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 #include <linux/ckrm.h>
  27
  28 #include <asm/uaccess.h>
  29 #include <asm/io.h>
  30 #include <asm/unistd.h>
  31
  32 #ifndef SET_UNALIGN_CTL
  33 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  34 #endif
  35 #ifndef GET_UNALIGN_CTL
  36 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  37 #endif
  38 #ifndef SET_FPEMU_CTL
  39 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
  40 #endif
  41 #ifndef GET_FPEMU_CTL
  42 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
  43 #endif
  44 #ifndef SET_FPEXC_CTL
  45 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
  46 #endif
  47 #ifndef GET_FPEXC_CTL
  48 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
  49 #endif
  50
  51 /*
  52  * this is where the system-wide overflow UID and GID are defined, for
  53  * architectures that now have 32-bit UID/GID but didn't in the past
  54  */
  55
  56 int overflowuid = DEFAULT_OVERFLOWUID;
  57 int overflowgid = DEFAULT_OVERFLOWGID;
  58
  59 #ifdef CONFIG_UID16
  60 EXPORT_SYMBOL(overflowuid);
  61 EXPORT_SYMBOL(overflowgid);
  62 #endif
  63
  64 /*
  65  * the same as above, but for filesystems which can only store a 16-bit
  66  * UID and GID. as such, this is needed on all architectures
  67  */
  68
  69 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
  70 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
  71
  72 EXPORT_SYMBOL(fs_overflowuid);
  73 EXPORT_SYMBOL(fs_overflowgid);
  74
  75 /*
  76  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
  77  */
  78
  79 int C_A_D = 1;
  80 int cad_pid = 1;
  81
  82 /*
  83  *      Notifier list for kernel code which wants to be called
  84  *      at shutdown. This is used to stop any idling DMA operations
  85  *      and the like.
  86  */
  87
  88 static struct notifier_block *reboot_notifier_list;
  89 rwlock_t notifier_lock = RW_LOCK_UNLOCKED;
  90
  91 /**
  92  *      notifier_chain_register - Add notifier to a notifier chain
  93  *      @list: Pointer to root list pointer
  94  *      @n: New entry in notifier chain
  95  *
  96  *      Adds a notifier to a notifier chain.
  97  *
  98  *      Currently always returns zero.
  99  */
 100
 101 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
 102 {
 103         write_lock(&notifier_lock);
 104         while(*list)
 105         {
 106                 if(n->priority > (*list)->priority)
 107                         break;
 108                 list= &((*list)->next);
 109         }
 110         n->next = *list;
 111         *list=n;
 112         write_unlock(&notifier_lock);
 113         return 0;
 114 }
 115
 116 EXPORT_SYMBOL(notifier_chain_register);
 117
 118 /**
 119  *      notifier_chain_unregister - Remove notifier from a notifier chain
 120  *      @nl: Pointer to root list pointer
 121  *      @n: New entry in notifier chain
 122  *
 123  *      Removes a notifier from a notifier chain.
 124  *
 125  *      Returns zero on success, or %-ENOENT on failure.
 126  */
 127
 128 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
 129 {
 130         write_lock(&notifier_lock);
 131         while((*nl)!=NULL)
 132         {
 133                 if((*nl)==n)
 134                 {
 135                         *nl=n->next;
 136                         write_unlock(&notifier_lock);
 137                         return 0;
 138                 }
 139                 nl=&((*nl)->next);
 140         }
 141         write_unlock(&notifier_lock);
 142         return -ENOENT;
 143 }
 144
 145 EXPORT_SYMBOL(notifier_chain_unregister);
 146
 147 /**
 148  *      notifier_call_chain - Call functions in a notifier chain
 149  *      @n: Pointer to root pointer of notifier chain
 150  *      @val: Value passed unmodified to notifier function
 151  *      @v: Pointer passed unmodified to notifier function
 152  *
 153  *      Calls each function in a notifier chain in turn.
 154  *
 155  *      If the return value of the notifier can be and'd
 156  *      with %NOTIFY_STOP_MASK, then notifier_call_chain
 157  *      will return immediately, with the return value of
 158  *      the notifier function which halted execution.
 159  *      Otherwise, the return value is the return value
 160  *      of the last notifier function called.
 161  */
 162
 163 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
 164 {
 165         int ret=NOTIFY_DONE;
 166         struct notifier_block *nb = *n;
 167
 168         while(nb)
 169         {
 170                 ret=nb->notifier_call(nb,val,v);
 171                 if(ret&NOTIFY_STOP_MASK)
 172                 {
 173                         return ret;
 174                 }
 175                 nb=nb->next;
 176         }
 177         return ret;
 178 }
 179
 180 EXPORT_SYMBOL(notifier_call_chain);
 181
 182 /**
 183  *      register_reboot_notifier - Register function to be called at reboot time
 184  *      @nb: Info about notifier function to be called
 185  *
 186  *      Registers a function with the list of functions
 187  *      to be called at reboot time.
 188  *
 189  *      Currently always returns zero, as notifier_chain_register
 190  *      always returns zero.
 191  */
 192
 193 int register_reboot_notifier(struct notifier_block * nb)
 194 {
 195         return notifier_chain_register(&reboot_notifier_list, nb);
 196 }
 197
 198 EXPORT_SYMBOL(register_reboot_notifier);
 199
 200 /**
 201  *      unregister_reboot_notifier - Unregister previously registered reboot notifier
 202  *      @nb: Hook to be unregistered
 203  *
 204  *      Unregisters a previously registered reboot
 205  *      notifier function.
 206  *
 207  *      Returns zero on success, or %-ENOENT on failure.
