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
  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 cond_syscall(sys_mbind)
 276 cond_syscall(sys_get_mempolicy)
 277 cond_syscall(sys_set_mempolicy)
 278 cond_syscall(compat_get_mempolicy)
 279
 280 /* arch-specific weak syscall entries */
 281 cond_syscall(sys_pciconfig_read)
 282 cond_syscall(sys_pciconfig_write)
 283 cond_syscall(sys_pciconfig_iobase)
 284
 285 static int set_one_prio(struct task_struct *p, int niceval, int error)
 286 {
 287         int no_nice;
 288
 289         if (p->uid != current->euid &&
 290                 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
 291                 error = -EPERM;
 292                 goto out;
 293         }
 294         if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
 295                 error = -EACCES;
 296                 goto out;
 297         }
 298         no_nice = security_task_setnice(p, niceval);
 299         if (no_nice) {
 300                 error = no_nice;
 301                 goto out;
 302         }
 303         if (error == -ESRCH)
 304                 error = 0;
 305         set_user_nice(p, niceval);
 306 out:
 307         return error;
 308 }
 309
 310 asmlinkage long sys_setpriority(int which, int who, int niceval)
 311 {
 312         struct task_struct *g, *p;
 313         struct user_struct *user;
 314         struct pid *pid;
 315         struct list_head *l;
 316         int error = -EINVAL;
 317
 318         if (which > 2 || which < 0)
 319                 goto out;
 320
 321         /* normalize: avoid signed division (rounding problems) */
 322         error = -ESRCH;
 323         if (niceval < -20)
 324                 niceval = -20;
 325         if (niceval > 19)
 326                 niceval = 19;
 327
 328         read_lock(&tasklist_lock);
 329         switch (which) {
 330                 case PRIO_PROCESS:
 331                         if (!who)
 332                                 who = current->pid;
 333                         p = find_task_by_pid(who);
 334                         if (p)
 335                                 error = set_one_prio(p, niceval, error);
 336                         break;
 337                 case PRIO_PGRP:
 338                         if (!who)
 339                                 who = process_group(current);
 340                         for_each_task_pid(who, PIDTYPE_PGID, p, l, pid)
 341                                 error = set_one_prio(p, niceval, error);
 342                         break;
 343                 case PRIO_USER:
 344                         if (!who)
 345                                 user = current->user;
 346                         else
 347                                 user = find_user(vx_current_xid(), who);
 348
 349                         if (!user)
 350                                 goto out_unlock;
 351
 352                         do_each_thread(g, p)
 353                                 if (p->uid == who)
 354                                         error = set_one_prio(p, niceval, error);
 355                         while_each_thread(g, p);
 356                         if (who)
 357                                 free_uid(user);         /* For find_user() */
 358                         break;
 359         }
 360 out_unlock:
 361         read_unlock(&tasklist_lock);
 362 out:
 363         return error;
 364 }
 365
 366 /*
 367  * Ugh. To avoid negative return values, "getpriority()" will
 368  * not return the normal nice-value, but a negated value that
 369  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 370  * to stay compatible.
 371  */
 372 asmlinkage long sys_getpriority(int which, int who)
 373 {
 374         struct task_struct *g, *p;
 375         struct list_head *l;
 376         struct pid *pid;
 377         struct user_struct *user;
 378         long niceval, retval = -ESRCH;
 379
 380         if (which > 2 || which < 0)
 381                 return -EINVAL;
 382
 383         read_lock(&tasklist_lock);
 384         switch (which) {
 385                 case PRIO_PROCESS:
 386                         if (!who)
 387                                 who = current->pid;
 388                         p = find_task_by_pid(who);
 389                         if (p) {
 390                                 niceval = 20 - task_nice(p);
 391                                 if (niceval > retval)
 392                                         retval = niceval;
 393                         }
 394                         break;
 395                 case PRIO_PGRP:
 396                         if (!who)
 397                                 who = process_group(current);
 398                         for_each_task_pid(who, PIDTYPE_PGID, p, l, pid) {
 399                                 niceval = 20 - task_nice(p);
 400                                 if (niceval > retval)
 401                                         retval = niceval;
 402                         }
 403                         break;
 404                 case PRIO_USER:
 405                         if (!who)
 406                                 user = current->user;
 407                         else
 408                                 user = find_user(vx_current_xid(), who);
 409
 410                         if (!user)
 411                                 goto out_unlock;
 412
 413                         do_each_thread(g, p)
 414                                 if (p->uid == who) {
 415                                         niceval = 20 - task_nice(p);
 416                                         if (niceval > retval)
 417                                                 retval = niceval;
 418                                 }
 419                         while_each_thread(g, p);
 420                         if (who)
 421                                 free_uid(user);         /* for find_user() */
 422                         break;
 423         }
 424 out_unlock:
 425         read_unlock(&tasklist_lock);
 426
 427         return retval;
 428 }
 429
 430 /*
 431  *      vshelper path is set via /proc/sys
 432  *      invoked by vserver sys_reboot(), with
 433  *      the following arguments
 434  *
 435  *      argv [0] = vshelper_path;
 436  *      argv [1] = action: "restart", "halt", "poweroff", ...
