -EPERM;
 844         }
 845         if (egid != (gid_t) -1) {
 846                 if (egid != current->egid)
 847                 {
 848                         current->mm->dumpable = suid_dumpable;
 849                         smp_wmb();
 850                 }
 851                 current->egid = egid;
 852         }
 853         current->fsgid = current->egid;
 854         if (rgid != (gid_t) -1)
 855                 current->gid = rgid;
 856         if (sgid != (gid_t) -1)
 857                 current->sgid = sgid;
 858
 859         key_fsgid_changed(current);
 860         return 0;
 861 }
 862
 863 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
 864 {
 865         int retval;
 866
 867         if (!(retval = put_user(current->gid, rgid)) &&
 868             !(retval = put_user(current->egid, egid)))
 869                 retval = put_user(current->sgid, sgid);
 870
 871         return retval;
 872 }
 873
 874
 875 /*
 876  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 877  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 878  * whatever uid it wants to). It normally shadows "euid", except when
 879  * explicitly set by setfsuid() or for access..
 880  */
 881 asmlinkage long sys_setfsuid(uid_t uid)
 882 {
 883         int old_fsuid;
 884
 885         old_fsuid = current->fsuid;
 886         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
 887                 return old_fsuid;
 888
 889         if (uid == current->uid || uid == current->euid ||
 890             uid == current->suid || uid == current->fsuid ||
 891             capable(CAP_SETUID))
 892         {
 893                 if (uid != old_fsuid)
 894                 {
 895                         current->mm->dumpable = suid_dumpable;
 896                         smp_wmb();
 897                 }
 898                 current->fsuid = uid;
 899         }
 900
 901         key_fsuid_changed(current);
 902
 903         security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
 904
 905         return old_fsuid;
 906 }
 907
 908 /*
 909  * Samma på svenska..
 910  */
 911 asmlinkage long sys_setfsgid(gid_t gid)
 912 {
 913         int old_fsgid;
 914
 915         old_fsgid = current->fsgid;
 916         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 917                 return old_fsgid;
 918
 919         if (gid == current->gid || gid == current->egid ||
 920             gid == current->sgid || gid == current->fsgid ||
 921             capable(CAP_SETGID))
 922         {
 923                 if (gid != old_fsgid)
 924                 {
 925                         current->mm->dumpable = suid_dumpable;
 926                         smp_wmb();
 927                 }
 928                 current->fsgid = gid;
 929                 key_fsgid_changed(current);
 930         }
 931         return old_fsgid;
 932 }
 933
 934 asmlinkage long sys_times(struct tms __user * tbuf)
 935 {
 936         /*
 937          *      In the SMP world we might just be unlucky and have one of
 938          *      the times increment as we use it. Since the value is an
 939          *      atomically safe type this is just fine. Conceptually its
 940          *      as if the syscall took an instant longer to occur.
 941          */
 942         if (tbuf) {
 943                 struct tms tmp;
 944                 cputime_t utime, stime, cutime, cstime;
 945
 946 #ifdef CONFIG_SMP
 947                 if (thread_group_empty(current)) {
 948                         /*
 949                          * Single thread case without the use of any locks.
 950                          *
 951                          * We may race with release_task if two threads are
 952                          * executing. However, release task first adds up the
 953                          * counters (__exit_signal) before  removing the task
 954                          * from the process tasklist (__unhash_process).
 955                          * __exit_signal also acquires and releases the
 956                          * siglock which results in the proper memory ordering
 957                          * so that the list modifications are always visible
 958                          * after the counters have been updated.
 959                          *
 960                          * If the counters have been updated by the second thread
 961                          * but the thread has not yet been removed from the list
 962                          * then the other branch will be executing which will
 963                          * block on tasklist_lock until the exit handling of the
 964                          * other task is finished.
 965                          *
 966                          * This also implies that the sighand->siglock cannot
 967                          * be held by another processor. So we can also
 968                          * skip acquiring that lock.