 208  */
 209
 210 int unregister_reboot_notifier(struct notifier_block * nb)
 211 {
 212         return notifier_chain_unregister(&reboot_notifier_list, nb);
 213 }
 214
 215 EXPORT_SYMBOL(unregister_reboot_notifier);
 216
 217 asmlinkage long sys_ni_syscall(void)
 218 {
 219         return -ENOSYS;
 220 }
 221
 222 cond_syscall(sys_nfsservctl)
 223 cond_syscall(sys_quotactl)
 224 cond_syscall(sys_acct)
 225 cond_syscall(sys_lookup_dcookie)
 226 cond_syscall(sys_swapon)
 227 cond_syscall(sys_swapoff)
 228 cond_syscall(sys_init_module)
 229 cond_syscall(sys_delete_module)
 230 cond_syscall(sys_socketpair)
 231 cond_syscall(sys_bind)
 232 cond_syscall(sys_listen)
 233 cond_syscall(sys_accept)
 234 cond_syscall(sys_connect)
 235 cond_syscall(sys_getsockname)
 236 cond_syscall(sys_getpeername)
 237 cond_syscall(sys_sendto)
 238 cond_syscall(sys_send)
 239 cond_syscall(sys_recvfrom)
 240 cond_syscall(sys_recv)
 241 cond_syscall(sys_socket)
 242 cond_syscall(sys_setsockopt)
 243 cond_syscall(sys_getsockopt)
 244 cond_syscall(sys_shutdown)
 245 cond_syscall(sys_sendmsg)
 246 cond_syscall(sys_recvmsg)
 247 cond_syscall(sys_socketcall)
 248 cond_syscall(sys_futex)
 249 cond_syscall(compat_sys_futex)
 250 cond_syscall(sys_epoll_create)
 251 cond_syscall(sys_epoll_ctl)
 252 cond_syscall(sys_epoll_wait)
 253 cond_syscall(sys_semget)
 254 cond_syscall(sys_semop)
 255 cond_syscall(sys_semtimedop)
 256 cond_syscall(sys_semctl)
 257 cond_syscall(sys_msgget)
 258 cond_syscall(sys_msgsnd)
 259 cond_syscall(sys_msgrcv)
 260 cond_syscall(sys_msgctl)
 261 cond_syscall(sys_shmget)
 262 cond_syscall(sys_shmdt)
 263 cond_syscall(sys_shmctl)
 264 cond_syscall(sys_mq_open)
 265 cond_syscall(sys_mq_unlink)
 266 cond_syscall(sys_mq_timedsend)
 267 cond_syscall(sys_mq_timedreceive)
 268 cond_syscall(sys_mq_notify)
 269 cond_syscall(sys_mq_getsetattr)
 270 cond_syscall(compat_sys_mq_open)
 271 cond_syscall(compat_sys_mq_timedsend)
 272 cond_syscall(compat_sys_mq_timedreceive)
 273 cond_syscall(compat_sys_mq_notify)
 274 cond_syscall(compat_sys_mq_getsetattr)
 275
 276 /* arch-specific weak syscall entries */
 277 cond_syscall(sys_pciconfig_read)
 278 cond_syscall(sys_pciconfig_write)
 279 cond_syscall(sys_pciconfig_iobase)
 280
 281 static int set_one_prio(struct task_struct *p, int niceval, int error)
 282 {
 283         int no_nice;
 284
 285         if (p->uid != current->euid &&
 286                 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
 287                 error = -EPERM;
 288                 goto out;
 289         }
 290         if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
 291                 error = -EACCES;
 292                 goto out;
 293         }
 294         no_nice = security_task_setnice(p, niceval);
 295         if (no_nice) {
 296                 error = no_nice;
 297                 goto out;
 298         }
 299         if (error == -ESRCH)
 300                 error = 0;
 301         set_user_nice(p, niceval);
 302 out:
 303         return error;
 304 }
 305
 306 asmlinkage long sys_setpriority(int which, int who, int niceval)
 307 {
 308         struct task_struct *g, *p;
 309         struct user_struct *user;
 310         struct pid *pid;
 311         struct list_head *l;
 312         int error = -EINVAL;
 313
 314         if (which > 2 || which < 0)
 315                 goto out;
 316
 317         /* normalize: avoid signed division (rounding problems) */
 318         error = -ESRCH;
 319         if (niceval < -20)
 320                 niceval = -20;
 321         if (niceval > 19)
 322                 niceval = 19;
 323
 324         read_lock(&tasklist_lock);
 325         switch (which) {
 326                 case PRIO_PROCESS:
 327                         if (!who)
 328                                 who = current->pid;
 329                         p = find_task_by_pid(who);
 330                         if (p)
 331                                 error = set_one_prio(p, niceval, error);
 332                         break;
 333                 case PRIO_PGRP:
 334                         if (!who)
 335                                 who = process_group(current);
 336                         for_each_task_pid(who, PIDTYPE_PGID, p, l, pid)
 337                                 error = set_one_prio(p, niceval, error);
 338                         break;
 339                 case PRIO_USER:
 340                         if (!who)
 341                                 user = current->user;
 342                         else
 343                                 user = find_user(who);
 344
 345                         if (!user)
 346                                 goto out_unlock;
 347
 348                         do_each_thread(g, p)
 349                                 if (p->uid == who)
 350                                         error = set_one_prio(p, niceval, error);
 351                         while_each_thread(g, p);
 352                         break;
 353         }
 354 out_unlock:
 355         read_unlock(&tasklist_lock);
 356 out:
 357         return error;
 358 }
 359
 360 /*
 361  * Ugh. To avoid negative return values, "getpriority()" will
 362  * not return the normal nice-value, but a negated value that
 363  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 364  * to stay compatible.
 365  */
 366 asmlinkage long sys_getpriority(int which, int who)
 367 {
 368         struct task_struct *g, *p;
 369         struct list_head *l;
 370         struct pid *pid;
 371         struct user_struct *user;
 372         long niceval, retval = -ESRCH;
 373
 374         if (which > 2 || which < 0)
 375                 return -EINVAL;
 376
 377         read_lock(&tasklist_lock);
 378         switch (which) {
 379                 case PRIO_PROCESS:
 380                         if (!who)
 381                                 who = current->pid;
 382                         p = find_task_by_pid(who);
 383                         if (p) {
 384                                 niceval = 20 - task_nice(p);
 385                                 if (niceval > retval)
 386                                         retval = niceval;
 387                         }
 388                         break;
 389                 case PRIO_PGRP:
 390                         if (!who)
 391                                 who = process_group(current);
 392                         for_each_task_pid(who, PIDTYPE_PGID, p, l, pid) {
 393                                 niceval = 20 - task_nice(p);
 394                                 if (niceval > retval)
 395                                         retval = niceval;
 396                         }
 397                         break;
 398                 case PRIO_USER:
 399                         if (!who)
 400                                 user = current->user;
 401                         else
 402                                 user = find_user(who);
 403
 404                         if (!user)
 405                                 goto out_unlock;
 406
 407                         do_each_thread(g, p)
 408                                 if (p->uid == who) {
 409                                         niceval = 20 - task_nice(p);
 410                                         if (niceval > retval)
 411                                                 retval = niceval;
 412                                 }
 413                         while_each_thread(g, p);
 414                         break;
 415         }
 416 out_unlock:
 417         read_unlock(&tasklist_lock);
 418
 419         return retval;
 420 }
 421
 422
 423 /*
 424  * Reboot system call: for obvious reasons only root may call it,
 425  * and even root needs to set up some magic numbers in the registers
 426  * so that some mistake won't make this reboot the whole machine.
