SAFE public projects git trees. - safe/jmp/linux-2.6/blob - 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/workqueue.h>
  20 #include <linux/device.h>
  21 #include <linux/key.h>
  22 #include <linux/times.h>
  23 #include <linux/posix-timers.h>
  24 #include <linux/security.h>
  25 #include <linux/dcookies.h>
  26 #include <linux/suspend.h>
  27 #include <linux/tty.h>
  28
  29 #include <linux/compat.h>
  30 #include <linux/syscalls.h>
  31
  32 #include <asm/uaccess.h>
  33 #include <asm/io.h>
  34 #include <asm/unistd.h>
  35
  36 #ifndef SET_UNALIGN_CTL
  37 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  38 #endif
  39 #ifndef GET_UNALIGN_CTL
  40 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  41 #endif
  42 #ifndef SET_FPEMU_CTL
  43 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
  44 #endif
  45 #ifndef GET_FPEMU_CTL
  46 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
  47 #endif
  48 #ifndef SET_FPEXC_CTL
  49 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
  50 #endif
  51 #ifndef GET_FPEXC_CTL
  52 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
  53 #endif
  54
  55 /*
  56  * this is where the system-wide overflow UID and GID are defined, for
  57  * architectures that now have 32-bit UID/GID but didn't in the past
  58  */
  59
  60 int overflowuid = DEFAULT_OVERFLOWUID;
  61 int overflowgid = DEFAULT_OVERFLOWGID;
  62
  63 #ifdef CONFIG_UID16
  64 EXPORT_SYMBOL(overflowuid);
  65 EXPORT_SYMBOL(overflowgid);
  66 #endif
  67
  68 /*
  69  * the same as above, but for filesystems which can only store a 16-bit
  70  * UID and GID. as such, this is needed on all architectures
  71  */
  72
  73 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
  74 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
  75
  76 EXPORT_SYMBOL(fs_overflowuid);
  77 EXPORT_SYMBOL(fs_overflowgid);
  78
  79 /*
  80  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
  81  */
  82
  83 int C_A_D = 1;
  84 int cad_pid = 1;
  85
  86 /*
  87  *      Notifier list for kernel code which wants to be called
  88  *      at shutdown. This is used to stop any idling DMA operations
  89  *      and the like.
  90  */
  91
  92 static struct notifier_block *reboot_notifier_list;
  93 static DEFINE_RWLOCK(notifier_lock);
  94
  95 /**
  96  *      notifier_chain_register - Add notifier to a notifier chain
  97  *      @list: Pointer to root list pointer
  98  *      @n: New entry in notifier chain
  99  *
 100  *      Adds a notifier to a notifier chain.
 101  *
 102  *      Currently always returns zero.
 103  */
 104
 105 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
 106 {
 107         write_lock(&notifier_lock);
 108         while(*list)
 109         {
 110                 if(n->priority > (*list)->priority)
 111                         break;
 112                 list= &((*list)->next);
 113         }
 114         n->next = *list;
 115         *list=n;
 116         write_unlock(&notifier_lock);
 117         return 0;
 118 }
 119
 120 EXPORT_SYMBOL(notifier_chain_register);
 121
 122 /**
 123  *      notifier_chain_unregister - Remove notifier from a notifier chain
 124  *      @nl: Pointer to root list pointer
 125  *      @n: New entry in notifier chain
 126  *
 127  *      Removes a notifier from a notifier chain.
 128  *
 129  *      Returns zero on success, or %-ENOENT on failure.
 130  */
 131
 132 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
 133 {
 134         write_lock(&notifier_lock);
 135         while((*nl)!=NULL)
 136         {
 137                 if((*nl)==n)
 138                 {
 139                         *nl=n->next;
 140                         write_unlock(&notifier_lock);
 141                         return 0;
 142                 }
 143                 nl=&((*nl)->next);
 144         }
 145         write_unlock(&notifier_lock);
 146         return -ENOENT;
 147 }
 148
 149 EXPORT_SYMBOL(notifier_chain_unregister);
 150
 151 /**
 152  *      notifier_call_chain - Call functions in a notifier chain
 153  *      @n: Pointer to root pointer of notifier chain
 154  *      @val: Value passed unmodified to notifier function
 155  *      @v: Pointer passed unmodified to notifier function
 156  *
 157  *      Calls each function in a notifier chain in turn.
 158  *
 159  *      If the return value of the notifier can be and'd
 160  *      with %NOTIFY_STOP_MASK, then notifier_call_chain
 161  *      will return immediately, with the return value of
 162  *      the notifier function which halted execution.
 163  *      Otherwise, the return value is the return value
 164  *      of the last notifier function called.
 165  */
 166
 167 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
 168 {
 169         int ret=NOTIFY_DONE;
 170         struct notifier_block *nb = *n;
 171
 172         while(nb)
 173         {
 174                 ret=nb->notifier_call(nb,val,v);
 175                 if(ret&NOTIFY_STOP_MASK)
 176                 {
 177                         return ret;
 178                 }
 179                 nb=nb->next;
 180         }
 181         return ret;
 182 }
 183
 184 EXPORT_SYMBOL(notifier_call_chain);
 185
 186 /**
 187  *      register_reboot_notifier - Register function to be called at reboot time
 188  *      @nb: Info about notifier function to be called
 189  *
 190  *      Registers a function with the list of functions
 191  *      to be called at reboot time.
 192  *
 193  *      Currently always returns zero, as notifier_chain_register
 194  *      always returns zero.
 195  */
 196
 197 int register_reboot_notifier(struct notifier_block * nb)
 198 {
 199         return notifier_chain_register(&reboot_notifier_list, nb);
 200 }
 201
 202 EXPORT_SYMBOL(register_reboot_notifier);
 203
 204 /**
 205  *      unregister_reboot_notifier - Unregister previously registered reboot notifier
 206  *      @nb: Hook to be unregistered
 207  *
 208  *      Unregisters a previously registered reboot
 209  *      notifier function.
 210  *
 211  *      Returns zero on success, or %-ENOENT on failure.