 969                          */
 970                         utime = cputime_add(current->signal->utime, current->utime);
 971                         stime = cputime_add(current->signal->utime, current->stime);
 972                         cutime = current->signal->cutime;
 973                         cstime = current->signal->cstime;
 974                 } else
 975 #endif
 976                 {
 977
 978                         /* Process with multiple threads */
 979                         struct task_struct *tsk = current;
 980                         struct task_struct *t;
 981
 982                         read_lock(&tasklist_lock);
 983                         utime = tsk->signal->utime;
 984                         stime = tsk->signal->stime;
 985                         t = tsk;
 986                         do {
 987                                 utime = cputime_add(utime, t->utime);
 988                                 stime = cputime_add(stime, t->stime);
 989                                 t = next_thread(t);
 990                         } while (t != tsk);
 991
 992                         /*
 993                          * While we have tasklist_lock read-locked, no dying thread
 994                          * can be updating current->signal->[us]time.  Instead,
 995                          * we got their counts included in the live thread loop.
 996                          * However, another thread can come in right now and
 997                          * do a wait call that updates current->signal->c[us]time.
 998                          * To make sure we always see that pair updated atomically,
 999                          * we take the siglock around fetching them.
1000                          */
1001                         spin_lock_irq(&tsk->sighand->siglock);
1002                         cutime = tsk->signal->cutime;
1003                         cstime = tsk->signal->cstime;
1004                         spin_unlock_irq(&tsk->sighand->siglock);
1005                         read_unlock(&tasklist_lock);
1006                 }
1007                 tmp.tms_utime = cputime_to_clock_t(utime);
1008                 tmp.tms_stime = cputime_to_clock_t(stime);
1009                 tmp.tms_cutime = cputime_to_clock_t(cutime);
1010                 tmp.tms_cstime = cputime_to_clock_t(cstime);
1011                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1012                         return -EFAULT;
1013         }
1014         return (long) jiffies_64_to_clock_t(get_jiffies_64());
1015 }
1016
1017 /*
1018  * This needs some heavy checking ...
1019  * I just haven't the stomach for it. I also don't fully
1020  * understand sessions/pgrp etc. Let somebody who does explain it.
1021  *
1022  * OK, I think I have the protection semantics right.... this is really
1023  * only important on a multi-user system anyway, to make sure one user
1024  * can't send a signal to a process owned by another.  -TYT, 12/12/91
1025  *
1026  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1027  * LBT 04.03.94
1028  */
1029
1030 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1031 {
1032         struct task_struct *p;
1033         int err = -EINVAL;
1034
1035         if (!pid)
1036                 pid = current->pid;
1037         if (!pgid)
1038                 pgid = pid;
1039         if (pgid < 0)
1040                 return -EINVAL;
1041
1042         /* From this point forward we keep holding onto the tasklist lock
1043          * so that our parent does not change from under us. -DaveM
1044          */
1045         write_lock_irq(&tasklist_lock);
1046
1047         err = -ESRCH;
1048         p = find_task_by_pid(pid);
1049         if (!p)
1050                 goto out;
1051
1052         err = -EINVAL;
1053         if (!thread_group_leader(p))
1054                 goto out;
1055
1056         if (p->parent == current || p->real_parent == current) {
1057                 err = -EPERM;
1058                 if (p->signal->session != current->signal->session)
1059                         goto out;
1060                 err = -EACCES;
1061                 if (p->did_exec)
1062                         goto out;
1063         } else {
1064                 err = -ESRCH;
1065                 if (p != current)
1066                         goto out;
1067         }
1068
1069         err = -EPERM;
1070         if (p->signal->leader)
1071                 goto out;
1072
1073         if (pgid != pid) {
1074                 struct task_struct *p;
1075
1076                 do_each_task_pid(pgid, PIDTYPE_PGID, p) {
