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 #include <linux/cn_proc.h>
  32
  33 #include <linux/compat.h>
  34 #include <linux/syscalls.h>
  35 #include <linux/kprobes.h>
  36
  37 #include <asm/uaccess.h>
  38 #include <asm/io.h>
  39 #include <asm/unistd.h>
  40
  41 #ifndef SET_UNALIGN_CTL
  42 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  43 #endif
  44 #ifndef GET_UNALIGN_CTL
  45 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  46 #endif
  47 #ifndef SET_FPEMU_CTL
  48 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
  49 #endif
  50 #ifndef GET_FPEMU_CTL
  51 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
  52 #endif
  53 #ifndef SET_FPEXC_CTL
  54 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
  55 #endif
  56 #ifndef GET_FPEXC_CTL
  57 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
  58 #endif
  59
  60 /*
  61  * this is where the system-wide overflow UID and GID are defined, for
  62  * architectures that now have 32-bit UID/GID but didn't in the past
  63  */
  64
  65 int overflowuid = DEFAULT_OVERFLOWUID;
  66 int overflowgid = DEFAULT_OVERFLOWGID;
  67
  68 #ifdef CONFIG_UID16
  69 EXPORT_SYMBOL(overflowuid);
  70 EXPORT_SYMBOL(overflowgid);
  71 #endif
  72
  73 /*
  74  * the same as above, but for filesystems which can only store a 16-bit
  75  * UID and GID. as such, this is needed on all architectures
  76  */
  77
  78 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
  79 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
  80
  81 EXPORT_SYMBOL(fs_overflowuid);
  82 EXPORT_SYMBOL(fs_overflowgid);
  83
  84 /*
  85  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
  86  */
  87
  88 int C_A_D = 1;
  89 int cad_pid = 1;
  90
  91 /*
  92  *      Notifier list for kernel code which wants to be called
  93  *      at shutdown. This is used to stop any idling DMA operations
  94  *      and the like.
  95  */
  96
  97 static struct notifier_block *reboot_notifier_list;
  98 static DEFINE_RWLOCK(notifier_lock);
  99
 100 /**
 101  *      notifier_chain_register - Add notifier to a notifier chain
 102  *      @list: Pointer to root list pointer
 103  *      @n: New entry in notifier chain
 104  *
 105  *      Adds a notifier to a notifier chain.
 106  *
 107  *      Currently always returns zero.
 108  */
 109
 110 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
 111 {
 112         write_lock(&notifier_lock);
 113         while(*list)
 114         {
 115                 if(n->priority > (*list)->priority)
 116                         break;
 117                 list= &((*list)->next);
 118         }
 119         n->next = *list;
 120         *list=n;
 121         write_unlock(&notifier_lock);
 122         return 0;
 123 }
 124
 125 EXPORT_SYMBOL(notifier_chain_register);
 126
 127 /**
 128  *      notifier_chain_unregister - Remove notifier from a notifier chain
 129  *      @nl: Pointer to root list pointer
 130  *      @n: New entry in notifier chain
 131  *
 132  *      Removes a notifier from a notifier chain.
 133  *
 134  *      Returns zero on success, or %-ENOENT on failure.
 135  */
 136
 137 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
 138 {
 139         write_lock(&notifier_lock);
 140         while((*nl)!=NULL)
 141         {
 142                 if((*nl)==n)
 143                 {
 144                         *nl=n->next;
 145                         write_unlock(&notifier_lock);
 146                         return 0;
 147                 }
 148                 nl=&((*nl)->next);
 149         }
 150         write_unlock(&notifier_lock);
 151         return -ENOENT;
 152 }
 153
 154 EXPORT_SYMBOL(notifier_chain_unregister);
 155
 156 /**
 157  *      notifier_call_chain - Call functions in a notifier chain
 158  *      @n: Pointer to root pointer of notifier chain
 159  *      @val: Value passed unmodified to notifier function
 160  *      @v: Pointer passed unmodified to notifier function
 161  *
 162  *      Calls each function in a notifier chain in turn.
 163  *
 164  *      If the return value of the notifier can be and'd
 165  *      with %NOTIFY_STOP_MASK, then notifier_call_chain
 166  *      will return immediately, with the return value of
 167  *      the notifier function which halted execution.
 168  *      Otherwise, the return value is the return value
 169  *      of the last notifier function called.
 170  */
 171
 172 int __kprobes notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
 173 {
 174         int ret=NOTIFY_DONE;
 175         struct notifier_block *nb = *n;
 176
 177         while(nb)
 178         {
 179                 ret=nb->notifier_call(nb,val,v);
 180                 if(ret&NOTIFY_STOP_MASK)
 181                 {
 182                         return ret;
 183                 }
 184                 nb=nb->next;
 185         }
 186         return ret;
 187 }
 188
 189 EXPORT_SYMBOL(notifier_call_chain);
 190
 191 /**
 192  *      register_reboot_notifier - Register function to be called at reboot time
 193  *      @nb: Info about notifier function to be called
 194  *
 195  *      Registers a function with the list of functions
 196  *      to be called at reboot time.
 197  *
 198  *      Currently always returns zero, as notifier_chain_register
 199  *      always returns zero.
 200  */
 201
 202 int register_reboot_notifier(struct notifier_block * nb)
 203 {
 204         return notifier_chain_register(&reboot_notifier_list, nb);
 205 }
 206
 207 EXPORT_SYMBOL(register_reboot_notifier);
 208
 209 /**
 210  *      unregister_reboot_notifier - Unregister previously registered reboot notifier
 211  *      @nb: Hook to be unregistered
 212  *
 213  *      Unregisters a previously registered reboot
 214  *      notifier function.
 215  *
 216  *      Returns zero on success, or %-ENOENT on failure.
 217  */
 218
 219 int unregister_reboot_notifier(struct notifier_block * nb)
 220 {
 221         return notifier_chain_unregister(&reboot_notifier_list, nb);
 222 }
 223
 224 EXPORT_SYMBOL(unregister_reboot_notifier);
 225
 226 static int set_one_prio(struct task_struct *p, int niceval, int error)
 227 {
 228         int no_nice;
 229
 230         if (p->uid != current->euid &&
 231                 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
 232                 error = -EPERM;
 233                 goto out;
 234         }
 235         if (niceval < task_nice(p) && !can_nice(p, niceval)) {
 236                 error = -EACCES;
 237                 goto out;
 238         }
 239         no_nice = security_task_setnice(p, niceval);
 240         if (no_nice) {
 241                 error = no_nice;
 242                 goto out;
 243         }
 244         if (error == -ESRCH)
 245                 error = 0;
 246         set_user_nice(p, niceval);
 247 out:
 248         return error;
 249 }
 250
 251 asmlinkage long sys_setpriority(int which, int who, int niceval)
 252 {
 253         struct task_struct *g, *p;
 254         struct user_struct *user;
 255         int error = -EINVAL;
 256
 257         if (which > 2 || which < 0)
 258                 goto out;
 259
 260         /* normalize: avoid signed division (rounding problems) */
 261         error = -ESRCH;
 262         if (niceval < -20)
 263                 niceval = -20;
 264         if (niceval > 19)
 265                 niceval = 19;
 266
 267         read_lock(&tasklist_lock);
 268         switch (which) {
 269                 case PRIO_PROCESS:
 270                         if (!who)
 271                                 who = current->pid;
 272                         p = find_task_by_pid(who);
 273                         if (p)
 274                                 error = set_one_prio(p, niceval, error);
 275                         break;
 276                 case PRIO_PGRP:
 277                         if (!who)
 278                                 who = process_group(current);
 279                         do_each_task_pid(who, PIDTYPE_PGID, p) {
 280                                 error = set_one_prio(p, niceval, error);
 281                         } while_each_task_pid(who, PIDTYPE_PGID, p);
 282                         break;
 283                 case PRIO_USER:
 284                         user = current->user;
 285                         if (!who)
 286                                 who = current->uid;
 287                         else
 288                                 if ((who != current->uid) && !(user = find_user(who)))
 289                                         goto out_unlock;        /* No processes for this user */
 290
 291                         do_each_thread(g, p)
 292                                 if (p->uid == who)
 293                                         error = set_one_prio(p, niceval, error);
 294                         while_each_thread(g, p);
 295                         if (who != current->uid)
 296                                 free_uid(user);         /* For find_user() */
 297                         break;
 298         }
 299 out_unlock:
 300         read_unlock(&tasklist_lock);
 301 out:
 302         return error;
 303 }
 304
 305 /*
 306  * Ugh. To avoid negative return values, "getpriority()" will
 307  * not return the normal nice-value, but a negated value that
 308  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 309  * to stay compatible.
