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 emergency_restart(void)
 365 {
 366         machine_emergency_restart();
 367 }
 368 EXPORT_SYMBOL_GPL(emergency_restart);
 369
 370 void kernel_restart(char *cmd)
 371 {
 372         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
 373         system_state = SYSTEM_RESTART;
 374         device_suspend(PMSG_FREEZE);
 375         device_shutdown();
 376         if (!cmd) {
 377                 printk(KERN_EMERG "Restarting system.\n");
 378         } else {
 379                 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
 380         }
 381         printk(".\n");
 382         machine_restart(cmd);
 383 }
 384 EXPORT_SYMBOL_GPL(kernel_restart);
 385
 386 void kernel_kexec(void)
 387 {
 388 #ifdef CONFIG_KEXEC
 389         struct kimage *image;
 390         image = xchg(&kexec_image, 0);
 391         if (!image) {
 392                 return;
 393         }
 394         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 395         system_state = SYSTEM_RESTART;
 396         device_suspend(PMSG_FREEZE);
 397         device_shutdown();
 398         printk(KERN_EMERG "Starting new kernel\n");
 399         machine_shutdown();
 400         machine_kexec(image);
 401 #endif
 402 }
 403 EXPORT_SYMBOL_GPL(kernel_kexec);
 404
 405 void kernel_halt(void)
 406 {
 407         notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
 408         system_state = SYSTEM_HALT;
 409         device_suspend(PMSG_SUSPEND);
 410         device_shutdown();
 411         printk(KERN_EMERG "System halted.\n");
 412         machine_halt();
 413 }
 414 EXPORT_SYMBOL_GPL(kernel_halt);
 415
 416 void kernel_power_off(void)
 417 {
 418         notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
 419         system_state = SYSTEM_POWER_OFF;
 420         device_suspend(PMSG_SUSPEND);
 421         device_shutdown();
 422         printk(KERN_EMERG "Power down.\n");
 423         machine_power_off();
 424 }
 425 EXPORT_SYMBOL_GPL(kernel_power_off);
 426
 427 /*
 428  * Reboot system call: for obvious reasons only root may call it,
 429  * and even root needs to set up some magic numbers in the registers
 430  * so that some mistake won't make this reboot the whole machine.
 431  * You can also set the meaning of the ctrl-alt-del-key here.
 432  *
 433  * reboot doesn't sync: do that yourself before calling this.
 434  */
 435 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
 436 {
 437         char buffer[256];
 438
 439         /* We only trust the superuser with rebooting the system. */
 440         if (!capable(CAP_SYS_BOOT))
 441                 return -EPERM;
 442
 443         /* For safety, we require "magic" arguments. */
 444         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 445             (magic2 != LINUX_REBOOT_MAGIC2 &&
 446                         magic2 != LINUX_REBOOT_MAGIC2A &&
 447                         magic2 != LINUX_REBOOT_MAGIC2B &&
 448                         magic2 != LINUX_REBOOT_MAGIC2C))
 449                 return -EINVAL;
 450
 451         lock_kernel();
 452         switch (cmd) {
 453         case LINUX_REBOOT_CMD_RESTART:
 454                 kernel_restart(NULL);
 455                 break;
 456
 457         case LINUX_REBOOT_CMD_CAD_ON:
 458                 C_A_D = 1;
 459                 break;
 460
 461         case LINUX_REBOOT_CMD_CAD_OFF:
 462                 C_A_D = 0;
 463                 break;
 464
 465         case LINUX_REBOOT_CMD_HALT:
 466                 kernel_halt();
 467                 unlock_kernel();
 468                 do_exit(0);
 469                 break;
 470
 471         case LINUX_REBOOT_CMD_POWER_OFF:
 472                 kernel_power_off();
 473                 unlock_kernel();
 474                 do_exit(0);
 475                 break;
 476
 477         case LINUX_REBOOT_CMD_RESTART2:
 478                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 479                         unlock_kernel();
 480                         return -EFAULT;
 481                 }
 482                 buffer[sizeof(buffer) - 1] = '\0';
 483
 484                 kernel_restart(buffer);
 485                 break;
 486
 487         case LINUX_REBOOT_CMD_KEXEC:
 488                 kernel_kexec();
 489                 unlock_kernel();
 490                 return -EINVAL;
 491
 492 #ifdef CONFIG_SOFTWARE_SUSPEND
 493         case LINUX_REBOOT_CMD_SW_SUSPEND:
 494                 {
 495                         int ret = software_suspend();
 496                         unlock_kernel();
 497                         return ret;
 498                 }
 499 #endif
 500
 501         default:
 502                 unlock_kernel();
 503                 return -EINVAL;
 504         }
 505         unlock_kernel();
 506         return 0;
 507 }
 508
 509 static void deferred_cad(void *dummy)
 510 {
 511         kernel_restart(NULL);
 512 }
 513
 514 /*
 515  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 516  * As it's called within an interrupt, it may NOT sync: the only choice
 517  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 518  */
 519 void ctrl_alt_del(void)
 520 {
 521         static DECLARE_WORK(cad_work, deferred_cad, NULL);
 522
 523         if (C_A_D)
 524                 schedule_work(&cad_work);
 525         else
 526                 kill_proc(cad_pid, SIGINT, 1);
 527 }
 528
 529
 530 /*
 531  * Unprivileged users may change the real gid to the effective gid
 532  * or vice versa.  (BSD-style)
 533  *
 534  * If you set the real gid at all, or set the effective gid to a value not
 535  * equal to the real gid, then the saved gid is set to the new effective gid.
 536  *
 537  * This makes it possible for a setgid program to completely drop its
 538  * privileges, which is often a useful assertion to make when you are doing
 539  * a security audit over a program.