1077                         if (p->signal->session == current->signal->session)
1078                                 goto ok_pgid;
1079                 } while_each_task_pid(pgid, PIDTYPE_PGID, p);
1080                 goto out;
1081         }
1082
1083 ok_pgid:
1084         err = security_task_setpgid(p, pgid);
1085         if (err)
1086                 goto out;
1087
1088         if (process_group(p) != pgid) {
1089                 detach_pid(p, PIDTYPE_PGID);
1090                 p->signal->pgrp = pgid;
1091                 attach_pid(p, PIDTYPE_PGID, pgid);
1092         }
1093
1094         err = 0;
1095 out:
1096         /* All paths lead to here, thus we are safe. -DaveM */
1097         write_unlock_irq(&tasklist_lock);
1098         return err;
1099 }
1100
1101 asmlinkage long sys_getpgid(pid_t pid)
1102 {
1103         if (!pid) {
1104                 return process_group(current);
1105         } else {
1106                 int retval;
1107                 struct task_struct *p;
1108
1109                 read_lock(&tasklist_lock);
1110                 p = find_task_by_pid(pid);
1111
1112                 retval = -ESRCH;
1113                 if (p) {
1114                         retval = security_task_getpgid(p);
1115                         if (!retval)
1116                                 retval = process_group(p);
1117                 }
1118                 read_unlock(&tasklist_lock);
1119                 return retval;
1120         }
1121 }
1122
1123 #ifdef __ARCH_WANT_SYS_GETPGRP
1124
1125 asmlinkage long sys_getpgrp(void)
1126 {
1127         /* SMP - assuming writes are word atomic this is fine */
1128         return process_group(current);
1129 }
1130
1131 #endif
1132
1133 asmlinkage long sys_getsid(pid_t pid)
1134 {
1135         if (!pid) {
1136                 return current->signal->session;
1137         } else {
1138                 int retval;
1139                 struct task_struct *p;
1140
1141                 read_lock(&tasklist_lock);
1142                 p = find_task_by_pid(pid);
1143
1144                 retval = -ESRCH;
1145                 if(p) {
1146                         retval = security_task_getsid(p);
1147                         if (!retval)
1148                                 retval = p->signal->session;
1149                 }
1150                 read_unlock(&tasklist_lock);
1151                 return retval;
1152         }
1153 }
1154
1155 asmlinkage long sys_setsid(void)
1156 {
1157         struct pid *pid;
1158         int err = -EPERM;
1159
1160         if (!thread_group_leader(current))
1161                 return -EINVAL;
1162
1163         down(&tty_sem);
1164         write_lock_irq(&tasklist_lock);
1165
1166         pid = find_pid(PIDTYPE_PGID, current->pid);
1167         if (pid)
1168                 goto out;
1169
1170         current->signal->leader = 1;
1171         __set_special_pids(current->pid, current->pid);
1172         current->signal->tty = NULL;
1173         current->signal->tty_old_pgrp = 0;
1174         err = process_group(current);
1175 out:
1176         write_unlock_irq(&tasklist_lock);
1177         up(&tty_sem);
1178         return err;
1179 }
1180
1181 /*
1182  * Supplementary group IDs
1183  */
1184
1185 /* init to 2 - one for init_task, one to ensure it is never freed */
1186 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1187
1188 struct group_info *groups_alloc(int gidsetsize)
1189 {
1190         struct group_info *group_info;
1191         int nblocks;
1192         int i;
1193
1194         nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1195         /* Make sure we always allocate at least one indirect block pointer */
1196         nblocks = nblocks ? : 1;
1197         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1198         if (!group_info)
1199                 return NULL;
1200         group_info->ngroups = gidsetsize;
1201         group_info->nblocks = nblocks;
1202         atomic_set(&group_info->usage, 1);
1203
1204         if (gidsetsize <= NGROUPS_SMALL) {
1205                 group_info->blocks[0] = group_info->small_block;
1206         } else {
1207                 for (i = 0; i < nblocks; i++) {
1208                         gid_t *b;
1209                         b = (void *)__get_free_page(GFP_USER);
1210                         if (!b)
1211                                 goto out_undo_partial_alloc;
1212                         group_info->blocks[i] = b;
1213                 }
1214         }
1215         return group_info;
1216
1217 out_undo_partial_alloc:
1218         while (--i >= 0) {
1219                 free_page((unsigned long)group_info->blocks[i]);
1220         }