 437  *      argv [2] = context identifier
 438  *      argv [3] = additional argument (restart2)
 439  *
 440  *      envp [*] = type-specific parameters
 441  */
 442 char vshelper_path[255] = "/sbin/vshelper";
 443
 444 long vs_reboot(unsigned int cmd, void * arg)
 445 {
 446         char id_buf[8], cmd_buf[32];
 447         char uid_buf[32], pid_buf[32];
 448         char buffer[256];
 449
 450         char *argv[] = {vshelper_path, NULL, id_buf, NULL, 0};
 451         char *envp[] = {"HOME=/", "TERM=linux",
 452                         "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
 453                         uid_buf, pid_buf, cmd_buf, 0};
 454
 455         snprintf(id_buf, sizeof(id_buf)-1, "%d", vx_current_xid());
 456
 457         snprintf(cmd_buf, sizeof(cmd_buf)-1, "VS_CMD=%08x", cmd);
 458         snprintf(uid_buf, sizeof(uid_buf)-1, "VS_UID=%d", current->uid);
 459         snprintf(pid_buf, sizeof(pid_buf)-1, "VS_PID=%d", current->pid);
 460
 461         switch (cmd) {
 462         case LINUX_REBOOT_CMD_RESTART:
 463                 argv[1] = "restart";
 464                 break;
 465
 466         case LINUX_REBOOT_CMD_HALT:
 467                 argv[1] = "halt";
 468                 break;
 469
 470         case LINUX_REBOOT_CMD_POWER_OFF:
 471                 argv[1] = "poweroff";
 472                 break;
 473
 474         case LINUX_REBOOT_CMD_SW_SUSPEND:
 475                 argv[1] = "swsusp";
 476                 break;
 477
 478         case LINUX_REBOOT_CMD_RESTART2:
 479                 if (strncpy_from_user(&buffer[0], (char *)arg, sizeof(buffer) - 1) < 0)
 480                         return -EFAULT;
 481                 argv[3] = buffer;
 482         default:
 483                 argv[1] = "restart2";
 484                 break;
 485         }
 486
 487         /* maybe we should wait ? */
 488         if (call_usermodehelper(*argv, argv, envp, 0)) {
 489                 printk( KERN_WARNING
 490                         "vs_reboot(): failed to exec (%s %s %s %s)\n",
 491                         vshelper_path, argv[1], argv[2], argv[3]);
 492                 return -EPERM;
 493         }
 494         return 0;
 495 }
 496
 497 /*
 498  * Reboot system call: for obvious reasons only root may call it,
 499  * and even root needs to set up some magic numbers in the registers
 500  * so that some mistake won't make this reboot the whole machine.
 501  * You can also set the meaning of the ctrl-alt-del-key here.
 502  *
 503  * reboot doesn't sync: do that yourself before calling this.
 504  */
 505 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
 506 {
 507         char buffer[256];
 508
 509         /* We only trust the superuser with rebooting the system. */
 510         if (!capable(CAP_SYS_BOOT))
 511                 return -EPERM;
 512
 513         /* For safety, we require "magic" arguments. */
 514         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 515             (magic2 != LINUX_REBOOT_MAGIC2 &&
 516                         magic2 != LINUX_REBOOT_MAGIC2A &&
 517                         magic2 != LINUX_REBOOT_MAGIC2B &&
 518                         magic2 != LINUX_REBOOT_MAGIC2C))
 519                 return -EINVAL;
 520
 521         if (!vx_check(0, VX_ADMIN|VX_WATCH))
 522                 return vs_reboot(cmd, arg);
 523
 524         lock_kernel();
 525         switch (cmd) {
 526         case LINUX_REBOOT_CMD_RESTART:
 527                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 528                 system_state = SYSTEM_RESTART;
 529                 device_shutdown();
 530                 printk(KERN_EMERG "Restarting system.\n");
 531                 machine_restart(NULL);
 532                 break;
 533
 534         case LINUX_REBOOT_CMD_CAD_ON:
 535                 C_A_D = 1;
 536                 break;
 537
 538         case LINUX_REBOOT_CMD_CAD_OFF:
 539                 C_A_D = 0;
 540                 break;
 541
 542         case LINUX_REBOOT_CMD_HALT:
 543                 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
 544                 system_state = SYSTEM_HALT;
 545                 device_shutdown();
 546                 printk(KERN_EMERG "System halted.\n");
 547                 machine_halt();
 548                 unlock_kernel();
 549                 do_exit(0);
 550                 break;
 551
 552         case LINUX_REBOOT_CMD_POWER_OFF:
 553                 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
 554                 system_state = SYSTEM_POWER_OFF;
 555                 device_shutdown();
 556                 printk(KERN_EMERG "Power down.\n");
 557                 machine_power_off();
 558                 unlock_kernel();
 559                 do_exit(0);
 560                 break;
 561
 562         case LINUX_REBOOT_CMD_RESTART2:
 563                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 564                         unlock_kernel();
 565                         return -EFAULT;
 566                 }
 567                 buffer[sizeof(buffer) - 1] = '\0';
 568
 569                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
 570                 system_state = SYSTEM_RESTART;
 571                 device_shutdown();
 572                 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
 573                 machine_restart(buffer);
 574                 break;
 575
 576 #ifdef CONFIG_SOFTWARE_SUSPEND
 577         case LINUX_REBOOT_CMD_SW_SUSPEND:
 578                 {
 579                         int ret = software_suspend();
 580                         unlock_kernel();
 581                         return ret;
 582                 }
 583 #endif
 584
 585         default:
 586                 unlock_kernel();
 587                 return -EINVAL;
 588         }
 589         unlock_kernel();
 590         return 0;
 591 }
 592
 593 static void deferred_cad(void *dummy)
 594 {
 595         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 596         machine_restart(NULL);
 597 }
 598
 599 /*
 600  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 601  * As it's called within an interrupt, it may NOT sync: the only choice
 602  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 603  */
 604 void ctrl_alt_del(void)
 605 {
 606         static DECLARE_WORK(cad_work, deferred_cad, NULL);
 607
 608         if (C_A_D)
 609                 schedule_work(&cad_work);
 610         else
 611                 kill_proc(cad_pid, SIGINT, 1);
 612 }
 613
 614
 615 /*
 616  * Unprivileged users may change the real gid to the effective gid
 617  * or vice versa.  (BSD-style)
 618  *
 619  * If you set the real gid at all, or set the effective gid to a value not
 620  * equal to the real gid, then the saved gid is set to the new effective gid.