 310  */
 311 asmlinkage long sys_getpriority(int which, int who)
 312 {
 313         struct task_struct *g, *p;
 314         struct user_struct *user;
 315         long niceval, retval = -ESRCH;
 316
 317         if (which > 2 || which < 0)
 318                 return -EINVAL;
 319
 320         read_lock(&tasklist_lock);
 321         switch (which) {
 322                 case PRIO_PROCESS:
 323                         if (!who)
 324                                 who = current->pid;
 325                         p = find_task_by_pid(who);
 326                         if (p) {
 327                                 niceval = 20 - task_nice(p);
 328                                 if (niceval > retval)
 329                                         retval = niceval;
 330                         }
 331                         break;
 332                 case PRIO_PGRP:
 333                         if (!who)
 334                                 who = process_group(current);
 335                         do_each_task_pid(who, PIDTYPE_PGID, p) {
 336                                 niceval = 20 - task_nice(p);
 337                                 if (niceval > retval)
 338                                         retval = niceval;
 339                         } while_each_task_pid(who, PIDTYPE_PGID, p);
 340                         break;
 341                 case PRIO_USER:
 342                         user = current->user;
 343                         if (!who)
 344                                 who = current->uid;
 345                         else
 346                                 if ((who != current->uid) && !(user = find_user(who)))
 347                                         goto out_unlock;        /* No processes for this user */
 348
 349                         do_each_thread(g, p)
 350                                 if (p->uid == who) {
 351                                         niceval = 20 - task_nice(p);
 352                                         if (niceval > retval)
 353                                                 retval = niceval;
 354                                 }
 355                         while_each_thread(g, p);
 356                         if (who != current->uid)
 357                                 free_uid(user);         /* for find_user() */
 358                         break;
 359         }
 360 out_unlock:
 361         read_unlock(&tasklist_lock);
 362
 363         return retval;
 364 }
 365
 366 /**
 367  *      emergency_restart - reboot the system
 368  *
 369  *      Without shutting down any hardware or taking any locks
 370  *      reboot the system.  This is called when we know we are in
 371  *      trouble so this is our best effort to reboot.  This is
 372  *      safe to call in interrupt context.
 373  */
 374 void emergency_restart(void)
 375 {
 376         machine_emergency_restart();
 377 }
 378 EXPORT_SYMBOL_GPL(emergency_restart);
 379
 380 void kernel_restart_prepare(char *cmd)
 381 {
 382         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
 383         system_state = SYSTEM_RESTART;
 384         device_shutdown();
 385 }
 386
 387 /**
 388  *      kernel_restart - reboot the system
 389  *      @cmd: pointer to buffer containing command to execute for restart
 390  *              or %NULL
 391  *
 392  *      Shutdown everything and perform a clean reboot.
 393  *      This is not safe to call in interrupt context.
 394  */
 395 void kernel_restart(char *cmd)
 396 {
 397         kernel_restart_prepare(cmd);
 398         if (!cmd) {
 399                 printk(KERN_EMERG "Restarting system.\n");
 400         } else {
 401                 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
 402         }
 403         printk(".\n");
 404         machine_restart(cmd);
 405 }
 406 EXPORT_SYMBOL_GPL(kernel_restart);
 407
 408 /**
 409  *      kernel_kexec - reboot the system
 410  *
 411  *      Move into place and start executing a preloaded standalone
 412  *      executable.  If nothing was preloaded return an error.
 413  */
 414 void kernel_kexec(void)
 415 {
 416 #ifdef CONFIG_KEXEC
 417         struct kimage *image;
 418         image = xchg(&kexec_image, 0);
 419         if (!image) {
 420                 return;
 421         }
 422         kernel_restart_prepare(NULL);
 423         printk(KERN_EMERG "Starting new kernel\n");
 424         machine_shutdown();
 425         machine_kexec(image);
 426 #endif
 427 }
 428 EXPORT_SYMBOL_GPL(kernel_kexec);
 429
 430 /**
 431  *      kernel_halt - halt the system
 432  *
 433  *      Shutdown everything and perform a clean system halt.
 434  */
 435 void kernel_halt_prepare(void)
 436 {
 437         notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
 438         system_state = SYSTEM_HALT;
 439         device_shutdown();
 440 }
 441 void kernel_halt(void)
 442 {
 443         kernel_halt_prepare();
 444         printk(KERN_EMERG "System halted.\n");
 445         machine_halt();
 446 }
 447 EXPORT_SYMBOL_GPL(kernel_halt);
 448
 449 /**
 450  *      kernel_power_off - power_off the system
 451  *
 452  *      Shutdown everything and perform a clean system power_off.
 453  */
 454 void kernel_power_off_prepare(void)
 455 {
 456         notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
 457         system_state = SYSTEM_POWER_OFF;
 458         device_shutdown();
 459 }
 460 void kernel_power_off(void)
 461 {
 462         kernel_power_off_prepare();
 463         printk(KERN_EMERG "Power down.\n");
 464         machine_power_off();
 465 }
 466 EXPORT_SYMBOL_GPL(kernel_power_off);
 467
 468 /*
 469  * Reboot system call: for obvious reasons only root may call it,
 470  * and even root needs to set up some magic numbers in the registers
 471  * so that some mistake won't make this reboot the whole machine.
 472  * You can also set the meaning of the ctrl-alt-del-key here.
 473  *
 474  * reboot doesn't sync: do that yourself before calling this.