1221         kfree(group_info);
1222         return NULL;
1223 }
1224
1225 EXPORT_SYMBOL(groups_alloc);
1226
1227 void groups_free(struct group_info *group_info)
1228 {
1229         if (group_info->blocks[0] != group_info->small_block) {
1230                 int i;
1231                 for (i = 0; i < group_info->nblocks; i++)
1232                         free_page((unsigned long)group_info->blocks[i]);
1233         }
1234         kfree(group_info);
1235 }
1236
1237 EXPORT_SYMBOL(groups_free);
1238
1239 /* export the group_info to a user-space array */
1240 static int groups_to_user(gid_t __user *grouplist,
1241     struct group_info *group_info)
1242 {
1243         int i;
1244         int count = group_info->ngroups;
1245
1246         for (i = 0; i < group_info->nblocks; i++) {
1247                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1248                 int off = i * NGROUPS_PER_BLOCK;
1249                 int len = cp_count * sizeof(*grouplist);
1250
1251                 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1252                         return -EFAULT;
1253
1254                 count -= cp_count;
1255         }
1256         return 0;
1257 }
1258
1259 /* fill a group_info from a user-space array - it must be allocated already */
1260 static int groups_from_user(struct group_info *group_info,
1261     gid_t __user *grouplist)
1262  {
1263         int i;
1264         int count = group_info->ngroups;
1265
1266         for (i = 0; i < group_info->nblocks; i++) {
1267                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1268                 int off = i * NGROUPS_PER_BLOCK;
1269                 int len = cp_count * sizeof(*grouplist);
1270
1271                 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1272                         return -EFAULT;
1273
1274                 count -= cp_count;
1275         }
1276         return 0;
1277 }
1278
1279 /* a simple Shell sort */
1280 static void groups_sort(struct group_info *group_info)
1281 {
1282         int base, max, stride;
1283         int gidsetsize = group_info->ngroups;
1284
1285         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1286                 ; /* nothing */
1287         stride /= 3;
1288
1289         while (stride) {
1290                 max = gidsetsize - stride;
1291                 for (base = 0; base < max; base++) {
1292                         int left = base;
1293                         int right = left + stride;
1294                         gid_t tmp = GROUP_AT(group_info, right);
1295
1296                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1297                                 GROUP_AT(group_info, right) =
1298                                     GROUP_AT(group_info, left);
1299                                 right = left;
1300                                 left -= stride;
1301                         }
1302                         GROUP_AT(group_info, right) = tmp;
1303                 }
1304                 stride /= 3;
1305         }
1306 }
1307
1308 /* a simple bsearch */
1309 int groups_search(struct group_info *group_info, gid_t grp)
1310 {
1311         int left, right;
1312
1313         if (!group_info)
1314                 return 0;
1315
1316         left = 0;
1317         right = group_info->ngroups;
1318         while (left < right) {
1319                 int mid = (left+right)/2;
1320                 int cmp = grp - GROUP_AT(group_info, mid);
1321                 if (cmp > 0)
1322                         left = mid + 1;
1323                 else if (cmp < 0)
1324                         right = mid;
1325                 else
1326                         return 1;
1327         }
1328         return 0;
1329 }
1330
1331 /* validate and set current->group_info */
1332 int set_current_groups(struct group_info *group_info)
1333 {
1334         int retval;
1335         struct group_info *old_info;
1336
1337         retval = security_task_setgroups(group_info);
1338         if (retval)
1339                 return retval;
1340
1341         groups_sort(group_info);
1342         get_group_info(group_info);
1343
1344         task_lock(current);
1345         old_info = current->group_info;
1346         current->group_info = group_info;
1347         task_unlock(current);
1348
1349         put_group_info(old_info);
1350
1351         return 0;
1352 }
1353
1354 EXPORT_SYMBOL(set_current_groups);
1355
1356 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1357 {
1358         int i = 0;
1359
1360         /*
1361          *      SMP: Nobody else can change our grouplist. Thus we are
1362          *      safe.
