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/compat.h>
   9 #include <linux/module.h>
  10 #include <linux/mm.h>
  11 #include <linux/utsname.h>
  12 #include <linux/mman.h>
  13 #include <linux/smp_lock.h>
  14 #include <linux/notifier.h>
  15 #include <linux/reboot.h>
  16 #include <linux/prctl.h>
  17 #include <linux/init.h>
  18 #include <linux/highuid.h>
  19 #include <linux/fs.h>
  20 #include <linux/workqueue.h>
  21 #include <linux/device.h>
  22 #include <linux/times.h>
  23 #include <linux/security.h>
  24 #include <linux/dcookies.h>
  25 #include <linux/suspend.h>
  26
  27 #include <asm/uaccess.h>
  28 #include <asm/io.h>
  29 #include <asm/unistd.h>
  30
  31 #ifndef SET_UNALIGN_CTL
  32 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  33 #endif
  34 #ifndef GET_UNALIGN_CTL
  35 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  36 #endif
  37 #ifndef SET_FPEMU_CTL
  38 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
  39 #endif
  40 #ifndef GET_FPEMU_CTL
  41 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
  42 #endif
  43 #ifndef SET_FPEXC_CTL
  44 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
  45 #endif
  46 #ifndef GET_FPEXC_CTL
  47 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
  48 #endif
  49
  50 /*
  51  * this is where the system-wide overflow UID and GID are defined, for
  52  * architectures that now have 32-bit UID/GID but didn't in the past
  53  */
  54
  55 int overflowuid = DEFAULT_OVERFLOWUID;
  56 int overflowgid = DEFAULT_OVERFLOWGID;
  57
  58 #ifdef CONFIG_UID16
  59 EXPORT_SYMBOL(overflowuid);
  60 EXPORT_SYMBOL(overflowgid);
  61 #endif
  62
  63 /*
  64  * the same as above, but for filesystems which can only store a 16-bit
  65  * UID and GID. as such, this is needed on all architectures
  66  */
  67
  68 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
  69 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
  70
  71 EXPORT_SYMBOL(fs_overflowuid);
  72 EXPORT_SYMBOL(fs_overflowgid);
  73
  74 /*
  75  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
  76  */
  77
  78 int C_A_D = 1;
  79 int cad_pid = 1;
  80
  81 /*
  82  *      Notifier list for kernel code which wants to be called
  83  *      at shutdown. This is used to stop any idling DMA operations
  84  *      and the like.
  85  */
  86
  87 static struct notifier_block *reboot_notifier_list;
  88 rwlock_t notifier_lock = RW_LOCK_UNLOCKED;
  89
  90 /**
  91  *      notifier_chain_register - Add notifier to a notifier chain
  92  *      @list: Pointer to root list pointer
  93  *      @n: New entry in notifier chain
  94  *
  95  *      Adds a notifier to a notifier chain.
  96  *
  97  *      Currently always returns zero.
  98  */
  99
 100 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
 101 {
 102         write_lock(&notifier_lock);
 103         while(*list)
 104         {
 105                 if(n->priority > (*list)->priority)
 106                         break;
 107                 list= &((*list)->next);
 108         }
 109         n->next = *list;
 110         *list=n;
 111         write_unlock(&notifier_lock);
 112         return 0;
 113 }
 114
 115 EXPORT_SYMBOL(notifier_chain_register);
 116
 117 /**
 118  *      notifier_chain_unregister - Remove notifier from a notifier chain
 119  *      @nl: Pointer to root list pointer
 120  *      @n: New entry in notifier chain
 121  *
 122  *      Removes a notifier from a notifier chain.
 123  *
 124  *      Returns zero on success, or %-ENOENT on failure.
 125  */
 126
 127 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
 128 {
 129         write_lock(&notifier_lock);
 130         while((*nl)!=NULL)
 131         {
 132                 if((*nl)==n)
 133                 {
 134                         *nl=n->next;
 135                         write_unlock(&notifier_lock);
 136                         return 0;
 137                 }
 138                 nl=&((*nl)->next);
 139         }
 140         write_unlock(&notifier_lock);
 141         return -ENOENT;
 142 }
 143
 144 EXPORT_SYMBOL(notifier_chain_unregister);
 145
 146 /**
 147  *      notifier_call_chain - Call functions in a notifier chain
 148  *      @n: Pointer to root pointer of notifier chain
 149  *      @val: Value passed unmodified to notifier function
 150  *      @v: Pointer passed unmodified to notifier function
 151  *
 152  *      Calls each function in a notifier chain in turn.
 153  *
 154  *      If the return value of the notifier can be and'd
 155  *      with %NOTIFY_STOP_MASK, then notifier_call_chain
 156  *      will return immediately, with the return value of
 157  *      the notifier function which halted execution.
 158  *      Otherwise, the return value is the return value
 159  *      of the last notifier function called.
 160  */
 161
 162 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
 163 {
 164         int ret=NOTIFY_DONE;
 165         struct notifier_block *nb = *n;
 166
 167         while(nb)
 168         {
 169                 ret=nb->notifier_call(nb,val,v);
 170                 if(ret&NOTIFY_STOP_MASK)
 171                 {
 172                         return ret;
 173                 }
 174                 nb=nb->next;
 175         }
 176         return ret;
 177 }
 178
 179 EXPORT_SYMBOL(notifier_call_chain);
 180
 181 /**
 182  *      register_reboot_notifier - Register function to be called at reboot time
 183  *      @nb: Info about notifier function to be called
 184  *
 185  *      Registers a function with the list of functions
 186  *      to be called at reboot time.
 187  *
 188  *      Currently always returns zero, as notifier_chain_register
 189  *      always returns zero.
 190  */
 191
 192 int register_reboot_notifier(struct notifier_block * nb)
 193 {
 194         return notifier_chain_register(&reboot_notifier_list, nb);
 195 }
 196
 197 EXPORT_SYMBOL(register_reboot_notifier);
 198
 199 /**
 200  *      unregister_reboot_notifier - Unregister previously registered reboot notifier
 201  *      @nb: Hook to be unregistered
 202  *
 203  *      Unregisters a previously registered reboot
 204  *      notifier function.
 205  *
 206  *      Returns zero on success, or %-ENOENT on failure.