 427  * You can also set the meaning of the ctrl-alt-del-key here.
 428  *
 429  * reboot doesn't sync: do that yourself before calling this.
 430  */
 431 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
 432 {
 433         char buffer[256];
 434
 435         /* We only trust the superuser with rebooting the system. */
 436         if (!capable(CAP_SYS_BOOT))
 437                 return -EPERM;
 438
 439         /* For safety, we require "magic" arguments. */
 440         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 441             (magic2 != LINUX_REBOOT_MAGIC2 &&
 442                         magic2 != LINUX_REBOOT_MAGIC2A &&
 443                         magic2 != LINUX_REBOOT_MAGIC2B &&
 444                         magic2 != LINUX_REBOOT_MAGIC2C))
 445                 return -EINVAL;
 446
 447         lock_kernel();
 448         switch (cmd) {
 449         case LINUX_REBOOT_CMD_RESTART:
 450                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 451                 system_state = SYSTEM_SHUTDOWN;
 452                 device_shutdown();
 453                 printk(KERN_EMERG "Restarting system.\n");
 454                 machine_restart(NULL);
 455                 break;
 456
 457         case LINUX_REBOOT_CMD_CAD_ON:
 458                 C_A_D = 1;
 459                 break;
 460
 461         case LINUX_REBOOT_CMD_CAD_OFF:
 462                 C_A_D = 0;
 463                 break;
 464
 465         case LINUX_REBOOT_CMD_HALT:
 466                 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
 467                 system_state = SYSTEM_SHUTDOWN;
 468                 device_shutdown();
 469                 printk(KERN_EMERG "System halted.\n");
 470                 machine_halt();
 471                 unlock_kernel();
 472                 do_exit(0);
 473                 break;
 474
 475         case LINUX_REBOOT_CMD_POWER_OFF:
 476                 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
 477                 system_state = SYSTEM_SHUTDOWN;
 478                 device_shutdown();
 479                 printk(KERN_EMERG "Power down.\n");
 480                 machine_power_off();
 481                 unlock_kernel();
 482                 do_exit(0);
 483                 break;
 484
 485         case LINUX_REBOOT_CMD_RESTART2:
 486                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 487                         unlock_kernel();
 488                         return -EFAULT;
 489                 }
 490                 buffer[sizeof(buffer) - 1] = '\0';
 491
 492                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
 493                 system_state = SYSTEM_SHUTDOWN;
 494                 device_shutdown();
 495                 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
 496                 machine_restart(buffer);
 497                 break;
 498
 499 #ifdef CONFIG_SOFTWARE_SUSPEND
 500         case LINUX_REBOOT_CMD_SW_SUSPEND:
 501                 {
 502                         int ret = software_suspend();
 503                         unlock_kernel();
 504                         return ret;
 505                 }
 506 #endif
 507
 508         default:
 509                 unlock_kernel();
 510                 return -EINVAL;
 511         }
 512         unlock_kernel();
 513         return 0;
 514 }
 515
 516 static void deferred_cad(void *dummy)
 517 {
 518         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 519         machine_restart(NULL);
 520 }
 521
 522 /*
 523  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 524  * As it's called within an interrupt, it may NOT sync: the only choice
 525  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 526  */
 527 void ctrl_alt_del(void)
 528 {
 529         static DECLARE_WORK(cad_work, deferred_cad, NULL);
 530
 531         if (C_A_D)
 532                 schedule_work(&cad_work);
 533         else
 534                 kill_proc(cad_pid, SIGINT, 1);
 535 }
 536
 537
 538 /*
 539  * Unprivileged users may change the real gid to the effective gid
 540  * or vice versa.  (BSD-style)
 541  *
 542  * If you set the real gid at all, or set the effective gid to a value not
 543  * equal to the real gid, then the saved gid is set to the new effective gid.
 544  *
 545  * This makes it possible for a setgid program to completely drop its
 546  * privileges, which is often a useful assertion to make when you are doing
 547  * a security audit over a program.
 548  *
 549  * The general idea is that a program which uses just setregid() will be
 550  * 100% compatible with BSD.  A program which uses just setgid() will be
 551  * 100% compatible with POSIX with saved IDs.
 552  *
 553  * SMP: There are not races, the GIDs are checked only by filesystem
 554  *      operations (as far as semantic preservation is concerned).
 555  */
 556 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 557 {
 558         int old_rgid = current->gid;
 559         int old_egid = current->egid;
 560         int new_rgid = old_rgid;
 561         int new_egid = old_egid;
 562         int retval;
 563
 564         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 565         if (retval)
 566                 return retval;
 567
 568         if (rgid != (gid_t) -1) {
 569                 if ((old_rgid == rgid) ||
 570                     (current->egid==rgid) ||
 571                     capable(CAP_SETGID))
 572                         new_rgid = rgid;
 573                 else
 574                         return -EPERM;
 575         }
 576         if (egid != (gid_t) -1) {
 577                 if ((old_rgid == egid) ||
 578                     (current->egid == egid) ||
 579                     (current->sgid == egid) ||
 580                     capable(CAP_SETGID))
 581                         new_egid = egid;
 582                 else {
 583                         return -EPERM;
 584                 }
 585         }
 586         if (new_egid != old_egid)
 587         {
 588                 current->mm->dumpable = 0;
 589                 wmb();
 590         }
 591         if (rgid != (gid_t) -1 ||
 592             (egid != (gid_t) -1 && egid != old_rgid))
 593                 current->sgid = new_egid;
 594         current->fsgid = new_egid;
 595         current->egid = new_egid;
 596         current->gid = new_rgid;
 597
 598         ckrm_cb_gid();
 599
 600         return 0;
 601 }
 602
 603 /*
 604  * setgid() is implemented like SysV w/ SAVED_IDS
 605  *
 606  * SMP: Same implicit races as above.