 621  *
 622  * This makes it possible for a setgid program to completely drop its
 623  * privileges, which is often a useful assertion to make when you are doing
 624  * a security audit over a program.
 625  *
 626  * The general idea is that a program which uses just setregid() will be
 627  * 100% compatible with BSD.  A program which uses just setgid() will be
 628  * 100% compatible with POSIX with saved IDs.
 629  *
 630  * SMP: There are not races, the GIDs are checked only by filesystem
 631  *      operations (as far as semantic preservation is concerned).
 632  */
 633 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 634 {
 635         int old_rgid = current->gid;
 636         int old_egid = current->egid;
 637         int new_rgid = old_rgid;
 638         int new_egid = old_egid;
 639         int retval;
 640
 641         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 642         if (retval)
 643                 return retval;
 644
 645         if (rgid != (gid_t) -1) {
 646                 if ((old_rgid == rgid) ||
 647                     (current->egid==rgid) ||
 648                     capable(CAP_SETGID))
 649                         new_rgid = rgid;
 650                 else
 651                         return -EPERM;
 652         }
 653         if (egid != (gid_t) -1) {
 654                 if ((old_rgid == egid) ||
 655                     (current->egid == egid) ||
 656                     (current->sgid == egid) ||
 657                     capable(CAP_SETGID))
 658                         new_egid = egid;
 659                 else {
 660                         return -EPERM;
 661                 }
 662         }
 663         if (new_egid != old_egid)
 664         {
 665                 current->mm->dumpable = 0;
 666                 wmb();
 667         }
 668         if (rgid != (gid_t) -1 ||
 669             (egid != (gid_t) -1 && egid != old_rgid))
 670                 current->sgid = new_egid;
 671         current->fsgid = new_egid;
 672         current->egid = new_egid;
 673         current->gid = new_rgid;
 674         return 0;
 675 }
 676
 677 /*
 678  * setgid() is implemented like SysV w/ SAVED_IDS
 679  *
 680  * SMP: Same implicit races as above.
 681  */
 682 asmlinkage long sys_setgid(gid_t gid)
 683 {
 684         int old_egid = current->egid;
 685         int retval;
 686
 687         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 688         if (retval)
 689                 return retval;
 690
 691         if (capable(CAP_SETGID))
 692         {
 693                 if(old_egid != gid)
 694                 {
 695                         current->mm->dumpable=0;
 696                         wmb();
 697                 }
 698                 current->gid = current->egid = current->sgid = current->fsgid = gid;
 699         }
 700         else if ((gid == current->gid) || (gid == current->sgid))
 701         {
 702                 if(old_egid != gid)
 703                 {
 704                         current->mm->dumpable=0;
 705                         wmb();
 706                 }
 707                 current->egid = current->fsgid = gid;
 708         }
 709         else
 710                 return -EPERM;
 711         return 0;
 712 }
 713
 714 static int set_user(uid_t new_ruid, int dumpclear)
 715 {
 716         struct user_struct *new_user;
 717
 718         new_user = alloc_uid(vx_current_xid(), new_ruid);
 719         if (!new_user)
 720                 return -EAGAIN;
 721
 722         if (atomic_read(&new_user->processes) >=
 723                                 current->rlim[RLIMIT_NPROC].rlim_cur &&
 724                         new_user != &root_user) {
 725                 free_uid(new_user);
 726                 return -EAGAIN;
 727         }
 728
 729         switch_uid(new_user);
 730
 731         if(dumpclear)
 732         {
 733                 current->mm->dumpable = 0;
 734                 wmb();
 735         }
 736         current->uid = new_ruid;
 737         return 0;
 738 }
 739
 740 /*
 741  * Unprivileged users may change the real uid to the effective uid
 742  * or vice versa.  (BSD-style)
 743  *
 744  * If you set the real uid at all, or set the effective uid to a value not
 745  * equal to the real uid, then the saved uid is set to the new effective uid.
 746  *
 747  * This makes it possible for a setuid program to completely drop its
 748  * privileges, which is often a useful assertion to make when you are doing
 749  * a security audit over a program.
 750  *
 751  * The general idea is that a program which uses just setreuid() will be
 752  * 100% compatible with BSD.  A program which uses just setuid() will be
 753  * 100% compatible with POSIX with saved IDs.
 754  */
 755 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 756 {
 757         int old_ruid, old_euid, old_suid, new_ruid, new_euid;
 758         int retval;
 759
 760         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 761         if (retval)
 762                 return retval;
 763
 764         new_ruid = old_ruid = current->uid;
 765         new_euid = old_euid = current->euid;
 766         old_suid = current->suid;
 767
 768         if (ruid != (uid_t) -1) {
 769                 new_ruid = ruid;
 770                 if ((old_ruid != ruid) &&
 771                     (current->euid != ruid) &&
 772                     !capable(CAP_SETUID))
 773                         return -EPERM;
 774         }
 775
 776         if (euid != (uid_t) -1) {
 777                 new_euid = euid;
 778                 if ((old_ruid != euid) &&
 779                     (current->euid != euid) &&
 780                     (current->suid != euid) &&
 781                     !capable(CAP_SETUID))
 782                         return -EPERM;
 783         }
 784
 785         if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
 786                 return -EAGAIN;
 787
 788         if (new_euid != old_euid)
 789         {
 790                 current->mm->dumpable=0;
 791                 wmb();
 792         }
 793         current->fsuid = current->euid = new_euid;
 794         if (ruid != (uid_t) -1 ||
 795             (euid != (uid_t) -1 && euid != old_ruid))
 796                 current->suid = current->euid;
 797         current->fsuid = current->euid;
 798
 799         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
 800 }
 801
 802
 803
 804 /*
 805  * setuid() is implemented like SysV with SAVED_IDS
 806  *
 807  * Note that SAVED_ID's is deficient in that a setuid root program
 808  * like sendmail, for example, cannot set its uid to be a normal
 809  * user and then switch back, because if you're root, setuid() sets
 810  * the saved uid too.  If you don't like this, blame the bright people
 811  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 812  * will allow a root program to temporarily drop privileges and be able to
 813  * regain them by swapping the real and effective uid.