 207  */
 208
 209 int unregister_reboot_notifier(struct notifier_block * nb)
 210 {
 211         return notifier_chain_unregister(&reboot_notifier_list, nb);
 212 }
 213
 214 EXPORT_SYMBOL(unregister_reboot_notifier);
 215
 216 asmlinkage long sys_ni_syscall(void)
 217 {
 218         return -ENOSYS;
 219 }
 220
 221 cond_syscall(sys_nfsservctl)
 222 cond_syscall(sys_quotactl)
 223 cond_syscall(sys_acct)
 224 cond_syscall(sys_lookup_dcookie)
 225 cond_syscall(sys_swapon)
 226 cond_syscall(sys_swapoff)
 227 cond_syscall(sys_init_module)
 228 cond_syscall(sys_delete_module)
 229 cond_syscall(sys_socketpair)
 230 cond_syscall(sys_bind)
 231 cond_syscall(sys_listen)
 232 cond_syscall(sys_accept)
 233 cond_syscall(sys_connect)
 234 cond_syscall(sys_getsockname)
 235 cond_syscall(sys_getpeername)
 236 cond_syscall(sys_sendto)
 237 cond_syscall(sys_send)
 238 cond_syscall(sys_recvfrom)
 239 cond_syscall(sys_recv)
 240 cond_syscall(sys_socket)
 241 cond_syscall(sys_setsockopt)
 242 cond_syscall(sys_getsockopt)
 243 cond_syscall(sys_shutdown)
 244 cond_syscall(sys_sendmsg)
 245 cond_syscall(sys_recvmsg)
 246 cond_syscall(sys_socketcall)
 247 cond_syscall(sys_futex)
 248 cond_syscall(compat_sys_futex)
 249 cond_syscall(sys_epoll_create)
 250 cond_syscall(sys_epoll_ctl)
 251 cond_syscall(sys_epoll_wait)
 252 cond_syscall(sys_semget)
 253 cond_syscall(sys_semop)
 254 cond_syscall(sys_semtimedop)
 255 cond_syscall(sys_semctl)
 256 cond_syscall(sys_msgget)
 257 cond_syscall(sys_msgsnd)
 258 cond_syscall(sys_msgrcv)
 259 cond_syscall(sys_msgctl)
 260 cond_syscall(sys_shmget)
 261 cond_syscall(sys_shmdt)
 262 cond_syscall(sys_shmctl)
 263 cond_syscall(sys_mq_open)
 264 cond_syscall(sys_mq_unlink)
 265 cond_syscall(sys_mq_timedsend)
 266 cond_syscall(sys_mq_timedreceive)
 267 cond_syscall(sys_mq_notify)
 268 cond_syscall(sys_mq_getsetattr)
 269 cond_syscall(compat_sys_mq_open)
 270 cond_syscall(compat_sys_mq_timedsend)
 271 cond_syscall(compat_sys_mq_timedreceive)
 272 cond_syscall(compat_sys_mq_notify)
 273 cond_syscall(compat_sys_mq_getsetattr)
 274
 275 /* arch-specific weak syscall entries */
 276 cond_syscall(sys_pciconfig_read)
 277 cond_syscall(sys_pciconfig_write)
 278 cond_syscall(sys_pciconfig_iobase)
 279
 280 static int set_one_prio(struct task_struct *p, int niceval, int error)
 281 {
 282         int no_nice;
 283
 284         if (p->uid != current->euid &&
 285                 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
 286                 error = -EPERM;
 287                 goto out;
 288         }
 289         if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
 290                 error = -EACCES;
 291                 goto out;
 292         }
 293         no_nice = security_task_setnice(p, niceval);
 294         if (no_nice) {
 295                 error = no_nice;
 296                 goto out;
 297         }
 298         if (error == -ESRCH)
 299                 error = 0;
 300         set_user_nice(p, niceval);
 301 out:
 302         return error;
 303 }
 304
 305 asmlinkage long sys_setpriority(int which, int who, int niceval)
 306 {
 307         struct task_struct *g, *p;
 308         struct user_struct *user;
 309         struct pid *pid;
 310         struct list_head *l;
 311         int error = -EINVAL;
 312
 313         if (which > 2 || which < 0)
 314                 goto out;
 315
 316         /* normalize: avoid signed division (rounding problems) */
 317         error = -ESRCH;
 318         if (niceval < -20)
 319                 niceval = -20;
 320         if (niceval > 19)
 321                 niceval = 19;
 322
 323         read_lock(&tasklist_lock);
 324         switch (which) {
 325                 case PRIO_PROCESS:
 326                         if (!who)
 327                                 who = current->pid;
 328                         p = find_task_by_pid(who);
 329                         if (p)
 330                                 error = set_one_prio(p, niceval, error);
 331                         break;
 332                 case PRIO_PGRP:
 333                         if (!who)
 334                                 who = process_group(current);
 335                         for_each_task_pid(who, PIDTYPE_PGID, p, l, pid)
 336                                 error = set_one_prio(p, niceval, error);
 337                         break;
 338                 case PRIO_USER:
 339                         if (!who)
 340                                 user = current->user;
 341                         else
 342                                 user = find_user(who);
 343
 344                         if (!user)
 345                                 goto out_unlock;
 346
 347                         do_each_thread(g, p)
 348                                 if (p->uid == who)
 349                                         error = set_one_prio(p, niceval, error);
 350                         while_each_thread(g, p);
 351                         break;
 352         }
 353 out_unlock:
 354         read_unlock(&tasklist_lock);
 355 out:
 356         return error;
 357 }
 358
 359 /*
 360  * Ugh. To avoid negative return values, "getpriority()" will
 361  * not return the normal nice-value, but a negated value that
 362  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 363  * to stay compatible.
 364  */
 365 asmlinkage long sys_getpriority(int which, int who)
 366 {
 367         struct task_struct *g, *p;
 368         struct list_head *l;
 369         struct pid *pid;
 370         struct user_struct *user;
 371         long niceval, retval = -ESRCH;
 372
 373         if (which > 2 || which < 0)
 374                 return -EINVAL;
 375
 376         read_lock(&tasklist_lock);
 377         switch (which) {
 378                 case PRIO_PROCESS:
 379                         if (!who)
 380                                 who = current->pid;
 381                         p = find_task_by_pid(who);
 382                         if (p) {
 383                                 niceval = 20 - task_nice(p);
 384                                 if (niceval > retval)
 385                                         retval = niceval;
 386                         }
 387                         break;
 388                 case PRIO_PGRP:
 389                         if (!who)
 390                                 who = process_group(current);
 391                         for_each_task_pid(who, PIDTYPE_PGID, p, l, pid) {
 392                                 niceval = 20 - task_nice(p);
 393                                 if (niceval > retval)
 394                                         retval = niceval;
 395                         }
 396                         break;
 397                 case PRIO_USER:
 398                         if (!who)
 399                                 user = current->user;
 400                         else
 401                                 user = find_user(who);
 402
 403                         if (!user)
 404                                 goto out_unlock;
 405
 406                         do_each_thread(g, p)
 407                                 if (p->uid == who) {
 408                                         niceval = 20 - task_nice(p);
 409                                         if (niceval > retval)
 410                                                 retval = niceval;
 411                                 }
 412                         while_each_thread(g, p);
 413                         break;
 414         }
 415 out_unlock:
 416         read_unlock(&tasklist_lock);
 417
 418         return retval;
 419 }
 420
 421
 422 /*
 423  * Reboot system call: for obvious reasons only root may call it,
 424  * and even root needs to set up some magic numbers in the registers
 425  * so that some mistake won't make this reboot the whole machine.