 607  */
 608 asmlinkage long sys_setgid(gid_t gid)
 609 {
 610         int old_egid = current->egid;
 611         int retval;
 612
 613         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 614         if (retval)
 615                 return retval;
 616
 617         if (capable(CAP_SETGID))
 618         {
 619                 if(old_egid != gid)
 620                 {
 621                         current->mm->dumpable=0;
 622                         wmb();
 623                 }
 624                 current->gid = current->egid = current->sgid = current->fsgid = gid;
 625         }
 626         else if ((gid == current->gid) || (gid == current->sgid))
 627         {
 628                 if(old_egid != gid)
 629                 {
 630                         current->mm->dumpable=0;
 631                         wmb();
 632                 }
 633                 current->egid = current->fsgid = gid;
 634         }
 635         else
 636                 return -EPERM;
 637
 638         ckrm_cb_gid();
 639
 640         return 0;
 641 }
 642
 643 static int set_user(uid_t new_ruid, int dumpclear)
 644 {
 645         struct user_struct *new_user;
 646
 647         new_user = alloc_uid(new_ruid);
 648         if (!new_user)
 649                 return -EAGAIN;
 650
 651         if (atomic_read(&new_user->processes) >=
 652                                 current->rlim[RLIMIT_NPROC].rlim_cur &&
 653                         new_user != &root_user) {
 654                 free_uid(new_user);
 655                 return -EAGAIN;
 656         }
 657
 658         switch_uid(new_user);
 659
 660         if(dumpclear)
 661         {
 662                 current->mm->dumpable = 0;
 663                 wmb();
 664         }
 665         current->uid = new_ruid;
 666         return 0;
 667 }
 668
 669 /*
 670  * Unprivileged users may change the real uid to the effective uid
 671  * or vice versa.  (BSD-style)
 672  *
 673  * If you set the real uid at all, or set the effective uid to a value not
 674  * equal to the real uid, then the saved uid is set to the new effective uid.
 675  *
 676  * This makes it possible for a setuid program to completely drop its
 677  * privileges, which is often a useful assertion to make when you are doing
 678  * a security audit over a program.
 679  *
 680  * The general idea is that a program which uses just setreuid() will be
 681  * 100% compatible with BSD.  A program which uses just setuid() will be
 682  * 100% compatible with POSIX with saved IDs.
 683  */
 684 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 685 {
 686         int old_ruid, old_euid, old_suid, new_ruid, new_euid;
 687         int retval;
 688
 689         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 690         if (retval)
 691                 return retval;
 692
 693         new_ruid = old_ruid = current->uid;
 694         new_euid = old_euid = current->euid;
 695         old_suid = current->suid;
 696
 697         if (ruid != (uid_t) -1) {
 698                 new_ruid = ruid;
 699                 if ((old_ruid != ruid) &&
 700                     (current->euid != ruid) &&
 701                     !capable(CAP_SETUID))
 702                         return -EPERM;
 703         }
 704
 705         if (euid != (uid_t) -1) {
 706                 new_euid = euid;
 707                 if ((old_ruid != euid) &&
 708                     (current->euid != euid) &&
 709                     (current->suid != euid) &&
 710                     !capable(CAP_SETUID))
 711                         return -EPERM;
 712         }
 713
 714         if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
 715                 return -EAGAIN;
 716
 717         if (new_euid != old_euid)
 718         {
 719                 current->mm->dumpable=0;
 720                 wmb();
 721         }
 722         current->fsuid = current->euid = new_euid;
 723         if (ruid != (uid_t) -1 ||
 724             (euid != (uid_t) -1 && euid != old_ruid))
 725                 current->suid = current->euid;
 726         current->fsuid = current->euid;
 727
 728         ckrm_cb_uid();
 729
 730         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
 731 }
 732
 733
 734
 735 /*
 736  * setuid() is implemented like SysV with SAVED_IDS
 737  *
 738  * Note that SAVED_ID's is deficient in that a setuid root program
 739  * like sendmail, for example, cannot set its uid to be a normal
 740  * user and then switch back, because if you're root, setuid() sets
 741  * the saved uid too.  If you don't like this, blame the bright people
 742  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 743  * will allow a root program to temporarily drop privileges and be able to
 744  * regain them by swapping the real and effective uid.
 745  */
 746 asmlinkage long sys_setuid(uid_t uid)
 747 {
 748         int old_euid = current->euid;
 749         int old_ruid, old_suid, new_ruid, new_suid;
 750         int retval;
 751
 752         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 753         if (retval)
 754                 return retval;
 755
 756         old_ruid = new_ruid = current->uid;
 757         old_suid = current->suid;
 758         new_suid = old_suid;
 759
 760         if (capable(CAP_SETUID)) {
 761                 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
 762                         return -EAGAIN;
 763                 new_suid = uid;
 764         } else if ((uid != current->uid) && (uid != new_suid))
 765                 return -EPERM;
 766
 767         if (old_euid != uid)
 768         {
 769                 current->mm->dumpable = 0;
 770                 wmb();
 771         }
 772         current->fsuid = current->euid = uid;
 773         current->suid = new_suid;
 774
 775         ckrm_cb_uid();
 776
 777         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
 778 }
 779
 780
 781 /*
 782  * This function implements a generic ability to update ruid, euid,
 783  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 784  */
 785 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 786 {
 787         int old_ruid = current->uid;
 788         int old_euid = current->euid;
 789         int old_suid = current->suid;
 790         int retval;
 791
 792         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 793         if (retval)
 794                 return retval;
 795
 796         if (!capable(CAP_SETUID)) {
 797                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
 798                     (ruid != current->euid) && (ruid != current->suid))
 799                         return -EPERM;
 800                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
 801                     (euid != current->euid) && (euid != current->suid))
 802                         return -EPERM;
 803                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
 804                     (suid != current->euid) && (suid != current->suid))
 805                         return -EPERM;
 806         }
 807         if (ruid != (uid_t) -1) {
 808                 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
 809                         return -EAGAIN;
 810         }
 811         if (euid != (uid_t) -1) {
 812                 if (euid != current->euid)
 813                 {
 814                         current->mm->dumpable = 0;
 815                         wmb();
 816                 }
 817                 current->euid = euid;
 818         }
 819         current->fsuid = current->euid;
 820         if (suid != (uid_t) -1)
 821                 current->suid = suid;
 822
 823         ckrm_cb_uid();
 824
 825         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
 826 }
 827
 828 asmlinkage long sys_getresuid(uid_t *ruid, uid_t *euid, uid_t *suid)
 829 {
 830         int retval;
 831
 832         if (!(retval = put_user(current->uid, ruid)) &&
 833             !(retval = put_user(current->euid, euid)))
 834                 retval = put_user(current->suid, suid);
 835
 836         return retval;
 837 }
 838
 839 /*
 840  * Same as above, but for rgid, egid, sgid.