 814  */
 815 asmlinkage long sys_setuid(uid_t uid)
 816 {
 817         int old_euid = current->euid;
 818         int old_ruid, old_suid, new_ruid, new_suid;
 819         int retval;
 820
 821         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 822         if (retval)
 823                 return retval;
 824
 825         old_ruid = new_ruid = current->uid;
 826         old_suid = current->suid;
 827         new_suid = old_suid;
 828
 829         if (capable(CAP_SETUID)) {
 830                 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
 831                         return -EAGAIN;
 832                 new_suid = uid;
 833         } else if ((uid != current->uid) && (uid != new_suid))
 834                 return -EPERM;
 835
 836         if (old_euid != uid)
 837         {
 838                 current->mm->dumpable = 0;
 839                 wmb();
 840         }
 841         current->fsuid = current->euid = uid;
 842         current->suid = new_suid;
 843
 844         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
 845 }
 846
 847
 848 /*
 849  * This function implements a generic ability to update ruid, euid,
 850  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 851  */
 852 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 853 {
 854         int old_ruid = current->uid;
 855         int old_euid = current->euid;
 856         int old_suid = current->suid;
 857         int retval;
 858
 859         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 860         if (retval)
 861                 return retval;
 862
 863         if (!capable(CAP_SETUID)) {
 864                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
 865                     (ruid != current->euid) && (ruid != current->suid))
 866                         return -EPERM;
 867                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
 868                     (euid != current->euid) && (euid != current->suid))
 869                         return -EPERM;
 870                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
 871                     (suid != current->euid) && (suid != current->suid))
 872                         return -EPERM;
 873         }
 874         if (ruid != (uid_t) -1) {
 875                 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
 876                         return -EAGAIN;
 877         }
 878         if (euid != (uid_t) -1) {
 879                 if (euid != current->euid)
 880                 {
 881                         current->mm->dumpable = 0;
 882                         wmb();
 883                 }
 884                 current->euid = euid;
 885         }
 886         current->fsuid = current->euid;
 887         if (suid != (uid_t) -1)
 888                 current->suid = suid;
 889
 890         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
 891 }
 892
 893 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
 894 {
 895         int retval;
 896
 897         if (!(retval = put_user(current->uid, ruid)) &&
 898             !(retval = put_user(current->euid, euid)))
 899                 retval = put_user(current->suid, suid);
 900
 901         return retval;
 902 }
 903
 904 /*
 905  * Same as above, but for rgid, egid, sgid.
 906  */
 907 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 908 {
 909         int retval;
 910
 911         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 912         if (retval)
 913                 return retval;
 914
 915         if (!capable(CAP_SETGID)) {
 916                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
 917                     (rgid != current->egid) && (rgid != current->sgid))
 918                         return -EPERM;
 919                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
 920                     (egid != current->egid) && (egid != current->sgid))
 921                         return -EPERM;
 922                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
 923                     (sgid != current->egid) && (sgid != current->sgid))
 924                         return -EPERM;
 925         }
 926         if (egid != (gid_t) -1) {
 927                 if (egid != current->egid)
 928                 {
 929                         current->mm->dumpable = 0;
 930                         wmb();
 931                 }
 932                 current->egid = egid;
 933         }
 934         current->fsgid = current->egid;
 935         if (rgid != (gid_t) -1)
 936                 current->gid = rgid;
 937         if (sgid != (gid_t) -1)
 938                 current->sgid = sgid;
 939         return 0;
 940 }
 941
 942 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
 943 {
 944         int retval;
 945
 946         if (!(retval = put_user(current->gid, rgid)) &&
 947             !(retval = put_user(current->egid, egid)))
 948                 retval = put_user(current->sgid, sgid);
 949
 950         return retval;
 951 }
 952
 953
 954 /*
 955  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 956  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 957  * whatever uid it wants to). It normally shadows "euid", except when
 958  * explicitly set by setfsuid() or for access..
 959  */
 960 asmlinkage long sys_setfsuid(uid_t uid)
 961 {
 962         int old_fsuid;
 963
 964         old_fsuid = current->fsuid;
 965         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
 966                 return old_fsuid;
 967
 968         if (uid == current->uid || uid == current->euid ||
 969             uid == current->suid || uid == current->fsuid ||
 970             capable(CAP_SETUID))
 971         {
 972                 if (uid != old_fsuid)
 973                 {
 974                         current->mm->dumpable = 0;
 975                         wmb();
 976                 }
 977                 current->fsuid = uid;
 978         }
 979
 980         security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
 981
 982         return old_fsuid;
 983 }
 984
 985 /*
 986  * Samma på svenska..
 987  */
 988 asmlinkage long sys_setfsgid(gid_t gid)
 989 {
 990         int old_fsgid;
 991
 992         old_fsgid = current->fsgid;
 993         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 994                 return old_fsgid;
 995
 996         if (gid == current->gid || gid == current->egid ||
 997             gid == current->sgid || gid == current->fsgid ||
 998             capable(CAP_SETGID))
 999         {
1000                 if (gid != old_fsgid)
1001                 {
1002                         current->mm->dumpable = 0;
1003                         wmb();
1004                 }
1005                 current->fsgid = gid;
1006         }
1007         return old_fsgid;
1008 }
1009
1010 asmlinkage long sys_times(struct tms __user * tbuf)
1011 {
1012         /*
1013          *      In the SMP world we might just be unlucky and have one of
1014          *      the times increment as we use it. Since the value is an
1015          *      atomically safe type this is just fine. Conceptually its
1016          *      as if the syscall took an instant longer to occur.