 426  * You can also set the meaning of the ctrl-alt-del-key here.
 427  *
 428  * reboot doesn't sync: do that yourself before calling this.
 429  */
 430 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
 431 {
 432         char buffer[256];
 433
 434         /* We only trust the superuser with rebooting the system. */
 435         if (!capable(CAP_SYS_BOOT))
 436                 return -EPERM;
 437
 438         /* For safety, we require "magic" arguments. */
 439         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 440             (magic2 != LINUX_REBOOT_MAGIC2 &&
 441                         magic2 != LINUX_REBOOT_MAGIC2A &&
 442                         magic2 != LINUX_REBOOT_MAGIC2B &&
 443                         magic2 != LINUX_REBOOT_MAGIC2C))
 444                 return -EINVAL;
 445
 446         lock_kernel();
 447         switch (cmd) {
 448         case LINUX_REBOOT_CMD_RESTART:
 449                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 450                 system_state = SYSTEM_SHUTDOWN;
 451                 device_shutdown();
 452                 printk(KERN_EMERG "Restarting system.\n");
 453                 machine_restart(NULL);
 454                 break;
 455
 456         case LINUX_REBOOT_CMD_CAD_ON:
 457                 C_A_D = 1;
 458                 break;
 459
 460         case LINUX_REBOOT_CMD_CAD_OFF:
 461                 C_A_D = 0;
 462                 break;
 463
 464         case LINUX_REBOOT_CMD_HALT:
 465                 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
 466                 system_state = SYSTEM_SHUTDOWN;
 467                 device_shutdown();
 468                 printk(KERN_EMERG "System halted.\n");
 469                 machine_halt();
 470                 unlock_kernel();
 471                 do_exit(0);
 472                 break;
 473
 474         case LINUX_REBOOT_CMD_POWER_OFF:
 475                 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
 476                 system_state = SYSTEM_SHUTDOWN;
 477                 device_shutdown();
 478                 printk(KERN_EMERG "Power down.\n");
 479                 machine_power_off();
 480                 unlock_kernel();
 481                 do_exit(0);
 482                 break;
 483
 484         case LINUX_REBOOT_CMD_RESTART2:
 485                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 486                         unlock_kernel();
 487                         return -EFAULT;
 488                 }
 489                 buffer[sizeof(buffer) - 1] = '\0';
 490
 491                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
 492                 system_state = SYSTEM_SHUTDOWN;
 493                 device_shutdown();
 494                 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
 495                 machine_restart(buffer);
 496                 break;
 497
 498 #ifdef CONFIG_SOFTWARE_SUSPEND
 499         case LINUX_REBOOT_CMD_SW_SUSPEND:
 500                 {
 501                         int ret = software_suspend();
 502                         unlock_kernel();
 503                         return ret;
 504                 }
 505 #endif
 506
 507         default:
 508                 unlock_kernel();
 509                 return -EINVAL;
 510         }
 511         unlock_kernel();
 512         return 0;
 513 }
 514
 515 static void deferred_cad(void *dummy)
 516 {
 517         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 518         machine_restart(NULL);
 519 }
 520
 521 /*
 522  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 523  * As it's called within an interrupt, it may NOT sync: the only choice
 524  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 525  */
 526 void ctrl_alt_del(void)
 527 {
 528         static DECLARE_WORK(cad_work, deferred_cad, NULL);
 529
 530         if (C_A_D)
 531                 schedule_work(&cad_work);
 532         else
 533                 kill_proc(cad_pid, SIGINT, 1);
 534 }
 535
 536
 537 /*
 538  * Unprivileged users may change the real gid to the effective gid
 539  * or vice versa.  (BSD-style)
 540  *
 541  * If you set the real gid at all, or set the effective gid to a value not
 542  * equal to the real gid, then the saved gid is set to the new effective gid.
 543  *
 544  * This makes it possible for a setgid program to completely drop its
 545  * privileges, which is often a useful assertion to make when you are doing
 546  * a security audit over a program.
 547  *
 548  * The general idea is that a program which uses just setregid() will be
 549  * 100% compatible with BSD.  A program which uses just setgid() will be
 550  * 100% compatible with POSIX with saved IDs.
 551  *
 552  * SMP: There are not races, the GIDs are checked only by filesystem
 553  *      operations (as far as semantic preservation is concerned).
 554  */
 555 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 556 {
 557         int old_rgid = current->gid;
 558         int old_egid = current->egid;
 559         int new_rgid = old_rgid;
 560         int new_egid = old_egid;
 561         int retval;
 562
 563         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 564         if (retval)
 565                 return retval;
 566
 567         if (rgid != (gid_t) -1) {
 568                 if ((old_rgid == rgid) ||
 569                     (current->egid==rgid) ||
 570                     capable(CAP_SETGID))
 571                         new_rgid = rgid;
 572                 else
 573                         return -EPERM;
 574         }
 575         if (egid != (gid_t) -1) {
 576                 if ((old_rgid == egid) ||
 577                     (current->egid == egid) ||
 578                     (current->sgid == egid) ||
 579                     capable(CAP_SETGID))
 580                         new_egid = egid;
 581                 else {
 582                         return -EPERM;
 583                 }
 584         }
 585         if (new_egid != old_egid)
 586         {
 587                 current->mm->dumpable = 0;
 588                 wmb();
 589         }
 590         if (rgid != (gid_t) -1 ||
 591             (egid != (gid_t) -1 && egid != old_rgid))
 592                 current->sgid = new_egid;
 593         current->fsgid = new_egid;
 594         current->egid = new_egid;
 595         current->gid = new_rgid;
 596         return 0;
 597 }
 598
 599 /*
 600  * setgid() is implemented like SysV w/ SAVED_IDS
 601  *
 602  * SMP: Same implicit races as above.
