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/kmod.h>
  16 #include <linux/reboot.h>
  17 #include <linux/prctl.h>
  18 #include <linux/init.h>
  19 #include <linux/highuid.h>
  20 #include <linux/fs.h>
  21 #include <linux/workqueue.h>
  22 #include <linux/device.h>
  23 #include <linux/times.h>
  24 #include <linux/security.h>
  25 #include <linux/dcookies.h>
  26 #include <linux/suspend.h>
  27 #include <linux/vs_base.h>
  28 #include <linux/vs_cvirt.h>
  29
  30 #include <asm/uaccess.h>
  31 #include <asm/io.h>
  32 #include <asm/unistd.h>
  33
  34 #ifndef SET_UNALIGN_CTL
  35 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  36 #endif
  37 #ifndef GET_UNALIGN_CTL
  38 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  39 #endif
  40 #ifndef SET_FPEMU_CTL
  41 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
  42 #endif
  43 #ifndef GET_FPEMU_CTL
  44 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
  45 #endif
  46 #ifndef SET_FPEXC_CTL
  47 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
  48 #endif
  49 #ifndef GET_FPEXC_CTL
  50 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
  51 #endif
  52
  53 /*
  54  * this is where the system-wide overflow UID and GID are defined, for
  55  * architectures that now have 32-bit UID/GID but didn't in the past
  56  */
  57
  58 int overflowuid = DEFAULT_OVERFLOWUID;
  59 int overflowgid = DEFAULT_OVERFLOWGID;
  60
  61 #ifdef CONFIG_UID16
  62 EXPORT_SYMBOL(overflowuid);
  63 EXPORT_SYMBOL(overflowgid);
  64 #endif
  65
  66 /*
  67  * the same as above, but for filesystems which can only store a 16-bit
  68  * UID and GID. as such, this is needed on all architectures
  69  */
  70
  71 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
  72 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
  73
  74 EXPORT_SYMBOL(fs_overflowuid);
  75 EXPORT_SYMBOL(fs_overflowgid);
  76
  77 /*
  78  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
  79  */
  80
  81 int C_A_D = 1;
  82 int cad_pid = 1;
  83
  84 /*
  85  *      Notifier list for kernel code which wants to be called
  86  *      at shutdown. This is used to stop any idling DMA operations
  87  *      and the like.
  88  */
  89
  90 static struct notifier_block *reboot_notifier_list;
  91 rwlock_t notifier_lock = RW_LOCK_UNLOCKED;
  92
  93 /**
  94  *      notifier_chain_register - Add notifier to a notifier chain
  95  *      @list: Pointer to root list pointer
  96  *      @n: New entry in notifier chain
  97  *
  98  *      Adds a notifier to a notifier chain.
  99  *
 100  *      Currently always returns zero.
 101  */
 102
 103 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
 104 {
 105         write_lock(&notifier_lock);
 106         while(*list)
 107         {
 108                 if(n->priority > (*list)->priority)
 109                         break;
 110                 list= &((*list)->next);
 111         }
 112         n->next = *list;
 113         *list=n;
 114         write_unlock(&notifier_lock);
 115         return 0;
 116 }
 117
 118 EXPORT_SYMBOL(notifier_chain_register);
 119
 120 /**
 121  *      notifier_chain_unregister - Remove notifier from a notifier chain
 122  *      @nl: Pointer to root list pointer
 123  *      @n: New entry in notifier chain
 124  *
 125  *      Removes a notifier from a notifier chain.
 126  *
 127  *      Returns zero on success, or %-ENOENT on failure.
 128  */
 129
 130 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
 131 {
 132         write_lock(&notifier_lock);
 133         while((*nl)!=NULL)
 134         {
 135                 if((*nl)==n)
 136                 {
 137                         *nl=n->next;
 138                         write_unlock(&notifier_lock);
 139                         return 0;
 140                 }
 141                 nl=&((*nl)->next);
 142         }
 143         write_unlock(&notifier_lock);
 144         return -ENOENT;
 145 }
 146
 147 EXPORT_SYMBOL(notifier_chain_unregister);
 148
 149 /**
 150  *      notifier_call_chain - Call functions in a notifier chain
 151  *      @n: Pointer to root pointer of notifier chain
 152  *      @val: Value passed unmodified to notifier function
 153  *      @v: Pointer passed unmodified to notifier function
 154  *
 155  *      Calls each function in a notifier chain in turn.
 156  *
 157  *      If the return value of the notifier can be and'd
 158  *      with %NOTIFY_STOP_MASK, then notifier_call_chain
 159  *      will return immediately, with the return value of
 160  *      the notifier function which halted execution.
 161  *      Otherwise, the return value is the return value
 162  *      of the last notifier function called.
 163  */
 164
 165 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
 166 {
 167         int ret=NOTIFY_DONE;
 168         struct notifier_block *nb = *n;
 169
 170         while(nb)
 171         {
 172                 ret=nb->notifier_call(nb,val,v);
 173                 if(ret&NOTIFY_STOP_MASK)
 174                 {
 175                         return ret;
 176                 }
 177                 nb=nb->next;
 178         }
 179         return ret;
 180 }
 181
 182 EXPORT_SYMBOL(notifier_call_chain);
 183
 184 /**
 185  *      register_reboot_notifier - Register function to be called at reboot time
 186  *      @nb: Info about notifier function to be called
 187  *
 188  *      Registers a function with the list of functions
 189  *      to be called at reboot time.
 190  *
 191  *      Currently always returns zero, as notifier_chain_register
 192  *      always returns zero.
 193  */
 194
 195 int register_reboot_notifier(struct notifier_block * nb)
 196 {
 197         return notifier_chain_register(&reboot_notifier_list, nb);
 198 }
 199
 200 EXPORT_SYMBOL(register_reboot_notifier);
 201
 202 /**
 203  *      unregister_reboot_notifier - Unregister previously registered reboot notifier
 204  *      @nb: Hook to be unregistered
 205  *
 206  *      Unregisters a previously registered reboot
 207  *      notifier function.
 208  *
 209  *      Returns zero on success, or %-ENOENT on failure.