1017          */
1018         if (tbuf) {
1019                 struct tms tmp;
1020                 tmp.tms_utime = jiffies_to_clock_t(current->utime);
1021                 tmp.tms_stime = jiffies_to_clock_t(current->stime);
1022                 tmp.tms_cutime = jiffies_to_clock_t(current->cutime);
1023                 tmp.tms_cstime = jiffies_to_clock_t(current->cstime);
1024                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1025                         return -EFAULT;
1026         }
1027         return (long) jiffies_64_to_clock_t(get_jiffies_64());
1028 }
1029
1030 /*
1031  * This needs some heavy checking ...
1032  * I just haven't the stomach for it. I also don't fully
1033  * understand sessions/pgrp etc. Let somebody who does explain it.
1034  *
1035  * OK, I think I have the protection semantics right.... this is really
1036  * only important on a multi-user system anyway, to make sure one user
1037  * can't send a signal to a process owned by another.  -TYT, 12/12/91
1038  *
1039  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1040  * LBT 04.03.94
1041  */
1042
1043 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1044 {
1045         struct task_struct *p;
1046         int err = -EINVAL;
1047
1048         if (!pid)
1049                 pid = current->pid;
1050         if (!pgid)
1051                 pgid = pid;
1052         if (pgid < 0)
1053                 return -EINVAL;
1054
1055         /* From this point forward we keep holding onto the tasklist lock
1056          * so that our parent does not change from under us. -DaveM
1057          */
1058         write_lock_irq(&tasklist_lock);
1059
1060         err = -ESRCH;
1061         p = find_task_by_pid(pid);
1062         if (!p)
1063                 goto out;
1064
1065         err = -EINVAL;
1066         if (!thread_group_leader(p))
1067                 goto out;
1068
1069         if (p->parent == current || p->real_parent == current) {
1070                 err = -EPERM;
1071                 if (p->signal->session != current->signal->session)
1072                         goto out;
1073                 err = -EACCES;
1074                 if (p->did_exec)
1075                         goto out;
1076         } else {
1077                 err = -ESRCH;
1078                 if (p != current)
1079                         goto out;
1080         }
1081
1082         err = -EPERM;
1083         if (p->signal->leader)
1084                 goto out;
1085
1086         if (pgid != pid) {
1087                 struct task_struct *p;
1088                 struct pid *pid;
1089                 struct list_head *l;
1090
1091                 for_each_task_pid(pgid, PIDTYPE_PGID, p, l, pid)
1092                         if (p->signal->session == current->signal->session)
1093                                 goto ok_pgid;
1094                 goto out;
1095         }
1096
1097 ok_pgid:
1098         err = security_task_setpgid(p, pgid);
1099         if (err)
1100                 goto out;
1101
1102         if (process_group(p) != pgid) {
1103                 detach_pid(p, PIDTYPE_PGID);
1104                 p->signal->pgrp = pgid;
1105                 attach_pid(p, PIDTYPE_PGID, pgid);
1106         }
1107
1108         err = 0;
1109 out:
1110         /* All paths lead to here, thus we are safe. -DaveM */
1111         write_unlock_irq(&tasklist_lock);
1112         return err;
1113 }
1114
1115 asmlinkage long sys_getpgid(pid_t pid)
1116 {
1117         if (!pid) {
1118                 return process_group(current);
1119         } else {
1120                 int retval;
1121                 struct task_struct *p;
1122
1123                 read_lock(&tasklist_lock);
1124                 p = find_task_by_pid(pid);
1125
1126                 retval = -ESRCH;
1127                 if (p) {
1128                         retval = security_task_getpgid(p);
1129                         if (!retval)
1130                                 retval = process_group(p);
1131                 }
1132                 read_unlock(&tasklist_lock);
1133                 return retval;
1134         }
1135 }
1136
1137 #ifdef __ARCH_WANT_SYS_GETPGRP
1138
1139 asmlinkage long sys_getpgrp(void)
1140 {
1141         /* SMP - assuming writes are word atomic this is fine */
1142         return process_group(current);
1143 }
1144
1145 #endif
1146
1147 asmlinkage long sys_getsid(pid_t pid)
1148 {
1149         if (!pid) {
1150                 return current->signal->session;
1151         } else {
1152                 int retval;
1153                 struct task_struct *p;
1154
1155                 read_lock(&tasklist_lock);
1156                 p = find_task_by_pid(pid);
1157
1158                 retval = -ESRCH;
1159                 if(p) {
1160                         retval = security_task_getsid(p);
1161                         if (!retval)
1162                                 retval = p->signal->session;
1163                 }
1164                 read_unlock(&tasklist_lock);
1165                 return retval;
1166         }
1167 }
1168
1169 asmlinkage long sys_setsid(void)
1170 {
1171         struct pid *pid;
1172         int err = -EPERM;
1173
1174         if (!thread_group_leader(current))
1175                 return -EINVAL;
1176
1177         write_lock_irq(&tasklist_lock);
1178
1179         pid = find_pid(PIDTYPE_PGID, current->pid);
1180         if (pid)
1181                 goto out;
1182
1183         current->signal->leader = 1;
1184         __set_special_pids(current->pid, current->pid);
1185         current->signal->tty = NULL;
1186         current->signal->tty_old_pgrp = 0;
1187         err = process_group(current);
1188 out:
1189         write_unlock_irq(&tasklist_lock);
1190         return err;
1191 }
1192
1193 /*
1194  * Supplementary group IDs
1195  */
1196
1197 /* init to 2 - one for init_task, one to ensure it is never freed */
1198 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1199
1200 struct group_info *groups_alloc(int gidsetsize)
1201 {
1202         struct group_info *group_info;
1203         int nblocks;
1204         int i;
1205
1206         nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1207         /* Make sure we always allocate at least one indirect block pointer */
1208         nblocks = nblocks ? : 1;
1209         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1210         if (!group_info)
