kernel/sys.c

   1 /*
   2  *  linux/kernel/sys.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6
   7 #include <linux/config.h>
   8 #include <linux/module.h>
   9 #include <linux/mm.h>
  10 #include <linux/utsname.h>
  11 #include <linux/mman.h>
  12 #include <linux/smp_lock.h>
  13 #include <linux/notifier.h>
  14 #include <linux/kmod.h>
  15 #include <linux/reboot.h>
  16 #include <linux/prctl.h>
  17 #include <linux/init.h>
  18 #include <linux/highuid.h>
  19 #include <linux/fs.h>
  20 #include <linux/workqueue.h>
  21 #include <linux/device.h>
  22 #include <linux/key.h>
  23 #include <linux/times.h>
  24 #include <linux/security.h>
  25 #include <linux/dcookies.h>
  26 #include <linux/suspend.h>
  27 #include <linux/ckrm_events.h>
  28 #include <linux/tty.h>
  29 #include <linux/vs_cvirt.h>
  30 #include <linux/compat.h>
  31 #include <linux/syscalls.h>
  32
  33 #include <asm/uaccess.h>
  34 #include <asm/io.h>
  35 #include <asm/unistd.h>
  36
  37 #ifndef SET_UNALIGN_CTL
  38 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  39 #endif
  40 #ifndef GET_UNALIGN_CTL
  41 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  42 #endif
  43 #ifndef SET_FPEMU_CTL
  44 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
  45 #endif
  46 #ifndef GET_FPEMU_CTL
  47 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
  48 #endif
  49 #ifndef SET_FPEXC_CTL
  50 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
  51 #endif
  52 #ifndef GET_FPEXC_CTL
  53 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
  54 #endif
  55
  56 /*
  57  * this is where the system-wide overflow UID and GID are defined, for
  58  * architectures that now have 32-bit UID/GID but didn't in the past
  59  */
  60
  61 int overflowuid = DEFAULT_OVERFLOWUID;
  62 int overflowgid = DEFAULT_OVERFLOWGID;
  63
  64 #ifdef CONFIG_UID16
  65 EXPORT_SYMBOL(overflowuid);
  66 EXPORT_SYMBOL(overflowgid);
  67 #endif
  68
  69 /*
  70  * the same as above, but for filesystems which can only store a 16-bit
  71  * UID and GID. as such, this is needed on all architectures
  72  */
  73
  74 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
  75 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
  76
  77 EXPORT_SYMBOL(fs_overflowuid);
  78 EXPORT_SYMBOL(fs_overflowgid);
  79
  80 /*
  81  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
  82  */
  83
  84 int C_A_D = 1;
  85 int cad_pid = 1;
  86
  87 /*
  88  *      Notifier list for kernel code which wants to be called
  89  *      at shutdown. This is used to stop any idling DMA operations
  90  *      and the like.
  91  */
  92
  93 static struct notifier_block *reboot_notifier_list;
  94 rwlock_t notifier_lock = RW_LOCK_UNLOCKED;
  95
  96 /**
  97  *      notifier_chain_register - Add notifier to a notifier chain
  98  *      @list: Pointer to root list pointer
  99  *      @n: New entry in notifier chain
 100  *
 101  *      Adds a notifier to a notifier chain.
 102  *
 103  *      Currently always returns zero.
 104  */
 105
 106 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
 107 {
 108         write_lock(&notifier_lock);
 109         while(*list)
 110         {
 111                 if(n->priority > (*list)->priority)
 112                         break;
 113                 list= &((*list)->next);
 114         }
 115         n->next = *list;
 116         *list=n;
 117         write_unlock(&notifier_lock);
 118         return 0;
 119 }
 120
 121 EXPORT_SYMBOL(notifier_chain_register);
 122
 123 /**
 124  *      notifier_chain_unregister - Remove notifier from a notifier chain
 125  *      @nl: Pointer to root list pointer
 126  *      @n: New entry in notifier chain
 127  *
 128  *      Removes a notifier from a notifier chain.
 129  *
 130  *      Returns zero on success, or %-ENOENT on failure.
 131  */
 132
 133 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
 134 {
 135         write_lock(&notifier_lock);
 136         while((*nl)!=NULL)
 137         {
 138                 if((*nl)==n)
 139                 {
 140                         *nl=n->next;
 141                         write_unlock(&notifier_lock);
 142                         return 0;
 143                 }
 144                 nl=&((*nl)->next);
 145         }
 146         write_unlock(&notifier_lock);
 147         return -ENOENT;
 148 }
 149
 150 EXPORT_SYMBOL(notifier_chain_unregister);
 151
 152 /**
 153  *      notifier_call_chain - Call functions in a notifier chain
 154  *      @n: Pointer to root pointer of notifier chain
 155  *      @val: Value passed unmodified to notifier function
 156  *      @v: Pointer passed unmodified to notifier function
 157  *
 158  *      Calls each function in a notifier chain in turn.