 210  */
 211
 212 int unregister_reboot_notifier(struct notifier_block * nb)
 213 {
 214         return notifier_chain_unregister(&reboot_notifier_list, nb);
 215 }
 216
 217 EXPORT_SYMBOL(unregister_reboot_notifier);
 218
 219 asmlinkage long sys_ni_syscall(void)
 220 {
 221         return -ENOSYS;
 222 }
 223
 224 cond_syscall(sys_nfsservctl)
 225 cond_syscall(sys_quotactl)
 226 cond_syscall(sys_acct)
 227 cond_syscall(sys_lookup_dcookie)
 228 cond_syscall(sys_swapon)
 229 cond_syscall(sys_swapoff)
 230 cond_syscall(sys_init_module)
 231 cond_syscall(sys_delete_module)
 232 cond_syscall(sys_socketpair)
 233 cond_syscall(sys_bind)
 234 cond_syscall(sys_listen)
 235 cond_syscall(sys_accept)
 236 cond_syscall(sys_connect)
 237 cond_syscall(sys_getsockname)
 238 cond_syscall(sys_getpeername)
 239 cond_syscall(sys_sendto)
 240 cond_syscall(sys_send)
 241 cond_syscall(sys_recvfrom)
 242 cond_syscall(sys_recv)
 243 cond_syscall(sys_socket)
 244 cond_syscall(sys_setsockopt)
 245 cond_syscall(sys_getsockopt)
 246 cond_syscall(sys_shutdown)
 247 cond_syscall(sys_sendmsg)
 248 cond_syscall(sys_recvmsg)
 249 cond_syscall(sys_socketcall)
 250 cond_syscall(sys_futex)
 251 cond_syscall(compat_sys_futex)
 252 cond_syscall(sys_epoll_create)
 253 cond_syscall(sys_epoll_ctl)
 254 cond_syscall(sys_epoll_wait)
 255 cond_syscall(sys_semget)
 256 cond_syscall(sys_semop)
 257 cond_syscall(sys_semtimedop)
 258 cond_syscall(sys_semctl)
 259 cond_syscall(sys_msgget)
 260 cond_syscall(sys_msgsnd)
 261 cond_syscall(sys_msgrcv)
 262 cond_syscall(sys_msgctl)
 263 cond_syscall(sys_shmget)
 264 cond_syscall(sys_shmdt)
 265 cond_syscall(sys_shmctl)
 266 cond_syscall(sys_mq_open)
 267 cond_syscall(sys_mq_unlink)
 268 cond_syscall(sys_mq_timedsend)
 269 cond_syscall(sys_mq_timedreceive)
 270 cond_syscall(sys_mq_notify)
 271 cond_syscall(sys_mq_getsetattr)
 272 cond_syscall(compat_sys_mq_open)
 273 cond_syscall(compat_sys_mq_timedsend)
 274 cond_syscall(compat_sys_mq_timedreceive)
 275 cond_syscall(compat_sys_mq_notify)
 276 cond_syscall(compat_sys_mq_getsetattr)
 277 cond_syscall(sys_mbind)
 278 cond_syscall(sys_get_mempolicy)
 279 cond_syscall(sys_set_mempolicy)
 280 cond_syscall(compat_mbind)
 281 cond_syscall(compat_get_mempolicy)
 282 cond_syscall(compat_set_mempolicy)
 283
 284 /* arch-specific weak syscall entries */
 285 cond_syscall(sys_pciconfig_read)
 286 cond_syscall(sys_pciconfig_write)
 287 cond_syscall(sys_pciconfig_iobase)
 288
 289 static int set_one_prio(struct task_struct *p, int niceval, int error)
 290 {
 291         int no_nice;
 292
 293         if (p->uid != current->euid &&
 294                 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
 295                 error = -EPERM;
 296                 goto out;
 297         }
 298         if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
 299                 if (vx_flags(VXF_IGNEG_NICE, 0))
 300                         error = 0;
 301                 else
 302                         error = -EACCES;
 303                 goto out;
 304         }
 305         no_nice = security_task_setnice(p, niceval);
 306         if (no_nice) {
 307                 error = no_nice;
 308                 goto out;
 309         }
 310         if (error == -ESRCH)
 311                 error = 0;
 312         set_user_nice(p, niceval);
 313 out:
 314         return error;
 315 }
 316
 317 asmlinkage long sys_setpriority(int which, int who, int niceval)
 318 {
 319         struct task_struct *g, *p;
 320         struct user_struct *user;
 321         int error = -EINVAL;
 322
 323         if (which > 2 || which < 0)
 324                 goto out;
 325
 326         /* normalize: avoid signed division (rounding problems) */
 327         error = -ESRCH;
 328         if (niceval < -20)
 329                 niceval = -20;
 330         if (niceval > 19)
 331                 niceval = 19;
 332
 333         read_lock(&tasklist_lock);
 334         switch (which) {
 335                 case PRIO_PROCESS:
 336                         if (!who)
 337                                 who = current->pid;
 338                         p = find_task_by_pid(who);
 339                         if (p)
 340                                 error = set_one_prio(p, niceval, error);
 341                         break;
 342                 case PRIO_PGRP:
 343                         if (!who)
 344                                 who = process_group(current);
 345                         do_each_task_pid(who, PIDTYPE_PGID, p) {
 346                                 error = set_one_prio(p, niceval, error);
 347                         } while_each_task_pid(who, PIDTYPE_PGID, p);
 348                         break;
 349                 case PRIO_USER:
 350                         if (!who)
 351                                 user = current->user;
 352                         else
 353                                 user = find_user(vx_current_xid(), who);
 354
 355                         if (!user)
 356                                 goto out_unlock;
 357
 358                         do_each_thread(g, p)
 359                                 if (p->uid == who)
 360                                         error = set_one_prio(p, niceval, error);
 361                         while_each_thread(g, p);
 362                         if (who)
 363                                 free_uid(user);         /* For find_user() */
 364                         break;
 365         }
 366 out_unlock:
 367         read_unlock(&tasklist_lock);
 368 out:
 369         return error;
 370 }
 371
 372 /*
 373  * Ugh. To avoid negative return values, "getpriority()" will
 374  * not return the normal nice-value, but a negated value that
 375  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 376  * to stay compatible.
 377  */
 378 asmlinkage long sys_getpriority(int which, int who)
 379 {
 380         struct task_struct *g, *p;
 381         struct user_struct *user;
 382         long niceval, retval = -ESRCH;
 383
 384         if (which > 2 || which < 0)
 385                 return -EINVAL;
 386
 387         read_lock(&tasklist_lock);
 388         switch (which) {
 389                 case PRIO_PROCESS:
 390                         if (!who)
 391                                 who = current->pid;
 392                         p = find_task_by_pid(who);
 393                         if (p) {
 394                                 niceval = 20 - task_nice(p);
 395                                 if (niceval > retval)
 396                                         retval = niceval;
 397                         }
 398                         break;
 399                 case PRIO_PGRP:
 400                         if (!who)
 401                                 who = process_group(current);
 402                         do_each_task_pid(who, PIDTYPE_PGID, p) {
 403                                 niceval = 20 - task_nice(p);
 404                                 if (niceval > retval)
 405                                         retval = niceval;
 406                         } while_each_task_pid(who, PIDTYPE_PGID, p);
 407                         break;
 408                 case PRIO_USER:
 409                         if (!who)
 410                                 user = current->user;
 411                         else
 412                                 user = find_user(vx_current_xid(), who);
 413
 414                         if (!user)
 415                                 goto out_unlock;
 416
 417                         do_each_thread(g, p)
 418                                 if (p->uid == who) {
 419                                         niceval = 20 - task_nice(p);
 420                                         if (niceval > retval)
 421                                                 retval = niceval;
 422                                 }
 423                         while_each_thread(g, p);
 424                         if (who)
 425                                 free_uid(user);         /* for find_user() */
 426                         break;
 427         }
 428 out_unlock:
 429         read_unlock(&tasklist_lock);
 430
 431         return retval;
 432 }
 433
 434 long vs_reboot(unsigned int, void *);
 435
 436 /*
 437  * Reboot system call: for obvious reasons only root may call it,
 438  * and even root needs to set up some magic numbers in the registers
 439  * so that some mistake won't make this reboot the whole machine.