 540  *
 541  * The general idea is that a program which uses just setregid() will be
 542  * 100% compatible with BSD.  A program which uses just setgid() will be
 543  * 100% compatible with POSIX with saved IDs.
 544  *
 545  * SMP: There are not races, the GIDs are checked only by filesystem
 546  *      operations (as far as semantic preservation is concerned).
 547  */
 548 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 549 {
 550         int old_rgid = current->gid;
 551         int old_egid = current->egid;
 552         int new_rgid = old_rgid;
 553         int new_egid = old_egid;
 554         int retval;
 555
 556         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 557         if (retval)
 558                 return retval;
 559
 560         if (rgid != (gid_t) -1) {
 561                 if ((old_rgid == rgid) ||
 562                     (current->egid==rgid) ||
 563                     capable(CAP_SETGID))
 564                         new_rgid = rgid;
 565                 else
 566                         return -EPERM;
 567         }
 568         if (egid != (gid_t) -1) {
 569                 if ((old_rgid == egid) ||
 570                     (current->egid == egid) ||
 571                     (current->sgid == egid) ||
 572                     capable(CAP_SETGID))
 573                         new_egid = egid;
 574                 else {
 575                         return -EPERM;
 576                 }
 577         }
 578         if (new_egid != old_egid)
 579         {
 580                 current->mm->dumpable = suid_dumpable;
 581                 smp_wmb();
 582         }
 583         if (rgid != (gid_t) -1 ||
 584             (egid != (gid_t) -1 && egid != old_rgid))
 585                 current->sgid = new_egid;
 586         current->fsgid = new_egid;
 587         current->egid = new_egid;
 588         current->gid = new_rgid;
 589         key_fsgid_changed(current);
 590         return 0;
 591 }
 592
 593 /*
 594  * setgid() is implemented like SysV w/ SAVED_IDS
 595  *
 596  * SMP: Same implicit races as above.
 597  */
 598 asmlinkage long sys_setgid(gid_t gid)
 599 {
 600         int old_egid = current->egid;
 601         int retval;
 602
 603         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 604         if (retval)
 605                 return retval;
 606
 607         if (capable(CAP_SETGID))
 608         {
 609                 if(old_egid != gid)
 610                 {
 611                         current->mm->dumpable = suid_dumpable;
 612                         smp_wmb();
 613                 }
 614                 current->gid = current->egid = current->sgid = current->fsgid = gid;
 615         }
 616         else if ((gid == current->gid) || (gid == current->sgid))
 617         {
 618                 if(old_egid != gid)
 619                 {
 620                         current->mm->dumpable = suid_dumpable;
 621                         smp_wmb();
 622                 }
 623                 current->egid = current->fsgid = gid;
 624         }
 625         else
 626                 return -EPERM;
 627
 628         key_fsgid_changed(current);
 629         return 0;
 630 }
 631
 632 static int set_user(uid_t new_ruid, int dumpclear)
 633 {
 634         struct user_struct *new_user;
 635
 636         new_user = alloc_uid(new_ruid);
 637         if (!new_user)
 638                 return -EAGAIN;
 639
 640         if (atomic_read(&new_user->processes) >=
 641                                 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
 642                         new_user != &root_user) {
 643                 free_uid(new_user);
 644                 return -EAGAIN;
 645         }
 646
 647         switch_uid(new_user);
 648
 649         if(dumpclear)
 650         {
 651                 current->mm->dumpable = suid_dumpable;
 652                 smp_wmb();
 653         }
 654         current->uid = new_ruid;
 655         return 0;
 656 }
 657
 658 /*
 659  * Unprivileged users may change the real uid to the effective uid
 660  * or vice versa.  (BSD-style)
 661  *
 662  * If you set the real uid at all, or set the effective uid to a value not
 663  * equal to the real uid, then the saved uid is set to the new effective uid.
 664  *
 665  * This makes it possible for a setuid program to completely drop its
 666  * privileges, which is often a useful assertion to make when you are doing
 667  * a security audit over a program.
 668  *
 669  * The general idea is that a program which uses just setreuid() will be
 670  * 100% compatible with BSD.  A program which uses just setuid() will be
 671  * 100% compatible with POSIX with saved IDs.
 672  */
 673 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 674 {
 675         int old_ruid, old_euid, old_suid, new_ruid, new_euid;
 676         int retval;
 677
 678         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 679         if (retval)
 680                 return retval;
 681
 682         new_ruid = old_ruid = current->uid;
 683         new_euid = old_euid = current->euid;
 684         old_suid = current->suid;
 685
 686         if (ruid != (uid_t) -1) {
 687                 new_ruid = ruid;
 688                 if ((old_ruid != ruid) &&
 689                     (current->euid != ruid) &&
 690                     !capable(CAP_SETUID))
 691                         return -EPERM;
 692         }
 693
 694         if (euid != (uid_t) -1) {
 695                 new_euid = euid;
 696                 if ((old_ruid != euid) &&
 697                     (current->euid != euid) &&
 698                     (current->suid != euid) &&
 699                     !capable(CAP_SETUID))
 700                         return -EPERM;
 701         }
 702
 703         if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
 704                 return -EAGAIN;
 705
 706         if (new_euid != old_euid)
 707         {
 708                 current->mm->dumpable = suid_dumpable;
 709                 smp_wmb();
 710         }
 711         current->fsuid = current->euid = new_euid;
 712         if (ruid != (uid_t) -1 ||
 713             (euid != (uid_t) -1 && euid != old_ruid))
 714                 current->suid = current->euid;
 715         current->fsuid = current->euid;
 716
 717         key_fsuid_changed(current);
 718
 719         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
 720 }
 721
 722
 723
 724 /*
 725  * setuid() is implemented like SysV with SAVED_IDS
 726  *
 727  * Note that SAVED_ID's is deficient in that a setuid root program
 728  * like sendmail, for example, cannot set its uid to be a normal
 729  * user and then switch back, because if you're root, setuid() sets
 730  * the saved uid too.  If you don't like this, blame the bright people
 731  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 732  * will allow a root program to temporarily drop privileges and be able to
 733  * regain them by swapping the real and effective uid.