1363          */
1364
1365         if (gidsetsize < 0)
1366                 return -EINVAL;
1367
1368         /* no need to grab task_lock here; it cannot change */
1369         get_group_info(current->group_info);
1370         i = current->group_info->ngroups;
1371         if (gidsetsize) {
1372                 if (i > gidsetsize) {
1373                         i = -EINVAL;
1374                         goto out;
1375                 }
1376                 if (groups_to_user(grouplist, current->group_info)) {
1377                         i = -EFAULT;
1378                         goto out;
1379                 }
1380         }
1381 out:
1382         put_group_info(current->group_info);
1383         return i;
1384 }
1385
1386 /*
1387  *      SMP: Our groups are copy-on-write. We can set them safely
1388  *      without another task interfering.
1389  */
1390
1391 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1392 {
1393         struct group_info *group_info;
1394         int retval;
1395
1396         if (!capable(CAP_SETGID))
1397                 return -EPERM;
1398         if ((unsigned)gidsetsize > NGROUPS_MAX)
1399                 return -EINVAL;
1400
1401         group_info = groups_alloc(gidsetsize);
1402         if (!group_info)
1403                 return -ENOMEM;
1404         retval = groups_from_user(group_info, grouplist);
1405         if (retval) {
1406                 put_group_info(group_info);
1407                 return retval;
1408         }
1409
1410         retval = set_current_groups(group_info);
1411         put_group_info(group_info);
1412
1413         return retval;
1414 }
1415
1416 /*
1417  * Check whether we're fsgid/egid or in the supplemental group..
1418  */
1419 int in_group_p(gid_t grp)
1420 {
1421         int retval = 1;
1422         if (grp != current->fsgid) {
1423                 get_group_info(current->group_info);
1424                 retval = groups_search(current->group_info, grp);
1425                 put_group_info(current->group_info);
1426         }
1427         return retval;
1428 }
1429
1430 EXPORT_SYMBOL(in_group_p);
1431
1432 int in_egroup_p(gid_t grp)
1433 {
1434         int retval = 1;
1435         if (grp != current->egid) {
1436                 get_group_info(current->group_info);
1437                 retval = groups_search(current->group_info, grp);
1438                 put_group_info(current->group_info);
1439         }
1440         return retval;
1441 }
1442
1443 EXPORT_SYMBOL(in_egroup_p);
1444
1445 DECLARE_RWSEM(uts_sem);
1446
1447 EXPORT_SYMBOL(uts_sem);
1448
1449 asmlinkage long sys_newuname(struct new_utsname __user * name)
1450 {
1451         int errno = 0;
1452
1453         down_read(&uts_sem);
1454         if (copy_to_user(name,&system_utsname,sizeof *name))
1455                 errno = -EFAULT;
1456         up_read(&uts_sem);
1457         return errno;
1458 }
1459
1460 asmlinkage long sys_sethostname(char __user *name, int len)
1461 {
1462         int errno;
1463         char tmp[__NEW_UTS_LEN];
1464
1465         if (!capable(CAP_SYS_ADMIN))
1466                 return -EPERM;
1467         if (len < 0 || len > __NEW_UTS_LEN)
1468                 return -EINVAL;
1469         down_write(&uts_sem);
1470         errno = -EFAULT;
1471         if (!copy_from_user(tmp, name, len)) {
1472                 memcpy(system_utsname.nodename, tmp, len);
1473                 system_utsname.nodename[len] = 0;
1474                 errno = 0;
1475         }
1476         up_write(&uts_sem);
1477         return errno;
1478 }
1479
1480 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1481
1482 asmlinkage long sys_gethostname(char __user *name, int len)
1483 {
1484         int i, errno;
1485
1486         if (len < 0)
1487                 return -EINVAL;
1488         down_read(&uts_sem);
1489         i = 1 + strlen(system_utsname.nodename);
1490         if (i > len)
1491                 i = len;
1492         errno = 0;
1493         if (copy_to_user(name, system_utsname.nodename, i))
1494                 errno = -EFAULT;
1495         up_read(&uts_sem);
1496         return errno;
1497 }
1498
1499 #endif
1500
1501 /*