 475  */
 476 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
 477 {
 478         char buffer[256];
 479
 480         /* We only trust the superuser with rebooting the system. */
 481         if (!capable(CAP_SYS_BOOT))
 482                 return -EPERM;
 483
 484         /* For safety, we require "magic" arguments. */
 485         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 486             (magic2 != LINUX_REBOOT_MAGIC2 &&
 487                         magic2 != LINUX_REBOOT_MAGIC2A &&
 488                         magic2 != LINUX_REBOOT_MAGIC2B &&
 489                         magic2 != LINUX_REBOOT_MAGIC2C))
 490                 return -EINVAL;
 491
 492         lock_kernel();
 493         switch (cmd) {
 494         case LINUX_REBOOT_CMD_RESTART:
 495                 kernel_restart(NULL);
 496                 break;
 497
 498         case LINUX_REBOOT_CMD_CAD_ON:
 499                 C_A_D = 1;
 500                 break;
 501
 502         case LINUX_REBOOT_CMD_CAD_OFF:
 503                 C_A_D = 0;
 504                 break;
 505
 506         case LINUX_REBOOT_CMD_HALT:
 507                 kernel_halt();
 508                 unlock_kernel();
 509                 do_exit(0);
 510                 break;
 511
 512         case LINUX_REBOOT_CMD_POWER_OFF:
 513                 kernel_power_off();
 514                 unlock_kernel();
 515                 do_exit(0);
 516                 break;
 517
 518         case LINUX_REBOOT_CMD_RESTART2:
 519                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 520                         unlock_kernel();
 521                         return -EFAULT;
 522                 }
 523                 buffer[sizeof(buffer) - 1] = '\0';
 524
 525                 kernel_restart(buffer);
 526                 break;
 527
 528         case LINUX_REBOOT_CMD_KEXEC:
 529                 kernel_kexec();
 530                 unlock_kernel();
 531                 return -EINVAL;
 532
 533 #ifdef CONFIG_SOFTWARE_SUSPEND
 534         case LINUX_REBOOT_CMD_SW_SUSPEND:
 535                 {
 536                         int ret = software_suspend();
 537                         unlock_kernel();
 538                         return ret;
 539                 }
 540 #endif
 541
 542         default:
 543                 unlock_kernel();
 544                 return -EINVAL;
 545         }
 546         unlock_kernel();
 547         return 0;
 548 }
 549
 550 static void deferred_cad(void *dummy)
 551 {
 552         kernel_restart(NULL);
 553 }
 554
 555 /*
 556  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 557  * As it's called within an interrupt, it may NOT sync: the only choice
 558  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 559  */
 560 void ctrl_alt_del(void)
 561 {
 562         static DECLARE_WORK(cad_work, deferred_cad, NULL);
 563
 564         if (C_A_D)
 565                 schedule_work(&cad_work);
 566         else
 567                 kill_proc(cad_pid, SIGINT, 1);
 568 }
 569
 570
 571 /*
 572  * Unprivileged users may change the real gid to the effective gid
 573  * or vice versa.  (BSD-style)
 574  *
 575  * If you set the real gid at all, or set the effective gid to a value not
 576  * equal to the real gid, then the saved gid is set to the new effective gid.
 577  *
 578  * This makes it possible for a setgid program to completely drop its
 579  * privileges, which is often a useful assertion to make when you are doing
 580  * a security audit over a program.
 581  *
 582  * The general idea is that a program which uses just setregid() will be
 583  * 100% compatible with BSD.  A program which uses just setgid() will be
 584  * 100% compatible with POSIX with saved IDs.
 585  *
 586  * SMP: There are not races, the GIDs are checked only by filesystem
 587  *      operations (as far as semantic preservation is concerned).
 588  */
 589 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 590 {
 591         int old_rgid = current->gid;
 592         int old_egid = current->egid;
 593         int new_rgid = old_rgid;
 594         int new_egid = old_egid;
 595         int retval;
 596
 597         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 598         if (retval)
 599                 return retval;
 600
 601         if (rgid != (gid_t) -1) {
 602                 if ((old_rgid == rgid) ||
 603                     (current->egid==rgid) ||
 604                     capable(CAP_SETGID))
 605                         new_rgid = rgid;
 606                 else
 607                         return -EPERM;
 608         }
 609         if (egid != (gid_t) -1) {
 610                 if ((old_rgid == egid) ||
 611                     (current->egid == egid) ||
 612                     (current->sgid == egid) ||
 613                     capable(CAP_SETGID))
 614                         new_egid = egid;
 615                 else {
 616                         return -EPERM;
 617                 }
 618         }
 619         if (new_egid != old_egid)
 620         {
 621                 current->mm->dumpable = suid_dumpable;
 622                 smp_wmb();
 623         }
 624         if (rgid != (gid_t) -1 ||
 625             (egid != (gid_t) -1 && egid != old_rgid))
 626                 current->sgid = new_egid;
 627         current->fsgid = new_egid;
 628         current->egid = new_egid;
 629         current->gid = new_rgid;
 630         key_fsgid_changed(current);
 631         proc_id_connector(current, PROC_EVENT_GID);
 632         return 0;
 633 }
 634
 635 /*
 636  * setgid() is implemented like SysV w/ SAVED_IDS
 637  *
 638  * SMP: Same implicit races as above.
 639  */
 640 asmlinkage long sys_setgid(gid_t gid)
 641 {
 642         int old_egid = current->egid;
 643         int retval;
 644
 645         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 646         if (retval)
 647                 return retval;
 648
 649         if (capable(CAP_SETGID))
 650         {
 651                 if(old_egid != gid)
 652                 {
 653                         current->mm->dumpable = suid_dumpable;
 654                         smp_wmb();
 655                 }
 656                 current->gid = current->egid = current->sgid = current->fsgid = gid;
 657         }
 658         else if ((gid == current->gid) || (gid == current->sgid))
 659         {
 660                 if(old_egid != gid)
 661                 {
 662                         current->mm->dumpable = suid_dumpable;
 663                         smp_wmb();
 664                 }
 665                 current->egid = current->fsgid = gid;
 666         }
 667         else
 668                 return -EPERM;
 669
 670         key_fsgid_changed(current);
 671         proc_id_connector(current, PROC_EVENT_GID);
 672         return 0;
 673 }
 674
 675 static int set_user(uid_t new_ruid, int dumpclear)
 676 {
 677         struct user_struct *new_user;
 678
 679         new_user = alloc_uid(new_ruid);
 680         if (!new_user)
 681                 return -EAGAIN;
 682
 683         if (atomic_read(&new_user->processes) >=
 684                                 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
 685                         new_user != &root_user) {
 686                 free_uid(new_user);
 687                 return -EAGAIN;
 688         }
 689
 690         switch_uid(new_user);
 691
 692         if(dumpclear)
 693         {
 694                 current->mm->dumpable = suid_dumpable;
 695                 smp_wmb();
 696         }
 697         current->uid = new_ruid;
 698         return 0;
 699 }
 700
 701 /*
 702  * Unprivileged users may change the real uid to the effective uid
 703  * or vice versa.  (BSD-style)
 704  *
 705  * If you set the real uid at all, or set the effective uid to a value not
 706  * equal to the real uid, then the saved uid is set to the new effective uid.
 707  *
 708  * This makes it possible for a setuid program to completely drop its
 709  * privileges, which is often a useful assertion to make when you are doing
 710  * a security audit over a program.
 711  *
 712  * The general idea is that a program which uses just setreuid() will be
 713  * 100% compatible with BSD.  A program which uses just setuid() will be
 714  * 100% compatible with POSIX with saved IDs.