 212  */
 213
 214 int unregister_reboot_notifier(struct notifier_block * nb)
 215 {
 216         return notifier_chain_unregister(&reboot_notifier_list, nb);
 217 }
 218
 219 EXPORT_SYMBOL(unregister_reboot_notifier);
 220
 221 static int set_one_prio(struct task_struct *p, int niceval, int error)
 222 {
 223         int no_nice;
 224
 225         if (p->uid != current->euid &&
 226                 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
 227                 error = -EPERM;
 228                 goto out;
 229         }
 230         if (niceval < task_nice(p) && !can_nice(p, niceval)) {
 231                 error = -EACCES;
 232                 goto out;
 233         }
 234         no_nice = security_task_setnice(p, niceval);
 235         if (no_nice) {
 236                 error = no_nice;
 237                 goto out;
 238         }
 239         if (error == -ESRCH)
 240                 error = 0;
 241         set_user_nice(p, niceval);
 242 out:
 243         return error;
 244 }
 245
 246 asmlinkage long sys_setpriority(int which, int who, int niceval)
 247 {
 248         struct task_struct *g, *p;
 249         struct user_struct *user;
 250         int error = -EINVAL;
 251
 252         if (which > 2 || which < 0)
 253                 goto out;
 254
 255         /* normalize: avoid signed division (rounding problems) */
 256         error = -ESRCH;
 257         if (niceval < -20)
 258                 niceval = -20;
 259         if (niceval > 19)
 260                 niceval = 19;
 261
 262         read_lock(&tasklist_lock);
 263         switch (which) {
 264                 case PRIO_PROCESS:
 265                         if (!who)
 266                                 who = current->pid;
 267                         p = find_task_by_pid(who);
 268                         if (p)
 269                                 error = set_one_prio(p, niceval, error);
 270                         break;
 271                 case PRIO_PGRP:
 272                         if (!who)
 273                                 who = process_group(current);
 274                         do_each_task_pid(who, PIDTYPE_PGID, p) {
 275                                 error = set_one_prio(p, niceval, error);
 276                         } while_each_task_pid(who, PIDTYPE_PGID, p);
 277                         break;
 278                 case PRIO_USER:
 279                         user = current->user;
 280                         if (!who)
 281                                 who = current->uid;
 282                         else
 283                                 if ((who != current->uid) && !(user = find_user(who)))
 284                                         goto out_unlock;        /* No processes for this user */
 285
 286                         do_each_thread(g, p)
 287                                 if (p->uid == who)
 288                                         error = set_one_prio(p, niceval, error);
 289                         while_each_thread(g, p);
 290                         if (who != current->uid)
 291                                 free_uid(user);         /* For find_user() */
 292                         break;
 293         }
 294 out_unlock:
 295         read_unlock(&tasklist_lock);
 296 out:
 297         return error;
 298 }
 299
 300 /*
 301  * Ugh. To avoid negative return values, "getpriority()" will
 302  * not return the normal nice-value, but a negated value that
 303  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 304  * to stay compatible.
 305  */
 306 asmlinkage long sys_getpriority(int which, int who)
 307 {
 308         struct task_struct *g, *p;
 309         struct user_struct *user;
 310         long niceval, retval = -ESRCH;
 311
 312         if (which > 2 || which < 0)
 313                 return -EINVAL;
 314
 315         read_lock(&tasklist_lock);
 316         switch (which) {
 317                 case PRIO_PROCESS:
 318                         if (!who)
 319                                 who = current->pid;
 320                         p = find_task_by_pid(who);
 321                         if (p) {
 322                                 niceval = 20 - task_nice(p);
 323                                 if (niceval > retval)
 324                                         retval = niceval;
 325                         }
 326                         break;
 327                 case PRIO_PGRP:
 328                         if (!who)
 329                                 who = process_group(current);
 330                         do_each_task_pid(who, PIDTYPE_PGID, p) {
 331                                 niceval = 20 - task_nice(p);
 332                                 if (niceval > retval)
 333                                         retval = niceval;
 334                         } while_each_task_pid(who, PIDTYPE_PGID, p);
 335                         break;
 336                 case PRIO_USER:
 337                         user = current->user;
 338                         if (!who)
 339                                 who = current->uid;
 340                         else
 341                                 if ((who != current->uid) && !(user = find_user(who)))
 342                                         goto out_unlock;        /* No processes for this user */
 343
 344                         do_each_thread(g, p)
 345                                 if (p->uid == who) {
 346                                         niceval = 20 - task_nice(p);
 347                                         if (niceval > retval)
 348                                                 retval = niceval;
 349                                 }
 350                         while_each_thread(g, p);
 351                         if (who != current->uid)
 352                                 free_uid(user);         /* for find_user() */
 353                         break;
 354         }
 355 out_unlock:
 356         read_unlock(&tasklist_lock);
 357
 358         return retval;
 359 }
 360
 361
 362 /*
 363  * Reboot system call: for obvious reasons only root may call it,
 364  * and even root needs to set up some magic numbers in the registers
 365  * so that some mistake won't make this reboot the whole machine.
 366  * You can also set the meaning of the ctrl-alt-del-key here.
 367  *
 368  * reboot doesn't sync: do that yourself before calling this.
 369  */
 370 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
 371 {
 372         char buffer[256];
 373
 374         /* We only trust the superuser with rebooting the system. */
 375         if (!capable(CAP_SYS_BOOT))
 376                 return -EPERM;
 377
 378         /* For safety, we require "magic" arguments. */
 379         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 380             (magic2 != LINUX_REBOOT_MAGIC2 &&
 381                         magic2 != LINUX_REBOOT_MAGIC2A &&
 382                         magic2 != LINUX_REBOOT_MAGIC2B &&
 383                         magic2 != LINUX_REBOOT_MAGIC2C))
 384                 return -EINVAL;
 385
 386         lock_kernel();
 387         switch (cmd) {
 388         case LINUX_REBOOT_CMD_RESTART:
 389                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 390                 system_state = SYSTEM_RESTART;
 391                 device_shutdown();
 392                 printk(KERN_EMERG "Restarting system.\n");
 393                 machine_restart(NULL);
 394                 break;
 395
 396         case LINUX_REBOOT_CMD_CAD_ON:
 397                 C_A_D = 1;
 398                 break;
 399
 400         case LINUX_REBOOT_CMD_CAD_OFF:
 401                 C_A_D = 0;
 402                 break;
 403
 404         case LINUX_REBOOT_CMD_HALT:
 405                 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
 406                 system_state = SYSTEM_HALT;
 407                 device_shutdown();
 408                 printk(KERN_EMERG "System halted.\n");
 409                 machine_halt();
 410                 unlock_kernel();
 411                 do_exit(0);
 412                 break;
 413
 414         case LINUX_REBOOT_CMD_POWER_OFF:
 415                 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
 416                 system_state = SYSTEM_POWER_OFF;
 417                 device_shutdown();
 418                 printk(KERN_EMERG "Power down.\n");
 419                 machine_power_off();
 420                 unlock_kernel();
 421                 do_exit(0);
 422                 break;
 423
 424         case LINUX_REBOOT_CMD_RESTART2:
 425                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 426                         unlock_kernel();
 427                         return -EFAULT;
 428                 }
 429                 buffer[sizeof(buffer) - 1] = '\0';
 430
 431                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
 432                 system_state = SYSTEM_RESTART;
 433                 device_shutdown();
 434                 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
 435                 machine_restart(buffer);
 436                 break;
 437
 438 #ifdef CONFIG_SOFTWARE_SUSPEND
 439         case LINUX_REBOOT_CMD_SW_SUSPEND:
 440                 {
 441                         int ret = software_suspend();
 442                         unlock_kernel();
 443                         return ret;
 444                 }
 445 #endif
 446
 447         default:
 448                 unlock_kernel();
 449                 return -EINVAL;
 450         }
 451         unlock_kernel();
 452         return 0;
 453 }
 454
 455 static void deferred_cad(void *dummy)
 456 {
 457         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 458         machine_restart(NULL);
 459 }
 460
 461 /*
 462  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 463  * As it's called within an interrupt, it may NOT sync: the only choice
 464  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 465  */
 466 void ctrl_alt_del(void)
 467 {
 468         static DECLARE_WORK(cad_work, deferred_cad, NULL);
 469
 470         if (C_A_D)
 471                 schedule_work(&cad_work);
 472         else
 473                 kill_proc(cad_pid, SIGINT, 1);
 474 }
 475
 476
 477 /*
 478  * Unprivileged users may change the real gid to the effective gid
 479  * or vice versa.  (BSD-style)
 480  *
 481  * If you set the real gid at all, or set the effective gid to a value not
 482  * equal to the real gid, then the saved gid is set to the new effective gid.