 841  */
 842 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 843 {
 844         int retval;
 845
 846         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 847         if (retval)
 848                 return retval;
 849
 850         if (!capable(CAP_SETGID)) {
 851                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
 852                     (rgid != current->egid) && (rgid != current->sgid))
 853                         return -EPERM;
 854                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
 855                     (egid != current->egid) && (egid != current->sgid))
 856                         return -EPERM;
 857                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
 858                     (sgid != current->egid) && (sgid != current->sgid))
 859                         return -EPERM;
 860         }
 861         if (egid != (gid_t) -1) {
 862                 if (egid != current->egid)
 863                 {
 864                         current->mm->dumpable = 0;
 865                         wmb();
 866                 }
 867                 current->egid = egid;
 868         }
 869         current->fsgid = current->egid;
 870         if (rgid != (gid_t) -1)
 871                 current->gid = rgid;
 872         if (sgid != (gid_t) -1)
 873                 current->sgid = sgid;
 874
 875         ckrm_cb_gid();
 876
 877         return 0;
 878 }
 879
 880 asmlinkage long sys_getresgid(gid_t *rgid, gid_t *egid, gid_t *sgid)
 881 {
 882         int retval;
 883
 884         if (!(retval = put_user(current->gid, rgid)) &&
 885             !(retval = put_user(current->egid, egid)))
 886                 retval = put_user(current->sgid, sgid);
 887
 888         return retval;
 889 }
 890
 891
 892 /*
 893  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 894  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 895  * whatever uid it wants to). It normally shadows "euid", except when
 896  * explicitly set by setfsuid() or for access..
 897  */
 898 asmlinkage long sys_setfsuid(uid_t uid)
 899 {
 900         int old_fsuid;
 901
 902         old_fsuid = current->fsuid;
 903         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
 904                 return old_fsuid;
 905
 906         if (uid == current->uid || uid == current->euid ||
 907             uid == current->suid || uid == current->fsuid ||
 908             capable(CAP_SETUID))
 909         {
 910                 if (uid != old_fsuid)
 911                 {
 912                         current->mm->dumpable = 0;
 913                         wmb();
 914                 }
 915                 current->fsuid = uid;
 916         }
 917
 918         security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
 919
 920         return old_fsuid;
 921 }
 922
 923 /*
 924  * Samma på svenska..
 925  */
 926 asmlinkage long sys_setfsgid(gid_t gid)
 927 {
 928         int old_fsgid;
 929
 930         old_fsgid = current->fsgid;
 931         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 932                 return old_fsgid;
 933
 934         if (gid == current->gid || gid == current->egid ||
 935             gid == current->sgid || gid == current->fsgid ||
 936             capable(CAP_SETGID))
 937         {
 938                 if (gid != old_fsgid)
 939                 {
 940                         current->mm->dumpable = 0;
 941                         wmb();
 942                 }
 943                 current->fsgid = gid;
 944         }
 945         return old_fsgid;
 946 }
 947
 948 asmlinkage long sys_times(struct tms __user * tbuf)
 949 {
 950         /*
 951          *      In the SMP world we might just be unlucky and have one of
 952          *      the times increment as we use it. Since the value is an
 953          *      atomically safe type this is just fine. Conceptually its
 954          *      as if the syscall took an instant longer to occur.
 955          */
 956         if (tbuf) {
 957                 struct tms tmp;
 958                 tmp.tms_utime = jiffies_to_clock_t(current->utime);
 959                 tmp.tms_stime = jiffies_to_clock_t(current->stime);
 960                 tmp.tms_cutime = jiffies_to_clock_t(current->cutime);
 961                 tmp.tms_cstime = jiffies_to_clock_t(current->cstime);
 962                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
 963                         return -EFAULT;
 964         }
 965         return (long) jiffies_64_to_clock_t(get_jiffies_64());
 966 }
 967
 968 /*
 969  * This needs some heavy checking ...
 970  * I just haven't the stomach for it. I also don't fully
 971  * understand sessions/pgrp etc. Let somebody who does explain it.
 972  *
 973  * OK, I think I have the protection semantics right.... this is really
 974  * only important on a multi-user system anyway, to make sure one user
 975  * can't send a signal to a process owned by another.  -TYT, 12/12/91
 976  *
 977  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
 978  * LBT 04.03.94
 979  */
 980
 981 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
 982 {
 983         struct task_struct *p;
 984         int err = -EINVAL;
 985
 986         if (!pid)
 987                 pid = current->pid;
 988         if (!pgid)
 989                 pgid = pid;
 990         if (pgid < 0)
 991                 return -EINVAL;
 992
 993         /* From this point forward we keep holding onto the tasklist lock
 994          * so that our parent does not change from under us. -DaveM
 995          */
 996         write_lock_irq(&tasklist_lock);
 997
 998         err = -ESRCH;
 999         p = find_task_by_pid(pid);
1000         if (!p)
1001                 goto out;
1002
1003         err = -EINVAL;
1004         if (!thread_group_leader(p))
1005                 goto out;
1006
1007         if (p->parent == current || p->real_parent == current) {
1008                 err = -EPERM;
1009                 if (p->signal->session != current->signal->session)
1010                         goto out;
1011                 err = -EACCES;
1012                 if (p->did_exec)
1013                         goto out;
1014         } else {
1015                 err = -ESRCH;
1016                 if (p != current)
1017                         goto out;
1018         }
1019
1020         err = -EPERM;
1021         if (p->signal->leader)
1022                 goto out;
1023
1024         if (pgid != pid) {
1025                 struct task_struct *p;
1026                 struct pid *pid;
1027                 struct list_head *l;
1028
1029                 for_each_task_pid(pgid, PIDTYPE_PGID, p, l, pid)
1030                         if (p->signal->session == current->signal->session)
1031                                 goto ok_pgid;
1032                 goto out;
1033         }
1034
1035 ok_pgid:
1036         err = security_task_setpgid(p, pgid);
1037         if (err)
1038                 goto out;
1039
1040         if (process_group(p) != pgid) {
1041                 detach_pid(p, PIDTYPE_PGID);
1042                 p->signal->pgrp = pgid;
1043                 attach_pid(p, PIDTYPE_PGID, pgid);
1044         }
1045
1046         err = 0;
1047 out:
1048         /* All paths lead to here, thus we are safe. -DaveM */
1049         write_unlock_irq(&tasklist_lock);
1050         return err;
1051 }
1052
1053 asmlinkage long sys_getpgid(pid_t pid)
1054 {
1055         if (!pid) {
1056                 return process_group(current);
1057         } else {
1058                 int retval;
1059                 struct task_struct *p;
1060
1061                 read_lock(&tasklist_lock);
1062                 p = find_task_by_pid(pid);
1063
1064                 retval = -ESRCH;
1065                 if (p) {
1066                         retval = security_task_getpgid(p);