 715  */
 716 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 717 {
 718         int old_ruid, old_euid, old_suid, new_ruid, new_euid;
 719         int retval;
 720
 721         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 722         if (retval)
 723                 return retval;
 724
 725         new_ruid = old_ruid = current->uid;
 726         new_euid = old_euid = current->euid;
 727         old_suid = current->suid;
 728
 729         if (ruid != (uid_t) -1) {
 730                 new_ruid = ruid;
 731                 if ((old_ruid != ruid) &&
 732                     (current->euid != ruid) &&
 733                     !capable(CAP_SETUID))
 734                         return -EPERM;
 735         }
 736
 737         if (euid != (uid_t) -1) {
 738                 new_euid = euid;
 739                 if ((old_ruid != euid) &&
 740                     (current->euid != euid) &&
 741                     (current->suid != euid) &&
 742                     !capable(CAP_SETUID))
 743                         return -EPERM;
 744         }
 745
 746         if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
 747                 return -EAGAIN;
 748
 749         if (new_euid != old_euid)
 750         {
 751                 current->mm->dumpable = suid_dumpable;
 752                 smp_wmb();
 753         }
 754         current->fsuid = current->euid = new_euid;
 755         if (ruid != (uid_t) -1 ||
 756             (euid != (uid_t) -1 && euid != old_ruid))
 757                 current->suid = current->euid;
 758         current->fsuid = current->euid;
 759
 760         key_fsuid_changed(current);
 761         proc_id_connector(current, PROC_EVENT_UID);
 762
 763         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
 764 }
 765
 766
 767
 768 /*
 769  * setuid() is implemented like SysV with SAVED_IDS
 770  *
 771  * Note that SAVED_ID's is deficient in that a setuid root program
 772  * like sendmail, for example, cannot set its uid to be a normal
 773  * user and then switch back, because if you're root, setuid() sets
 774  * the saved uid too.  If you don't like this, blame the bright people
 775  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 776  * will allow a root program to temporarily drop privileges and be able to
 777  * regain them by swapping the real and effective uid.
 778  */
 779 asmlinkage long sys_setuid(uid_t uid)
 780 {
 781         int old_euid = current->euid;
 782         int old_ruid, old_suid, new_ruid, new_suid;
 783         int retval;
 784
 785         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 786         if (retval)
 787                 return retval;
 788
 789         old_ruid = new_ruid = current->uid;
 790         old_suid = current->suid;
 791         new_suid = old_suid;
 792
 793         if (capable(CAP_SETUID)) {
 794                 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
 795                         return -EAGAIN;
 796                 new_suid = uid;
 797         } else if ((uid != current->uid) && (uid != new_suid))
 798                 return -EPERM;
 799
 800         if (old_euid != uid)
 801         {
 802                 current->mm->dumpable = suid_dumpable;
 803                 smp_wmb();
 804         }
 805         current->fsuid = current->euid = uid;
 806         current->suid = new_suid;
 807
 808         key_fsuid_changed(current);
 809         proc_id_connector(current, PROC_EVENT_UID);
 810
 811         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
 812 }
 813
 814
 815 /*
 816  * This function implements a generic ability to update ruid, euid,
 817  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 818  */
 819 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 820 {
 821         int old_ruid = current->uid;
 822         int old_euid = current->euid;
 823         int old_suid = current->suid;
 824         int retval;
 825
 826         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 827         if (retval)
 828                 return retval;
 829
 830         if (!capable(CAP_SETUID)) {
 831                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
 832                     (ruid != current->euid) && (ruid != current->suid))
 833                         return -EPERM;
 834                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
 835                     (euid != current->euid) && (euid != current->suid))
 836                         return -EPERM;
 837                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
 838                     (suid != current->euid) && (suid != current->suid))
 839                         return -EPERM;
 840         }
 841         if (ruid != (uid_t) -1) {
 842                 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
 843                         return -EAGAIN;
 844         }
 845         if (euid != (uid_t) -1) {
 846                 if (euid != current->euid)
 847                 {
 848                         current->mm->dumpable = suid_dumpable;
 849                         smp_wmb();
 850                 }
 851                 current->euid = euid;
 852         }
 853         current->fsuid = current->euid;
 854         if (suid != (uid_t) -1)
 855                 current->suid = suid;
 856
 857         key_fsuid_changed(current);
 858         proc_id_connector(current, PROC_EVENT_UID);
 859
 860         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
 861 }
 862
 863 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
 864 {
 865         int retval;
 866
 867         if (!(retval = put_user(current->uid, ruid)) &&
 868             !(retval = put_user(current->euid, euid)))
 869                 retval = put_user(current->suid, suid);
 870
 871         return retval;
 872 }
 873
 874 /*
 875  * Same as above, but for rgid, egid, sgid.
 876  */
 877 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 878 {
 879         int retval;
 880
 881         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 882         if (retval)
 883                 return retval;
 884
 885         if (!capable(CAP_SETGID)) {
 886                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
 887                     (rgid != current->egid) && (rgid != current->sgid))
 888                         return -EPERM;
 889                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
 890                     (egid != current->egid) && (egid != current->sgid))
 891                         return -EPERM;
 892                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
 893                     (sgid != current->egid) && (sgid != current->sgid))
 894                         return -EPERM;
 895         }
 896         if (egid != (gid_t) -1) {
 897                 if (egid != current->egid)
 898                 {
 899                         current->mm->dumpable = suid_dumpable;
 900                         smp_wmb();
 901                 }
 902                 current->egid = egid;
 903         }
 904         current->fsgid = current->egid;
 905         if (rgid != (gid_t) -1)
 906                 current->gid = rgid;
 907         if (sgid != (gid_t) -1)
 908                 current->sgid = sgid;
 909
 910         key_fsgid_changed(current);
 911         proc_id_connector(current, PROC_EVENT_GID);
 912         return 0;
 913 }
 914
 915 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
 916 {
 917         int retval;
 918
 919         if (!(retval = put_user(current->gid, rgid)) &&
 920             !(retval = put_user(current->egid, egid)))
 921                 retval = put_user(current->sgid, sgid);
 922
 923         return retval;
 924 }
 925
 926
 927 /*
 928  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 929  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 930  * whatever uid it wants to). It normally shadows "euid", except when
 931  * explicitly set by setfsuid() or for access..
 932  */
 933 asmlinkage long sys_setfsuid(uid_t uid)
 934 {
 935         int old_fsuid;
 936
 937         old_fsuid = current->fsuid;
 938         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
 939                 return old_fsuid;
 940
 941         if (uid == current->uid || uid == current->euid ||
 942             uid == current->suid || uid == current->fsuid ||
 943             capable(CAP_SETUID))
 944         {
 945                 if (uid != old_fsuid)
 946                 {
 947                         current->mm->dumpable = suid_dumpable;
 948                         smp_wmb();
 949                 }
 950                 current->fsuid = uid;
 951         }
 952
 953         key_fsuid_changed(current);
 954         proc_id_connector(current, PROC_EVENT_UID);
 955
 956         security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
 957
 958         return old_fsuid;
 959 }
 960
 961 /*
 962  * Samma på svenska..