 483  *
 484  * This makes it possible for a setgid program to completely drop its
 485  * privileges, which is often a useful assertion to make when you are doing
 486  * a security audit over a program.
 487  *
 488  * The general idea is that a program which uses just setregid() will be
 489  * 100% compatible with BSD.  A program which uses just setgid() will be
 490  * 100% compatible with POSIX with saved IDs.
 491  *
 492  * SMP: There are not races, the GIDs are checked only by filesystem
 493  *      operations (as far as semantic preservation is concerned).
 494  */
 495 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 496 {
 497         int old_rgid = current->gid;
 498         int old_egid = current->egid;
 499         int new_rgid = old_rgid;
 500         int new_egid = old_egid;
 501         int retval;
 502
 503         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 504         if (retval)
 505                 return retval;
 506
 507         if (rgid != (gid_t) -1) {
 508                 if ((old_rgid == rgid) ||
 509                     (current->egid==rgid) ||
 510                     capable(CAP_SETGID))
 511                         new_rgid = rgid;
 512                 else
 513                         return -EPERM;
 514         }
 515         if (egid != (gid_t) -1) {
 516                 if ((old_rgid == egid) ||
 517                     (current->egid == egid) ||
 518                     (current->sgid == egid) ||
 519                     capable(CAP_SETGID))
 520                         new_egid = egid;
 521                 else {
 522                         return -EPERM;
 523                 }
 524         }
 525         if (new_egid != old_egid)
 526         {
 527                 current->mm->dumpable = 0;
 528                 smp_wmb();
 529         }
 530         if (rgid != (gid_t) -1 ||
 531             (egid != (gid_t) -1 && egid != old_rgid))
 532                 current->sgid = new_egid;
 533         current->fsgid = new_egid;
 534         current->egid = new_egid;
 535         current->gid = new_rgid;
 536         key_fsgid_changed(current);
 537         return 0;
 538 }
 539
 540 /*
 541  * setgid() is implemented like SysV w/ SAVED_IDS
 542  *
 543  * SMP: Same implicit races as above.
 544  */
 545 asmlinkage long sys_setgid(gid_t gid)
 546 {
 547         int old_egid = current->egid;
 548         int retval;
 549
 550         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 551         if (retval)
 552                 return retval;
 553
 554         if (capable(CAP_SETGID))
 555         {
 556                 if(old_egid != gid)
 557                 {
 558                         current->mm->dumpable=0;
 559                         smp_wmb();
 560                 }
 561                 current->gid = current->egid = current->sgid = current->fsgid = gid;
 562         }
 563         else if ((gid == current->gid) || (gid == current->sgid))
 564         {
 565                 if(old_egid != gid)
 566                 {
 567                         current->mm->dumpable=0;
 568                         smp_wmb();
 569                 }
 570                 current->egid = current->fsgid = gid;
 571         }
 572         else
 573                 return -EPERM;
 574
 575         key_fsgid_changed(current);
 576         return 0;
 577 }
 578
 579 static int set_user(uid_t new_ruid, int dumpclear)
 580 {
 581         struct user_struct *new_user;
 582
 583         new_user = alloc_uid(new_ruid);
 584         if (!new_user)
 585                 return -EAGAIN;
 586
 587         if (atomic_read(&new_user->processes) >=
 588                                 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
 589                         new_user != &root_user) {
 590                 free_uid(new_user);
 591                 return -EAGAIN;
 592         }
 593
 594         switch_uid(new_user);
 595
 596         if(dumpclear)
 597         {
 598                 current->mm->dumpable = 0;
 599                 smp_wmb();
 600         }
 601         current->uid = new_ruid;
 602         return 0;
 603 }
 604
 605 /*
 606  * Unprivileged users may change the real uid to the effective uid
 607  * or vice versa.  (BSD-style)
 608  *
 609  * If you set the real uid at all, or set the effective uid to a value not
 610  * equal to the real uid, then the saved uid is set to the new effective uid.
 611  *
 612  * This makes it possible for a setuid program to completely drop its
 613  * privileges, which is often a useful assertion to make when you are doing
 614  * a security audit over a program.
 615  *
 616  * The general idea is that a program which uses just setreuid() will be
 617  * 100% compatible with BSD.  A program which uses just setuid() will be
 618  * 100% compatible with POSIX with saved IDs.
 619  */
 620 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 621 {
 622         int old_ruid, old_euid, old_suid, new_ruid, new_euid;
 623         int retval;
 624
 625         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 626         if (retval)
 627                 return retval;
 628
 629         new_ruid = old_ruid = current->uid;
 630         new_euid = old_euid = current->euid;
 631         old_suid = current->suid;
 632
 633         if (ruid != (uid_t) -1) {
 634                 new_ruid = ruid;
 635                 if ((old_ruid != ruid) &&
 636                     (current->euid != ruid) &&
 637                     !capable(CAP_SETUID))
 638                         return -EPERM;
 639         }
 640
 641         if (euid != (uid_t) -1) {
 642                 new_euid = euid;
 643                 if ((old_ruid != euid) &&
 644                     (current->euid != euid) &&
 645                     (current->suid != euid) &&
 646                     !capable(CAP_SETUID))
 647                         return -EPERM;
 648         }
 649
 650         if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
 651                 return -EAGAIN;
 652
 653         if (new_euid != old_euid)
 654         {
 655                 current->mm->dumpable=0;
 656                 smp_wmb();
 657         }
 658         current->fsuid = current->euid = new_euid;
 659         if (ruid != (uid_t) -1 ||
 660             (euid != (uid_t) -1 && euid != old_ruid))
 661                 current->suid = current->euid;
 662         current->fsuid = current->euid;
 663
 664         key_fsuid_changed(current);
 665
 666         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
 667 }
 668
 669
 670
 671 /*
 672  * setuid() is implemented like SysV with SAVED_IDS
 673  *
 674  * Note that SAVED_ID's is deficient in that a setuid root program
 675  * like sendmail, for example, cannot set its uid to be a normal
 676  * user and then switch back, because if you're root, setuid() sets
 677  * the saved uid too.  If you don't like this, blame the bright people
 678  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 679  * will allow a root program to temporarily drop privileges and be able to
 680  * regain them by swapping the real and effective uid.