 440  * You can also set the meaning of the ctrl-alt-del-key here.
 441  *
 442  * reboot doesn't sync: do that yourself before calling this.
 443  */
 444 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
 445 {
 446         char buffer[256];
 447
 448         /* We only trust the superuser with rebooting the system. */
 449         if (!capable(CAP_SYS_BOOT))
 450                 return -EPERM;
 451
 452         /* For safety, we require "magic" arguments. */
 453         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 454             (magic2 != LINUX_REBOOT_MAGIC2 &&
 455                         magic2 != LINUX_REBOOT_MAGIC2A &&
 456                         magic2 != LINUX_REBOOT_MAGIC2B &&
 457                         magic2 != LINUX_REBOOT_MAGIC2C))
 458                 return -EINVAL;
 459
 460         if (!vx_check(0, VX_ADMIN|VX_WATCH))
 461                 return vs_reboot(cmd, arg);
 462
 463         lock_kernel();
 464         switch (cmd) {
 465         case LINUX_REBOOT_CMD_RESTART:
 466                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 467                 system_state = SYSTEM_RESTART;
 468                 device_shutdown();
 469                 printk(KERN_EMERG "Restarting system.\n");
 470                 machine_restart(NULL);
 471                 break;
 472
 473         case LINUX_REBOOT_CMD_CAD_ON:
 474                 C_A_D = 1;
 475                 break;
 476
 477         case LINUX_REBOOT_CMD_CAD_OFF:
 478                 C_A_D = 0;
 479                 break;
 480
 481         case LINUX_REBOOT_CMD_HALT:
 482                 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
 483                 system_state = SYSTEM_HALT;
 484                 device_shutdown();
 485                 printk(KERN_EMERG "System halted.\n");
 486                 machine_halt();
 487                 unlock_kernel();
 488                 do_exit(0);
 489                 break;
 490
 491         case LINUX_REBOOT_CMD_POWER_OFF:
 492                 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
 493                 system_state = SYSTEM_POWER_OFF;
 494                 device_shutdown();
 495                 printk(KERN_EMERG "Power down.\n");
 496                 machine_power_off();
 497                 unlock_kernel();
 498                 do_exit(0);
 499                 break;
 500
 501         case LINUX_REBOOT_CMD_RESTART2:
 502                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 503                         unlock_kernel();
 504                         return -EFAULT;
 505                 }
 506                 buffer[sizeof(buffer) - 1] = '\0';
 507
 508                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
 509                 system_state = SYSTEM_RESTART;
 510                 device_shutdown();
 511                 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
 512                 machine_restart(buffer);
 513                 break;
 514
 515 #ifdef CONFIG_SOFTWARE_SUSPEND
 516         case LINUX_REBOOT_CMD_SW_SUSPEND:
 517                 {
 518                         int ret = software_suspend();
 519                         unlock_kernel();
 520                         return ret;
 521                 }
 522 #endif
 523
 524         default:
 525                 unlock_kernel();
 526                 return -EINVAL;
 527         }
 528         unlock_kernel();
 529         return 0;
 530 }
 531
 532 static void deferred_cad(void *dummy)
 533 {
 534         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 535         machine_restart(NULL);
 536 }
 537
 538 /*
 539  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 540  * As it's called within an interrupt, it may NOT sync: the only choice
 541  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 542  */
 543 void ctrl_alt_del(void)
 544 {
 545         static DECLARE_WORK(cad_work, deferred_cad, NULL);
 546
 547         if (C_A_D)
 548                 schedule_work(&cad_work);
 549         else
 550                 kill_proc(cad_pid, SIGINT, 1);
 551 }
 552
 553
 554 /*
 555  * Unprivileged users may change the real gid to the effective gid
 556  * or vice versa.  (BSD-style)
 557  *
 558  * If you set the real gid at all, or set the effective gid to a value not
 559  * equal to the real gid, then the saved gid is set to the new effective gid.
 560  *
 561  * This makes it possible for a setgid program to completely drop its
 562  * privileges, which is often a useful assertion to make when you are doing
 563  * a security audit over a program.
 564  *
 565  * The general idea is that a program which uses just setregid() will be
 566  * 100% compatible with BSD.  A program which uses just setgid() will be
 567  * 100% compatible with POSIX with saved IDs.
 568  *
 569  * SMP: There are not races, the GIDs are checked only by filesystem
 570  *      operations (as far as semantic preservation is concerned).
 571  */
 572 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 573 {
 574         int old_rgid = current->gid;
 575         int old_egid = current->egid;
 576         int new_rgid = old_rgid;
 577         int new_egid = old_egid;
 578         int retval;
 579
 580         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 581         if (retval)
 582                 return retval;
 583
 584         if (rgid != (gid_t) -1) {
 585                 if ((old_rgid == rgid) ||
 586                     (current->egid==rgid) ||
 587                     capable(CAP_SETGID))
 588                         new_rgid = rgid;
 589                 else
 590                         return -EPERM;
 591         }
 592         if (egid != (gid_t) -1) {
 593                 if ((old_rgid == egid) ||
 594                     (current->egid == egid) ||
 595                     (current->sgid == egid) ||
 596                     capable(CAP_SETGID))
 597                         new_egid = egid;
 598                 else {
 599                         return -EPERM;
 600                 }
 601         }
 602         if (new_egid != old_egid)
 603         {
 604                 current->mm->dumpable = 0;
 605                 wmb();
 606         }
 607         if (rgid != (gid_t) -1 ||
 608             (egid != (gid_t) -1 && egid != old_rgid))
 609                 current->sgid = new_egid;
 610         current->fsgid = new_egid;
 611         current->egid = new_egid;
 612         current->gid = new_rgid;
 613         return 0;
 614 }
 615
 616 /*
 617  * setgid() is implemented like SysV w/ SAVED_IDS
 618  *
 619  * SMP: Same implicit races as above.