 963  */
 964 asmlinkage long sys_setfsgid(gid_t gid)
 965 {
 966         int old_fsgid;
 967
 968         old_fsgid = current->fsgid;
 969         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 970                 return old_fsgid;
 971
 972         if (gid == current->gid || gid == current->egid ||
 973             gid == current->sgid || gid == current->fsgid ||
 974             capable(CAP_SETGID))
 975         {
 976                 if (gid != old_fsgid)
 977                 {
 978                         current->mm->dumpable = suid_dumpable;
 979                         smp_wmb();
 980                 }
 981                 current->fsgid = gid;
 982                 key_fsgid_changed(current);
 983                 proc_id_connector(current, PROC_EVENT_GID);
 984         }
 985         return old_fsgid;
 986 }
 987
 988 asmlinkage long sys_times(struct tms __user * tbuf)
 989 {
 990         /*
 991          *      In the SMP world we might just be unlucky and have one of
 992          *      the times increment as we use it. Since the value is an
 993          *      atomically safe type this is just fine. Conceptually its
 994          *      as if the syscall took an instant longer to occur.
 995          */
 996         if (tbuf) {
 997                 struct tms tmp;
 998                 cputime_t utime, stime, cutime, cstime;
 999
1000 #ifdef CONFIG_SMP
1001                 if (thread_group_empty(current)) {
1002                         /*
1003                          * Single thread case without the use of any locks.
1004                          *
1005                          * We may race with release_task if two threads are
1006                          * executing. However, release task first adds up the
1007                          * counters (__exit_signal) before  removing the task
1008                          * from the process tasklist (__unhash_process).
1009                          * __exit_signal also acquires and releases the
1010                          * siglock which results in the proper memory ordering
1011                          * so that the list modifications are always visible
1012                          * after the counters have been updated.
1013                          *
1014                          * If the counters have been updated by the second thread
1015                          * but the thread has not yet been removed from the list
1016                          * then the other branch will be executing which will
1017                          * block on tasklist_lock until the exit handling of the
1018                          * other task is finished.
1019                          *
1020                          * This also implies that the sighand->siglock cannot
1021                          * be held by another processor. So we can also
1022                          * skip acquiring that lock.
1023                          */
1024                         utime = cputime_add(current->signal->utime, current->utime);
1025                         stime = cputime_add(current->signal->utime, current->stime);
1026                         cutime = current->signal->cutime;
1027                         cstime = current->signal->cstime;
1028                 } else
1029 #endif
1030                 {
1031
1032                         /* Process with multiple threads */
1033                         struct task_struct *tsk = current;
1034                         struct task_struct *t;
1035
1036                         read_lock(&tasklist_lock);
1037                         utime = tsk->signal->utime;
1038                         stime = tsk->signal->stime;
1039                         t = tsk;
1040                         do {
1041                                 utime = cputime_add(utime, t->utime);
1042                                 stime = cputime_add(stime, t->stime);
1043                                 t = next_thread(t);
1044                         } while (t != tsk);
1045
1046                         /*
1047                          * While we have tasklist_lock read-locked, no dying thread
1048                          * can be updating current->signal->[us]time.  Instead,
1049                          * we got their counts included in the live thread loop.
1050                          * However, another thread can come in right now and
1051                          * do a wait call that updates current->signal->c[us]time.
1052                          * To make sure we always see that pair updated atomically,
1053                          * we take the siglock around fetching them.
1054                          */
1055                         spin_lock_irq(&tsk->sighand->siglock);
1056                         cutime = tsk->signal->cutime;
1057                         cstime = tsk->signal->cstime;
1058                         spin_unlock_irq(&tsk->sighand->siglock);
1059                         read_unlock(&tasklist_lock);
1060                 }
1061                 tmp.tms_utime = cputime_to_clock_t(utime);
1062                 tmp.tms_stime = cputime_to_clock_t(stime);
1063                 tmp.tms_cutime = cputime_to_clock_t(cutime);
1064                 tmp.tms_cstime = cputime_to_clock_t(cstime);
1065                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1066                         return -EFAULT;
1067         }
1068         return (long) jiffies_64_to_clock_t(get_jiffies_64());
1069 }
1070
1071 /*
1072  * This needs some heavy checking ...
1073  * I just haven't the stomach for it. I also don't fully
1074  * understand sessions/pgrp etc. Let somebody who does explain it.
1075  *
1076  * OK, I think I have the protection semantics right.... this is really
1077  * only important on a multi-user system anyway, to make sure one user
1078  * can't send a signal to a process owned by another.  -TYT, 12/12/91
1079  *
1080  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1081  * LBT 04.03.94
1082  */
1083
1084 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1085 {
1086         struct task_struct *p;
1087         int err = -EINVAL;
1088
1089         if (!pid)
1090                 pid = current->pid;
1091         if (!pgid)
1092                 pgid = pid;
1093         if (pgid < 0)
1094                 return -EINVAL;
1095
1096         /* From this point forward we keep holding onto the tasklist lock
1097          * so that our parent does not change from under us. -DaveM
1098          */
1099         write_lock_irq(&tasklist_lock);
1100
1101         err = -ESRCH;
1102         p = find_task_by_pid(pid);
1103         if (!p)
1104                 goto out;
1105
1106         err = -EINVAL;
1107         if (!thread_group_leader(p))
1108                 goto out;
1109
1110         if (p->parent == current || p->real_parent == current) {
1111                 err = -EPERM;
1112                 if (p->signal->session != current->signal->session)
1113                         goto out;
1114                 err = -EACCES;
1115                 if (p->did_exec)
1116                         goto out;
1117         } else {
1118                 err = -ESRCH;
1119                 if (p != current)
1120                         goto out;
1121         }
1122
1123         err = -EPERM;
1124         if (p->signal->leader)
1125                 goto out;
1126
1127         if (pgid != pid) {
1128                 struct task_struct *p;
1129
1130                 do_each_task_pid(pgid, PIDTYPE_PGID, p) {
1131                         if (p->signal->session == current->signal->session)
1132                                 goto ok_pgid;
1133                 } while_each_task_pid(pgid, PIDTYPE_PGID, p);
1134                 goto out;
1135         }
1136
1137 ok_pgid:
1138         err = security_task_setpgid(p, pgid);
1139         if (err)
1140                 goto out;
1141
1142         if (process_group(p) != pgid) {
1143                 detach_pid(p, PIDTYPE_PGID);
1144                 p->signal->pgrp = pgid;
1145                 attach_pid(p, PIDTYPE_PGID, pgid);
1146         }
1147
1148         err = 0;
1149 out:
1150         /* All paths lead to here, thus we are safe. -DaveM */
1151         write_unlock_irq(&tasklist_lock);
1152         return err;
1153 }
1154
1155 asmlinkage long sys_getpgid(pid_t pid)
1156 {
1157         if (!pid) {
1158                 return process_group(current);
1159         } else {
1160                 int retval;
1161                 struct task_struct *p;
1162
1163                 read_lock(&tasklist_lock);
1164                 p = find_task_by_pid(pid);
1165
1166                 retval = -ESRCH;
1167                 if (p) {
1168                         retval = security_task_getpgid(p);
1169                         if (!retval)
1170                                 retval = process_group(p);