 681  */
 682 asmlinkage long sys_setuid(uid_t uid)
 683 {
 684         int old_euid = current->euid;
 685         int old_ruid, old_suid, new_ruid, new_suid;
 686         int retval;
 687
 688         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 689         if (retval)
 690                 return retval;
 691
 692         old_ruid = new_ruid = current->uid;
 693         old_suid = current->suid;
 694         new_suid = old_suid;
 695
 696         if (capable(CAP_SETUID)) {
 697                 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
 698                         return -EAGAIN;
 699                 new_suid = uid;
 700         } else if ((uid != current->uid) && (uid != new_suid))
 701                 return -EPERM;
 702
 703         if (old_euid != uid)
 704         {
 705                 current->mm->dumpable = 0;
 706                 smp_wmb();
 707         }
 708         current->fsuid = current->euid = uid;
 709         current->suid = new_suid;
 710
 711         key_fsuid_changed(current);
 712
 713         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
 714 }
 715
 716
 717 /*
 718  * This function implements a generic ability to update ruid, euid,
 719  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 720  */
 721 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 722 {
 723         int old_ruid = current->uid;
 724         int old_euid = current->euid;
 725         int old_suid = current->suid;
 726         int retval;
 727
 728         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 729         if (retval)
 730                 return retval;
 731
 732         if (!capable(CAP_SETUID)) {
 733                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
 734                     (ruid != current->euid) && (ruid != current->suid))
 735                         return -EPERM;
 736                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
 737                     (euid != current->euid) && (euid != current->suid))
 738                         return -EPERM;
 739                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
 740                     (suid != current->euid) && (suid != current->suid))
 741                         return -EPERM;
 742         }
 743         if (ruid != (uid_t) -1) {
 744                 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
 745                         return -EAGAIN;
 746         }
 747         if (euid != (uid_t) -1) {
 748                 if (euid != current->euid)
 749                 {
 750                         current->mm->dumpable = 0;
 751                         smp_wmb();
 752                 }
 753                 current->euid = euid;
 754         }
 755         current->fsuid = current->euid;
 756         if (suid != (uid_t) -1)
 757                 current->suid = suid;
 758
 759         key_fsuid_changed(current);
 760
 761         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
 762 }
 763
 764 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
 765 {
 766         int retval;
 767
 768         if (!(retval = put_user(current->uid, ruid)) &&
 769             !(retval = put_user(current->euid, euid)))
 770                 retval = put_user(current->suid, suid);
 771
 772         return retval;
 773 }
 774
 775 /*
 776  * Same as above, but for rgid, egid, sgid.
 777  */
 778 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 779 {
 780         int retval;
 781
 782         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 783         if (retval)
 784                 return retval;
 785
 786         if (!capable(CAP_SETGID)) {
 787                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
 788                     (rgid != current->egid) && (rgid != current->sgid))
 789                         return -EPERM;
 790                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
 791                     (egid != current->egid) && (egid != current->sgid))
 792                         return -EPERM;
 793                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
 794                     (sgid != current->egid) && (sgid != current->sgid))
 795                         return -EPERM;
 796         }
 797         if (egid != (gid_t) -1) {
 798                 if (egid != current->egid)
 799                 {
 800                         current->mm->dumpable = 0;
 801                         smp_wmb();
 802                 }
 803                 current->egid = egid;
 804         }
 805         current->fsgid = current->egid;
 806         if (rgid != (gid_t) -1)
 807                 current->gid = rgid;
 808         if (sgid != (gid_t) -1)
 809                 current->sgid = sgid;
 810
 811         key_fsgid_changed(current);
 812         return 0;
 813 }
 814
 815 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
 816 {
 817         int retval;
 818
 819         if (!(retval = put_user(current->gid, rgid)) &&
 820             !(retval = put_user(current->egid, egid)))
 821                 retval = put_user(current->sgid, sgid);
 822
 823         return retval;
 824 }
 825
 826
 827 /*
 828  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 829  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 830  * whatever uid it wants to). It normally shadows "euid", except when
 831  * explicitly set by setfsuid() or for access..
 832  */
 833 asmlinkage long sys_setfsuid(uid_t uid)
 834 {
 835         int old_fsuid;
 836
 837         old_fsuid = current->fsuid;
 838         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
 839                 return old_fsuid;
 840
 841         if (uid == current->uid || uid == current->euid ||
 842             uid == current->suid || uid == current->fsuid ||
 843             capable(CAP_SETUID))
 844         {
 845                 if (uid != old_fsuid)
 846                 {
 847                         current->mm->dumpable = 0;
 848                         smp_wmb();
 849                 }
 850                 current->fsuid = uid;
 851         }
 852
 853         key_fsuid_changed(current);
 854
 855         security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
 856
 857         return old_fsuid;
 858 }
 859
 860 /*
 861  * Samma på svenska..
 862  */
 863 asmlinkage long sys_setfsgid(gid_t gid)
 864 {
 865         int old_fsgid;
 866
 867         old_fsgid = current->fsgid;
 868         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 869                 return old_fsgid;
 870
 871         if (gid == current->gid || gid == current->egid ||
 872             gid == current->sgid || gid == current->fsgid ||
 873             capable(CAP_SETGID))
 874         {
 875                 if (gid != old_fsgid)
 876                 {
 877                         current->mm->dumpable = 0;
 878                         smp_wmb();
 879                 }
 880                 current->fsgid = gid;
 881                 key_fsgid_changed(current);
 882         }
 883         return old_fsgid;
 884 }
 885
 886 asmlinkage long sys_times(struct tms __user * tbuf)
 887 {
 888         /*
 889          *      In the SMP world we might just be unlucky and have one of
 890          *      the times increment as we use it. Since the value is an
 891          *      atomically safe type this is just fine. Conceptually its
 892          *      as if the syscall took an instant longer to occur.