1067                         if (!retval)
1068                                 retval = process_group(p);
1069                 }
1070                 read_unlock(&tasklist_lock);
1071                 return retval;
1072         }
1073 }
1074
1075 asmlinkage long sys_getpgrp(void)
1076 {
1077         /* SMP - assuming writes are word atomic this is fine */
1078         return process_group(current);
1079 }
1080
1081 asmlinkage long sys_getsid(pid_t pid)
1082 {
1083         if (!pid) {
1084                 return current->signal->session;
1085         } else {
1086                 int retval;
1087                 struct task_struct *p;
1088
1089                 read_lock(&tasklist_lock);
1090                 p = find_task_by_pid(pid);
1091
1092                 retval = -ESRCH;
1093                 if(p) {
1094                         retval = security_task_getsid(p);
1095                         if (!retval)
1096                                 retval = p->signal->session;
1097                 }
1098                 read_unlock(&tasklist_lock);
1099                 return retval;
1100         }
1101 }
1102
1103 asmlinkage long sys_setsid(void)
1104 {
1105         struct pid *pid;
1106         int err = -EPERM;
1107
1108         if (!thread_group_leader(current))
1109                 return -EINVAL;
1110
1111         write_lock_irq(&tasklist_lock);
1112
1113         pid = find_pid(PIDTYPE_PGID, current->pid);
1114         if (pid)
1115                 goto out;
1116
1117         current->signal->leader = 1;
1118         __set_special_pids(current->pid, current->pid);
1119         current->signal->tty = NULL;
1120         current->signal->tty_old_pgrp = 0;
1121         err = process_group(current);
1122 out:
1123         write_unlock_irq(&tasklist_lock);
1124         return err;
1125 }
1126
1127 /*
1128  * Supplementary group IDs
1129  */
1130
1131 /* init to 2 - one for init_task, one to ensure it is never freed */
1132 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1133
1134 struct group_info *groups_alloc(int gidsetsize)
1135 {
1136         struct group_info *group_info;
1137         int nblocks;
1138         int i;
1139
1140         nblocks = (gidsetsize/NGROUPS_PER_BLOCK) +
1141             (gidsetsize%NGROUPS_PER_BLOCK?1:0);
1142         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *),
1143             GFP_USER);
1144         if (!group_info)
1145                 return NULL;
1146         group_info->ngroups = gidsetsize;
1147         group_info->nblocks = nblocks;
1148         atomic_set(&group_info->usage, 1);
1149
1150         if (gidsetsize <= NGROUPS_SMALL) {
1151                 group_info->blocks[0] = group_info->small_block;
1152         } else {
1153                 for (i = 0; i < nblocks; i++) {
1154                         gid_t *b;
1155                         b = (void *)__get_free_page(GFP_USER);
1156                         if (!b)
1157                                 goto out_undo_partial_alloc;
1158                         group_info->blocks[i] = b;
1159                 }
1160         }
1161         return group_info;
1162
1163 out_undo_partial_alloc:
1164         while (--i >= 0) {
1165                 free_page((unsigned long)group_info->blocks[i]);
1166         }
1167         kfree(group_info);
1168         return NULL;
1169 }
1170
1171 EXPORT_SYMBOL(groups_alloc);
1172
1173 void groups_free(struct group_info *group_info)
1174 {
1175         if (group_info->blocks[0] != group_info->small_block) {
1176                 int i;
1177                 for (i = 0; i < group_info->nblocks; i++)
1178                         free_page((unsigned long)group_info->blocks[i]);
1179         }
1180         kfree(group_info);
1181 }
1182
1183 EXPORT_SYMBOL(groups_free);
1184
1185 /* export the group_info to a user-space array */
1186 static int groups_to_user(gid_t __user *grouplist,
1187     struct group_info *group_info)
1188 {
1189         int i;
1190         int count = group_info->ngroups;
1191
1192         for (i = 0; i < group_info->nblocks; i++) {
1193                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1194                 int off = i * NGROUPS_PER_BLOCK;
1195                 int len = cp_count * sizeof(*grouplist);
1196
1197                 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1198                         return -EFAULT;
1199
1200                 count -= cp_count;
1201         }
1202         return 0;
1203 }
1204
1205 /* fill a group_info from a user-space array - it must be allocated already */
1206 static int groups_from_user(struct group_info *group_info,
1207     gid_t __user *grouplist)
1208  {
1209         int i;
1210         int count = group_info->ngroups;
1211
1212         for (i = 0; i < group_info->nblocks; i++) {
1213                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1214                 int off = i * NGROUPS_PER_BLOCK;
1215                 int len = cp_count * sizeof(*grouplist);
1216
1217                 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1218                         return -EFAULT;
1219
1220                 count -= cp_count;
1221         }
1222         return 0;
1223 }
1224
1225 /* a simple shell-metzner sort */
1226 static void groups_sort(struct group_info *group_info)
1227 {
1228         int base, max, stride;
1229         int gidsetsize = group_info->ngroups;
1230
1231         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1232                 ; /* nothing */
1233         stride /= 3;
1234
1235         while (stride) {
1236                 max = gidsetsize - stride;
1237                 for (base = 0; base < max; base++) {
1238                         int left = base;
1239                         int right = left + stride;
1240                         gid_t tmp = GROUP_AT(group_info, right);
1241
1242                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1243                                 GROUP_AT(group_info, right) =
1244                                     GROUP_AT(group_info, left);
1245                                 right = left;
1246                                 left -= stride;
1247                         }
1248                         GROUP_AT(group_info, right) = tmp;
1249                 }
1250                 stride /= 3;
1251         }
1252 }
1253
1254 /* a simple bsearch */
1255 static int groups_search(struct group_info *group_info, gid_t grp)
1256 {
1257         int left, right;
1258
1259         if (!group_info)
1260                 return 0;
1261
1262         left = 0;
1263         right = group_info->ngroups;
1264         while (left < right) {
1265                 int mid = (left+right)/2;
1266                 int cmp = grp - GROUP_AT(group_info, mid);
1267                 if (cmp > 0)
1268                         left = mid + 1;
1269                 else if (cmp < 0)
1270                         right = mid;
1271                 else
1272                         return 1;
1273         }
1274         return 0;
1275 }
1276
1277 /* validate and set current->group_info */
1278 int set_current_groups(struct group_info *group_info)
1279 {
1280         int retval;
1281         struct group_info *old_info;
1282
1283         retval = security_task_setgroups(group_info);
1284         if (retval)
1285                 return retval;
1286
1287         groups_sort(group_info);
1288         get_group_info(group_info);
1289         old_info = current->group_info;
1290         current->group_info = group_info;
1291         put_group_info(old_info);
1292
1293         return 0;
1294 }
1295
1296 EXPORT_SYMBOL(set_current_groups);
1297
1298 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1299 {
1300         int i = 0;
1301
1302         /*
1303          *      SMP: Nobody else can change our grouplist. Thus we are
1304          *      safe.