1171                 }
1172                 read_unlock(&tasklist_lock);
1173                 return retval;
1174         }
1175 }
1176
1177 #ifdef __ARCH_WANT_SYS_GETPGRP
1178
1179 asmlinkage long sys_getpgrp(void)
1180 {
1181         /* SMP - assuming writes are word atomic this is fine */
1182         return process_group(current);
1183 }
1184
1185 #endif
1186
1187 asmlinkage long sys_getsid(pid_t pid)
1188 {
1189         if (!pid) {
1190                 return current->signal->session;
1191         } else {
1192                 int retval;
1193                 struct task_struct *p;
1194
1195                 read_lock(&tasklist_lock);
1196                 p = find_task_by_pid(pid);
1197
1198                 retval = -ESRCH;
1199                 if(p) {
1200                         retval = security_task_getsid(p);
1201                         if (!retval)
1202                                 retval = p->signal->session;
1203                 }
1204                 read_unlock(&tasklist_lock);
1205                 return retval;
1206         }
1207 }
1208
1209 asmlinkage long sys_setsid(void)
1210 {
1211         struct pid *pid;
1212         int err = -EPERM;
1213
1214         if (!thread_group_leader(current))
1215                 return -EINVAL;
1216
1217         down(&tty_sem);
1218         write_lock_irq(&tasklist_lock);
1219
1220         pid = find_pid(PIDTYPE_PGID, current->pid);
1221         if (pid)
1222                 goto out;
1223
1224         current->signal->leader = 1;
1225         __set_special_pids(current->pid, current->pid);
1226         current->signal->tty = NULL;
1227         current->signal->tty_old_pgrp = 0;
1228         err = process_group(current);
1229 out:
1230         write_unlock_irq(&tasklist_lock);
1231         up(&tty_sem);
1232         return err;
1233 }
1234
1235 /*
1236  * Supplementary group IDs
1237  */
1238
1239 /* init to 2 - one for init_task, one to ensure it is never freed */
1240 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1241
1242 struct group_info *groups_alloc(int gidsetsize)
1243 {
1244         struct group_info *group_info;
1245         int nblocks;
1246         int i;
1247
1248         nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1249         /* Make sure we always allocate at least one indirect block pointer */
1250         nblocks = nblocks ? : 1;
1251         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1252         if (!group_info)
1253                 return NULL;
1254         group_info->ngroups = gidsetsize;
1255         group_info->nblocks = nblocks;
1256         atomic_set(&group_info->usage, 1);
1257
1258         if (gidsetsize <= NGROUPS_SMALL) {
1259                 group_info->blocks[0] = group_info->small_block;
1260         } else {
1261                 for (i = 0; i < nblocks; i++) {
1262                         gid_t *b;
1263                         b = (void *)__get_free_page(GFP_USER);
1264                         if (!b)
1265                                 goto out_undo_partial_alloc;
1266                         group_info->blocks[i] = b;
1267                 }
1268         }
1269         return group_info;
1270
1271 out_undo_partial_alloc:
1272         while (--i >= 0) {
1273                 free_page((unsigned long)group_info->blocks[i]);
1274         }
1275         kfree(group_info);
1276         return NULL;
1277 }
1278
1279 EXPORT_SYMBOL(groups_alloc);
1280
1281 void groups_free(struct group_info *group_info)
1282 {
1283         if (group_info->blocks[0] != group_info->small_block) {
1284                 int i;
1285                 for (i = 0; i < group_info->nblocks; i++)
1286                         free_page((unsigned long)group_info->blocks[i]);
1287         }
1288         kfree(group_info);
1289 }
1290
1291 EXPORT_SYMBOL(groups_free);
1292
1293 /* export the group_info to a user-space array */
1294 static int groups_to_user(gid_t __user *grouplist,
1295     struct group_info *group_info)
1296 {
1297         int i;
1298         int count = group_info->ngroups;
1299
1300         for (i = 0; i < group_info->nblocks; i++) {
1301                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1302                 int off = i * NGROUPS_PER_BLOCK;
1303                 int len = cp_count * sizeof(*grouplist);
1304
1305                 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1306                         return -EFAULT;
1307
1308                 count -= cp_count;
1309         }
1310         return 0;
1311 }
1312
1313 /* fill a group_info from a user-space array - it must be allocated already */
1314 static int groups_from_user(struct group_info *group_info,
1315     gid_t __user *grouplist)
1316  {
1317         int i;
1318         int count = group_info->ngroups;
1319
1320         for (i = 0; i < group_info->nblocks; i++) {
1321                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1322                 int off = i * NGROUPS_PER_BLOCK;
1323                 int len = cp_count * sizeof(*grouplist);
1324
1325                 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1326                         return -EFAULT;
1327
1328                 count -= cp_count;
1329         }
1330         return 0;
1331 }
1332
1333 /* a simple Shell sort */
1334 static void groups_sort(struct group_info *group_info)
1335 {
1336         int base, max, stride;
1337         int gidsetsize = group_info->ngroups;
1338
1339         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1340                 ; /* nothing */
1341         stride /= 3;
1342
1343         while (stride) {
1344                 max = gidsetsize - stride;
1345                 for (base = 0; base < max; base++) {
1346                         int left = base;
1347                         int right = left + stride;
1348                         gid_t tmp = GROUP_AT(group_info, right);
1349
1350                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1351                                 GROUP_AT(group_info, right) =
1352                                     GROUP_AT(group_info, left);
1353                                 right = left;
1354                                 left -= stride;
1355                         }
1356                         GROUP_AT(group_info, right) = tmp;
1357                 }
1358                 stride /= 3;
1359         }
1360 }
1361
1362 /* a simple bsearch */
1363 int groups_search(struct group_info *group_info, gid_t grp)
1364 {
1365         int left, right;
1366
1367         if (!group_info)
1368                 return 0;
1369
1370         left = 0;
1371         right = group_info->ngroups;
1372         while (left < right) {
1373                 int mid = (left+right)/2;
1374                 int cmp = grp - GROUP_AT(group_info, mid);
1375                 if (cmp > 0)
1376                         left = mid + 1;
1377                 else if (cmp < 0)
1378                         right = mid;
1379                 else
1380                         return 1;
1381         }
1382         return 0;
1383 }
1384
1385 /* validate and set current->group_info */
1386 int set_current_groups(struct group_info *group_info)
1387 {
1388         int retval;
1389         struct group_info *old_info;
1390
1391         retval = security_task_setgroups(group_info);
1392         if (retval)
1393                 return retval;
1394
1395         groups_sort(group_info);
1396         get_group_info(group_info);
1397
1398         task_lock(current);
1399         old_info = current->group_info;
1400         current->group_info = group_info;
1401         task_unlock(current);
1402
1403         put_group_info(old_info);
1404
1405         return 0;
1406 }
1407
1408 EXPORT_SYMBOL(set_current_groups);
1409
1410 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1411 {
1412         int i = 0;
1413
1414         /*
1415          *      SMP: Nobody else can change our grouplist. Thus we are
1416          *      safe.
1417          */
1418
1419         if (gidsetsize < 0)
1420                 return -EINVAL;
1421
1422         /* no need to grab task_lock here; it cannot change */
1423         get_group_info(current->group_info);
1424         i = current->group_info->ngroups;
1425         if (gidsetsize) {
1426                 if (i > gidsetsize) {
1427                         i = -EINVAL;
1428                         goto out;
1429                 }
1430                 if (groups_to_user(grouplist, current->group_info)) {
1431                         i = -EFAULT;
1432                         goto out;
1433                 }
1434         }
1435 out:
1436         put_group_info(current->group_info);
1437         return i;
1438 }
1439
1440 /*
1441  *      SMP: Our groups are copy-on-write. We can set them safely
1442  *      without another task interfering.