 734  */
 735 asmlinkage long sys_setuid(uid_t uid)
 736 {
 737         int old_euid = current->euid;
 738         int old_ruid, old_suid, new_ruid, new_suid;
 739         int retval;
 740
 741         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 742         if (retval)
 743                 return retval;
 744
 745         old_ruid = new_ruid = current->uid;
 746         old_suid = current->suid;
 747         new_suid = old_suid;
 748
 749         if (capable(CAP_SETUID)) {
 750                 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
 751                         return -EAGAIN;
 752                 new_suid = uid;
 753         } else if ((uid != current->uid) && (uid != new_suid))
 754                 return -EPERM;
 755
 756         if (old_euid != uid)
 757         {
 758                 current->mm->dumpable = suid_dumpable;
 759                 smp_wmb();
 760         }
 761         current->fsuid = current->euid = uid;
 762         current->suid = new_suid;
 763
 764         key_fsuid_changed(current);
 765
 766         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
 767 }
 768
 769
 770 /*
 771  * This function implements a generic ability to update ruid, euid,
 772  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 773  */
 774 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 775 {
 776         int old_ruid = current->uid;
 777         int old_euid = current->euid;
 778         int old_suid = current->suid;
 779         int retval;
 780
 781         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 782         if (retval)
 783                 return retval;
 784
 785         if (!capable(CAP_SETUID)) {
 786                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
 787                     (ruid != current->euid) && (ruid != current->suid))
 788                         return -EPERM;
 789                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
 790                     (euid != current->euid) && (euid != current->suid))
 791                         return -EPERM;
 792                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
 793                     (suid != current->euid) && (suid != current->suid))
 794                         return -EPERM;
 795         }
 796         if (ruid != (uid_t) -1) {
 797                 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
 798                         return -EAGAIN;
 799         }
 800         if (euid != (uid_t) -1) {
 801                 if (euid != current->euid)
 802                 {
 803                         current->mm->dumpable = suid_dumpable;
 804                         smp_wmb();
 805                 }
 806                 current->euid = euid;
 807         }
 808         current->fsuid = current->euid;
 809         if (suid != (uid_t) -1)
 810                 current->suid = suid;
 811
 812         key_fsuid_changed(current);
 813
 814         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
 815 }
 816
 817 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
 818 {
 819         int retval;
 820
 821         if (!(retval = put_user(current->uid, ruid)) &&
 822             !(retval = put_user(current->euid, euid)))
 823                 retval = put_user(current->suid, suid);
 824
 825         return retval;
 826 }
 827
 828 /*
 829  * Same as above, but for rgid, egid, sgid.
 830  */
 831 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 832 {
 833         int retval;
 834
 835         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 836         if (retval)
 837                 return retval;
 838
 839         if (!capable(CAP_SETGID)) {
 840                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
 841                     (rgid != current->egid) && (rgid != current->sgid))
 842                         return -EPERM;
 843                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
 844                     (egid != current->egid) && (egid != current->sgid))
 845                         return -EPERM;
 846                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
 847                     (sgid != current->egid) && (sgid != current->sgid))
 848                         return -EPERM;
 849         }
 850         if (egid != (gid_t) -1) {
 851                 if (egid != current->egid)
 852                 {
 853                         current->mm->dumpable = suid_dumpable;
 854                         smp_wmb();
 855                 }
 856                 current->egid = egid;
 857         }
 858         current->fsgid = current->egid;
 859         if (rgid != (gid_t) -1)
 860                 current->gid = rgid;
 861         if (sgid != (gid_t) -1)
 862                 current->sgid = sgid;
 863
 864         key_fsgid_changed(current);
 865         return 0;
 866 }
 867
 868 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
 869 {
 870         int retval;
 871
 872         if (!(retval = put_user(current->gid, rgid)) &&
 873             !(retval = put_user(current->egid, egid)))
 874                 retval = put_user(current->sgid, sgid);
 875
 876         return retval;
 877 }
 878
 879
 880 /*
 881  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 882  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 883  * whatever uid it wants to). It normally shadows "euid", except when
 884  * explicitly set by setfsuid() or for access..
 885  */
 886 asmlinkage long sys_setfsuid(uid_t uid)
 887 {
 888         int old_fsuid;
 889
 890         old_fsuid = current->fsuid;
 891         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
 892                 return old_fsuid;
 893
 894         if (uid == current->uid || uid == current->euid ||
 895             uid == current->suid || uid == current->fsuid ||
 896             capable(CAP_SETUID))
 897         {
 898                 if (uid != old_fsuid)
 899                 {
 900                         current->mm->dumpable = suid_dumpable;
 901                         smp_wmb();
 902                 }
 903                 current->fsuid = uid;
 904         }
 905
 906         key_fsuid_changed(current);
 907
 908         security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
 909
 910         return old_fsuid;
 911 }
 912
 913 /*
 914  * Samma på svenska..
 915  */
 916 asmlinkage long sys_setfsgid(gid_t gid)
 917 {
 918         int old_fsgid;
 919
 920         old_fsgid = current->fsgid;
 921         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 922                 return old_fsgid;
 923
 924         if (gid == current->gid || gid == current->egid ||
 925             gid == current->sgid || gid == current->fsgid ||
 926             capable(CAP_SETGID))
 927         {
 928                 if (gid != old_fsgid)
 929                 {
 930                         current->mm->dumpable = suid_dumpable;
 931                         smp_wmb();
 932                 }
 933                 current->fsgid = gid;
 934                 key_fsgid_changed(current);
 935         }
 936         return old_fsgid;
 937 }
 938
 939 asmlinkage long sys_times(struct tms __user * tbuf)
 940 {
 941         /*
 942          *      In the SMP world we might just be unlucky and have one of
 943          *      the times increment as we use it. Since the value is an
 944          *      atomically safe type this is just fine. Conceptually its
 945          *      as if the syscall took an instant longer to occur.