 603  */
 604 asmlinkage long sys_setgid(gid_t gid)
 605 {
 606         int old_egid = current->egid;
 607         int retval;
 608
 609         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 610         if (retval)
 611                 return retval;
 612
 613         if (capable(CAP_SETGID))
 614         {
 615                 if(old_egid != gid)
 616                 {
 617                         current->mm->dumpable=0;
 618                         wmb();
 619                 }
 620                 current->gid = current->egid = current->sgid = current->fsgid = gid;
 621         }
 622         else if ((gid == current->gid) || (gid == current->sgid))
 623         {
 624                 if(old_egid != gid)
 625                 {
 626                         current->mm->dumpable=0;
 627                         wmb();
 628                 }
 629                 current->egid = current->fsgid = gid;
 630         }
 631         else
 632                 return -EPERM;
 633         return 0;
 634 }
 635
 636 static int set_user(uid_t new_ruid, int dumpclear)
 637 {
 638         struct user_struct *new_user;
 639
 640         new_user = alloc_uid(new_ruid);
 641         if (!new_user)
 642                 return -EAGAIN;
 643
 644         if (atomic_read(&new_user->processes) >=
 645                                 current->rlim[RLIMIT_NPROC].rlim_cur &&
 646                         new_user != &root_user) {
 647                 free_uid(new_user);
 648                 return -EAGAIN;
 649         }
 650
 651         switch_uid(new_user);
 652
 653         if(dumpclear)
 654         {
 655                 current->mm->dumpable = 0;
 656                 wmb();
 657         }
 658         current->uid = new_ruid;
 659         return 0;
 660 }
 661
 662 /*
 663  * Unprivileged users may change the real uid to the effective uid
 664  * or vice versa.  (BSD-style)
 665  *
 666  * If you set the real uid at all, or set the effective uid to a value not
 667  * equal to the real uid, then the saved uid is set to the new effective uid.
 668  *
 669  * This makes it possible for a setuid program to completely drop its
 670  * privileges, which is often a useful assertion to make when you are doing
 671  * a security audit over a program.
 672  *
 673  * The general idea is that a program which uses just setreuid() will be
 674  * 100% compatible with BSD.  A program which uses just setuid() will be
 675  * 100% compatible with POSIX with saved IDs.
 676  */
 677 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 678 {
 679         int old_ruid, old_euid, old_suid, new_ruid, new_euid;
 680         int retval;
 681
 682         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 683         if (retval)
 684                 return retval;
 685
 686         new_ruid = old_ruid = current->uid;
 687         new_euid = old_euid = current->euid;
 688         old_suid = current->suid;
 689
 690         if (ruid != (uid_t) -1) {
 691                 new_ruid = ruid;
 692                 if ((old_ruid != ruid) &&
 693                     (current->euid != ruid) &&
 694                     !capable(CAP_SETUID))
 695                         return -EPERM;
 696         }
 697
 698         if (euid != (uid_t) -1) {
 699                 new_euid = euid;
 700                 if ((old_ruid != euid) &&
 701                     (current->euid != euid) &&
 702                     (current->suid != euid) &&
 703                     !capable(CAP_SETUID))
 704                         return -EPERM;
 705         }
 706
 707         if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
 708                 return -EAGAIN;
 709
 710         if (new_euid != old_euid)
 711         {
 712                 current->mm->dumpable=0;
 713                 wmb();
 714         }
 715         current->fsuid = current->euid = new_euid;
 716         if (ruid != (uid_t) -1 ||
 717             (euid != (uid_t) -1 && euid != old_ruid))
 718                 current->suid = current->euid;
 719         current->fsuid = current->euid;
 720
 721         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
 722 }
 723
 724
 725
 726 /*
 727  * setuid() is implemented like SysV with SAVED_IDS
 728  *
 729  * Note that SAVED_ID's is deficient in that a setuid root program
 730  * like sendmail, for example, cannot set its uid to be a normal
 731  * user and then switch back, because if you're root, setuid() sets
 732  * the saved uid too.  If you don't like this, blame the bright people
 733  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 734  * will allow a root program to temporarily drop privileges and be able to
 735  * regain them by swapping the real and effective uid.
 736  */
 737 asmlinkage long sys_setuid(uid_t uid)
 738 {
 739         int old_euid = current->euid;
 740         int old_ruid, old_suid, new_ruid, new_suid;
 741         int retval;
 742
 743         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 744         if (retval)
 745                 return retval;
 746
 747         old_ruid = new_ruid = current->uid;
 748         old_suid = current->suid;
 749         new_suid = old_suid;
 750
 751         if (capable(CAP_SETUID)) {
 752                 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
 753                         return -EAGAIN;
 754                 new_suid = uid;
 755         } else if ((uid != current->uid) && (uid != new_suid))
 756                 return -EPERM;
 757
 758         if (old_euid != uid)
 759         {
 760                 current->mm->dumpable = 0;
 761                 wmb();
 762         }
 763         current->fsuid = current->euid = uid;
 764         current->suid = new_suid;
 765
 766         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
 767 }
 768
 769
 770 /*
 771  * This function implements a generic ability to update ruid, euid,
 772  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 773  */
 774 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 775 {
 776         int old_ruid = current->uid;
 777         int old_euid = current->euid;
 778         int old_suid = current->suid;
 779         int retval;
 780
 781         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 782         if (retval)
 783                 return retval;
 784
 785         if (!capable(CAP_SETUID)) {
 786                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
 787                     (ruid != current->euid) && (ruid != current->suid))
 788                         return -EPERM;
 789                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
 790                     (euid != current->euid) && (euid != current->suid))
 791                         return -EPERM;
 792                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
 793                     (suid != current->euid) && (suid != current->suid))
 794                         return -EPERM;
 795         }
 796         if (ruid != (uid_t) -1) {
 797                 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
 798                         return -EAGAIN;
 799         }
 800         if (euid != (uid_t) -1) {
 801                 if (euid != current->euid)
 802                 {
 803                         current->mm->dumpable = 0;
 804                         wmb();
 805                 }
 806                 current->euid = euid;
 807         }
 808         current->fsuid = current->euid;
 809         if (suid != (uid_t) -1)
 810                 current->suid = suid;
 811
 812         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
 813 }
 814
 815 asmlinkage long sys_getresuid(uid_t *ruid, uid_t *euid, uid_t *suid)
 816 {
 817         int retval;
 818
 819         if (!(retval = put_user(current->uid, ruid)) &&
 820             !(retval = put_user(current->euid, euid)))
 821                 retval = put_user(current->suid, suid);
 822
 823         return retval;
 824 }
 825
 826 /*
 827  * Same as above, but for rgid, egid, sgid.