1443  */
1444
1445 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1446 {
1447         struct group_info *group_info;
1448         int retval;
1449
1450         if (!capable(CAP_SETGID))
1451                 return -EPERM;
1452         if ((unsigned)gidsetsize > NGROUPS_MAX)
1453                 return -EINVAL;
1454
1455         group_info = groups_alloc(gidsetsize);
1456         if (!group_info)
1457                 return -ENOMEM;
1458         retval = groups_from_user(group_info, grouplist);
1459         if (retval) {
1460                 put_group_info(group_info);
1461                 return retval;
1462         }
1463
1464         retval = set_current_groups(group_info);
1465         put_group_info(group_info);
1466
1467         return retval;
1468 }
1469
1470 /*
1471  * Check whether we're fsgid/egid or in the supplemental group..
1472  */
1473 int in_group_p(gid_t grp)
1474 {
1475         int retval = 1;
1476         if (grp != current->fsgid) {
1477                 get_group_info(current->group_info);
1478                 retval = groups_search(current->group_info, grp);
1479                 put_group_info(current->group_info);
1480         }
1481         return retval;
1482 }
1483
1484 EXPORT_SYMBOL(in_group_p);
1485
1486 int in_egroup_p(gid_t grp)
1487 {
1488         int retval = 1;
1489         if (grp != current->egid) {
1490                 get_group_info(current->group_info);
1491                 retval = groups_search(current->group_info, grp);
1492                 put_group_info(current->group_info);
1493         }
1494         return retval;
1495 }
1496
1497 EXPORT_SYMBOL(in_egroup_p);
1498
1499 DECLARE_RWSEM(uts_sem);
1500
1501 EXPORT_SYMBOL(uts_sem);
1502
1503 asmlinkage long sys_newuname(struct new_utsname __user * name)
1504 {
1505         int errno = 0;
1506
1507         down_read(&uts_sem);
1508         if (copy_to_user(name,&system_utsname,sizeof *name))
1509                 errno = -EFAULT;
1510         up_read(&uts_sem);
1511         return errno;
1512 }
1513
1514 asmlinkage long sys_sethostname(char __user *name, int len)
1515 {
1516         int errno;
1517         char tmp[__NEW_UTS_LEN];
1518
1519         if (!capable(CAP_SYS_ADMIN))
1520                 return -EPERM;
1521         if (len < 0 || len > __NEW_UTS_LEN)
1522                 return -EINVAL;
1523         down_write(&uts_sem);
1524         errno = -EFAULT;
1525         if (!copy_from_user(tmp, name, len)) {
1526                 memcpy(system_utsname.nodename, tmp, len);
1527                 system_utsname.nodename[len] = 0;
1528                 errno = 0;
1529         }
1530         up_write(&uts_sem);
1531         return errno;
1532 }
1533
1534 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1535
1536 asmlinkage long sys_gethostname(char __user *name, int len)
1537 {
1538         int i, errno;
1539
1540         if (len < 0)
1541                 return -EINVAL;
1542         down_read(&uts_sem);
1543         i = 1 + strlen(system_utsname.nodename);
1544         if (i > len)
1545                 i = len;
1546         errno = 0;
1547         if (copy_to_user(name, system_utsname.nodename, i))
1548                 errno = -EFAULT;
1549         up_read(&uts_sem);
1550         return errno;
1551 }
1552
1553 #endif
1554
1555 /*
1556  * Only setdomainname; getdomainname can be implemented by calling
1557  * uname()
1558  */
1559 asmlinkage long sys_setdomainname(char __user *name, int len)
1560 {
1561         int errno;
1562         char tmp[__NEW_UTS_LEN];
1563
1564         if (!capable(CAP_SYS_ADMIN))
1565                 return -EPERM;
1566         if (len < 0 || len > __NEW_UTS_LEN)
1567                 return -EINVAL;
1568
1569         down_write(&uts_sem);
1570         errno = -EFAULT;
1571         if (!copy_from_user(tmp, name, len)) {
1572                 memcpy(system_utsname.domainname, tmp, len);
1573                 system_utsname.domainname[len] = 0;
1574                 errno = 0;
1575         }
1576         up_write(&uts_sem);
1577         return errno;
1578 }
1579
1580 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1581 {
1582         if (resource >= RLIM_NLIMITS)
1583                 return -EINVAL;
1584         else {
1585                 struct rlimit value;
1586                 task_lock(current->group_leader);
1587                 value = current->signal->rlim[resource];
1588                 task_unlock(current->group_leader);
1589                 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1590         }
1591 }
1592
1593 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1594
1595 /*
1596  *      Back compatibility for getrlimit. Needed for some apps.
1597  */
1598
1599 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1600 {
1601         struct rlimit x;
1602         if (resource >= RLIM_NLIMITS)
1603                 return -EINVAL;
1604
1605         task_lock(current->group_leader);
1606         x = current->signal->rlim[resource];
1607         task_unlock(current->group_leader);
1608         if(x.rlim_cur > 0x7FFFFFFF)
1609                 x.rlim_cur = 0x7FFFFFFF;
1610         if(x.rlim_max > 0x7FFFFFFF)
1611                 x.rlim_max = 0x7FFFFFFF;
1612         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1613 }
1614
1615 #endif
1616
1617 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1618 {
1619         struct rlimit new_rlim, *old_rlim;
1620         int retval;
1621
1622         if (resource >= RLIM_NLIMITS)
1623                 return -EINVAL;
1624         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1625                 return -EFAULT;
1626        if (new_rlim.rlim_cur > new_rlim.rlim_max)
1627                return -EINVAL;
1628         old_rlim = current->signal->rlim + resource;
1629         if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1630             !capable(CAP_SYS_RESOURCE))
1631                 return -EPERM;
1632         if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1633                         return -EPERM;
1634
1635         retval = security_task_setrlimit(resource, &new_rlim);
1636         if (retval)
1637                 return retval;
1638
1639         task_lock(current->group_leader);
1640         *old_rlim = new_rlim;
1641         task_unlock(current->group_leader);
1642
1643         if (resource == RLIMIT_CPU && new_rlim.rlim_cur != RLIM_INFINITY &&
1644             (cputime_eq(current->signal->it_prof_expires, cputime_zero) ||
1645              new_rlim.rlim_cur <= cputime_to_secs(
1646                      current->signal->it_prof_expires))) {
1647                 cputime_t cputime = secs_to_cputime(new_rlim.rlim_cur);
1648                 read_lock(&tasklist_lock);
1649                 spin_lock_irq(&current->sighand->siglock);
1650                 set_process_cpu_timer(current, CPUCLOCK_PROF,
1651                                       &cputime, NULL);
1652                 spin_unlock_irq(&current->sighand->siglock);
1653                 read_unlock(&tasklist_lock);
1654         }
1655
1656         return 0;
1657 }
1658
1659 /*
1660  * It would make sense to put struct rusage in the task_struct,
1661  * except that would make the task_struct be *really big*.  After
1662  * task_struct gets moved into malloc'ed memory, it would
1663  * make sense to do this.  It will make moving the rest of the information
1664  * a lot simpler!  (Which we're not doing right now because we're not
1665  * measuring them yet).