 946          */
 947         if (tbuf) {
 948                 struct tms tmp;
 949                 cputime_t utime, stime, cutime, cstime;
 950
 951 #ifdef CONFIG_SMP
 952                 if (thread_group_empty(current)) {
 953                         /*
 954                          * Single thread case without the use of any locks.
 955                          *
 956                          * We may race with release_task if two threads are
 957                          * executing. However, release task first adds up the
 958                          * counters (__exit_signal) before  removing the task
 959                          * from the process tasklist (__unhash_process).
 960                          * __exit_signal also acquires and releases the
 961                          * siglock which results in the proper memory ordering
 962                          * so that the list modifications are always visible
 963                          * after the counters have been updated.
 964                          *
 965                          * If the counters have been updated by the second thread
 966                          * but the thread has not yet been removed from the list
 967                          * then the other branch will be executing which will
 968                          * block on tasklist_lock until the exit handling of the
 969                          * other task is finished.
 970                          *
 971                          * This also implies that the sighand->siglock cannot
 972                          * be held by another processor. So we can also
 973                          * skip acquiring that lock.
 974                          */
 975                         utime = cputime_add(current->signal->utime, current->utime);
 976                         stime = cputime_add(current->signal->utime, current->stime);
 977                         cutime = current->signal->cutime;
 978                         cstime = current->signal->cstime;
 979                 } else
 980 #endif
 981                 {
 982
 983                         /* Process with multiple threads */
 984                         struct task_struct *tsk = current;
 985                         struct task_struct *t;
 986
 987                         read_lock(&tasklist_lock);
 988                         utime = tsk->signal->utime;
 989                         stime = tsk->signal->stime;
 990                         t = tsk;
 991                         do {
 992                                 utime = cputime_add(utime, t->utime);
 993                                 stime = cputime_add(stime, t->stime);
 994                                 t = next_thread(t);
 995                         } while (t != tsk);
 996
 997                         /*
 998                          * While we have tasklist_lock read-locked, no dying thread
 999                          * can be updating current->signal->[us]time.  Instead,
1000                          * we got their counts included in the live thread loop.
1001                          * However, another thread can come in right now and
1002                          * do a wait call that updates current->signal->c[us]time.
1003                          * To make sure we always see that pair updated atomically,
1004                          * we take the siglock around fetching them.
1005                          */
1006                         spin_lock_irq(&tsk->sighand->siglock);
1007                         cutime = tsk->signal->cutime;
1008                         cstime = tsk->signal->cstime;
1009                         spin_unlock_irq(&tsk->sighand->siglock);
1010                         read_unlock(&tasklist_lock);
1011                 }
1012                 tmp.tms_utime = cputime_to_clock_t(utime);
1013                 tmp.tms_stime = cputime_to_clock_t(stime);
1014                 tmp.tms_cutime = cputime_to_clock_t(cutime);
1015                 tmp.tms_cstime = cputime_to_clock_t(cstime);
1016                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1017                         return -EFAULT;
1018         }
1019         return (long) jiffies_64_to_clock_t(get_jiffies_64());
1020 }
1021
1022 /*
1023  * This needs some heavy checking ...
1024  * I just haven't the stomach for it. I also don't fully
1025  * understand sessions/pgrp etc. Let somebody who does explain it.
1026  *
1027  * OK, I think I have the protection semantics right.... this is really
1028  * only important on a multi-user system anyway, to make sure one user
1029  * can't send a signal to a process owned by another.  -TYT, 12/12/91
1030  *
1031  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1032  * LBT 04.03.94
1033  */
1034
1035 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1036 {
1037         struct task_struct *p;
1038         int err = -EINVAL;
1039
1040         if (!pid)
1041                 pid = current->pid;
1042         if (!pgid)
1043                 pgid = pid;
1044         if (pgid < 0)
1045                 return -EINVAL;
1046
1047         /* From this point forward we keep holding onto the tasklist lock
1048          * so that our parent does not change from under us. -DaveM
1049          */
1050         write_lock_irq(&tasklist_lock);
1051
1052         err = -ESRCH;
1053         p = find_task_by_pid(pid);
1054         if (!p)
1055                 goto out;
1056
1057         err = -EINVAL;
1058         if (!thread_group_leader(p))
1059                 goto out;
1060
1061         if (p->parent == current || p->real_parent == current) {
1062                 err = -EPERM;
1063                 if (p->signal->session != current->signal->session)
1064                         goto out;
1065                 err = -EACCES;
1066                 if (p->did_exec)
1067                         goto out;
1068         } else {
1069                 err = -ESRCH;
1070                 if (p != current)
1071                         goto out;
1072         }
1073
1074         err = -EPERM;
1075         if (p->signal->leader)
1076                 goto out;
1077