 620  */
 621 asmlinkage long sys_setgid(gid_t gid)
 622 {
 623         int old_egid = current->egid;
 624         int retval;
 625
 626         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 627         if (retval)
 628                 return retval;
 629
 630         if (capable(CAP_SETGID))
 631         {
 632                 if(old_egid != gid)
 633                 {
 634                         current->mm->dumpable=0;
 635                         wmb();
 636                 }
 637                 current->gid = current->egid = current->sgid = current->fsgid = gid;
 638         }
 639         else if ((gid == current->gid) || (gid == current->sgid))
 640         {
 641                 if(old_egid != gid)
 642                 {
 643                         current->mm->dumpable=0;
 644                         wmb();
 645                 }
 646                 current->egid = current->fsgid = gid;
 647         }
 648         else
 649                 return -EPERM;
 650         return 0;
 651 }
 652
 653 static int set_user(uid_t new_ruid, int dumpclear)
 654 {
 655         struct user_struct *new_user;
 656
 657         new_user = alloc_uid(vx_current_xid(), new_ruid);
 658         if (!new_user)
 659                 return -EAGAIN;
 660
 661         if (atomic_read(&new_user->processes) >=
 662                                 current->rlim[RLIMIT_NPROC].rlim_cur &&
 663                         new_user != &root_user) {
 664                 free_uid(new_user);
 665                 return -EAGAIN;
 666         }
 667
 668         switch_uid(new_user);
 669
 670         if(dumpclear)
 671         {
 672                 current->mm->dumpable = 0;
 673                 wmb();
 674         }
 675         current->uid = new_ruid;
 676         return 0;
 677 }
 678
 679 /*
 680  * Unprivileged users may change the real uid to the effective uid
 681  * or vice versa.  (BSD-style)
 682  *
 683  * If you set the real uid at all, or set the effective uid to a value not
 684  * equal to the real uid, then the saved uid is set to the new effective uid.
 685  *
 686  * This makes it possible for a setuid program to completely drop its
 687  * privileges, which is often a useful assertion to make when you are doing
 688  * a security audit over a program.
 689  *
 690  * The general idea is that a program which uses just setreuid() will be
 691  * 100% compatible with BSD.  A program which uses just setuid() will be
 692  * 100% compatible with POSIX with saved IDs.
 693  */
 694 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 695 {
 696         int old_ruid, old_euid, old_suid, new_ruid, new_euid;
 697         int retval;
 698
 699         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 700         if (retval)
 701                 return retval;
 702
 703         new_ruid = old_ruid = current->uid;
 704         new_euid = old_euid = current->euid;
 705         old_suid = current->suid;
 706
 707         if (ruid != (uid_t) -1) {
 708                 new_ruid = ruid;
 709                 if ((old_ruid != ruid) &&
 710                     (current->euid != ruid) &&
 711                     !capable(CAP_SETUID))
 712                         return -EPERM;
 713         }
 714
 715         if (euid != (uid_t) -1) {
 716                 new_euid = euid;
 717                 if ((old_ruid != euid) &&
 718                     (current->euid != euid) &&
 719                     (current->suid != euid) &&
 720                     !capable(CAP_SETUID))
 721                         return -EPERM;
 722         }
 723
 724         if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
 725                 return -EAGAIN;
 726
 727         if (new_euid != old_euid)
 728         {
 729                 current->mm->dumpable=0;
 730                 wmb();
 731         }
 732         current->fsuid = current->euid = new_euid;
 733         if (ruid != (uid_t) -1 ||
 734             (euid != (uid_t) -1 && euid != old_ruid))
 735                 current->suid = current->euid;
 736         current->fsuid = current->euid;
 737
 738         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
 739 }
 740
 741
 742
 743 /*
 744  * setuid() is implemented like SysV with SAVED_IDS
 745  *
 746  * Note that SAVED_ID's is deficient in that a setuid root program
 747  * like sendmail, for example, cannot set its uid to be a normal
 748  * user and then switch back, because if you're root, setuid() sets
 749  * the saved uid too.  If you don't like this, blame the bright people
 750  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 751  * will allow a root program to temporarily drop privileges and be able to
 752  * regain them by swapping the real and effective uid.
 753  */
 754 asmlinkage long sys_setuid(uid_t uid)
 755 {
 756         int old_euid = current->euid;
 757         int old_ruid, old_suid, new_ruid, new_suid;
 758         int retval;
 759
 760         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 761         if (retval)
 762                 return retval;
 763
 764         old_ruid = new_ruid = current->uid;
 765         old_suid = current->suid;
 766         new_suid = old_suid;
 767
 768         if (capable(CAP_SETUID)) {
 769                 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
 770                         return -EAGAIN;
 771                 new_suid = uid;
 772         } else if ((uid != current->uid) && (uid != new_suid))
 773                 return -EPERM;
 774
 775         if (old_euid != uid)
 776         {
 777                 current->mm->dumpable = 0;
 778                 wmb();
 779         }
 780         current->fsuid = current->euid = uid;
 781         current->suid = new_suid;
 782
 783         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
 784 }
 785
 786
 787 /*
 788  * This function implements a generic ability to update ruid, euid,
 789  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 790  */
 791 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 792 {
 793         int old_ruid = current->uid;
 794         int old_euid = current->euid;
 795         int old_suid = current->suid;
 796         int retval;
 797
 798         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 799         if (retval)
 800                 return retval;
 801
 802         if (!capable(CAP_SETUID)) {
 803                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
 804                     (ruid != current->euid) && (ruid != current->suid))
 805                         return -EPERM;
 806                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
 807                     (euid != current->euid) && (euid != current->suid))
 808                         return -EPERM;
 809                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
 810                     (suid != current->euid) && (suid != current->suid))
 811                         return -EPERM;
 812         }
 813         if (ruid != (uid_t) -1) {
 814                 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
 815                         return -EAGAIN;
 816         }
 817         if (euid != (uid_t) -1) {
 818                 if (euid != current->euid)
 819                 {
 820                         current->mm->dumpable = 0;
 821                         wmb();
 822                 }
 823                 current->euid = euid;
 824         }
 825         current->fsuid = current->euid;
 826         if (suid != (uid_t) -1)
 827                 current->suid = suid;
 828
 829         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
 830 }
 831
 832 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
 833 {
 834         int retval;
 835
 836         if (!(retval = put_user(current->uid, ruid)) &&
 837             !(retval = put_user(current->euid, euid)))
 838                 retval = put_user(current->suid, suid);
 839
 840         return retval;
 841 }
 842
 843 /*
 844  * Same as above, but for rgid, egid, sgid.
 845  */
 846 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 847 {
 848         int retval;
 849
 850         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 851         if (retval)
 852                 return retval;
 853
 854         if (!capable(CAP_SETGID)) {
 855                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
 856                     (rgid != current->egid) && (rgid != current->sgid))
 857                         return -EPERM;
 858                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
 859                     (egid != current->egid) && (egid != current->sgid))
 860                         return -EPERM;
 861                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
 862                     (sgid != current->egid) && (sgid != current->sgid))
 863                         return -EPERM;
 864         }
 865         if (egid != (gid_t) -1) {
 866                 if (egid != current->egid)
 867                 {
 868                         current->mm->dumpable = 0;
 869                         wmb();
 870                 }
 871                 current->egid = egid;
 872         }
 873         current->fsgid = current->egid;
 874         if (rgid != (gid_t) -1)
 875                 current->gid = rgid;
 876         if (sgid != (gid_t) -1)
 877                 current->sgid = sgid;
 878         return 0;
 879 }
 880
 881 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
 882 {
 883         int retval;
 884
 885         if (!(retval = put_user(current->gid, rgid)) &&
 886             !(retval = put_user(current->egid, egid)))
 887                 retval = put_user(current->sgid, sgid);
 888
 889         return retval;
 890 }
 891
 892
 893 /*
 894  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 895  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 896  * whatever uid it wants to). It normally shadows "euid", except when
 897  * explicitly set by setfsuid() or for access..