 893          */
 894         if (tbuf) {
 895                 struct tms tmp;
 896                 struct task_struct *tsk = current;
 897                 struct task_struct *t;
 898                 cputime_t utime, stime, cutime, cstime;
 899
 900                 read_lock(&tasklist_lock);
 901                 utime = tsk->signal->utime;
 902                 stime = tsk->signal->stime;
 903                 t = tsk;
 904                 do {
 905                         utime = cputime_add(utime, t->utime);
 906                         stime = cputime_add(stime, t->stime);
 907                         t = next_thread(t);
 908                 } while (t != tsk);
 909
 910                 /*
 911                  * While we have tasklist_lock read-locked, no dying thread
 912                  * can be updating current->signal->[us]time.  Instead,
 913                  * we got their counts included in the live thread loop.
 914                  * However, another thread can come in right now and
 915                  * do a wait call that updates current->signal->c[us]time.
 916                  * To make sure we always see that pair updated atomically,
 917                  * we take the siglock around fetching them.
 918                  */
 919                 spin_lock_irq(&tsk->sighand->siglock);
 920                 cutime = tsk->signal->cutime;
 921                 cstime = tsk->signal->cstime;
 922                 spin_unlock_irq(&tsk->sighand->siglock);
 923                 read_unlock(&tasklist_lock);
 924
 925                 tmp.tms_utime = cputime_to_clock_t(utime);
 926                 tmp.tms_stime = cputime_to_clock_t(stime);
 927                 tmp.tms_cutime = cputime_to_clock_t(cutime);
 928                 tmp.tms_cstime = cputime_to_clock_t(cstime);
 929                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
 930                         return -EFAULT;
 931         }
 932         return (long) jiffies_64_to_clock_t(get_jiffies_64());
 933 }
 934
 935 /*
 936  * This needs some heavy checking ...
 937  * I just haven't the stomach for it. I also don't fully
 938  * understand sessions/pgrp etc. Let somebody who does explain it.
 939  *
 940  * OK, I think I have the protection semantics right.... this is really
 941  * only important on a multi-user system anyway, to make sure one user
 942  * can't send a signal to a process owned by another.  -TYT, 12/12/91
 943  *
 944  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
 945  * LBT 04.03.94
 946  */
 947
 948 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
 949 {
 950         struct task_struct *p;
 951         int err = -EINVAL;
 952
 953         if (!pid)
 954                 pid = current->pid;
 955         if (!pgid)
 956                 pgid = pid;
 957         if (pgid < 0)
 958                 return -EINVAL;
 959
 960         /* From this point forward we keep holding onto the tasklist lock
 961          * so that our parent does not change from under us. -DaveM
 962          */
 963         write_lock_irq(&tasklist_lock);
 964
 965         err = -ESRCH;
 966         p = find_task_by_pid(pid);
 967         if (!p)
 968                 goto out;
 969
 970         err = -EINVAL;
 971         if (!thread_group_leader(p))
 972                 goto out;
 973
 974         if (p->parent == current || p->real_parent == current) {
 975                 err = -EPERM;
 976                 if (p->signal->session != current->signal->session)
 977                         goto out;
 978                 err = -EACCES;
 979                 if (p->did_exec)
 980                         goto out;
 981         } else {
 982                 err = -ESRCH;
 983                 if (p != current)
 984                         goto out;
 985         }
 986
 987         err = -EPERM;
 988         if (p->signal->leader)
 989                 goto out;
 990
 991         if (pgid != pid) {
 992                 struct task_struct *p;
 993
 994                 do_each_task_pid(pgid, PIDTYPE_PGID, p) {
 995                         if (p->signal->session == current->signal->session)
 996                                 goto ok_pgid;
 997                 } while_each_task_pid(pgid, PIDTYPE_PGID, p);
 998                 goto out;
 999         }
1000
1001 ok_pgid:
1002         err = security_task_setpgid(p, pgid);
1003         if (err)
1004                 goto out;
1005
1006         if (process_group(p) != pgid) {
1007                 detach_pid(p, PIDTYPE_PGID);
1008                 p->signal->pgrp = pgid;
1009                 attach_pid(p, PIDTYPE_PGID, pgid);
1010         }
1011
1012         err = 0;
1013 out:
1014         /* All paths lead to here, thus we are safe. -DaveM */
1015         write_unlock_irq(&tasklist_lock);
1016         return err;
1017 }
1018
1019 asmlinkage long sys_getpgid(pid_t pid)
1020 {
1021         if (!pid) {
1022                 return process_group(current);
1023         } else {
1024                 int retval;
1025                 struct task_struct *p;
1026
1027                 read_lock(&tasklist_lock);
1028                 p = find_task_by_pid(pid);
1029
1030                 retval = -ESRCH;
1031                 if (p) {
1032                         retval = security_task_getpgid(p);
1033                         if (!retval)
1034                                 retval = process_group(p);
1035                 }
1036                 read_unlock(&tasklist_lock);
1037                 return retval;
1038         }
1039 }
1040
1041 #ifdef __ARCH_WANT_SYS_GETPGRP
1042
1043 asmlinkage long sys_getpgrp(void)
1044 {
1045         /* SMP - assuming writes are word atomic this is fine */
1046         return process_group(current);
1047 }
1048
1049 #endif
1050
1051 asmlinkage long sys_getsid(pid_t pid)
1052 {
1053         if (!pid) {
1054                 return current->signal->session;
1055         } else {
1056                 int retval;
1057                 struct task_struct *p;
1058
1059                 read_lock(&tasklist_lock);
1060                 p = find_task_by_pid(pid);
1061
1062                 retval = -ESRCH;
1063                 if(p) {
1064                         retval = security_task_getsid(p);
1065                         if (!retval)
1066                                 retval = p->signal->session;
1067                 }
1068                 read_unlock(&tasklist_lock);
1069                 return retval;
1070         }
1071 }
1072
1073 asmlinkage long sys_setsid(void)
1074 {
1075         struct pid *pid;
1076         int err = -EPERM;
1077
1078         if (!thread_group_leader(current))
1079                 return -EINVAL;
1080
1081         down(&tty_sem);
1082         write_lock_irq(&tasklist_lock);
1083
1084         pid = find_pid(PIDTYPE_PGID, current->pid);
1085         if (pid)
1086                 goto out;
1087
1088         current->signal->leader = 1;
1089         __set_special_pids(current->pid, current->pid);
1090         current->signal->tty = NULL;
1091         current->signal->tty_old_pgrp = 0;
1092         err = process_group(current);
1093 out:
1094         write_unlock_irq(&tasklist_lock);
1095         up(&tty_sem);
1096         return err;
1097 }
1098
1099 /*
1100  * Supplementary group IDs
1101  */
1102
1103 /* init to 2 - one for init_task, one to ensure it is never freed */
1104 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1105
1106 struct group_info *groups_alloc(int gidsetsize)
1107 {
1108         struct group_info *group_info;
1109         int nblocks;