1078         if (pgid != pid) {
1079                 struct task_struct *p;
1080
1081                 do_each_task_pid(pgid, PIDTYPE_PGID, p) {
1082                         if (p->signal->session == current->signal->session)
1083                                 goto ok_pgid;
1084                 } while_each_task_pid(pgid, PIDTYPE_PGID, p);
1085                 goto out;
1086         }
1087
1088 ok_pgid:
1089         err = security_task_setpgid(p, pgid);
1090         if (err)
1091                 goto out;
1092
1093         if (process_group(p) != pgid) {
1094                 detach_pid(p, PIDTYPE_PGID);
1095                 p->signal->pgrp = pgid;
1096                 attach_pid(p, PIDTYPE_PGID, pgid);
1097         }
1098
1099         err = 0;
1100 out:
1101         /* All paths lead to here, thus we are safe. -DaveM */
1102         write_unlock_irq(&tasklist_lock);
1103         return err;
1104 }
1105
1106 asmlinkage long sys_getpgid(pid_t pid)
1107 {
1108         if (!pid) {
1109                 return process_group(current);
1110         } else {
1111                 int retval;
1112                 struct task_struct *p;
1113
1114                 read_lock(&tasklist_lock);
1115                 p = find_task_by_pid(pid);
1116
1117                 retval = -ESRCH;
1118                 if (p) {
1119                         retval = security_task_getpgid(p);
1120                         if (!retval)
1121                                 retval = process_group(p);
1122                 }
1123                 read_unlock(&tasklist_lock);
1124                 return retval;
1125         }
1126 }
1127
1128 #ifdef __ARCH_WANT_SYS_GETPGRP
1129
1130 asmlinkage long sys_getpgrp(void)
1131 {
1132         /* SMP - assuming writes are word atomic this is fine */
1133         return process_group(current);
1134 }
1135
1136 #endif
1137
1138 asmlinkage long sys_getsid(pid_t pid)
1139 {
1140         if (!pid) {
1141                 return current->signal->session;
1142         } else {
1143                 int retval;
1144                 struct task_struct *p;
1145
1146                 read_lock(&tasklist_lock);
1147                 p = find_task_by_pid(pid);
1148
1149                 retval = -ESRCH;
1150                 if(p) {
1151                         retval = security_task_getsid(p);
1152                         if (!retval)
1153                                 retval = p->signal->session;
1154                 }
1155                 read_unlock(&tasklist_lock);
1156                 return retval;
1157         }
1158 }
1159
1160 asmlinkage long sys_setsid(void)
1161 {
1162         struct pid *pid;
1163         int err = -EPERM;
1164
1165         if (!thread_group_leader(current))
1166                 return -EINVAL;
1167
1168         down(&tty_sem);
1169         write_lock_irq(&tasklist_lock);
1170
1171         pid = find_pid(PIDTYPE_PGID, current->pid);
1172         if (pid)
1173                 goto out;
1174
1175         current->signal->leader = 1;
1176         __set_special_pids(current->pid, current->pid);
1177         current->signal->tty = NULL;
1178         current->signal->tty_old_pgrp = 0;
1179         err = process_group(current);
1180 out:
1181         write_unlock_irq(&tasklist_lock);
1182         up(&tty_sem);
1183         return err;
1184 }
1185
1186 /*
1187  * Supplementary group IDs
1188  */
1189
1190 /* init to 2 - one for init_task, one to ensure it is never freed */
1191 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1192
1193 struct group_info *groups_alloc(int gidsetsize)
1194 {
1195         struct group_info *group_info;
1196         int nblocks;
1197         int i;
1198
1199         nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1200         /* Make sure we always allocate at least one indirect block pointer */
1201         nblocks = nblocks ? : 1;
1202         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1203         if (!group_info)
1204                 return NULL;
1205         group_info->ngroups = gidsetsize;
1206         group_info->nblocks = nblocks;
1207         atomic_set(&group_info->usage, 1);
1208
1209         if (gidsetsize <= NGROUPS_SMALL) {
1210                 group_info->blocks[0] = group_info->small_block;
1211         } else {
1212                 for (i = 0; i < nblocks; i++) {
1213                         gid_t *b;
1214                         b = (void *)__get_free_page(GFP_USER);
1215                         if (!b)
1216                                 goto out_undo_partial_alloc;
1217                         group_info->blocks[i] = b;
1218                 }
1219         }
1220         return group_info;
1221
1222 out_undo_partial_alloc:
1223         while (--i >= 0) {
1224                 free_page((unsigned long)group_info->blocks[i]);
1225         }
1226         kfree(group_info);
1227         return NULL;
1228 }
1229
1230 EXPORT_SYMBOL(groups_alloc);
1231
1232 void groups_free(struct group_info *group_info)
1233 {
1234         if (group_info->blocks[0] != group_info->small_block) {
1235                 int i;
1236                 for (i = 0; i < group_info->nblocks; i++)
1237                         free_page((unsigned long)group_info->blocks[i]);
1238         }
1239         kfree(group_info);
1240 }
1241
1242 EXPORT_SYMBOL(groups_free);
1243
1244 /* export the group_info to a user-space array */
1245 static int groups_to_user(gid_t __user *grouplist,
1246     struct group_info *group_info)
1247 {
1248         int i;
1249         int count = group_info->ngroups;
1250
1251         for (i = 0; i < group_info->nblocks; i++) {
1252                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1253                 int off = i * NGROUPS_PER_BLOCK;
1254                 int len = cp_count * sizeof(*grouplist);
1255
1256                 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1257                         return -EFAULT;
1258
1259                 count -= cp_count;
1260         }
1261         return 0;
1262 }
1263
1264 /* fill a group_info from a user-space array - it must be allocated already */
1265 static int groups_from_user(struct group_info *group_info,