 898  */
 899 asmlinkage long sys_setfsuid(uid_t uid)
 900 {
 901         int old_fsuid;
 902
 903         old_fsuid = current->fsuid;
 904         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
 905                 return old_fsuid;
 906
 907         if (uid == current->uid || uid == current->euid ||
 908             uid == current->suid || uid == current->fsuid ||
 909             capable(CAP_SETUID))
 910         {
 911                 if (uid != old_fsuid)
 912                 {
 913                         current->mm->dumpable = 0;
 914                         wmb();
 915                 }
 916                 current->fsuid = uid;
 917         }
 918
 919         security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
 920
 921         return old_fsuid;
 922 }
 923
 924 /*
 925  * Samma på svenska..
 926  */
 927 asmlinkage long sys_setfsgid(gid_t gid)
 928 {
 929         int old_fsgid;
 930
 931         old_fsgid = current->fsgid;
 932         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 933                 return old_fsgid;
 934
 935         if (gid == current->gid || gid == current->egid ||
 936             gid == current->sgid || gid == current->fsgid ||
 937             capable(CAP_SETGID))
 938         {
 939                 if (gid != old_fsgid)
 940                 {
 941                         current->mm->dumpable = 0;
 942                         wmb();
 943                 }
 944                 current->fsgid = gid;
 945         }
 946         return old_fsgid;
 947 }
 948
 949 asmlinkage long sys_times(struct tms __user * tbuf)
 950 {
 951         /*
 952          *      In the SMP world we might just be unlucky and have one of
 953          *      the times increment as we use it. Since the value is an
 954          *      atomically safe type this is just fine. Conceptually its
 955          *      as if the syscall took an instant longer to occur.
 956          */
 957         if (tbuf) {
 958                 struct tms tmp;
 959                 struct task_struct *tsk = current;
 960                 struct task_struct *t;
 961                 unsigned long utime, stime, cutime, cstime;
 962
 963                 read_lock(&tasklist_lock);
 964                 utime = tsk->signal->utime;
 965                 stime = tsk->signal->stime;
 966                 t = tsk;
 967                 do {
 968                         utime += t->utime;
 969                         stime += t->stime;
 970                         t = next_thread(t);
 971                 } while (t != tsk);
 972
 973                 /*
 974                  * While we have tasklist_lock read-locked, no dying thread
 975                  * can be updating current->signal->[us]time.  Instead,
 976                  * we got their counts included in the live thread loop.
 977                  * However, another thread can come in right now and
 978                  * do a wait call that updates current->signal->c[us]time.
 979                  * To make sure we always see that pair updated atomically,
 980                  * we take the siglock around fetching them.
 981                  */
 982                 spin_lock_irq(&tsk->sighand->siglock);
 983                 cutime = tsk->signal->cutime;
 984                 cstime = tsk->signal->cstime;
 985                 spin_unlock_irq(&tsk->sighand->siglock);
 986                 read_unlock(&tasklist_lock);
 987
 988                 tmp.tms_utime = jiffies_to_clock_t(utime);
 989                 tmp.tms_stime = jiffies_to_clock_t(stime);
 990                 tmp.tms_cutime = jiffies_to_clock_t(cutime);
 991                 tmp.tms_cstime = jiffies_to_clock_t(cstime);
 992                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
 993                         return -EFAULT;
 994         }
 995         return (long) jiffies_64_to_clock_t(get_jiffies_64());
 996 }
 997
 998 /*
 999  * This needs some heavy checking ...
1000  * I just haven't the stomach for it. I also don't fully
1001  * understand sessions/pgrp etc. Let somebody who does explain it.
1002  *
1003  * OK, I think I have the protection semantics right.... this is really
1004  * only important on a multi-user system anyway, to make sure one user
1005  * can't send a signal to a process owned by another.  -TYT, 12/12/91
1006  *
1007  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1008  * LBT 04.03.94
1009  */
1010
1011 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1012 {
1013         struct task_struct *p;
1014         int err = -EINVAL;
1015         pid_t rpgid;
1016
1017         if (!pid)
1018                 pid = vx_map_pid(current->pid);
1019         if (!pgid)
1020                 pgid = pid;
1021         if (pgid < 0)
1022                 return -EINVAL;
1023
1024         rpgid = vx_rmap_pid(pgid);
1025
1026         /* From this point forward we keep holding onto the tasklist lock
1027          * so that our parent does not change from under us. -DaveM
1028          */
1029         write_lock_irq(&tasklist_lock);
1030
1031         err = -ESRCH;
1032         p = find_task_by_pid(pid);
1033         if (!p)
1034                 goto out;
1035
1036         err = -EINVAL;
1037         if (!thread_group_leader(p))
1038                 goto out;
1039
1040         if (p->parent == current || p->real_parent == current) {
1041                 err = -EPERM;
1042                 if (p->signal->session != current->signal->session)
1043                         goto out;
1044                 err = -EACCES;
1045                 if (p->did_exec)
1046                         goto out;
1047         } else {
1048                 err = -ESRCH;
1049                 if (p != current)
1050                         goto out;
1051         }
1052
1053         err = -EPERM;
1054         if (p->signal->leader)
1055                 goto out;
1056
1057         if (pgid != pid) {
1058                 struct task_struct *p;
1059
1060                 do_each_task_pid(rpgid, PIDTYPE_PGID, p) {
1061                         if (p->signal->session == current->signal->session)
1062                                 goto ok_pgid;
1063                 } while_each_task_pid(rpgid, PIDTYPE_PGID, p);
1064                 goto out;
1065         }
1066
1067 ok_pgid:
1068         err = security_task_setpgid(p, rpgid);
1069         if (err)
1070                 goto out;
1071
1072         if (process_group(p) != rpgid) {
1073                 detach_pid(p, PIDTYPE_PGID);
1074                 p->signal->pgrp = rpgid;
1075                 attach_pid(p, PIDTYPE_PGID, rpgid);
1076         }
1077
1078         err = 0;
1079 out:
1080         /* All paths lead to here, thus we are safe. -DaveM */
1081         write_unlock_irq(&tasklist_lock);
1082         return err;
1083 }
1084
1085 asmlinkage long sys_getpgid(pid_t pid)
1086 {
1087         if (!pid) {
1088                 return vx_rmap_pid(process_group(current));
1089         } else {
1090                 int retval;
1091                 struct task_struct *p;
1092
1093                 read_lock(&tasklist_lock);
1094                 p = find_task_by_pid(pid);
1095
1096                 retval = -ESRCH;
1097                 if (p) {
1098                         retval = security_task_getpgid(p);
1099                         if (!retval)
1100                                 retval = vx_rmap_pid(process_group(p));
1101                 }
1102                 read_unlock(&tasklist_lock);
1103                 return retval;
1104         }
1105 }
1106
1107 #ifdef __ARCH_WANT_SYS_GETPGRP
1108
1109 asmlinkage long sys_getpgrp(void)
1110 {
1111         /* SMP - assuming writes are word atomic this is fine */