1110         int i;
1111
1112         nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1113         /* Make sure we always allocate at least one indirect block pointer */
1114         nblocks = nblocks ? : 1;
1115         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1116         if (!group_info)
1117                 return NULL;
1118         group_info->ngroups = gidsetsize;
1119         group_info->nblocks = nblocks;
1120         atomic_set(&group_info->usage, 1);
1121
1122         if (gidsetsize <= NGROUPS_SMALL) {
1123                 group_info->blocks[0] = group_info->small_block;
1124         } else {
1125                 for (i = 0; i < nblocks; i++) {
1126                         gid_t *b;
1127                         b = (void *)__get_free_page(GFP_USER);
1128                         if (!b)
1129                                 goto out_undo_partial_alloc;
1130                         group_info->blocks[i] = b;
1131                 }
1132         }
1133         return group_info;
1134
1135 out_undo_partial_alloc:
1136         while (--i >= 0) {
1137                 free_page((unsigned long)group_info->blocks[i]);
1138         }
1139         kfree(group_info);
1140         return NULL;
1141 }
1142
1143 EXPORT_SYMBOL(groups_alloc);
1144
1145 void groups_free(struct group_info *group_info)
1146 {
1147         if (group_info->blocks[0] != group_info->small_block) {
1148                 int i;
1149                 for (i = 0; i < group_info->nblocks; i++)
1150                         free_page((unsigned long)group_info->blocks[i]);
1151         }
1152         kfree(group_info);
1153 }
1154
1155 EXPORT_SYMBOL(groups_free);
1156
1157 /* export the group_info to a user-space array */
1158 static int groups_to_user(gid_t __user *grouplist,
1159     struct group_info *group_info)
1160 {
1161         int i;
1162         int count = group_info->ngroups;
1163
1164         for (i = 0; i < group_info->nblocks; i++) {
1165                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1166                 int off = i * NGROUPS_PER_BLOCK;
1167                 int len = cp_count * sizeof(*grouplist);
1168
1169                 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1170                         return -EFAULT;
1171
1172                 count -= cp_count;
1173         }
1174         return 0;
1175 }
1176
1177 /* fill a group_info from a user-space array - it must be allocated already */
1178 static int groups_from_user(struct group_info *group_info,
1179     gid_t __user *grouplist)
1180  {
1181         int i;
1182         int count = group_info->ngroups;
1183
1184         for (i = 0; i < group_info->nblocks; i++) {
1185                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1186                 int off = i * NGROUPS_PER_BLOCK;
1187                 int len = cp_count * sizeof(*grouplist);
1188
1189                 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1190                         return -EFAULT;
1191
1192                 count -= cp_count;
1193         }
1194         return 0;
1195 }
1196
1197 /* a simple shell-metzner sort */
1198 static void groups_sort(struct group_info *group_info)
1199 {
1200         int base, max, stride;
1201         int gidsetsize = group_info->ngroups;
1202
1203         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1204                 ; /* nothing */
1205         stride /= 3;
1206
1207         while (stride) {
1208                 max = gidsetsize - stride;
1209                 for (base = 0; base < max; base++) {
1210                         int left = base;
1211                         int right = left + stride;
1212                         gid_t tmp = GROUP_AT(group_info, right);
1213
1214                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1215                                 GROUP_AT(group_info, right) =
1216                                     GROUP_AT(group_info, left);
1217                                 right = left;
1218                                 left -= stride;
1219                         }
1220                         GROUP_AT(group_info, right) = tmp;
1221                 }
1222                 stride /= 3;
1223         }
1224 }
1225
1226 /* a simple bsearch */
1227 static int groups_search(struct group_info *group_info, gid_t grp)
1228 {
1229         int left, right;
1230
1231         if (!group_info)
1232                 return 0;
1233
1234         left = 0;
1235         right = group_info->ngroups;
1236         while (left < right) {
1237                 int mid = (left+right)/2;
1238                 int cmp = grp - GROUP_AT(group_info, mid);
1239                 if (cmp > 0)
1240                         left = mid + 1;
1241                 else if (cmp < 0)
1242                         right = mid;
1243                 else
1244                         return 1;
1245         }
1246         return 0;
1247 }
1248
1249 /* validate and set current->group_info */
1250 int set_current_groups(struct group_info *group_info)
1251 {
1252         int retval;
1253         struct group_info *old_info;
1254
1255         retval = security_task_setgroups(group_info);
1256         if (retval)
1257                 return retval;
1258
1259         groups_sort(group_info);
1260         get_group_info(group_info);
1261
1262         task_lock(current);
1263         old_info = current->group_info;
1264         current->group_info = group_info;
1265         task_unlock(current);
1266
1267         put_group_info(old_info);
1268
1269         return 0;
1270 }
1271
1272 EXPORT_SYMBOL(set_current_groups);
1273
1274 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1275 {
1276         int i = 0;
1277
1278         /*
1279          *      SMP: Nobody else can change our grouplist. Thus we are
1280          *      safe.
1281          */
1282
1283         if (gidsetsize < 0)
1284                 return -EINVAL;
1285
1286         /* no need to grab task_lock here; it cannot change */
1287         get_group_info(current->group_info);
1288         i = current->group_info->ngroups;
1289         if (gidsetsize) {
1290                 if (i > gidsetsize) {
1291                         i = -EINVAL;
1292                         goto out;
1293                 }
1294                 if (groups_to_user(grouplist, current->group_info)) {
1295                         i = -EFAULT;
1296                         goto out;
1297                 }
1298         }
1299 out:
1300         put_group_info(current->group_info);
1301         return i;
1302 }
1303
1304 /*
1305  *      SMP: Our groups are copy-on-write. We can set them safely
1306  *      without another task interfering.
1307  */
1308
1309 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1310 {
1311         struct group_info *group_info;
1312         int retval;
1313
1314         if (!capable(CAP_SETGID))
1315                 return -EPERM;
1316         if ((unsigned)gidsetsize > NGROUPS_MAX)
1317                 return -EINVAL;
1318
1319         group_info = groups_alloc(gidsetsize);
1320         if (!group_info)
1321                 return -ENOMEM;
1322         retval = groups_from_user(group_info, grouplist);
1323         if (retval) {
1324                 put_group_info(group_info);
1325                 return retval;
1326         }
1327
1328         retval = set_current_groups(group_info);
1329         put_group_info(group_info);
1330
1331         return retval;
1332 }
1333
1334 /*
1335  * Check whether we're fsgid/egid or in the supplemental group..