1305          */
1306
1307         if (gidsetsize < 0)
1308                 return -EINVAL;
1309
1310         get_group_info(current->group_info);
1311         i = current->group_info->ngroups;
1312         if (gidsetsize) {
1313                 if (i > gidsetsize) {
1314                         i = -EINVAL;
1315                         goto out;
1316                 }
1317                 if (groups_to_user(grouplist, current->group_info)) {
1318                         i = -EFAULT;
1319                         goto out;
1320                 }
1321         }
1322 out:
1323         put_group_info(current->group_info);
1324         return i;
1325 }
1326
1327 /*
1328  *      SMP: Our groups are copy-on-write. We can set them safely
1329  *      without another task interfering.
1330  */
1331
1332 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1333 {
1334         struct group_info *group_info;
1335         int retval;
1336
1337         if (!capable(CAP_SETGID))
1338                 return -EPERM;
1339         if ((unsigned)gidsetsize > NGROUPS_MAX)
1340                 return -EINVAL;
1341
1342         group_info = groups_alloc(gidsetsize);
1343         if (!group_info)
1344                 return -ENOMEM;
1345         retval = groups_from_user(group_info, grouplist);
1346         if (retval) {
1347                 put_group_info(group_info);
1348                 return retval;
1349         }
1350
1351         retval = set_current_groups(group_info);
1352         put_group_info(group_info);
1353
1354         return retval;
1355 }
1356
1357 /*
1358  * Check whether we're fsgid/egid or in the supplemental group..
1359  */
1360 int in_group_p(gid_t grp)
1361 {
1362         int retval = 1;
1363         if (grp != current->fsgid) {
1364                 get_group_info(current->group_info);
1365                 retval = groups_search(current->group_info, grp);
1366                 put_group_info(current->group_info);
1367         }
1368         return retval;
1369 }
1370
1371 EXPORT_SYMBOL(in_group_p);
1372
1373 int in_egroup_p(gid_t grp)
1374 {
1375         int retval = 1;
1376         if (grp != current->egid) {
1377                 get_group_info(current->group_info);
1378                 retval = groups_search(current->group_info, grp);
1379                 put_group_info(current->group_info);
1380         }
1381         return retval;
1382 }
1383
1384 EXPORT_SYMBOL(in_egroup_p);
1385
1386 DECLARE_RWSEM(uts_sem);
1387
1388 EXPORT_SYMBOL(uts_sem);
1389
1390 asmlinkage long sys_newuname(struct new_utsname __user * name)
1391 {
1392         int errno = 0;
1393
1394         down_read(&uts_sem);
1395         if (copy_to_user(name,&system_utsname,sizeof *name))
1396                 errno = -EFAULT;
1397         up_read(&uts_sem);
1398         return errno;
1399 }
1400
1401 asmlinkage long sys_sethostname(char __user *name, int len)
1402 {
1403         int errno;
1404         char tmp[__NEW_UTS_LEN];
1405
1406         if (!capable(CAP_SYS_ADMIN))
1407                 return -EPERM;
1408         if (len < 0 || len > __NEW_UTS_LEN)
1409                 return -EINVAL;
1410         down_write(&uts_sem);
1411         errno = -EFAULT;
1412         if (!copy_from_user(tmp, name, len)) {
1413                 memcpy(system_utsname.nodename, tmp, len);
1414                 system_utsname.nodename[len] = 0;
1415                 errno = 0;
1416         }
1417         up_write(&uts_sem);
1418         return errno;
1419 }
1420
1421 asmlinkage long sys_gethostname(char __user *name, int len)
1422 {
1423         int i, errno;
1424
1425         if (len < 0)
1426                 return -EINVAL;
1427         down_read(&uts_sem);
1428         i = 1 + strlen(system_utsname.nodename);
1429         if (i > len)
1430                 i = len;
1431         errno = 0;
1432         if (copy_to_user(name, system_utsname.nodename, i))
1433                 errno = -EFAULT;
1434         up_read(&uts_sem);
1435         return errno;
1436 }
1437
1438 /*
1439  * Only setdomainname; getdomainname can be implemented by calling
1440  * uname()
1441  */
1442 asmlinkage long sys_setdomainname(char __user *name, int len)
1443 {
1444         int errno;
1445         char tmp[__NEW_UTS_LEN];
1446
1447         if (!capable(CAP_SYS_ADMIN))
1448                 return -EPERM;
1449         if (len < 0 || len > __NEW_UTS_LEN)
1450                 return -EINVAL;
1451
1452         down_write(&uts_sem);
1453         errno = -EFAULT;
1454         if (!copy_from_user(tmp, name, len)) {
1455                 memcpy(system_utsname.domainname, tmp, len);
1456                 system_utsname.domainname[len] = 0;
1457                 errno = 0;
1458         }
1459         up_write(&uts_sem);
1460         return errno;
1461 }
1462
1463 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1464 {
1465         if (resource >= RLIM_NLIMITS)
1466                 return -EINVAL;
1467         else
1468                 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
1469                         ? -EFAULT : 0;
1470 }
1471
1472 #if defined(COMPAT_RLIM_OLD_INFINITY) || !(defined(CONFIG_IA64) || defined(CONFIG_V850))
1473
1474 /*
1475  *      Back compatibility for getrlimit. Needed for some apps.