1666  *
1667  * This expects to be called with tasklist_lock read-locked or better,
1668  * and the siglock not locked.  It may momentarily take the siglock.
1669  *
1670  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1671  * races with threads incrementing their own counters.  But since word
1672  * reads are atomic, we either get new values or old values and we don't
1673  * care which for the sums.  We always take the siglock to protect reading
1674  * the c* fields from p->signal from races with exit.c updating those
1675  * fields when reaping, so a sample either gets all the additions of a
1676  * given child after it's reaped, or none so this sample is before reaping.
1677  */
1678
1679 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1680 {
1681         struct task_struct *t;
1682         unsigned long flags;
1683         cputime_t utime, stime;
1684
1685         memset((char *) r, 0, sizeof *r);
1686
1687         if (unlikely(!p->signal))
1688                 return;
1689
1690         switch (who) {
1691                 case RUSAGE_CHILDREN:
1692                         spin_lock_irqsave(&p->sighand->siglock, flags);
1693                         utime = p->signal->cutime;
1694                         stime = p->signal->cstime;
1695                         r->ru_nvcsw = p->signal->cnvcsw;
1696                         r->ru_nivcsw = p->signal->cnivcsw;
1697                         r->ru_minflt = p->signal->cmin_flt;
1698                         r->ru_majflt = p->signal->cmaj_flt;
1699                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1700                         cputime_to_timeval(utime, &r->ru_utime);
1701                         cputime_to_timeval(stime, &r->ru_stime);
1702                         break;
1703                 case RUSAGE_SELF:
1704                         spin_lock_irqsave(&p->sighand->siglock, flags);
1705                         utime = stime = cputime_zero;
1706                         goto sum_group;
1707                 case RUSAGE_BOTH:
1708                         spin_lock_irqsave(&p->sighand->siglock, flags);
1709                         utime = p->signal->cutime;
1710                         stime = p->signal->cstime;
1711                         r->ru_nvcsw = p->signal->cnvcsw;
1712                         r->ru_nivcsw = p->signal->cnivcsw;
1713                         r->ru_minflt = p->signal->cmin_flt;
1714                         r->ru_majflt = p->signal->cmaj_flt;
1715                 sum_group:
1716                         utime = cputime_add(utime, p->signal->utime);
1717                         stime = cputime_add(stime, p->signal->stime);
1718                         r->ru_nvcsw += p->signal->nvcsw;
1719                         r->ru_nivcsw += p->signal->nivcsw;
1720                         r->ru_minflt += p->signal->min_flt;
1721                         r->ru_majflt += p->signal->maj_flt;
1722                         t = p;
1723                         do {
1724                                 utime = cputime_add(utime, t->utime);
1725                                 stime = cputime_add(stime, t->stime);
1726                                 r->ru_nvcsw += t->nvcsw;
1727                                 r->ru_nivcsw += t->nivcsw;
1728                                 r->ru_minflt += t->min_flt;
1729                                 r->ru_majflt += t->maj_flt;
1730                                 t = next_thread(t);
1731                         } while (t != p);
1732                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1733                         cputime_to_timeval(utime, &r->ru_utime);
1734                         cputime_to_timeval(stime, &r->ru_stime);
1735                         break;
1736                 default:
1737                         BUG();
1738         }
1739 }
1740
1741 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1742 {
1743         struct rusage r;
1744         read_lock(&tasklist_lock);
1745         k_getrusage(p, who, &r);
1746         read_unlock(&tasklist_lock);
1747         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1748 }
1749
1750 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1751 {
1752         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1753                 return -EINVAL;
1754         return getrusage(current, who, ru);
1755 }
1756
1757 asmlinkage long sys_umask(int mask)
1758 {
1759         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1760         return mask;
1761 }
1762
1763 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1764                           unsigned long arg4, unsigned long arg5)
1765 {
1766         long error;
1767
1768         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1769         if (error)
1770                 return error;
1771
1772         switch (option) {
1773                 case PR_SET_PDEATHSIG:
1774                         if (!valid_signal(arg2)) {
1775                                 error = -EINVAL;
1776                                 break;
1777                         }
1778                         current->pdeath_signal = arg2;
1779                         break;
1780                 case PR_GET_PDEATHSIG:
1781                         error = put_user(current->pdeath_signal, (int __user *)arg2);
1782                         break;
1783                 case PR_GET_DUMPABLE:
1784                         error = current->mm->dumpable;
1785                         break;
1786                 case PR_SET_DUMPABLE:
1787                         if (arg2 < 0 || arg2 > 2) {
1788                                 error = -EINVAL;
1789                                 break;
1790                         }
1791                         current->mm->dumpable = arg2;
1792                         break;
1793
1794                 case PR_SET_UNALIGN:
1795                         error = SET_UNALIGN_CTL(current, arg2);
1796                         break;
1797                 case PR_GET_UNALIGN:
1798                         error = GET_UNALIGN_CTL(current, arg2);
1799                         break;
1800                 case PR_SET_FPEMU:
1801                         error = SET_FPEMU_CTL(current, arg2);
1802                         break;
1803                 case PR_GET_FPEMU:
1804                         error = GET_FPEMU_CTL(current, arg2);
1805                         break;
1806                 case PR_SET_FPEXC:
1807                         error = SET_FPEXC_CTL(current, arg2);
1808                         break;
1809                 case PR_GET_FPEXC:
1810                         error = GET_FPEXC_CTL(current, arg2);
1811                         break;
1812                 case PR_GET_TIMING:
1813                         error = PR_TIMING_STATISTICAL;
1814                         break;
1815                 case PR_SET_TIMING:
1816                         if (arg2 == PR_TIMING_STATISTICAL)
1817                                 error = 0;
1818                         else
1819                                 error = -EINVAL;
1820                         break;
1821
1822                 case PR_GET_KEEPCAPS:
1823                         if (current->keep_capabilities)
1824                                 error = 1;
1825                         break;
1826                 case PR_SET_KEEPCAPS:
1827                         if (arg2 != 0 && arg2 != 1) {
1828                                 error = -EINVAL;
1829                                 break;
1830                         }
1831                         current->keep_capabilities = arg2;
1832                         break;
1833                 case PR_SET_NAME: {
1834                         struct task_struct *me = current;
1835                         unsigned char ncomm[sizeof(me->comm)];
1836
1837                         ncomm[sizeof(me->comm)-1] = 0;
1838                         if (strncpy_from_user(ncomm, (char __user *)arg2,
1839                                                 sizeof(me->comm)-1) < 0)
1840                                 return -EFAULT;
1841                         set_task_comm(me, ncomm);
1842                         return 0;
1843                 }
1844                 case PR_GET_NAME: {
1845                         struct task_struct *me = current;
1846                         unsigned char tcomm[sizeof(me->comm)];
1847
1848                         get_task_comm(tcomm, me);
1849                         if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
1850                                 return -EFAULT;
1851                         return 0;
1852                 }
1853                 default:
1854                         error = -EINVAL;
1855                         break;
1856         }
1857         return error;
1858 }