1336  */
1337 int in_group_p(gid_t grp)
1338 {
1339         int retval = 1;
1340         if (grp != current->fsgid) {
1341                 get_group_info(current->group_info);
1342                 retval = groups_search(current->group_info, grp);
1343                 put_group_info(current->group_info);
1344         }
1345         return retval;
1346 }
1347
1348 EXPORT_SYMBOL(in_group_p);
1349
1350 int in_egroup_p(gid_t grp)
1351 {
1352         int retval = 1;
1353         if (grp != current->egid) {
1354                 get_group_info(current->group_info);
1355                 retval = groups_search(current->group_info, grp);
1356                 put_group_info(current->group_info);
1357         }
1358         return retval;
1359 }
1360
1361 EXPORT_SYMBOL(in_egroup_p);
1362
1363 DECLARE_RWSEM(uts_sem);
1364
1365 EXPORT_SYMBOL(uts_sem);
1366
1367 asmlinkage long sys_newuname(struct new_utsname __user * name)
1368 {
1369         int errno = 0;
1370
1371         down_read(&uts_sem);
1372         if (copy_to_user(name,&system_utsname,sizeof *name))
1373                 errno = -EFAULT;
1374         up_read(&uts_sem);
1375         return errno;
1376 }
1377
1378 asmlinkage long sys_sethostname(char __user *name, int len)
1379 {
1380         int errno;
1381         char tmp[__NEW_UTS_LEN];
1382
1383         if (!capable(CAP_SYS_ADMIN))
1384                 return -EPERM;
1385         if (len < 0 || len > __NEW_UTS_LEN)
1386                 return -EINVAL;
1387         down_write(&uts_sem);
1388         errno = -EFAULT;
1389         if (!copy_from_user(tmp, name, len)) {
1390                 memcpy(system_utsname.nodename, tmp, len);
1391                 system_utsname.nodename[len] = 0;
1392                 errno = 0;
1393         }
1394         up_write(&uts_sem);
1395         return errno;
1396 }
1397
1398 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1399
1400 asmlinkage long sys_gethostname(char __user *name, int len)
1401 {
1402         int i, errno;
1403
1404         if (len < 0)
1405                 return -EINVAL;
1406         down_read(&uts_sem);
1407         i = 1 + strlen(system_utsname.nodename);
1408         if (i > len)
1409                 i = len;
1410         errno = 0;
1411         if (copy_to_user(name, system_utsname.nodename, i))
1412                 errno = -EFAULT;
1413         up_read(&uts_sem);
1414         return errno;
1415 }
1416
1417 #endif
1418
1419 /*
1420  * Only setdomainname; getdomainname can be implemented by calling
1421  * uname()
1422  */
1423 asmlinkage long sys_setdomainname(char __user *name, int len)
1424 {
1425         int errno;
1426         char tmp[__NEW_UTS_LEN];
1427
1428         if (!capable(CAP_SYS_ADMIN))
1429                 return -EPERM;
1430         if (len < 0 || len > __NEW_UTS_LEN)
1431                 return -EINVAL;
1432
1433         down_write(&uts_sem);
1434         errno = -EFAULT;
1435         if (!copy_from_user(tmp, name, len)) {
1436                 memcpy(system_utsname.domainname, tmp, len);
1437                 system_utsname.domainname[len] = 0;
1438                 errno = 0;
1439         }
1440         up_write(&uts_sem);
1441         return errno;
1442 }
1443
1444 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1445 {
1446         if (resource >= RLIM_NLIMITS)
1447                 return -EINVAL;
1448         else {
1449                 struct rlimit value;
1450                 task_lock(current->group_leader);
1451                 value = current->signal->rlim[resource];
1452                 task_unlock(current->group_leader);
1453                 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1454         }
1455 }
1456
1457 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1458
1459 /*
1460  *      Back compatibility for getrlimit. Needed for some apps.
1461  */
1462
1463 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1464 {
1465         struct rlimit x;
1466         if (resource >= RLIM_NLIMITS)
1467                 return -EINVAL;
1468
1469         task_lock(current->group_leader);
1470         x = current->signal->rlim[resource];
1471         task_unlock(current->group_leader);
1472         if(x.rlim_cur > 0x7FFFFFFF)
1473                 x.rlim_cur = 0x7FFFFFFF;
1474         if(x.rlim_max > 0x7FFFFFFF)
1475                 x.rlim_max = 0x7FFFFFFF;
1476         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1477 }
1478
1479 #endif
1480
1481 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1482 {
1483         struct rlimit new_rlim, *old_rlim;
1484         int retval;
1485
1486         if (resource >= RLIM_NLIMITS)
1487                 return -EINVAL;
1488         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1489                 return -EFAULT;
1490        if (new_rlim.rlim_cur > new_rlim.rlim_max)
1491                return -EINVAL;
1492         old_rlim = current->signal->rlim + resource;
1493         if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1494             !capable(CAP_SYS_RESOURCE))
1495                 return -EPERM;
1496         if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1497                         return -EPERM;
1498
1499         retval = security_task_setrlimit(resource, &new_rlim);
1500         if (retval)
1501                 return retval;
1502
1503         task_lock(current->group_leader);
1504         *old_rlim = new_rlim;
1505         task_unlock(current->group_leader);
1506
1507         if (resource == RLIMIT_CPU && new_rlim.rlim_cur != RLIM_INFINITY &&
1508             (cputime_eq(current->signal->it_prof_expires, cputime_zero) ||
1509              new_rlim.rlim_cur <= cputime_to_secs(
1510                      current->signal->it_prof_expires))) {
1511                 cputime_t cputime = secs_to_cputime(new_rlim.rlim_cur);
1512                 read_lock(&tasklist_lock);
1513                 spin_lock_irq(&current->sighand->siglock);
1514                 set_process_cpu_timer(current, CPUCLOCK_PROF,
1515                                       &cputime, NULL);
1516                 spin_unlock_irq(&current->sighand->siglock);
1517                 read_unlock(&tasklist_lock);
1518         }
1519
1520         return 0;
1521 }
1522
1523 /*
1524  * It would make sense to put struct rusage in the task_struct,
1525  * except that would make the task_struct be *really big*.  After
1526  * task_struct gets moved into malloc'ed memory, it would
1527  * make sense to do this.  It will make moving the rest of the information
1528  * a lot simpler!  (Which we're not doing right now because we're not
1529  * measuring them yet).
1530  *
1531  * This expects to be called with tasklist_lock read-locked or better,
1532  * and the siglock not locked.  It may momentarily take the siglock.
1533  *
1534  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1535  * races with threads incrementing their own counters.  But since word
1536  * reads are atomic, we either get new values or old values and we don't
1537  * care which for the sums.  We always take the siglock to protect reading
1538  * the c* fields from p->signal from races with exit.c updating those
1539  * fields when reaping, so a sample either gets all the additions of a
1540  * given child after it's reaped, or none so this sample is before reaping.