1502  * Only setdomainname; getdomainname can be implemented by calling
1503  * uname()
1504  */
1505 asmlinkage long sys_setdomainname(char __user *name, int len)
1506 {
1507         int errno;
1508         char tmp[__NEW_UTS_LEN];
1509
1510         if (!capable(CAP_SYS_ADMIN))
1511                 return -EPERM;
1512         if (len < 0 || len > __NEW_UTS_LEN)
1513                 return -EINVAL;
1514
1515         down_write(&uts_sem);
1516         errno = -EFAULT;
1517         if (!copy_from_user(tmp, name, len)) {
1518                 memcpy(system_utsname.domainname, tmp, len);
1519                 system_utsname.domainname[len] = 0;
1520                 errno = 0;
1521         }
1522         up_write(&uts_sem);
1523         return errno;
1524 }
1525
1526 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1527 {
1528         if (resource >= RLIM_NLIMITS)
1529                 return -EINVAL;
1530         else {
1531                 struct rlimit value;
1532                 task_lock(current->group_leader);
1533                 value = current->signal->rlim[resource];
1534                 task_unlock(current->group_leader);
1535                 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1536         }
1537 }
1538
1539 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1540
1541 /*
1542  *      Back compatibility for getrlimit. Needed for some apps.
1543  */
1544
1545 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1546 {
1547         struct rlimit x;
1548         if (resource >= RLIM_NLIMITS)
1549                 return -EINVAL;
1550
1551         task_lock(current->group_leader);
1552         x = current->signal->rlim[resource];
1553         task_unlock(current->group_leader);
1554         if(x.rlim_cur > 0x7FFFFFFF)
1555                 x.rlim_cur = 0x7FFFFFFF;
1556         if(x.rlim_max > 0x7FFFFFFF)
1557                 x.rlim_max = 0x7FFFFFFF;
1558         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1559 }
1560
1561 #endif
1562
1563 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1564 {
1565         struct rlimit new_rlim, *old_rlim;
1566         int retval;
1567
1568         if (resource >= RLIM_NLIMITS)
1569                 return -EINVAL;
1570         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1571                 return -EFAULT;
1572        if (new_rlim.rlim_cur > new_rlim.rlim_max)
1573                return -EINVAL;
1574         old_rlim = current->signal->rlim + resource;
1575         if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1576             !capable(CAP_SYS_RESOURCE))
1577                 return -EPERM;
1578         if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1579                         return -EPERM;
1580
1581         retval = security_task_setrlimit(resource, &new_rlim);
1582         if (retval)
1583                 return retval;
1584
1585         task_lock(current->group_leader);
1586         *old_rlim = new_rlim;
1587         task_unlock(current->group_leader);
1588
1589         if (resource == RLIMIT_CPU && new_rlim.rlim_cur != RLIM_INFINITY &&
1590             (cputime_eq(current->signal->it_prof_expires, cputime_zero) ||
1591              new_rlim.rlim_cur <= cputime_to_secs(
1592                      current->signal->it_prof_expires))) {
1593                 cputime_t cputime = secs_to_cputime(new_rlim.rlim_cur);
1594                 read_lock(&tasklist_lock);
1595                 spin_lock_irq(&current->sighand->siglock);
1596                 set_process_cpu_timer(current, CPUCLOCK_PROF,
1597                                       &cputime, NULL);
1598                 spin_unlock_irq(&current->sighand->siglock);
1599                 read_unlock(&tasklist_lock);
1600         }
1601
1602         return 0;
1603 }
1604
1605 /*
1606  * It would make sense to put struct rusage in the task_struct,
1607  * except that would make the task_struct be *really big*.  After
1608  * task_struct gets moved into malloc'ed memory, it would
1609  * make sense to do this.  It will make moving the rest of the information
1610  * a lot simpler!  (Which we're not doing right now because we're not
1611  * measuring them yet).