 159  *
 160  *      If the return value of the notifier can be and'd
 161  *      with %NOTIFY_STOP_MASK, then notifier_call_chain
 162  *      will return immediately, with the return value of
 163  *      the notifier function which halted execution.
 164  *      Otherwise, the return value is the return value
 165  *      of the last notifier function called.
 166  */
 167
 168 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
 169 {
 170         int ret=NOTIFY_DONE;
 171         struct notifier_block *nb = *n;
 172
 173         while(nb)
 174         {
 175                 ret=nb->notifier_call(nb,val,v);
 176                 if(ret&NOTIFY_STOP_MASK)
 177                 {
 178                         return ret;
 179                 }
 180                 nb=nb->next;
 181         }
 182         return ret;
 183 }
 184
 185 EXPORT_SYMBOL(notifier_call_chain);
 186
 187 /**
 188  *      register_reboot_notifier - Register function to be called at reboot time
 189  *      @nb: Info about notifier function to be called
 190  *
 191  *      Registers a function with the list of functions
 192  *      to be called at reboot time.
 193  *
 194  *      Currently always returns zero, as notifier_chain_register
 195  *      always returns zero.
 196  */
 197
 198 int register_reboot_notifier(struct notifier_block * nb)
 199 {
 200         return notifier_chain_register(&reboot_notifier_list, nb);
 201 }
 202
 203 EXPORT_SYMBOL(register_reboot_notifier);
 204
 205 /**
 206  *      unregister_reboot_notifier - Unregister previously registered reboot notifier
 207  *      @nb: Hook to be unregistered
 208  *
 209  *      Unregisters a previously registered reboot
 210  *      notifier function.
 211  *
 212  *      Returns zero on success, or %-ENOENT on failure.
 213  */
 214
 215 int unregister_reboot_notifier(struct notifier_block * nb)
 216 {
 217         return notifier_chain_unregister(&reboot_notifier_list, nb);
 218 }
 219
 220 EXPORT_SYMBOL(unregister_reboot_notifier);
 221 static int set_one_prio(struct task_struct *p, int niceval, int error)
 222 {
 223         int no_nice;
 224
 225         if (p->uid != current->euid &&
 226                 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
 227                 error = -EPERM;
 228                 goto out;
 229         }
 230         if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
 231                 if (vx_flags(VXF_IGNEG_NICE, 0))
 232                         error = 0;
 233                 else
 234                         error = -EACCES;
 235                 goto out;
 236         }
 237         no_nice = security_task_setnice(p, niceval);
 238         if (no_nice) {
 239                 error = no_nice;
 240                 goto out;
 241         }
 242         if (error == -ESRCH)
 243                 error = 0;
 244         set_user_nice(p, niceval);
 245 out:
 246         return error;
 247 }
 248
 249 asmlinkage long sys_setpriority(int which, int who, int niceval)
 250 {
 251         struct task_struct *g, *p;
 252         struct user_struct *user;
 253         int error = -EINVAL;
 254
 255         if (which > 2 || which < 0)
 256                 goto out;
 257
 258         /* normalize: avoid signed division (rounding problems) */
 259         error = -ESRCH;
 260         if (niceval < -20)
 261                 niceval = -20;
 262         if (niceval > 19)
 263                 niceval = 19;
 264
 265         read_lock(&tasklist_lock);
 266         switch (which) {
 267                 case PRIO_PROCESS:
 268                         if (!who)
 269                                 who = current->pid;
 270                         p = find_task_by_pid(who);
 271                         if (p)
 272                                 error = set_one_prio(p, niceval, error);
 273                         break;
 274                 case PRIO_PGRP:
 275                         if (!who)
 276                                 who = process_group(current);
 277                         do_each_task_pid(who, PIDTYPE_PGID, p) {
 278                                 error = set_one_prio(p, niceval, error);
 279                         } while_each_task_pid(who, PIDTYPE_PGID, p);
 280                         break;
 281                 case PRIO_USER:
 282                         user = current->user;
 283                         if (!who)
 284                                 who = current->uid;
 285                         else
 286                                 if ((who != current->uid) &&
 287                                         !(user = find_user(vx_current_xid(), who)))
 288                                         goto out_unlock;        /* No processes for this user */
 289
 290                         do_each_thread(g, p)
 291                                 if (p->uid == who)
 292                                         error = set_one_prio(p, niceval, error);
 293                         while_each_thread(g, p);
 294                         if (who != current->uid)
 295                                 free_uid(user);         /* For find_user() */
 296                         break;
 297         }
 298 out_unlock:
 299         read_unlock(&tasklist_lock);
 300 out:
 301         return error;
 302 }
 303
 304 /*
 305  * Ugh. To avoid negative return values, "getpriority()" will
 306  * not return the normal nice-value, but a negated value that
 307  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 308  * to stay compatible.