 828  */
 829 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 830 {
 831         int retval;
 832
 833         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 834         if (retval)
 835                 return retval;
 836
 837         if (!capable(CAP_SETGID)) {
 838                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
 839                     (rgid != current->egid) && (rgid != current->sgid))
 840                         return -EPERM;
 841                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
 842                     (egid != current->egid) && (egid != current->sgid))
 843                         return -EPERM;
 844                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
 845                     (sgid != current->egid) && (sgid != current->sgid))
 846                         return -EPERM;
 847         }
 848         if (egid != (gid_t) -1) {
 849                 if (egid != current->egid)
 850                 {
 851                         current->mm->dumpable = 0;
 852                         wmb();
 853                 }
 854                 current->egid = egid;
 855         }
 856         current->fsgid = current->egid;
 857         if (rgid != (gid_t) -1)
 858                 current->gid = rgid;
 859         if (sgid != (gid_t) -1)
 860                 current->sgid = sgid;
 861         return 0;
 862 }
 863
 864 asmlinkage long sys_getresgid(gid_t *rgid, gid_t *egid, gid_t *sgid)
 865 {
 866         int retval;
 867
 868         if (!(retval = put_user(current->gid, rgid)) &&
 869             !(retval = put_user(current->egid, egid)))
 870                 retval = put_user(current->sgid, sgid);
 871
 872         return retval;
 873 }
 874
 875
 876 /*
 877  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 878  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 879  * whatever uid it wants to). It normally shadows "euid", except when
 880  * explicitly set by setfsuid() or for access..
 881  */
 882 asmlinkage long sys_setfsuid(uid_t uid)
 883 {
 884         int old_fsuid;
 885
 886         old_fsuid = current->fsuid;
 887         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
 888                 return old_fsuid;
 889
 890         if (uid == current->uid || uid == current->euid ||
 891             uid == current->suid || uid == current->fsuid ||
 892             capable(CAP_SETUID))
 893         {
 894                 if (uid != old_fsuid)
 895                 {
 896                         current->mm->dumpable = 0;
 897                         wmb();
 898                 }
 899                 current->fsuid = uid;
 900         }
 901
 902         security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
 903
 904         return old_fsuid;
 905 }
 906
 907 /*
 908  * Samma på svenska..
 909  */
 910 asmlinkage long sys_setfsgid(gid_t gid)
 911 {
 912         int old_fsgid;
 913
 914         old_fsgid = current->fsgid;
 915         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 916                 return old_fsgid;
 917
 918         if (gid == current->gid || gid == current->egid ||
 919             gid == current->sgid || gid == current->fsgid ||
 920             capable(CAP_SETGID))
 921         {
 922                 if (gid != old_fsgid)
 923                 {
 924                         current->mm->dumpable = 0;
 925                         wmb();
 926                 }
 927                 current->fsgid = gid;
 928         }
 929         return old_fsgid;
 930 }
 931
 932 asmlinkage long sys_times(struct tms __user * tbuf)
 933 {
 934         /*
 935          *      In the SMP world we might just be unlucky and have one of
 936          *      the times increment as we use it. Since the value is an
 937          *      atomically safe type this is just fine. Conceptually its
 938          *      as if the syscall took an instant longer to occur.
 939          */
 940         if (tbuf) {
 941                 struct tms tmp;
 942                 tmp.tms_utime = jiffies_to_clock_t(current->utime);
 943                 tmp.tms_stime = jiffies_to_clock_t(current->stime);
 944                 tmp.tms_cutime = jiffies_to_clock_t(current->cutime);
 945                 tmp.tms_cstime = jiffies_to_clock_t(current->cstime);
 946                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
 947                         return -EFAULT;
 948         }
 949         return (long) jiffies_64_to_clock_t(get_jiffies_64());
 950 }
 951
 952 /*
 953  * This needs some heavy checking ...
 954  * I just haven't the stomach for it. I also don't fully
 955  * understand sessions/pgrp etc. Let somebody who does explain it.
 956  *
 957  * OK, I think I have the protection semantics right.... this is really
 958  * only important on a multi-user system anyway, to make sure one user
 959  * can't send a signal to a process owned by another.  -TYT, 12/12/91
 960  *
 961  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
 962  * LBT 04.03.94
 963  */
 964
 965 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
 966 {
 967         struct task_struct *p;
 968         int err = -EINVAL;
 969
 970         if (!pid)
 971                 pid = current->pid;
 972         if (!pgid)
 973                 pgid = pid;
 974         if (pgid < 0)
 975                 return -EINVAL;
 976
 977         /* From this point forward we keep holding onto the tasklist lock
 978          * so that our parent does not change from under us. -DaveM
 979          */
 980         write_lock_irq(&tasklist_lock);
 981
 982         err = -ESRCH;
 983         p = find_task_by_pid(pid);
 984         if (!p)
 985                 goto out;
 986
 987         err = -EINVAL;
 988         if (!thread_group_leader(p))
 989                 goto out;
 990
 991         if (p->parent == current || p->real_parent == current) {
 992                 err = -EPERM;
 993                 if (p->signal->session != current->signal->session)
 994                         goto out;
 995                 err = -EACCES;
 996                 if (p->did_exec)
 997                         goto out;
 998         } else {
 999                 err = -ESRCH;
1000                 if (p != current)
1001                         goto out;
1002         }
1003
1004         err = -EPERM;
1005         if (p->signal->leader)
1006                 goto out;
1007
1008         if (pgid != pid) {
1009                 struct task_struct *p;
1010                 struct pid *pid;
1011                 struct list_head *l;
1012
1013                 for_each_task_pid(pgid, PIDTYPE_PGID, p, l, pid)
1014                         if (p->signal->session == current->signal->session)
1015                                 goto ok_pgid;
1016                 goto out;
1017         }
1018
1019 ok_pgid:
1020         err = security_task_setpgid(p, pgid);
1021         if (err)
1022                 goto out;
1023
1024         if (process_group(p) != pgid) {
1025                 detach_pid(p, PIDTYPE_PGID);
1026                 p->signal->pgrp = pgid;
1027                 attach_pid(p, PIDTYPE_PGID, pgid);
1028         }
1029
1030         err = 0;
1031 out:
1032         /* All paths lead to here, thus we are safe. -DaveM */
1033         write_unlock_irq(&tasklist_lock);
1034         return err;
1035 }
1036
1037 asmlinkage long sys_getpgid(pid_t pid)
1038 {
1039         if (!pid) {
1040                 return process_group(current);
1041         } else {
1042                 int retval;
1043                 struct task_struct *p;
1044
1045                 read_lock(&tasklist_lock);
1046                 p = find_task_by_pid(pid);
1047
1048                 retval = -ESRCH;
1049                 if (p) {
1050                         retval = security_task_getpgid(p);