1612  *
1613  * This expects to be called with tasklist_lock read-locked or better,
1614  * and the siglock not locked.  It may momentarily take the siglock.
1615  *
1616  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1617  * races with threads incrementing their own counters.  But since word
1618  * reads are atomic, we either get new values or old values and we don't
1619  * care which for the sums.  We always take the siglock to protect reading
1620  * the c* fields from p->signal from races with exit.c updating those
1621  * fields when reaping, so a sample either gets all the additions of a
1622  * given child after it's reaped, or none so this sample is before reaping.
1623  */
1624
1625 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1626 {
1627         struct task_struct *t;
1628         unsigned long flags;
1629         cputime_t utime, stime;
1630
1631         memset((char *) r, 0, sizeof *r);
1632
1633         if (unlikely(!p->signal))
1634                 return;
1635
1636         switch (who) {
1637                 case RUSAGE_CHILDREN:
1638                         spin_lock_irqsave(&p->sighand->siglock, flags);
1639                         utime = p->signal->cutime;
1640                         stime = p->signal->cstime;
1641                         r->ru_nvcsw = p->signal->cnvcsw;
1642                         r->ru_nivcsw = p->signal->cnivcsw;
1643                         r->ru_minflt = p->signal->cmin_flt;
1644                         r->ru_majflt = p->signal->cmaj_flt;
1645                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1646                         cputime_to_timeval(utime, &r->ru_utime);
1647                         cputime_to_timeval(stime, &r->ru_stime);
1648                         break;
1649                 case RUSAGE_SELF:
1650                         spin_lock_irqsave(&p->sighand->siglock, flags);
1651                         utime = stime = cputime_zero;
1652                         goto sum_group;
1653                 case RUSAGE_BOTH:
1654                         spin_lock_irqsave(&p->sighand->siglock, flags);
1655                         utime = p->signal->cutime;
1656                         stime = p->signal->cstime;
1657                         r->ru_nvcsw = p->signal->cnvcsw;
1658                         r->ru_nivcsw = p->signal->cnivcsw;
1659                         r->ru_minflt = p->signal->cmin_flt;
1660                         r->ru_majflt = p->signal->cmaj_flt;
1661                 sum_group:
1662                         utime = cputime_add(utime, p->signal->utime);
1663                         stime = cputime_add(stime, p->signal->stime);
1664                         r->ru_nvcsw += p->signal->nvcsw;
1665                         r->ru_nivcsw += p->signal->nivcsw;
1666                         r->ru_minflt += p->signal->min_flt;
1667                         r->ru_majflt += p->signal->maj_flt;
1668                         t = p;
1669                         do {
1670                                 utime = cputime_add(utime, t->utime);
1671                                 stime = cputime_add(stime, t->stime);
1672                                 r->ru_nvcsw += t->nvcsw;
1673                                 r->ru_nivcsw += t->nivcsw;
1674                                 r->ru_minflt += t->min_flt;
1675                                 r->ru_majflt += t->maj_flt;
1676                                 t = next_thread(t);
1677                         } while (t != p);
1678                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1679                         cputime_to_timeval(utime, &r->ru_utime);
1680                         cputime_to_timeval(stime, &r->ru_stime);
1681                         break;
1682                 default:
1683                         BUG();
1684         }
1685 }
1686
1687 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1688 {
1689         struct rusage r;
1690         read_lock(&tasklist_lock);
1691         k_getrusage(p, who, &r);
1692         read_unlock(&tasklist_lock);
1693         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1694 }
1695
1696 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1697 {
1698         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1699                 return -EINVAL;
1700         return getrusage(current, who, ru);
1701 }
1702
1703 asmlinkage long sys_umask(int mask)
1704 {