1112         return process_group(current);
1113 }
1114
1115 #endif
1116
1117 asmlinkage long sys_getsid(pid_t pid)
1118 {
1119         if (!pid) {
1120                 return current->signal->session;
1121         } else {
1122                 int retval;
1123                 struct task_struct *p;
1124
1125                 read_lock(&tasklist_lock);
1126                 p = find_task_by_pid(pid);
1127
1128                 retval = -ESRCH;
1129                 if(p) {
1130                         retval = security_task_getsid(p);
1131                         if (!retval)
1132                                 retval = p->signal->session;
1133                 }
1134                 read_unlock(&tasklist_lock);
1135                 return retval;
1136         }
1137 }
1138
1139 asmlinkage long sys_setsid(void)
1140 {
1141         struct pid *pid;
1142         int err = -EPERM;
1143
1144         if (!thread_group_leader(current))
1145                 return -EINVAL;
1146
1147         write_lock_irq(&tasklist_lock);
1148
1149         pid = find_pid(PIDTYPE_PGID, current->pid);
1150         if (pid)
1151                 goto out;
1152
1153         current->signal->leader = 1;
1154         __set_special_pids(current->pid, current->pid);
1155         current->signal->tty = NULL;
1156         current->signal->tty_old_pgrp = 0;
1157         err = process_group(current);
1158 out:
1159         write_unlock_irq(&tasklist_lock);
1160         return err;
1161 }
1162
1163 /*
1164  * Supplementary group IDs
1165  */
1166
1167 /* init to 2 - one for init_task, one to ensure it is never freed */
1168 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1169
1170 struct group_info *groups_alloc(int gidsetsize)
1171 {
1172         struct group_info *group_info;
1173         int nblocks;
1174         int i;
1175
1176         nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1177         /* Make sure we always allocate at least one indirect block pointer */
1178         nblocks = nblocks ? : 1;
1179         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1180         if (!group_info)
1181                 return NULL;
1182         group_info->ngroups = gidsetsize;
1183         group_info->nblocks = nblocks;
1184         atomic_set(&group_info->usage, 1);
1185
1186         if (gidsetsize <= NGROUPS_SMALL) {
1187                 group_info->blocks[0] = group_info->small_block;
1188         } else {
1189                 for (i = 0; i < nblocks; i++) {
1190                         gid_t *b;
1191                         b = (void *)__get_free_page(GFP_USER);
1192                         if (!b)
1193                                 goto out_undo_partial_alloc;
1194                         group_info->blocks[i] = b;
1195                 }
1196         }
1197         return group_info;
1198
1199 out_undo_partial_alloc:
1200         while (--i >= 0) {
1201                 free_page((unsigned long)group_info->blocks[i]);
1202         }
1203         kfree(group_info);
1204         return NULL;
1205 }
1206
1207 EXPORT_SYMBOL(groups_alloc);
1208
1209 void groups_free(struct group_info *group_info)
1210 {
1211         if (group_info->blocks[0] != group_info->small_block) {
1212                 int i;
1213                 for (i = 0; i < group_info->nblocks; i++)
1214                         free_page((unsigned long)group_info->blocks[i]);
1215         }
1216         kfree(group_info);
1217 }
1218
1219 EXPORT_SYMBOL(groups_free);
1220
1221 /* export the group_info to a user-space array */
1222 static int groups_to_user(gid_t __user *grouplist,
1223     struct group_info *group_info)
1224 {
1225         int i;
1226         int count = group_info->ngroups;
1227
1228         for (i = 0; i < group_info->nblocks; i++) {
1229                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1230                 int off = i * NGROUPS_PER_BLOCK;
1231                 int len = cp_count * sizeof(*grouplist);
1232
1233                 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1234                         return -EFAULT;
1235
1236                 count -= cp_count;
1237         }
1238         return 0;
1239 }
1240
1241 /* fill a group_info from a user-space array - it must be allocated already */
1242 static int groups_from_user(struct group_info *group_info,
1243     gid_t __user *grouplist)
1244  {
1245         int i;
1246         int count = group_info->ngroups;
1247
1248         for (i = 0; i < group_info->nblocks; i++) {
1249                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1250                 int off = i * NGROUPS_PER_BLOCK;
1251                 int len = cp_count * sizeof(*grouplist);
1252
1253                 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1254                         return -EFAULT;
1255
1256                 count -= cp_count;
1257         }
1258         return 0;
1259 }
1260
1261 /* a simple shell-metzner sort */
1262 static void groups_sort(struct group_info *group_info)
1263 {
1264         int base, max, stride;
1265         int gidsetsize = group_info->ngroups;
1266
1267         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1268                 ; /* nothing */
1269         stride /= 3;
1270
1271         while (stride) {
1272                 max = gidsetsize - stride;
1273                 for (base = 0; base < max; base++) {
1274                         int left = base;
1275                         int right = left + stride;
1276                         gid_t tmp = GROUP_AT(group_info, right);
1277
1278                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1279                                 GROUP_AT(group_info, right) =
1280                                     GROUP_AT(group_info, left);
1281                                 right = left;
1282                                 left -= stride;
1283                         }
1284                         GROUP_AT(group_info, right) = tmp;
1285                 }
1286                 stride /= 3;
1287         }
1288 }
1289
1290 /* a simple bsearch */
1291 static int groups_search(struct group_info *group_info, gid_t grp)
1292 {
1293         int left, right;
1294
1295         if (!group_info)
1296                 return 0;
1297
1298         left = 0;
1299         right = group_info->ngroups;
1300         while (left < right) {
1301                 int mid = (left+right)/2;
1302                 int cmp = grp - GROUP_AT(group_info, mid);
1303                 if (cmp > 0)
1304                         left = mid + 1;
1305                 else if (cmp < 0)
1306                         right = mid;
1307                 else
1308                         return 1;
1309         }
1310         return 0;
1311 }
1312
1313 /* validate and set current->group_info */
1314 int set_current_groups(struct group_info *group_info)
1315 {
1316         int retval;
1317         struct group_info *old_info;
1318
1319         retval = security_task_setgroups(group_info);
1320         if (retval)
1321                 return retval;
1322
1323         groups_sort(group_info);
1324         get_group_info(group_info);
1325
1326         task_lock(current);
1327         old_info = current->group_info;
1328         current->group_info = group_info;
1329         task_unlock(current);
1330
1331         put_group_info(old_info);
1332
1333         return 0;
1334 }
1335
1336 EXPORT_SYMBOL(set_current_groups);
1337
1338 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1339 {
1340         int i = 0;
1341
1342         /*
1343          *      SMP: Nobody else can change our grouplist. Thus we are
1344          *      safe.