1541  */
1542
1543 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1544 {
1545         struct task_struct *t;
1546         unsigned long flags;
1547         cputime_t utime, stime;
1548
1549         memset((char *) r, 0, sizeof *r);
1550
1551         if (unlikely(!p->signal))
1552                 return;
1553
1554         switch (who) {
1555                 case RUSAGE_CHILDREN:
1556                         spin_lock_irqsave(&p->sighand->siglock, flags);
1557                         utime = p->signal->cutime;
1558                         stime = p->signal->cstime;
1559                         r->ru_nvcsw = p->signal->cnvcsw;
1560                         r->ru_nivcsw = p->signal->cnivcsw;
1561                         r->ru_minflt = p->signal->cmin_flt;
1562                         r->ru_majflt = p->signal->cmaj_flt;
1563                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1564                         cputime_to_timeval(utime, &r->ru_utime);
1565                         cputime_to_timeval(stime, &r->ru_stime);
1566                         break;
1567                 case RUSAGE_SELF:
1568                         spin_lock_irqsave(&p->sighand->siglock, flags);
1569                         utime = stime = cputime_zero;
1570                         goto sum_group;
1571                 case RUSAGE_BOTH:
1572                         spin_lock_irqsave(&p->sighand->siglock, flags);
1573                         utime = p->signal->cutime;
1574                         stime = p->signal->cstime;
1575                         r->ru_nvcsw = p->signal->cnvcsw;
1576                         r->ru_nivcsw = p->signal->cnivcsw;
1577                         r->ru_minflt = p->signal->cmin_flt;
1578                         r->ru_majflt = p->signal->cmaj_flt;
1579                 sum_group:
1580                         utime = cputime_add(utime, p->signal->utime);
1581                         stime = cputime_add(stime, p->signal->stime);
1582                         r->ru_nvcsw += p->signal->nvcsw;
1583                         r->ru_nivcsw += p->signal->nivcsw;
1584                         r->ru_minflt += p->signal->min_flt;
1585                         r->ru_majflt += p->signal->maj_flt;
1586                         t = p;
1587                         do {
1588                                 utime = cputime_add(utime, t->utime);
1589                                 stime = cputime_add(stime, t->stime);
1590                                 r->ru_nvcsw += t->nvcsw;
1591                                 r->ru_nivcsw += t->nivcsw;
1592                                 r->ru_minflt += t->min_flt;
1593                                 r->ru_majflt += t->maj_flt;
1594                                 t = next_thread(t);
1595                         } while (t != p);
1596                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1597                         cputime_to_timeval(utime, &r->ru_utime);
1598                         cputime_to_timeval(stime, &r->ru_stime);
1599                         break;
1600                 default:
1601                         BUG();
1602         }
1603 }
1604
1605 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1606 {
1607         struct rusage r;
1608         read_lock(&tasklist_lock);
1609         k_getrusage(p, who, &r);
1610         read_unlock(&tasklist_lock);
1611         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1612 }
1613
1614 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1615 {
1616         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1617                 return -EINVAL;
1618         return getrusage(current, who, ru);
1619 }
1620
1621 asmlinkage long sys_umask(int mask)
1622 {
1623         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1624         return mask;
1625 }
1626
1627 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1628                           unsigned long arg4, unsigned long arg5)
1629 {
1630         long error;
1631         int sig;
1632
1633         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1634         if (error)
1635                 return error;
1636
1637         switch (option) {
1638                 case PR_SET_PDEATHSIG:
1639                         sig = arg2;
1640                         if (sig < 0 || sig > _NSIG) {
1641                                 error = -EINVAL;
1642                                 break;
1643                         }
1644                         current->pdeath_signal = sig;
1645                         break;
1646                 case PR_GET_PDEATHSIG:
1647                         error = put_user(current->pdeath_signal, (int __user *)arg2);
1648                         break;
1649                 case PR_GET_DUMPABLE:
1650                         if (current->mm->dumpable)
1651                                 error = 1;
1652                         break;
1653                 case PR_SET_DUMPABLE:
1654                         if (arg2 != 0 && arg2 != 1) {
1655                                 error = -EINVAL;
1656                                 break;
1657                         }
1658                         current->mm->dumpable = arg2;
1659                         break;
1660
1661                 case PR_SET_UNALIGN:
1662                         error = SET_UNALIGN_CTL(current, arg2);
1663                         break;
1664                 case PR_GET_UNALIGN:
1665                         error = GET_UNALIGN_CTL(current, arg2);
1666                         break;
1667                 case PR_SET_FPEMU:
1668                         error = SET_FPEMU_CTL(current, arg2);
1669                         break;
1670                 case PR_GET_FPEMU:
1671                         error = GET_FPEMU_CTL(current, arg2);
1672                         break;
1673                 case PR_SET_FPEXC:
1674                         error = SET_FPEXC_CTL(current, arg2);
1675                         break;
1676                 case PR_GET_FPEXC:
1677                         error = GET_FPEXC_CTL(current, arg2);
1678                         break;
1679                 case PR_GET_TIMING:
1680                         error = PR_TIMING_STATISTICAL;
1681                         break;
1682                 case PR_SET_TIMING:
1683                         if (arg2 == PR_TIMING_STATISTICAL)
1684                                 error = 0;
1685                         else
1686                                 error = -EINVAL;
1687                         break;
1688
1689                 case PR_GET_KEEPCAPS:
1690                         if (current->keep_capabilities)
1691                                 error = 1;
1692                         break;
1693                 case PR_SET_KEEPCAPS:
1694                         if (arg2 != 0 && arg2 != 1) {
1695                                 error = -EINVAL;
1696                                 break;
1697                         }
1698                         current->keep_capabilities = arg2;
1699                         break;
1700                 case PR_SET_NAME: {
1701                         struct task_struct *me = current;
1702                         unsigned char ncomm[sizeof(me->comm)];
1703
1704                         ncomm[sizeof(me->comm)-1] = 0;
1705                         if (strncpy_from_user(ncomm, (char __user *)arg2,
1706                                                 sizeof(me->comm)-1) < 0)
1707                                 return -EFAULT;
1708                         set_task_comm(me, ncomm);
1709                         return 0;
1710                 }
1711                 case PR_GET_NAME: {
1712                         struct task_struct *me = current;
1713                         unsigned char tcomm[sizeof(me->comm)];
1714
1715                         get_task_comm(tcomm, me);
1716                         if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
1717                                 return -EFAULT;
1718                         return 0;
1719                 }
1720                 default:
1721                         error = -EINVAL;
1722                         break;
1723         }
1724         return error;
1725 }