1051                         if (!retval)
1052                                 retval = process_group(p);
1053                 }
1054                 read_unlock(&tasklist_lock);
1055                 return retval;
1056         }
1057 }
1058
1059 asmlinkage long sys_getpgrp(void)
1060 {
1061         /* SMP - assuming writes are word atomic this is fine */
1062         return process_group(current);
1063 }
1064
1065 asmlinkage long sys_getsid(pid_t pid)
1066 {
1067         if (!pid) {
1068                 return current->signal->session;
1069         } else {
1070                 int retval;
1071                 struct task_struct *p;
1072
1073                 read_lock(&tasklist_lock);
1074                 p = find_task_by_pid(pid);
1075
1076                 retval = -ESRCH;
1077                 if(p) {
1078                         retval = security_task_getsid(p);
1079                         if (!retval)
1080                                 retval = p->signal->session;
1081                 }
1082                 read_unlock(&tasklist_lock);
1083                 return retval;
1084         }
1085 }
1086
1087 asmlinkage long sys_setsid(void)
1088 {
1089         struct pid *pid;
1090         int err = -EPERM;
1091
1092         if (!thread_group_leader(current))
1093                 return -EINVAL;
1094
1095         write_lock_irq(&tasklist_lock);
1096
1097         pid = find_pid(PIDTYPE_PGID, current->pid);
1098         if (pid)
1099                 goto out;
1100
1101         current->signal->leader = 1;
1102         __set_special_pids(current->pid, current->pid);
1103         current->signal->tty = NULL;
1104         current->signal->tty_old_pgrp = 0;
1105         err = process_group(current);
1106 out:
1107         write_unlock_irq(&tasklist_lock);
1108         return err;
1109 }
1110
1111 /*
1112  * Supplementary group IDs
1113  */
1114
1115 /* init to 2 - one for init_task, one to ensure it is never freed */
1116 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1117
1118 struct group_info *groups_alloc(int gidsetsize)
1119 {
1120         struct group_info *group_info;
1121         int nblocks;
1122         int i;
1123
1124         nblocks = (gidsetsize/NGROUPS_PER_BLOCK) +
1125             (gidsetsize%NGROUPS_PER_BLOCK?1:0);
1126         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *),
1127             GFP_USER);
1128         if (!group_info)
1129                 return NULL;
1130         group_info->ngroups = gidsetsize;
1131         group_info->nblocks = nblocks;
1132         atomic_set(&group_info->usage, 1);
1133
1134         if (gidsetsize <= NGROUPS_SMALL) {
1135                 group_info->blocks[0] = group_info->small_block;
1136         } else {
1137                 for (i = 0; i < nblocks; i++) {
1138                         gid_t *b;
1139                         b = (void *)__get_free_page(GFP_USER);
1140                         if (!b)
1141                                 goto out_undo_partial_alloc;
1142                         group_info->blocks[i] = b;
1143                 }
1144         }
1145         return group_info;
1146
1147 out_undo_partial_alloc:
1148         while (--i >= 0) {
1149                 free_page((unsigned long)group_info->blocks[i]);
1150         }
1151         kfree(group_info);
1152         return NULL;
1153 }
1154
1155 EXPORT_SYMBOL(groups_alloc);
1156
1157 void groups_free(struct group_info *group_info)
1158 {
1159         if (group_info->blocks[0] != group_info->small_block) {
1160                 int i;
1161                 for (i = 0; i < group_info->nblocks; i++)
1162                         free_page((unsigned long)group_info->blocks[i]);
1163         }
1164         kfree(group_info);
1165 }
1166
1167 EXPORT_SYMBOL(groups_free);
1168
1169 /* export the group_info to a user-space array */
1170 static int groups_to_user(gid_t __user *grouplist,
1171     struct group_info *group_info)
1172 {
1173         int i;
1174         int count = group_info->ngroups;
1175
1176         for (i = 0; i < group_info->nblocks; i++) {
1177                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1178                 int off = i * NGROUPS_PER_BLOCK;
1179                 int len = cp_count * sizeof(*grouplist);
1180
1181                 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1182                         return -EFAULT;
1183
1184                 count -= cp_count;
1185         }
1186         return 0;
1187 }
1188
1189 /* fill a group_info from a user-space array - it must be allocated already */
1190 static int groups_from_user(struct group_info *group_info,
1191     gid_t __user *grouplist)
1192  {
1193         int i;
1194         int count = group_info->ngroups;
1195
1196         for (i = 0; i < group_info->nblocks; i++) {
1197                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1198                 int off = i * NGROUPS_PER_BLOCK;
1199                 int len = cp_count * sizeof(*grouplist);
1200
1201                 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1202                         return -EFAULT;
1203
1204                 count -= cp_count;
1205         }
1206         return 0;
1207 }
1208
1209 /* a simple shell-metzner sort */
1210 static void groups_sort(struct group_info *group_info)
1211 {
1212         int base, max, stride;
1213         int gidsetsize = group_info->ngroups;
1214
1215         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1216                 ; /* nothing */
1217         stride /= 3;
1218
1219         while (stride) {
1220                 max = gidsetsize - stride;
1221                 for (base = 0; base < max; base++) {
1222                         int left = base;
1223                         int right = left + stride;
1224                         gid_t tmp = GROUP_AT(group_info, right);
1225
1226                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1227                                 GROUP_AT(group_info, right) =
1228                                     GROUP_AT(group_info, left);
1229                                 right = left;
1230                                 left -= stride;
1231                         }
1232                         GROUP_AT(group_info, right) = tmp;
1233                 }
1234                 stride /= 3;
1235         }
1236 }
1237
1238 /* a simple bsearch */
1239 static int groups_search(struct group_info *group_info, gid_t grp)
1240 {
1241         int left, right;
1242
1243         if (!group_info)
1244                 return 0;
1245
1246         left = 0;
1247         right = group_info->ngroups;
1248         while (left < right) {
1249                 int mid = (left+right)/2;
1250                 int cmp = grp - GROUP_AT(group_info, mid);
1251                 if (cmp > 0)
1252                         left = mid + 1;
1253                 else if (cmp < 0)
1254                         right = mid;
1255                 else
1256                         return 1;
1257         }
1258         return 0;
1259 }
1260
1261 /* validate and set current->group_info */
1262 int set_current_groups(struct group_info *group_info)
1263 {
1264         int retval;
1265         struct group_info *old_info;
1266
1267         retval = security_task_setgroups(group_info);
1268         if (retval)
1269                 return retval;
1270
1271         groups_sort(group_info);
1272         get_group_info(group_info);
1273         old_info = current->group_info;
1274         current->group_info = group_info;
1275         put_group_info(old_info);
1276
1277         return 0;
1278 }
1279
1280 EXPORT_SYMBOL(set_current_groups);
1281
1282 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1283 {
1284         int i = 0;
1285
1286         /*
1287          *      SMP: Nobody else can change our grouplist. Thus we are
1288          *      safe.