1345          */
1346
1347         if (gidsetsize < 0)
1348                 return -EINVAL;
1349
1350         /* no need to grab task_lock here; it cannot change */
1351         get_group_info(current->group_info);
1352         i = current->group_info->ngroups;
1353         if (gidsetsize) {
1354                 if (i > gidsetsize) {
1355                         i = -EINVAL;
1356                         goto out;
1357                 }
1358                 if (groups_to_user(grouplist, current->group_info)) {
1359                         i = -EFAULT;
1360                         goto out;
1361                 }
1362         }
1363 out:
1364         put_group_info(current->group_info);
1365         return i;
1366 }
1367
1368 /*
1369  *      SMP: Our groups are copy-on-write. We can set them safely
1370  *      without another task interfering.
1371  */
1372
1373 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1374 {
1375         struct group_info *group_info;
1376         int retval;
1377
1378         if (!capable(CAP_SETGID))
1379                 return -EPERM;
1380         if ((unsigned)gidsetsize > NGROUPS_MAX)
1381                 return -EINVAL;
1382
1383         group_info = groups_alloc(gidsetsize);
1384         if (!group_info)
1385                 return -ENOMEM;
1386         retval = groups_from_user(group_info, grouplist);
1387         if (retval) {
1388                 put_group_info(group_info);
1389                 return retval;
1390         }
1391
1392         retval = set_current_groups(group_info);
1393         put_group_info(group_info);
1394
1395         return retval;
1396 }
1397
1398 /*
1399  * Check whether we're fsgid/egid or in the supplemental group..
1400  */
1401 int in_group_p(gid_t grp)
1402 {
1403         int retval = 1;
1404         if (grp != current->fsgid) {
1405                 get_group_info(current->group_info);
1406                 retval = groups_search(current->group_info, grp);
1407                 put_group_info(current->group_info);
1408         }
1409         return retval;
1410 }
1411
1412 EXPORT_SYMBOL(in_group_p);
1413
1414 int in_egroup_p(gid_t grp)
1415 {
1416         int retval = 1;
1417         if (grp != current->egid) {
1418                 get_group_info(current->group_info);
1419                 retval = groups_search(current->group_info, grp);
1420                 put_group_info(current->group_info);
1421         }
1422         return retval;
1423 }
1424
1425 EXPORT_SYMBOL(in_egroup_p);
1426
1427 DECLARE_RWSEM(uts_sem);
1428
1429 EXPORT_SYMBOL(uts_sem);
1430
1431 asmlinkage long sys_newuname(struct new_utsname __user * name)
1432 {
1433         int errno = 0;
1434
1435         down_read(&uts_sem);
1436         if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1437                 errno = -EFAULT;
1438         up_read(&uts_sem);
1439         return errno;
1440 }
1441
1442 asmlinkage long sys_sethostname(char __user *name, int len)
1443 {
1444         int errno;
1445         char tmp[__NEW_UTS_LEN];
1446
1447         if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1448                 return -EPERM;
1449         if (len < 0 || len > __NEW_UTS_LEN)
1450                 return -EINVAL;
1451         down_write(&uts_sem);
1452         errno = -EFAULT;
1453         if (!copy_from_user(tmp, name, len)) {
1454                 char *ptr = vx_new_uts(nodename);
1455
1456                 memcpy(ptr, tmp, len);
1457                 ptr[len] = 0;
1458                 errno = 0;
1459         }
1460         up_write(&uts_sem);
1461         return errno;
1462 }
1463
1464 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1465
1466 asmlinkage long sys_gethostname(char __user *name, int len)
1467 {
1468         int i, errno;
1469         char *ptr;
1470
1471         if (len < 0)
1472                 return -EINVAL;
1473         down_read(&uts_sem);
1474         ptr = vx_new_uts(nodename);
1475         i = 1 + strlen(ptr);
1476         if (i > len)
1477                 i = len;
1478         errno = 0;
1479         if (copy_to_user(name, ptr, i))
1480                 errno = -EFAULT;
1481         up_read(&uts_sem);
1482         return errno;
1483 }
1484
1485 #endif
1486
1487 /*
1488  * Only setdomainname; getdomainname can be implemented by calling
1489  * uname()
1490  */
1491 asmlinkage long sys_setdomainname(char __user *name, int len)
1492 {
1493         int errno;
1494         char tmp[__NEW_UTS_LEN];
1495
1496         if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1497                 return -EPERM;
1498         if (len < 0 || len > __NEW_UTS_LEN)
1499                 return -EINVAL;
1500
1501         down_write(&uts_sem);
1502         errno = -EFAULT;
1503         if (!copy_from_user(tmp, name, len)) {
1504                 char *ptr = vx_new_uts(domainname);
1505
1506                 memcpy(ptr, tmp, len);
1507                 ptr[len] = 0;
1508                 errno = 0;
1509         }
1510         up_write(&uts_sem);
1511         return errno;
1512 }
1513
1514 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1515 {
1516         if (resource >= RLIM_NLIMITS)
1517                 return -EINVAL;
1518         else
1519                 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
1520                         ? -EFAULT : 0;
1521 }
1522
1523 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1524
1525 /*
1526  *      Back compatibility for getrlimit. Needed for some apps.
1527  */
1528
1529 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1530 {
1531         struct rlimit x;
1532         if (resource >= RLIM_NLIMITS)
1533                 return -EINVAL;
1534
1535         memcpy(&x, current->rlim + resource, sizeof(*rlim));
1536         if(x.rlim_cur > 0x7FFFFFFF)
1537                 x.rlim_cur = 0x7FFFFFFF;
1538         if(x.rlim_max > 0x7FFFFFFF)
1539                 x.rlim_max = 0x7FFFFFFF;
1540         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1541 }
1542
1543 #endif
1544
1545 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1546 {
1547         struct rlimit new_rlim, *old_rlim;
1548         int retval;
1549
1550         if (resource >= RLIM_NLIMITS)
1551                 return -EINVAL;
1552         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1553                 return -EFAULT;
1554        if (new_rlim.rlim_cur > new_rlim.rlim_max)
1555                return -EINVAL;
1556         old_rlim = current->rlim + resource;
1557         if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
1558              (new_rlim.rlim_max > old_rlim->rlim_max)) &&
1559             !capable(CAP_SYS_RESOURCE) && !vx_ccaps(VXC_SET_RLIMIT))
1560                 return -EPERM;
1561         if (resource == RLIMIT_NOFILE) {
1562                 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
1563                         return -EPERM;
1564         }
1565
1566         retval = security_task_setrlimit(resource, &new_rlim);
1567         if (retval)
1568                 return retval;
1569
1570         *old_rlim = new_rlim;
1571         return 0;
1572 }
1573
1574 /*
1575  * It would make sense to put struct rusage in the task_struct,
1576  * except that would make the task_struct be *really big*.  After
1577  * task_struct gets moved into malloc'ed memory, it would
1578  * make sense to do this.  It will make moving the rest of the information
1579  * a lot simpler!  (Which we're not doing right now because we're not
1580  * measuring them yet).