1266     gid_t __user *grouplist)
1267  {
1268         int i;
1269         int count = group_info->ngroups;
1270
1271         for (i = 0; i < group_info->nblocks; i++) {
1272                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1273                 int off = i * NGROUPS_PER_BLOCK;
1274                 int len = cp_count * sizeof(*grouplist);
1275
1276                 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1277                         return -EFAULT;
1278
1279                 count -= cp_count;
1280         }
1281         return 0;
1282 }
1283
1284 /* a simple Shell sort */
1285 static void groups_sort(struct group_info *group_info)
1286 {
1287         int base, max, stride;
1288         int gidsetsize = group_info->ngroups;
1289
1290         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1291                 ; /* nothing */
1292         stride /= 3;
1293
1294         while (stride) {
1295                 max = gidsetsize - stride;
1296                 for (base = 0; base < max; base++) {
1297                         int left = base;
1298                         int right = left + stride;
1299                         gid_t tmp = GROUP_AT(group_info, right);
1300
1301                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1302                                 GROUP_AT(group_info, right) =
1303                                     GROUP_AT(group_info, left);
1304                                 right = left;
1305                                 left -= stride;
1306                         }
1307                         GROUP_AT(group_info, right) = tmp;
1308                 }
1309                 stride /= 3;
1310         }
1311 }
1312
1313 /* a simple bsearch */
1314 int groups_search(struct group_info *group_info, gid_t grp)
1315 {
1316         int left, right;
1317
1318         if (!group_info)
1319                 return 0;
1320
1321         left = 0;
1322         right = group_info->ngroups;
1323         while (left < right) {
1324                 int mid = (left+right)/2;
1325                 int cmp = grp - GROUP_AT(group_info, mid);
1326                 if (cmp > 0)
1327                         left = mid + 1;
1328                 else if (cmp < 0)
1329                         right = mid;
1330                 else
1331                         return 1;
1332         }
1333         return 0;
1334 }
1335
1336 /* validate and set current->group_info */
1337 int set_current_groups(struct group_info *group_info)
1338 {
1339         int retval;
1340         struct group_info *old_info;
1341
1342         retval = security_task_setgroups(group_info);
1343         if (retval)
1344                 return retval;
1345
1346         groups_sort(group_info);
1347         get_group_info(group_info);
1348
1349         task_lock(current);
1350         old_info = current->group_info;
1351         current->group_info = group_info;
1352         task_unlock(current);
1353
1354         put_group_info(old_info);
1355
1356         return 0;
1357 }
1358
1359 EXPORT_SYMBOL(set_current_groups);
1360
1361 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1362 {
1363         int i = 0;
1364
1365         /*
1366          *      SMP: Nobody else can change our grouplist. Thus we are
1367          *      safe.
1368          */
1369
1370         if (gidsetsize < 0)
1371                 return -EINVAL;
1372
1373         /* no need to grab task_lock here; it cannot change */
1374         get_group_info(current->group_info);
1375         i = current->group_info->ngroups;
1376         if (gidsetsize) {
1377                 if (i > gidsetsize) {
1378                         i = -EINVAL;
1379                         goto out;
1380                 }
1381                 if (groups_to_user(grouplist, current->group_info)) {
1382                         i = -EFAULT;
1383                         goto out;
1384                 }
1385         }
1386 out:
1387         put_group_info(current->group_info);
1388         return i;
1389 }
1390
1391 /*
1392  *      SMP: Our groups are copy-on-write. We can set them safely
1393  *      without another task interfering.
1394  */
1395
1396 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1397 {
1398         struct group_info *group_info;
1399         int retval;
1400
1401         if (!capable(CAP_SETGID))
1402                 return -EPERM;
1403         if ((unsigned)gidsetsize > NGROUPS_MAX)
1404                 return -EINVAL;
1405
1406         group_info = groups_alloc(gidsetsize);
1407         if (!group_info)
1408                 return -ENOMEM;
1409         retval = groups_from_user(group_info, grouplist);
1410         if (retval) {
1411                 put_group_info(group_info);
1412                 return retval;
1413         }
1414
1415         retval = set_current_groups(group_info);
1416         put_group_info(group_info);
1417
1418         return retval;
1419 }
1420
1421 /*
1422  * Check whether we're fsgid/egid or in the supplemental group..
1423  */
1424 int in_group_p(gid_t grp)
1425 {
1426         int retval = 1;
1427         if (grp != current->fsgid) {
1428                 get_group_info(current->group_info);
1429                 retval = groups_search(current->group_info, grp);
1430                 put_group_info(current->group_info);
1431         }
1432         return retval;
1433 }
1434
1435 EXPORT_SYMBOL(in_group_p);
1436
1437 int in_egroup_p(gid_t grp)
1438 {
1439         int retval = 1;
1440         if (grp != current->egid) {
1441                 get_group_info(current->group_info);
1442                 retval = groups_search(current->group_info, grp);
1443                 put_group_info(current->group_info);
1444         }
1445         return retval;
1446 }
1447
1448 EXPORT_SYMBOL(in_egroup_p);
1449
1450 DECLARE_RWSEM(uts_sem);
1451
1452 EXPORT_SYMBOL(uts_sem);
1453
1454 asmlinkage long sys_newuname(struct new_utsname __user * name)
1455 {
1456         int errno = 0;
1457
1458         down_read(&uts_sem);
1459         if (copy_to_user(name,&system_utsname,sizeof *name))
1460                 errno = -EFAULT;
1461         up_read(&uts_sem);