1211                 return NULL;
1212         group_info->ngroups = gidsetsize;
1213         group_info->nblocks = nblocks;
1214         atomic_set(&group_info->usage, 1);
1215
1216         if (gidsetsize <= NGROUPS_SMALL) {
1217                 group_info->blocks[0] = group_info->small_block;
1218         } else {
1219                 for (i = 0; i < nblocks; i++) {
1220                         gid_t *b;
1221                         b = (void *)__get_free_page(GFP_USER);
1222                         if (!b)
1223                                 goto out_undo_partial_alloc;
1224                         group_info->blocks[i] = b;
1225                 }
1226         }
1227         return group_info;
1228
1229 out_undo_partial_alloc:
1230         while (--i >= 0) {
1231                 free_page((unsigned long)group_info->blocks[i]);
1232         }
1233         kfree(group_info);
1234         return NULL;
1235 }
1236
1237 EXPORT_SYMBOL(groups_alloc);
1238
1239 void groups_free(struct group_info *group_info)
1240 {
1241         if (group_info->blocks[0] != group_info->small_block) {
1242                 int i;
1243                 for (i = 0; i < group_info->nblocks; i++)
1244                         free_page((unsigned long)group_info->blocks[i]);
1245         }
1246         kfree(group_info);
1247 }
1248
1249 EXPORT_SYMBOL(groups_free);
1250
1251 /* export the group_info to a user-space array */
1252 static int groups_to_user(gid_t __user *grouplist,
1253     struct group_info *group_info)
1254 {
1255         int i;
1256         int count = group_info->ngroups;
1257
1258         for (i = 0; i < group_info->nblocks; i++) {
1259                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1260                 int off = i * NGROUPS_PER_BLOCK;
1261                 int len = cp_count * sizeof(*grouplist);
1262
1263                 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1264                         return -EFAULT;
1265
1266                 count -= cp_count;
1267         }
1268         return 0;
1269 }
1270
1271 /* fill a group_info from a user-space array - it must be allocated already */
1272 static int groups_from_user(struct group_info *group_info,
1273     gid_t __user *grouplist)
1274  {
1275         int i;
1276         int count = group_info->ngroups;
1277
1278         for (i = 0; i < group_info->nblocks; i++) {
1279                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1280                 int off = i * NGROUPS_PER_BLOCK;
1281                 int len = cp_count * sizeof(*grouplist);
1282
1283                 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1284                         return -EFAULT;
1285
1286                 count -= cp_count;
1287         }
1288         return 0;
1289 }
1290
1291 /* a simple shell-metzner sort */
1292 static void groups_sort(struct group_info *group_info)
1293 {
1294         int base, max, stride;
1295         int gidsetsize = group_info->ngroups;
1296
1297         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1298                 ; /* nothing */
1299         stride /= 3;
1300
1301         while (stride) {
1302                 max = gidsetsize - stride;
1303                 for (base = 0; base < max; base++) {
1304                         int left = base;
1305                         int right = left + stride;
1306                         gid_t tmp = GROUP_AT(group_info, right);
1307
1308                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1309                                 GROUP_AT(group_info, right) =
1310                                     GROUP_AT(group_info, left);
1311                                 right = left;
1312                                 left -= stride;
1313                         }
1314                         GROUP_AT(group_info, right) = tmp;
1315                 }
1316                 stride /= 3;
1317         }
1318 }
1319
1320 /* a simple bsearch */
1321 static int groups_search(struct group_info *group_info, gid_t grp)
1322 {
1323         int left, right;
1324
1325         if (!group_info)
1326                 return 0;
1327
1328         left = 0;
1329         right = group_info->ngroups;
1330         while (left < right) {
1331                 int mid = (left+right)/2;
1332                 int cmp = grp - GROUP_AT(group_info, mid);
1333                 if (cmp > 0)
1334                         left = mid + 1;
1335                 else if (cmp < 0)
1336                         right = mid;
1337                 else
1338                         return 1;
1339         }
1340         return 0;
1341 }
1342
1343 /* validate and set current->group_info */
1344 int set_current_groups(struct group_info *group_info)
1345 {
1346         int retval;
1347         struct group_info *old_info;
1348
1349         retval = security_task_setgroups(group_info);
1350         if (retval)
1351                 return retval;
1352
1353         groups_sort(group_info);
1354         get_group_info(group_info);
1355
1356         task_lock(current);
1357         old_info = current->group_info;
1358         current->group_info = group_info;
1359         task_unlock(current);
1360
1361         put_group_info(old_info);
1362
1363         return 0;
1364 }
1365
1366 EXPORT_SYMBOL(set_current_groups);
1367
1368 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1369 {
1370         int i = 0;
1371
1372         /*
1373          *      SMP: Nobody else can change our grouplist. Thus we are
1374          *      safe.
1375          */
1376
1377         if (gidsetsize < 0)
1378                 return -EINVAL;
1379
1380         /* no need to grab task_lock here; it cannot change */
1381         get_group_info(current->group_info);
1382         i = current->group_info->ngroups;
1383         if (gidsetsize) {
1384                 if (i > gidsetsize) {
1385                         i = -EINVAL;
1386                         goto out;
1387                 }
1388                 if (groups_to_user(grouplist, current->group_info)) {
1389                         i = -EFAULT;
1390                         goto out;
1391                 }
1392         }
1393 out:
1394         put_group_info(current->group_info);
1395         return i;
1396 }
1397
1398 /*
1399  *      SMP: Our groups are copy-on-write. We can set them safely
1400  *      without another task interfering.