1289          */
1290
1291         if (gidsetsize < 0)
1292                 return -EINVAL;
1293
1294         get_group_info(current->group_info);
1295         i = current->group_info->ngroups;
1296         if (gidsetsize) {
1297                 if (i > gidsetsize) {
1298                         i = -EINVAL;
1299                         goto out;
1300                 }
1301                 if (groups_to_user(grouplist, current->group_info)) {
1302                         i = -EFAULT;
1303                         goto out;
1304                 }
1305         }
1306 out:
1307         put_group_info(current->group_info);
1308         return i;
1309 }
1310
1311 /*
1312  *      SMP: Our groups are copy-on-write. We can set them safely
1313  *      without another task interfering.
1314  */
1315
1316 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1317 {
1318         struct group_info *group_info;
1319         int retval;
1320
1321         if (!capable(CAP_SETGID))
1322                 return -EPERM;
1323         if ((unsigned)gidsetsize > NGROUPS_MAX)
1324                 return -EINVAL;
1325
1326         group_info = groups_alloc(gidsetsize);
1327         if (!group_info)
1328                 return -ENOMEM;
1329         retval = groups_from_user(group_info, grouplist);
1330         if (retval) {
1331                 put_group_info(group_info);
1332                 return retval;
1333         }
1334
1335         retval = set_current_groups(group_info);
1336         put_group_info(group_info);
1337
1338         return retval;
1339 }
1340
1341 /*
1342  * Check whether we're fsgid/egid or in the supplemental group..
1343  */
1344 int in_group_p(gid_t grp)
1345 {
1346         int retval = 1;
1347         if (grp != current->fsgid) {
1348                 get_group_info(current->group_info);
1349                 retval = groups_search(current->group_info, grp);
1350                 put_group_info(current->group_info);
1351         }
1352         return retval;
1353 }
1354
1355 EXPORT_SYMBOL(in_group_p);
1356
1357 int in_egroup_p(gid_t grp)
1358 {
1359         int retval = 1;
1360         if (grp != current->egid) {
1361                 get_group_info(current->group_info);
1362                 retval = groups_search(current->group_info, grp);
1363                 put_group_info(current->group_info);
1364         }
1365         return retval;
1366 }
1367
1368 EXPORT_SYMBOL(in_egroup_p);
1369
1370 DECLARE_RWSEM(uts_sem);
1371
1372 EXPORT_SYMBOL(uts_sem);
1373
1374 asmlinkage long sys_newuname(struct new_utsname __user * name)
1375 {
1376         int errno = 0;
1377
1378         down_read(&uts_sem);
1379         if (copy_to_user(name,&system_utsname,sizeof *name))
1380                 errno = -EFAULT;
1381         up_read(&uts_sem);
1382         return errno;
1383 }
1384
1385 asmlinkage long sys_sethostname(char __user *name, int len)
1386 {
1387         int errno;
1388         char tmp[__NEW_UTS_LEN];
1389
1390         if (!capable(CAP_SYS_ADMIN))
1391                 return -EPERM;
1392         if (len < 0 || len > __NEW_UTS_LEN)
1393                 return -EINVAL;
1394         down_write(&uts_sem);
1395         errno = -EFAULT;
1396         if (!copy_from_user(tmp, name, len)) {
1397                 memcpy(system_utsname.nodename, tmp, len);
1398                 system_utsname.nodename[len] = 0;
1399                 errno = 0;
1400         }
1401         up_write(&uts_sem);
1402         return errno;
1403 }
1404
1405 asmlinkage long sys_gethostname(char __user *name, int len)
1406 {
1407         int i, errno;
1408
1409         if (len < 0)
1410                 return -EINVAL;
1411         down_read(&uts_sem);
1412         i = 1 + strlen(system_utsname.nodename);
1413         if (i > len)
1414                 i = len;
1415         errno = 0;
1416         if (copy_to_user(name, system_utsname.nodename, i))
1417                 errno = -EFAULT;
1418         up_read(&uts_sem);
1419         return errno;
1420 }
1421
1422 /*
1423  * Only setdomainname; getdomainname can be implemented by calling
1424  * uname()
1425  */
1426 asmlinkage long sys_setdomainname(char __user *name, int len)
1427 {
1428         int errno;
1429         char tmp[__NEW_UTS_LEN];
1430
1431         if (!capable(CAP_SYS_ADMIN))
1432                 return -EPERM;
1433         if (len < 0 || len > __NEW_UTS_LEN)
1434                 return -EINVAL;
1435
1436         down_write(&uts_sem);
1437         errno = -EFAULT;
1438         if (!copy_from_user(tmp, name, len)) {
1439                 memcpy(system_utsname.domainname, tmp, len);
1440                 system_utsname.domainname[len] = 0;
1441                 errno = 0;
1442         }
1443         up_write(&uts_sem);
1444         return errno;
1445 }
1446
1447 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1448 {
1449         if (resource >= RLIM_NLIMITS)
1450                 return -EINVAL;
1451         else
1452                 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
1453                         ? -EFAULT : 0;
1454 }
1455
1456 #if defined(COMPAT_RLIM_OLD_INFINITY) || !(defined(CONFIG_IA64) || defined(CONFIG_V850))
1457
1458 /*
1459  *      Back compatibility for getrlimit. Needed for some apps.