1462         return errno;
1463 }
1464
1465 asmlinkage long sys_sethostname(char __user *name, int len)
1466 {
1467         int errno;
1468         char tmp[__NEW_UTS_LEN];
1469
1470         if (!capable(CAP_SYS_ADMIN))
1471                 return -EPERM;
1472         if (len < 0 || len > __NEW_UTS_LEN)
1473                 return -EINVAL;
1474         down_write(&uts_sem);
1475         errno = -EFAULT;
1476         if (!copy_from_user(tmp, name, len)) {
1477                 memcpy(system_utsname.nodename, tmp, len);
1478                 system_utsname.nodename[len] = 0;
1479                 errno = 0;
1480         }
1481         up_write(&uts_sem);
1482         return errno;
1483 }
1484
1485 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1486
1487 asmlinkage long sys_gethostname(char __user *name, int len)
1488 {
1489         int i, errno;
1490
1491         if (len < 0)
1492                 return -EINVAL;
1493         down_read(&uts_sem);
1494         i = 1 + strlen(system_utsname.nodename);
1495         if (i > len)
1496                 i = len;
1497         errno = 0;
1498         if (copy_to_user(name, system_utsname.nodename, i))
1499                 errno = -EFAULT;
1500         up_read(&uts_sem);
1501         return errno;
1502 }
1503
1504 #endif
1505
1506 /*
1507  * Only setdomainname; getdomainname can be implemented by calling
1508  * uname()
1509  */
1510 asmlinkage long sys_setdomainname(char __user *name, int len)
1511 {
1512         int errno;
1513         char tmp[__NEW_UTS_LEN];
1514
1515         if (!capable(CAP_SYS_ADMIN))
1516                 return -EPERM;
1517         if (len < 0 || len > __NEW_UTS_LEN)
1518                 return -EINVAL;
1519
1520         down_write(&uts_sem);
1521         errno = -EFAULT;
1522         if (!copy_from_user(tmp, name, len)) {
1523                 memcpy(system_utsname.domainname, tmp, len);
1524                 system_utsname.domainname[len] = 0;
1525                 errno = 0;
1526         }
1527         up_write(&uts_sem);
1528         return errno;
1529 }
1530
1531 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1532 {
1533         if (resource >= RLIM_NLIMITS)
1534                 return -EINVAL;
1535         else {
1536                 struct rlimit value;
1537                 task_lock(current->group_leader);
1538                 value = current->signal->rlim[resource];
1539                 task_unlock(current->group_leader);
1540                 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1541         }
1542 }
1543
1544 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1545
1546 /*
1547  *      Back compatibility for getrlimit. Needed for some apps.
1548  */
1549
1550 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1551 {
1552         struct rlimit x;
1553         if (resource >= RLIM_NLIMITS)
1554                 return -EINVAL;
1555
1556         task_lock(current->group_leader);
1557         x = current->signal->rlim[resource];
1558         task_unlock(current->group_leader);
1559         if(x.rlim_cur > 0x7FFFFFFF)
1560                 x.rlim_cur = 0x7FFFFFFF;
1561         if(x.rlim_max > 0x7FFFFFFF)
1562                 x.rlim_max = 0x7FFFFFFF;
1563         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1564 }
1565
1566 #endif
1567
1568 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1569 {
1570         struct rlimit new_rlim, *old_rlim;
1571         int retval;
1572
1573         if (resource >= RLIM_NLIMITS)
1574                 return -EINVAL;
1575         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1576                 return -EFAULT;
1577        if (new_rlim.rlim_cur > new_rlim.rlim_max)
1578                return -EINVAL;
1579         old_rlim = current->signal->rlim + resource;
1580         if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1581             !capable(CAP_SYS_RESOURCE))
1582                 return -EPERM;
1583         if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1584                         return -EPERM;
1585
1586         retval = security_task_setrlimit(resource, &new_rlim);
1587         if (retval)
1588                 return retval;
1589
1590         task_lock(current->group_leader);
1591         *old_rlim = new_rlim;
1592         task_unlock(current->group_leader);
1593
1594         if (resource == RLIMIT_CPU && new_rlim.rlim_cur != RLIM_INFINITY &&
1595             (cputime_eq(current->signal->it_prof_expires, cputime_zero) ||
1596              new_rlim.rlim_cur <= cputime_to_secs(
1597                      current->signal->it_prof_expires))) {
1598                 cputime_t cputime = secs_to_cputime(new_rlim.rlim_cur);
1599                 read_lock(&tasklist_lock);
1600                 spin_lock_irq(&current->sighand->siglock);
1601                 set_process_cpu_timer(current, CPUCLOCK_PROF,
1602                                       &cputime, NULL);
1603                 spin_unlock_irq(&current->sighand->siglock);
1604                 read_unlock(&tasklist_lock);
1605         }
1606
1607         return 0;
1608 }
1609
1610 /*
1611  * It would make sense to put struct rusage in the task_struct,
1612  * except that would make the task_struct be *really big*.  After
1613  * task_struct gets moved into malloc'ed memory, it would
1614  * make sense to do this.  It will make moving the rest of the information
1615  * a lot simpler!  (Which we're not doing right now because we're not
1616  * measuring them yet).
1617  *
1618  * This expects to be called with tasklist_lock read-locked or better,
1619  * and the siglock not locked.  It may momentarily take the siglock.
1620  *
1621  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1622  * races with threads incrementing their own counters.  But since word
1623  * reads are atomic, we either get new values or old values and we don't
1624  * care which for the sums.  We always take the siglock to protect reading
1625  * the c* fields from p->signal from races with exit.c updating those
1626  * fields when reaping, so a sample either gets all the additions of a
1627  * given child after it's reaped, or none so this sample is before reaping.