1401  */
1402
1403 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1404 {
1405         struct group_info *group_info;
1406         int retval;
1407
1408         if (!capable(CAP_SETGID))
1409                 return -EPERM;
1410         if ((unsigned)gidsetsize > NGROUPS_MAX)
1411                 return -EINVAL;
1412
1413         group_info = groups_alloc(gidsetsize);
1414         if (!group_info)
1415                 return -ENOMEM;
1416         retval = groups_from_user(group_info, grouplist);
1417         if (retval) {
1418                 put_group_info(group_info);
1419                 return retval;
1420         }
1421
1422         retval = set_current_groups(group_info);
1423         put_group_info(group_info);
1424
1425         return retval;
1426 }
1427
1428 /*
1429  * Check whether we're fsgid/egid or in the supplemental group..
1430  */
1431 int in_group_p(gid_t grp)
1432 {
1433         int retval = 1;
1434         if (grp != current->fsgid) {
1435                 get_group_info(current->group_info);
1436                 retval = groups_search(current->group_info, grp);
1437                 put_group_info(current->group_info);
1438         }
1439         return retval;
1440 }
1441
1442 EXPORT_SYMBOL(in_group_p);
1443
1444 int in_egroup_p(gid_t grp)
1445 {
1446         int retval = 1;
1447         if (grp != current->egid) {
1448                 get_group_info(current->group_info);
1449                 retval = groups_search(current->group_info, grp);
1450                 put_group_info(current->group_info);
1451         }
1452         return retval;
1453 }
1454
1455 EXPORT_SYMBOL(in_egroup_p);
1456
1457 DECLARE_RWSEM(uts_sem);
1458
1459 EXPORT_SYMBOL(uts_sem);
1460
1461 asmlinkage long sys_newuname(struct new_utsname __user * name)
1462 {
1463         int errno = 0;
1464
1465         down_read(&uts_sem);
1466         if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1467                 errno = -EFAULT;
1468         up_read(&uts_sem);
1469         return errno;
1470 }
1471
1472 asmlinkage long sys_sethostname(char __user *name, int len)
1473 {
1474         int errno;
1475         char tmp[__NEW_UTS_LEN];
1476
1477         if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1478                 return -EPERM;
1479         if (len < 0 || len > __NEW_UTS_LEN)
1480                 return -EINVAL;
1481         down_write(&uts_sem);
1482         errno = -EFAULT;
1483         if (!copy_from_user(tmp, name, len)) {
1484                 char *ptr = vx_new_uts(nodename);
1485
1486                 memcpy(ptr, tmp, len);
1487                 ptr[len] = 0;
1488                 errno = 0;
1489         }
1490         up_write(&uts_sem);
1491         return errno;
1492 }
1493
1494 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1495
1496 asmlinkage long sys_gethostname(char __user *name, int len)
1497 {
1498         int i, errno;
1499         char *ptr;
1500
1501         if (len < 0)
1502                 return -EINVAL;
1503         down_read(&uts_sem);
1504         ptr = vx_new_uts(nodename);
1505         i = 1 + strlen(ptr);
1506         if (i > len)
1507                 i = len;
1508         errno = 0;
1509         if (copy_to_user(name, ptr, i))
1510                 errno = -EFAULT;
1511         up_read(&uts_sem);
1512         return errno;
1513 }
1514
1515 #endif
1516
1517 /*
1518  * Only setdomainname; getdomainname can be implemented by calling
1519  * uname()
1520  */
1521 asmlinkage long sys_setdomainname(char __user *name, int len)
1522 {
1523         int errno;
1524         char tmp[__NEW_UTS_LEN];
1525
1526         if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1527                 return -EPERM;
1528         if (len < 0 || len > __NEW_UTS_LEN)
1529                 return -EINVAL;
1530
1531         down_write(&uts_sem);
1532         errno = -EFAULT;
1533         if (!copy_from_user(tmp, name, len)) {
1534                 char *ptr = vx_new_uts(domainname);
1535
1536                 memcpy(ptr, tmp, len);
1537                 ptr[len] = 0;
1538                 errno = 0;
1539         }
1540         up_write(&uts_sem);
1541         return errno;
1542 }
1543
1544 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1545 {
1546         if (resource >= RLIM_NLIMITS)
1547                 return -EINVAL;
1548         else
1549                 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
1550                         ? -EFAULT : 0;
1551 }
1552
1553 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1554
1555 /*
1556  *      Back compatibility for getrlimit. Needed for some apps.
1557  */
1558
1559 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1560 {
1561         struct rlimit x;
1562         if (resource >= RLIM_NLIMITS)
1563                 return -EINVAL;
1564
1565         memcpy(&x, current->rlim + resource, sizeof(*rlim));
1566         if(x.rlim_cur > 0x7FFFFFFF)
1567                 x.rlim_cur = 0x7FFFFFFF;
1568         if(x.rlim_max > 0x7FFFFFFF)
1569                 x.rlim_max = 0x7FFFFFFF;
1570         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1571 }
1572
1573 #endif
1574
1575 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1576 {
1577         struct rlimit new_rlim, *old_rlim;
1578         int retval;
1579
1580         if (resource >= RLIM_NLIMITS)
1581                 return -EINVAL;
1582         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1583                 return -EFAULT;
1584        if (new_rlim.rlim_cur > new_rlim.rlim_max)
1585                return -EINVAL;
1586         old_rlim = current->rlim + resource;
1587         if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
1588              (new_rlim.rlim_max > old_rlim->rlim_max)) &&
1589             !capable(CAP_SYS_RESOURCE) && vx_ccaps(VXC_SET_RLIMIT))
1590                 return -EPERM;
1591         if (resource == RLIMIT_NOFILE) {
1592                 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
1593                         return -EPERM;
1594         }
1595
1596         retval = security_task_setrlimit(resource, &new_rlim);
1597         if (retval)
1598                 return retval;
1599
1600         *old_rlim = new_rlim;
1601         return 0;
1602 }
1603
1604 /*
1605  * It would make sense to put struct rusage in the task_struct,
1606  * except that would make the task_struct be *really big*.  After
1607  * task_struct gets moved into malloc'ed memory, it would
1608  * make sense to do this.  It will make moving the rest of the information
1609  * a lot simpler!  (Which we're not doing right now because we're not
1610  * measuring them yet).