1460  */
1461
1462 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1463 {
1464         struct rlimit x;
1465         if (resource >= RLIM_NLIMITS)
1466                 return -EINVAL;
1467
1468         memcpy(&x, current->rlim + resource, sizeof(*rlim));
1469         if(x.rlim_cur > 0x7FFFFFFF)
1470                 x.rlim_cur = 0x7FFFFFFF;
1471         if(x.rlim_max > 0x7FFFFFFF)
1472                 x.rlim_max = 0x7FFFFFFF;
1473         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1474 }
1475
1476 #endif
1477
1478 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1479 {
1480         struct rlimit new_rlim, *old_rlim;
1481         int retval;
1482
1483         if (resource >= RLIM_NLIMITS)
1484                 return -EINVAL;
1485         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1486                 return -EFAULT;
1487        if (new_rlim.rlim_cur > new_rlim.rlim_max)
1488                return -EINVAL;
1489         old_rlim = current->rlim + resource;
1490         if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
1491              (new_rlim.rlim_max > old_rlim->rlim_max)) &&
1492             !capable(CAP_SYS_RESOURCE))
1493                 return -EPERM;
1494         if (resource == RLIMIT_NOFILE) {
1495                 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
1496                         return -EPERM;
1497         }
1498
1499         retval = security_task_setrlimit(resource, &new_rlim);
1500         if (retval)
1501                 return retval;
1502
1503         *old_rlim = new_rlim;
1504         return 0;
1505 }
1506
1507 /*
1508  * It would make sense to put struct rusage in the task_struct,
1509  * except that would make the task_struct be *really big*.  After
1510  * task_struct gets moved into malloc'ed memory, it would
1511  * make sense to do this.  It will make moving the rest of the information
1512  * a lot simpler!  (Which we're not doing right now because we're not
1513  * measuring them yet).
1514  *
1515  * This is SMP safe.  Either we are called from sys_getrusage on ourselves
1516  * below (we know we aren't going to exit/disappear and only we change our
1517  * rusage counters), or we are called from wait4() on a process which is
1518  * either stopped or zombied.  In the zombied case the task won't get
1519  * reaped till shortly after the call to getrusage(), in both cases the
1520  * task being examined is in a frozen state so the counters won't change.
1521  */
1522 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1523 {
1524         struct rusage r;
1525
1526         memset((char *) &r, 0, sizeof(r));
1527         switch (who) {
1528                 case RUSAGE_SELF:
1529                         jiffies_to_timeval(p->utime, &r.ru_utime);
1530                         jiffies_to_timeval(p->stime, &r.ru_stime);
1531                         r.ru_nvcsw = p->nvcsw;
1532                         r.ru_nivcsw = p->nivcsw;
1533                         r.ru_minflt = p->min_flt;
1534                         r.ru_majflt = p->maj_flt;
1535                         break;
1536                 case RUSAGE_CHILDREN:
1537                         jiffies_to_timeval(p->cutime, &r.ru_utime);
1538                         jiffies_to_timeval(p->cstime, &r.ru_stime);
1539                         r.ru_nvcsw = p->cnvcsw;
1540                         r.ru_nivcsw = p->cnivcsw;
1541                         r.ru_minflt = p->cmin_flt;
1542                         r.ru_majflt = p->cmaj_flt;
1543                         break;
1544                 default:
1545                         jiffies_to_timeval(p->utime + p->cutime, &r.ru_utime);
1546                         jiffies_to_timeval(p->stime + p->cstime, &r.ru_stime);
1547                         r.ru_nvcsw = p->nvcsw + p->cnvcsw;
1548                         r.ru_nivcsw = p->nivcsw + p->cnivcsw;
1549                         r.ru_minflt = p->min_flt + p->cmin_flt;
1550                         r.ru_majflt = p->maj_flt + p->cmaj_flt;
1551                         break;
1552         }
1553         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1554 }
1555
1556 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1557 {
1558         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1559                 return -EINVAL;
1560         return getrusage(current, who, ru);
1561 }
1562
1563 asmlinkage long sys_umask(int mask)
1564 {
1565         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1566         return mask;
1567 }
1568
1569 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1570                           unsigned long arg4, unsigned long arg5)
1571 {
1572         int error;
1573         int sig;
1574
1575         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1576         if (error)
1577                 return error;
1578
1579         switch (option) {
1580                 case PR_SET_PDEATHSIG:
1581                         sig = arg2;
1582                         if (sig < 0 || sig > _NSIG) {
1583                                 error = -EINVAL;
1584                                 break;
1585                         }
1586                         current->pdeath_signal = sig;
1587                         break;
1588                 case PR_GET_PDEATHSIG:
1589                         error = put_user(current->pdeath_signal, (int __user *)arg2);
1590                         break;
1591                 case PR_GET_DUMPABLE:
1592                         if (current->mm->dumpable)
1593                                 error = 1;
1594                         break;
1595                 case PR_SET_DUMPABLE:
1596                         if (arg2 != 0 && arg2 != 1) {
1597                                 error = -EINVAL;
1598                                 break;
1599                         }
1600                         current->mm->dumpable = arg2;
1601                         break;
1602
1603                 case PR_SET_UNALIGN:
1604                         error = SET_UNALIGN_CTL(current, arg2);
1605                         break;
1606                 case PR_GET_UNALIGN:
1607                         error = GET_UNALIGN_CTL(current, arg2);
1608                         break;
1609                 case PR_SET_FPEMU:
1610                         error = SET_FPEMU_CTL(current, arg2);
1611                         break;
1612                 case PR_GET_FPEMU:
1613                         error = GET_FPEMU_CTL(current, arg2);
1614                         break;
1615                 case PR_SET_FPEXC:
1616                         error = SET_FPEXC_CTL(current, arg2);
1617                         break;
1618                 case PR_GET_FPEXC:
1619                         error = GET_FPEXC_CTL(current, arg2);
1620                         break;
1621                 case PR_GET_TIMING:
1622                         error = PR_TIMING_STATISTICAL;
1623                         break;
1624                 case PR_SET_TIMING:
1625                         if (arg2 == PR_TIMING_STATISTICAL)
1626                                 error = 0;
1627                         else
1628                                 error = -EINVAL;
1629                         break;
1630
1631                 case PR_GET_KEEPCAPS:
1632                         if (current->keep_capabilities)
1633                                 error = 1;
1634                         break;
1635                 case PR_SET_KEEPCAPS:
1636                         if (arg2 != 0 && arg2 != 1) {
1637                                 error = -EINVAL;
1638                                 break;
1639                         }
1640                         current->keep_capabilities = arg2;
1641                         break;
1642                 default:
1643                         error = -EINVAL;
1644                         break;
1645         }
1646         return error;
1647 }