1581  *
1582  * This expects to be called with tasklist_lock read-locked or better,
1583  * and the siglock not locked.  It may momentarily take the siglock.
1584  *
1585  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1586  * races with threads incrementing their own counters.  But since word
1587  * reads are atomic, we either get new values or old values and we don't
1588  * care which for the sums.  We always take the siglock to protect reading
1589  * the c* fields from p->signal from races with exit.c updating those
1590  * fields when reaping, so a sample either gets all the additions of a
1591  * given child after it's reaped, or none so this sample is before reaping.
1592  */
1593
1594 void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1595 {
1596         struct task_struct *t;
1597         unsigned long flags;
1598         unsigned long utime, stime;
1599
1600         memset((char *) r, 0, sizeof *r);
1601
1602         if (unlikely(!p->signal))
1603                 return;
1604
1605         switch (who) {
1606                 case RUSAGE_CHILDREN:
1607                         spin_lock_irqsave(&p->sighand->siglock, flags);
1608                         utime = p->signal->cutime;
1609                         stime = p->signal->cstime;
1610                         r->ru_nvcsw = p->signal->cnvcsw;
1611                         r->ru_nivcsw = p->signal->cnivcsw;
1612                         r->ru_minflt = p->signal->cmin_flt;
1613                         r->ru_majflt = p->signal->cmaj_flt;
1614                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1615                         jiffies_to_timeval(utime, &r->ru_utime);
1616                         jiffies_to_timeval(stime, &r->ru_stime);
1617                         break;
1618                 case RUSAGE_SELF:
1619                         spin_lock_irqsave(&p->sighand->siglock, flags);
1620                         utime = stime = 0;
1621                         goto sum_group;
1622                 case RUSAGE_BOTH:
1623                         spin_lock_irqsave(&p->sighand->siglock, flags);
1624                         utime = p->signal->cutime;
1625                         stime = p->signal->cstime;
1626                         r->ru_nvcsw = p->signal->cnvcsw;
1627                         r->ru_nivcsw = p->signal->cnivcsw;
1628                         r->ru_minflt = p->signal->cmin_flt;
1629                         r->ru_majflt = p->signal->cmaj_flt;
1630                 sum_group:
1631                         utime += p->signal->utime;
1632                         stime += p->signal->stime;
1633                         r->ru_nvcsw += p->signal->nvcsw;
1634                         r->ru_nivcsw += p->signal->nivcsw;
1635                         r->ru_minflt += p->signal->min_flt;
1636                         r->ru_majflt += p->signal->maj_flt;
1637                         t = p;
1638                         do {
1639                                 utime += t->utime;
1640                                 stime += t->stime;
1641                                 r->ru_nvcsw += t->nvcsw;
1642                                 r->ru_nivcsw += t->nivcsw;
1643                                 r->ru_minflt += t->min_flt;
1644                                 r->ru_majflt += t->maj_flt;
1645                                 t = next_thread(t);
1646                         } while (t != p);
1647                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1648                         jiffies_to_timeval(utime, &r->ru_utime);
1649                         jiffies_to_timeval(stime, &r->ru_stime);
1650                         break;
1651                 default:
1652                         BUG();
1653         }
1654 }
1655
1656 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1657 {
1658         struct rusage r;
1659         read_lock(&tasklist_lock);
1660         k_getrusage(p, who, &r);
1661         read_unlock(&tasklist_lock);
1662         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1663 }
1664
1665 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1666 {
1667         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1668                 return -EINVAL;
1669         return getrusage(current, who, ru);
1670 }
1671
1672 asmlinkage long sys_umask(int mask)
1673 {
1674         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1675         return mask;
1676 }
1677
1678 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1679                           unsigned long arg4, unsigned long arg5)
1680 {
1681         int error;
1682         int sig;
1683
1684         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1685         if (error)
1686                 return error;
1687
1688         switch (option) {
1689                 case PR_SET_PDEATHSIG:
1690                         sig = arg2;
1691                         if (sig < 0 || sig > _NSIG) {
1692                                 error = -EINVAL;
1693                                 break;
1694                         }
1695                         current->pdeath_signal = sig;
1696                         break;
1697                 case PR_GET_PDEATHSIG:
1698                         error = put_user(current->pdeath_signal, (int __user *)arg2);
1699                         break;
1700                 case PR_GET_DUMPABLE:
1701                         if (current->mm->dumpable)
1702                                 error = 1;
1703                         break;
1704                 case PR_SET_DUMPABLE:
1705                         if (arg2 != 0 && arg2 != 1) {
1706                                 error = -EINVAL;
1707                                 break;
1708                         }
1709                         current->mm->dumpable = arg2;
1710                         break;
1711
1712                 case PR_SET_UNALIGN:
1713                         error = SET_UNALIGN_CTL(current, arg2);
1714                         break;
1715                 case PR_GET_UNALIGN:
1716                         error = GET_UNALIGN_CTL(current, arg2);
1717                         break;
1718                 case PR_SET_FPEMU:
1719                         error = SET_FPEMU_CTL(current, arg2);
1720                         break;
1721                 case PR_GET_FPEMU:
1722                         error = GET_FPEMU_CTL(current, arg2);
1723                         break;
1724                 case PR_SET_FPEXC:
1725                         error = SET_FPEXC_CTL(current, arg2);
1726                         break;
1727                 case PR_GET_FPEXC:
1728                         error = GET_FPEXC_CTL(current, arg2);
1729                         break;
1730                 case PR_GET_TIMING:
1731                         error = PR_TIMING_STATISTICAL;
1732                         break;
1733                 case PR_SET_TIMING:
1734                         if (arg2 == PR_TIMING_STATISTICAL)
1735                                 error = 0;
1736                         else
1737                                 error = -EINVAL;
1738                         break;
1739
1740                 case PR_GET_KEEPCAPS:
1741                         if (current->keep_capabilities)
1742                                 error = 1;
1743                         break;
1744                 case PR_SET_KEEPCAPS:
1745                         if (arg2 != 0 && arg2 != 1) {
1746                                 error = -EINVAL;
1747                                 break;
1748                         }
1749                         current->keep_capabilities = arg2;
1750                         break;
1751                 case PR_SET_NAME: {
1752                         struct task_struct *me = current;
1753                         unsigned char ncomm[sizeof(me->comm)];
1754
1755                         ncomm[sizeof(me->comm)-1] = 0;
1756                         if (strncpy_from_user(ncomm, (char __user *)arg2,
1757                                                 sizeof(me->comm)-1) < 0)
1758                                 return -EFAULT;
1759                         set_task_comm(me, ncomm);
1760                         return 0;
1761                 }
1762                 default:
1763                         error = -EINVAL;
1764                         break;
1765         }
1766         return error;
1767 }