 309  */
 310 asmlinkage long sys_getpriority(int which, int who)
 311 {
 312         struct task_struct *g, *p;
 313         struct user_struct *user;
 314         long niceval, retval = -ESRCH;
 315
 316         if (which > 2 || which < 0)
 317                 return -EINVAL;
 318
 319         read_lock(&tasklist_lock);
 320         switch (which) {
 321                 case PRIO_PROCESS:
 322                         if (!who)
 323                                 who = current->pid;
 324                         p = find_task_by_pid(who);
 325                         if (p) {
 326                                 niceval = 20 - task_nice(p);
 327                                 if (niceval > retval)
 328                                         retval = niceval;
 329                         }
 330                         break;
 331                 case PRIO_PGRP:
 332                         if (!who)
 333                                 who = process_group(current);
 334                         do_each_task_pid(who, PIDTYPE_PGID, p) {
 335                                 niceval = 20 - task_nice(p);
 336                                 if (niceval > retval)
 337                                         retval = niceval;
 338                         } while_each_task_pid(who, PIDTYPE_PGID, p);
 339                         break;
 340                 case PRIO_USER:
 341                         user = current->user;
 342                         if (!who)
 343                                 who = current->uid;
 344                         else
 345                                 if ((who != current->uid) &&
 346                                         !(user = find_user(vx_current_xid(), who)))
 347                                         goto out_unlock;        /* No processes for this user */
 348
 349                         do_each_thread(g, p)
 350                                 if (p->uid == who) {
 351                                         niceval = 20 - task_nice(p);
 352                                         if (niceval > retval)
 353                                                 retval = niceval;
 354                                 }
 355                         while_each_thread(g, p);
 356                         if (who != current->uid)
 357                                 free_uid(user);         /* for find_user() */
 358                         break;
 359         }
 360 out_unlock:
 361         read_unlock(&tasklist_lock);
 362
 363         key_fsgid_changed(current);
 364         return 0;
 365 }
 366
 367 long vs_reboot(unsigned int, void *);
 368
 369 /*
 370  * Reboot system call: for obvious reasons only root may call it,
 371  * and even root needs to set up some magic numbers in the registers
 372  * so that some mistake won't make this reboot the whole machine.
 373  * You can also set the meaning of the ctrl-alt-del-key here.
 374  *
 375  * reboot doesn't sync: do that yourself before calling this.
 376  */
 377 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
 378 {
 379         char buffer[256];
 380
 381         /* We only trust the superuser with rebooting the system. */
 382         if (!capable(CAP_SYS_BOOT))
 383                 return -EPERM;
 384
 385         /* For safety, we require "magic" arguments. */
 386         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 387             (magic2 != LINUX_REBOOT_MAGIC2 &&
 388                         magic2 != LINUX_REBOOT_MAGIC2A &&
 389                         magic2 != LINUX_REBOOT_MAGIC2B &&
 390                         magic2 != LINUX_REBOOT_MAGIC2C))
 391                 return -EINVAL;
 392
 393         if (!vx_check(0, VX_ADMIN|VX_WATCH))
 394                 return vs_reboot(cmd, arg);
 395
 396         lock_kernel();
 397         switch (cmd) {
 398         case LINUX_REBOOT_CMD_RESTART:
 399                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 400                 system_state = SYSTEM_RESTART;
 401                 device_shutdown();
 402                 printk(KERN_EMERG "Restarting system.\n");
 403                 machine_restart(NULL);
 404                 break;
 405
 406         case LINUX_REBOOT_CMD_CAD_ON:
 407                 C_A_D = 1;
 408                 break;
 409
 410         case LINUX_REBOOT_CMD_CAD_OFF:
 411                 C_A_D = 0;
 412                 break;
 413
 414         case LINUX_REBOOT_CMD_HALT:
 415                 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
 416                 system_state = SYSTEM_HALT;
 417                 device_shutdown();
 418                 printk(KERN_EMERG "System halted.\n");
 419                 machine_halt();
 420                 unlock_kernel();
 421                 do_exit(0);
 422                 break;
 423
 424         case LINUX_REBOOT_CMD_POWER_OFF:
 425                 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
 426                 system_state = SYSTEM_POWER_OFF;
 427                 device_shutdown();
 428                 printk(KERN_EMERG "Power down.\n");
 429                 machine_power_off();
 430                 unlock_kernel();
 431                 do_exit(0);
 432                 break;
 433
 434         case LINUX_REBOOT_CMD_RESTART2:
 435                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 436                         unlock_kernel();