1705         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1706         return mask;
1707 }
1708
1709 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1710                           unsigned long arg4, unsigned long arg5)
1711 {
1712         long error;
1713         int sig;
1714
1715         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1716         if (error)
1717                 return error;
1718
1719         switch (option) {
1720                 case PR_SET_PDEATHSIG:
1721                         sig = arg2;
1722                         if (!valid_signal(sig)) {
1723                                 error = -EINVAL;
1724                                 break;
1725                         }
1726                         current->pdeath_signal = sig;
1727                         break;
1728                 case PR_GET_PDEATHSIG:
1729                         error = put_user(current->pdeath_signal, (int __user *)arg2);
1730                         break;
1731                 case PR_GET_DUMPABLE:
1732                         if (current->mm->dumpable)
1733                                 error = 1;
1734                         break;
1735                 case PR_SET_DUMPABLE:
1736                         if (arg2 < 0 || arg2 > 2) {
1737                                 error = -EINVAL;
1738                                 break;
1739                         }
1740                         current->mm->dumpable = arg2;
1741                         break;
1742
1743                 case PR_SET_UNALIGN:
1744                         error = SET_UNALIGN_CTL(current, arg2);
1745                         break;
1746                 case PR_GET_UNALIGN:
1747                         error = GET_UNALIGN_CTL(current, arg2);
1748                         break;
1749                 case PR_SET_FPEMU:
1750                         error = SET_FPEMU_CTL(current, arg2);
1751                         break;
1752                 case PR_GET_FPEMU:
1753                         error = GET_FPEMU_CTL(current, arg2);
1754                         break;
1755                 case PR_SET_FPEXC:
1756                         error = SET_FPEXC_CTL(current, arg2);
1757                         break;
1758                 case PR_GET_FPEXC:
1759                         error = GET_FPEXC_CTL(current, arg2);
1760                         break;
1761                 case PR_GET_TIMING:
1762                         error = PR_TIMING_STATISTICAL;
1763                         break;
1764                 case PR_SET_TIMING:
1765                         if (arg2 == PR_TIMING_STATISTICAL)
1766                                 error = 0;
1767                         else
1768                                 error = -EINVAL;
1769                         break;
1770
1771                 case PR_GET_KEEPCAPS:
1772                         if (current->keep_capabilities)
1773                                 error = 1;
1774                         break;
1775                 case PR_SET_KEEPCAPS:
1776                         if (arg2 != 0 && arg2 != 1) {
1777                                 error = -EINVAL;
1778                                 break;
1779                         }
1780                         current->keep_capabilities = arg2;
1781                         break;
1782                 case PR_SET_NAME: {
1783                         struct task_struct *me = current;
1784                         unsigned char ncomm[sizeof(me->comm)];
1785
1786                         ncomm[sizeof(me->comm)-1] = 0;
1787                         if (strncpy_from_user(ncomm, (char __user *)arg2,
1788                                                 sizeof(me->comm)-1) < 0)
1789                                 return -EFAULT;
1790                         set_task_comm(me, ncomm);
1791                         return 0;
1792                 }
1793                 case PR_GET_NAME: {
1794                         struct task_struct *me = current;
1795                         unsigned char tcomm[sizeof(me->comm)];
1796
1797                         get_task_comm(tcomm, me);
1798                         if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
1799                                 return -EFAULT;
1800                         return 0;
1801                 }
1802                 default:
1803                         error = -EINVAL;
1804                         break;
1805         }
1806         return error;
1807 }