1611  *
1612  * This is SMP safe.  Either we are called from sys_getrusage on ourselves
1613  * below (we know we aren't going to exit/disappear and only we change our
1614  * rusage counters), or we are called from wait4() on a process which is
1615  * either stopped or zombied.  In the zombied case the task won't get
1616  * reaped till shortly after the call to getrusage(), in both cases the
1617  * task being examined is in a frozen state so the counters won't change.
1618  */
1619 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1620 {
1621         struct rusage r;
1622
1623         memset((char *) &r, 0, sizeof(r));
1624         switch (who) {
1625                 case RUSAGE_SELF:
1626                         jiffies_to_timeval(p->utime, &r.ru_utime);
1627                         jiffies_to_timeval(p->stime, &r.ru_stime);
1628                         r.ru_nvcsw = p->nvcsw;
1629                         r.ru_nivcsw = p->nivcsw;
1630                         r.ru_minflt = p->min_flt;
1631                         r.ru_majflt = p->maj_flt;
1632                         break;
1633                 case RUSAGE_CHILDREN:
1634                         jiffies_to_timeval(p->cutime, &r.ru_utime);
1635                         jiffies_to_timeval(p->cstime, &r.ru_stime);
1636                         r.ru_nvcsw = p->cnvcsw;
1637                         r.ru_nivcsw = p->cnivcsw;
1638                         r.ru_minflt = p->cmin_flt;
1639                         r.ru_majflt = p->cmaj_flt;
1640                         break;
1641                 default:
1642                         jiffies_to_timeval(p->utime + p->cutime, &r.ru_utime);
1643                         jiffies_to_timeval(p->stime + p->cstime, &r.ru_stime);
1644                         r.ru_nvcsw = p->nvcsw + p->cnvcsw;
1645                         r.ru_nivcsw = p->nivcsw + p->cnivcsw;
1646                         r.ru_minflt = p->min_flt + p->cmin_flt;
1647                         r.ru_majflt = p->maj_flt + p->cmaj_flt;
1648                         break;
1649         }
1650         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1651 }
1652
1653 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1654 {
1655         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1656                 return -EINVAL;
1657         return getrusage(current, who, ru);
1658 }
1659
1660 asmlinkage long sys_umask(int mask)
1661 {
1662         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1663         return mask;
1664 }
1665
1666 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1667                           unsigned long arg4, unsigned long arg5)
1668 {
1669         int error;
1670         int sig;
1671
1672         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1673         if (error)
1674                 return error;
1675
1676         switch (option) {
1677                 case PR_SET_PDEATHSIG:
1678                         sig = arg2;
1679                         if (sig < 0 || sig > _NSIG) {
1680                                 error = -EINVAL;
1681                                 break;
1682                         }
1683                         current->pdeath_signal = sig;
1684                         break;
1685                 case PR_GET_PDEATHSIG:
1686                         error = put_user(current->pdeath_signal, (int __user *)arg2);
1687                         break;
1688                 case PR_GET_DUMPABLE:
1689                         if (current->mm->dumpable)
1690                                 error = 1;
1691                         break;
1692                 case PR_SET_DUMPABLE:
1693                         if (arg2 != 0 && arg2 != 1) {
1694                                 error = -EINVAL;
1695                                 break;
1696                         }
1697                         current->mm->dumpable = arg2;
1698                         break;
1699
1700                 case PR_SET_UNALIGN:
1701                         error = SET_UNALIGN_CTL(current, arg2);
1702                         break;
1703                 case PR_GET_UNALIGN:
1704                         error = GET_UNALIGN_CTL(current, arg2);
1705                         break;
1706                 case PR_SET_FPEMU:
1707                         error = SET_FPEMU_CTL(current, arg2);
1708                         break;
1709                 case PR_GET_FPEMU:
1710                         error = GET_FPEMU_CTL(current, arg2);
1711                         break;
1712                 case PR_SET_FPEXC:
1713                         error = SET_FPEXC_CTL(current, arg2);
1714                         break;
1715                 case PR_GET_FPEXC:
1716                         error = GET_FPEXC_CTL(current, arg2);
1717                         break;
1718                 case PR_GET_TIMING:
1719                         error = PR_TIMING_STATISTICAL;
1720                         break;
1721                 case PR_SET_TIMING:
1722                         if (arg2 == PR_TIMING_STATISTICAL)
1723                                 error = 0;
1724                         else
1725                                 error = -EINVAL;
1726                         break;
1727
1728                 case PR_GET_KEEPCAPS:
1729                         if (current->keep_capabilities)
1730                                 error = 1;
1731                         break;
1732                 case PR_SET_KEEPCAPS:
1733                         if (arg2 != 0 && arg2 != 1) {
1734                                 error = -EINVAL;
1735                                 break;
1736                         }
1737                         current->keep_capabilities = arg2;
1738                         break;
1739                 default:
1740                         error = -EINVAL;
1741                         break;
1742         }
1743         return error;
1744 }