1476  */
1477
1478 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1479 {
1480         struct rlimit x;
1481         if (resource >= RLIM_NLIMITS)
1482                 return -EINVAL;
1483
1484         memcpy(&x, current->rlim + resource, sizeof(*rlim));
1485         if(x.rlim_cur > 0x7FFFFFFF)
1486                 x.rlim_cur = 0x7FFFFFFF;
1487         if(x.rlim_max > 0x7FFFFFFF)
1488                 x.rlim_max = 0x7FFFFFFF;
1489         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1490 }
1491
1492 #endif
1493
1494 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1495 {
1496         struct rlimit new_rlim, *old_rlim;
1497         int retval;
1498
1499         if (resource >= RLIM_NLIMITS)
1500                 return -EINVAL;
1501         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1502                 return -EFAULT;
1503        if (new_rlim.rlim_cur > new_rlim.rlim_max)
1504                return -EINVAL;
1505         old_rlim = current->rlim + resource;
1506         if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
1507              (new_rlim.rlim_max > old_rlim->rlim_max)) &&
1508             !capable(CAP_SYS_RESOURCE))
1509                 return -EPERM;
1510         if (resource == RLIMIT_NOFILE) {
1511                 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
1512                         return -EPERM;
1513         }
1514
1515         retval = security_task_setrlimit(resource, &new_rlim);
1516         if (retval)
1517                 return retval;
1518
1519         *old_rlim = new_rlim;
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 is SMP safe.  Either we are called from sys_getrusage on ourselves
1532  * below (we know we aren't going to exit/disappear and only we change our
1533  * rusage counters), or we are called from wait4() on a process which is
1534  * either stopped or zombied.  In the zombied case the task won't get
1535  * reaped till shortly after the call to getrusage(), in both cases the
1536  * task being examined is in a frozen state so the counters won't change.
1537  */
1538 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1539 {
1540         struct rusage r;
1541
1542         memset((char *) &r, 0, sizeof(r));
1543         switch (who) {
1544                 case RUSAGE_SELF:
1545                         jiffies_to_timeval(p->utime, &r.ru_utime);
1546                         jiffies_to_timeval(p->stime, &r.ru_stime);
1547                         r.ru_nvcsw = p->nvcsw;
1548                         r.ru_nivcsw = p->nivcsw;
1549                         r.ru_minflt = p->min_flt;
1550                         r.ru_majflt = p->maj_flt;
1551                         break;
1552                 case RUSAGE_CHILDREN:
1553                         jiffies_to_timeval(p->cutime, &r.ru_utime);
1554                         jiffies_to_timeval(p->cstime, &r.ru_stime);
1555                         r.ru_nvcsw = p->cnvcsw;
1556                         r.ru_nivcsw = p->cnivcsw;
1557                         r.ru_minflt = p->cmin_flt;
1558                         r.ru_majflt = p->cmaj_flt;
1559                         break;
1560                 default:
1561                         jiffies_to_timeval(p->utime + p->cutime, &r.ru_utime);
1562                         jiffies_to_timeval(p->stime + p->cstime, &r.ru_stime);
1563                         r.ru_nvcsw = p->nvcsw + p->cnvcsw;
1564                         r.ru_nivcsw = p->nivcsw + p->cnivcsw;
1565                         r.ru_minflt = p->min_flt + p->cmin_flt;
1566                         r.ru_majflt = p->maj_flt + p->cmaj_flt;
1567                         break;
1568         }
1569         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1570 }
1571
1572 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1573 {
1574         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1575                 return -EINVAL;
1576         return getrusage(current, who, ru);
1577 }
1578
1579 asmlinkage long sys_umask(int mask)
1580 {
1581         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1582         return mask;
1583 }
1584
1585 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1586                           unsigned long arg4, unsigned long arg5)
1587 {
1588         int error;
1589         int sig;
1590
1591         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1592         if (error)
1593                 return error;
1594
1595         switch (option) {
1596                 case PR_SET_PDEATHSIG:
1597                         sig = arg2;
1598                         if (sig < 0 || sig > _NSIG) {
1599                                 error = -EINVAL;
1600                                 break;
1601                         }
1602                         current->pdeath_signal = sig;
1603                         break;
1604                 case PR_GET_PDEATHSIG:
1605                         error = put_user(current->pdeath_signal, (int __user *)arg2);
1606                         break;
1607                 case PR_GET_DUMPABLE:
1608                         if (current->mm->dumpable)
1609                                 error = 1;
1610                         break;
1611                 case PR_SET_DUMPABLE:
1612                         if (arg2 != 0 && arg2 != 1) {
1613                                 error = -EINVAL;
1614                                 break;
1615                         }
1616                         current->mm->dumpable = arg2;
1617                         break;
1618
1619                 case PR_SET_UNALIGN:
1620                         error = SET_UNALIGN_CTL(current, arg2);
1621                         break;
1622                 case PR_GET_UNALIGN:
1623                         error = GET_UNALIGN_CTL(current, arg2);
1624                         break;
1625                 case PR_SET_FPEMU:
1626                         error = SET_FPEMU_CTL(current, arg2);
1627                         break;
1628                 case PR_GET_FPEMU:
1629                         error = GET_FPEMU_CTL(current, arg2);
1630                         break;
1631                 case PR_SET_FPEXC:
1632                         error = SET_FPEXC_CTL(current, arg2);
1633                         break;
1634                 case PR_GET_FPEXC:
1635                         error = GET_FPEXC_CTL(current, arg2);
1636                         break;
1637                 case PR_GET_TIMING:
1638                         error = PR_TIMING_STATISTICAL;
1639                         break;
1640                 case PR_SET_TIMING:
1641                         if (arg2 == PR_TIMING_STATISTICAL)
1642                                 error = 0;
1643                         else
1644                                 error = -EINVAL;
1645                         break;
1646
1647                 case PR_GET_KEEPCAPS:
1648                         if (current->keep_capabilities)
1649                                 error = 1;
1650                         break;
1651                 case PR_SET_KEEPCAPS:
1652                         if (arg2 != 0 && arg2 != 1) {
1653                                 error = -EINVAL;
1654                                 break;
1655                         }
1656                         current->keep_capabilities = arg2;
1657                         break;
1658                 default:
1659                         error = -EINVAL;
1660                         break;
1661         }
1662         return error;
1663 }