1628  */
1629
1630 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1631 {
1632         struct task_struct *t;
1633         unsigned long flags;
1634         cputime_t utime, stime;
1635
1636         memset((char *) r, 0, sizeof *r);
1637
1638         if (unlikely(!p->signal))
1639                 return;
1640
1641         switch (who) {
1642                 case RUSAGE_CHILDREN:
1643                         spin_lock_irqsave(&p->sighand->siglock, flags);
1644                         utime = p->signal->cutime;
1645                         stime = p->signal->cstime;
1646                         r->ru_nvcsw = p->signal->cnvcsw;
1647                         r->ru_nivcsw = p->signal->cnivcsw;
1648                         r->ru_minflt = p->signal->cmin_flt;
1649                         r->ru_majflt = p->signal->cmaj_flt;
1650                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1651                         cputime_to_timeval(utime, &r->ru_utime);
1652                         cputime_to_timeval(stime, &r->ru_stime);
1653                         break;
1654                 case RUSAGE_SELF:
1655                         spin_lock_irqsave(&p->sighand->siglock, flags);
1656                         utime = stime = cputime_zero;
1657                         goto sum_group;
1658                 case RUSAGE_BOTH:
1659                         spin_lock_irqsave(&p->sighand->siglock, flags);
1660                         utime = p->signal->cutime;
1661                         stime = p->signal->cstime;
1662                         r->ru_nvcsw = p->signal->cnvcsw;
1663                         r->ru_nivcsw = p->signal->cnivcsw;
1664                         r->ru_minflt = p->signal->cmin_flt;
1665                         r->ru_majflt = p->signal->cmaj_flt;
1666                 sum_group:
1667                         utime = cputime_add(utime, p->signal->utime);
1668                         stime = cputime_add(stime, p->signal->stime);
1669                         r->ru_nvcsw += p->signal->nvcsw;
1670                         r->ru_nivcsw += p->signal->nivcsw;
1671                         r->ru_minflt += p->signal->min_flt;
1672                         r->ru_majflt += p->signal->maj_flt;
1673                         t = p;
1674                         do {
1675                                 utime = cputime_add(utime, t->utime);
1676                                 stime = cputime_add(stime, t->stime);
1677                                 r->ru_nvcsw += t->nvcsw;
1678                                 r->ru_nivcsw += t->nivcsw;
1679                                 r->ru_minflt += t->min_flt;
1680                                 r->ru_majflt += t->maj_flt;
1681                                 t = next_thread(t);
1682                         } while (t != p);
1683                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1684                         cputime_to_timeval(utime, &r->ru_utime);
1685                         cputime_to_timeval(stime, &r->ru_stime);
1686                         break;
1687                 default:
1688                         BUG();
1689         }
1690 }
1691
1692 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1693 {
1694         struct rusage r;
1695         read_lock(&tasklist_lock);
1696         k_getrusage(p, who, &r);
1697         read_unlock(&tasklist_lock);
1698         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1699 }
1700
1701 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1702 {
1703         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1704                 return -EINVAL;
1705         return getrusage(current, who, ru);
1706 }
1707
1708 asmlinkage long sys_umask(int mask)
1709 {
1710         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1711         return mask;
1712 }
1713
1714 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1715                           unsigned long arg4, unsigned long arg5)
1716 {
1717         long error;
1718         int sig;
1719
1720         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1721         if (error)
1722                 return error;
1723
1724         switch (option) {
1725                 case PR_SET_PDEATHSIG:
1726                         sig = arg2;
1727                         if (!valid_signal(sig)) {
1728                                 error = -EINVAL;
1729                                 break;
1730                         }
1731                         current->pdeath_signal = sig;
1732                         break;
1733                 case PR_GET_PDEATHSIG:
1734                         error = put_user(current->pdeath_signal, (int __user *)arg2);
1735                         break;
1736                 case PR_GET_DUMPABLE:
1737                         if (current->mm->dumpable)
1738                                 error = 1;
1739                         break;
1740                 case PR_SET_DUMPABLE:
1741                         if (arg2 < 0 || arg2 > 2) {
1742                                 error = -EINVAL;
1743                                 break;
1744                         }
1745                         current->mm->dumpable = arg2;
1746                         break;
1747
1748                 case PR_SET_UNALIGN:
1749                         error = SET_UNALIGN_CTL(current, arg2);
1750                         break;
1751                 case PR_GET_UNALIGN:
1752                         error = GET_UNALIGN_CTL(current, arg2);
1753                         break;
1754                 case PR_SET_FPEMU:
1755                         error = SET_FPEMU_CTL(current, arg2);
1756                         break;
1757                 case PR_GET_FPEMU:
1758                         error = GET_FPEMU_CTL(current, arg2);
1759                         break;
1760                 case PR_SET_FPEXC:
1761                         error = SET_FPEXC_CTL(current, arg2);
1762                         break;
1763                 case PR_GET_FPEXC:
1764                         error = GET_FPEXC_CTL(current, arg2);
1765                         break;
1766                 case PR_GET_TIMING:
1767                         error = PR_TIMING_STATISTICAL;
1768                         break;
1769                 case PR_SET_TIMING:
1770                         if (arg2 == PR_TIMING_STATISTICAL)
1771                                 error = 0;
1772                         else
1773                                 error = -EINVAL;
1774                         break;
1775
1776                 case PR_GET_KEEPCAPS:
1777                         if (current->keep_capabilities)
1778                                 error = 1;
1779                         break;
1780                 case PR_SET_KEEPCAPS:
1781                         if (arg2 != 0 && arg2 != 1) {
1782                                 error = -EINVAL;
1783                                 break;
1784                         }
1785                         current->keep_capabilities = arg2;
1786                         break;
1787                 case PR_SET_NAME: {
1788                         struct task_struct *me = current;
1789                         unsigned char ncomm[sizeof(me->comm)];
1790
1791                         ncomm[sizeof(me->comm)-1] = 0;
1792                         if (strncpy_from_user(ncomm, (char __user *)arg2,
1793                                                 sizeof(me->comm)-1) < 0)
1794                                 return -EFAULT;
1795                         set_task_comm(me, ncomm);
1796                         return 0;
1797                 }
1798                 case PR_GET_NAME: {
1799                         struct task_struct *me = current;
1800                         unsigned char tcomm[sizeof(me->comm)];
1801
1802                         get_task_comm(tcomm, me);
1803                         if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
1804                                 return -EFAULT;
1805                         return 0;
1806                 }
1807                 default:
1808                         error = -EINVAL;
1809                         break;
1810         }
1811         return error;
1812 }