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