 437                         return -EFAULT;
 438                 }
 439                 buffer[sizeof(buffer) - 1] = '\0';
 440
 441                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
 442                 system_state = SYSTEM_RESTART;
 443                 device_shutdown();
 444                 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
 445                 machine_restart(buffer);
 446                 break;
 447
 448 #ifdef CONFIG_KEXEC
 449         case LINUX_REBOOT_CMD_KEXEC:
 450         {
 451                 struct kimage *image;
 452                 image = xchg(&kexec_image, 0);
 453                 if (!image) {
 454                         unlock_kernel();
 455                         return -EINVAL;
 456                 }
 457                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 458                 system_state = SYSTEM_RESTART;
 459                 device_shutdown();
 460                 system_state = SYSTEM_BOOTING;
 461                 printk(KERN_EMERG "Starting new kernel\n");
 462                 machine_shutdown();
 463                 machine_kexec(image);
 464                 break;
 465         }
 466 #endif
 467
 468 #ifdef CONFIG_SOFTWARE_SUSPEND
 469         case LINUX_REBOOT_CMD_SW_SUSPEND:
 470                 {
 471                         int ret = software_suspend();
 472                         unlock_kernel();
 473                         return ret;
 474                 }
 475 #endif
 476
 477         default:
 478                 unlock_kernel();
 479                 return -EINVAL;
 480         }
 481         unlock_kernel();
 482         return 0;
 483 }
 484
 485 static void deferred_cad(void *dummy)
 486 {
 487         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 488         machine_restart(NULL);
 489 }
 490
 491 /*
 492  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 493  * As it's called within an interrupt, it may NOT sync: the only choice
 494  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 495  */
 496 void ctrl_alt_del(void)
 497 {
 498         static DECLARE_WORK(cad_work, deferred_cad, NULL);
 499
 500         if (C_A_D)
 501                 schedule_work(&cad_work);
 502         else
 503                 kill_proc(cad_pid, SIGINT, 1);
 504 }
 505
 506
 507
 508 /*
 509  * Unprivileged users may change the real gid to the effective gid
 510  * or vice versa.  (BSD-style)
 511  *
 512  * If you set the real gid at all, or set the effective gid to a value not
 513  * equal to the real gid, then the saved gid is set to the new effective gid.
 514  *
 515  * This makes it possible for a setgid program to completely drop its
 516  * privileges, which is often a useful assertion to make when you are doing
 517  * a security audit over a program.
 518  *
 519  * The general idea is that a program which uses just setregid() will be
 520  * 100% compatible with BSD.  A program which uses just setgid() will be
 521  * 100% compatible with POSIX with saved IDs.
 522  *
 523  * SMP: There are not races, the GIDs are checked only by filesystem
 524  *      operations (as far as semantic preservation is concerned).
 525  */
 526 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 527 {
 528         int old_rgid = current->gid;
 529         int old_egid = current->egid;
 530         int new_rgid = old_rgid;
 531         int new_egid = old_egid;
 532         int retval;
 533
 534         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 535         if (retval)
 536                 return retval;
 537
 538         if (rgid != (gid_t) -1) {
 539                 if ((old_rgid == rgid) ||
 540                     (current->egid==rgid) ||
 541                     capable(CAP_SETGID))
 542                         new_rgid = rgid;
 543                 else
 544                         return -EPERM;
 545         }
 546         if (egid != (gid_t) -1) {
 547                 if ((old_rgid == egid) ||
 548                     (current->egid == egid) ||
 549                     (current->sgid == egid) ||
 550                     capable(CAP_SETGID))
 551                         new_egid = egid;
 552                 else {
 553                         return -EPERM;
 554                 }
 555         }
 556         if (new_egid != old_egid)
 557         {
 558                 current->mm->dumpable = suid_dumpable;
 559                 wmb();
 560         }
 561         if (rgid != (gid_t) -1 ||
 562             (egid != (gid_t) -1 && egid != old_rgid))
 563                 current->sgid = new_egid;
 564         current->fsgid = new_egid;
 565         current->egid = new_egid;
 566         current->gid = new_rgid;
 567
 568         key_fsgid_changed(current);
 569         ckrm_cb_gid();
 570         return 0;
 571 }
 572
 573 /*
 574  * setgid() is implemented like SysV w/ SAVED_IDS
 575  *
 576  * SMP: Same implicit races as above.
 577  */
 578 asmlinkage long sys_setgid(gid_t gid)
 579 {
 580         int old_egid = current->egid;
 581         int retval;
 582
 583         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 584         if (retval)
 585                 return retval;
 586
 587         if (capable(CAP_SETGID))
 588         {
 589                 if(old_egid != gid)
 590                 {
 591                         current->mm->dumpable = suid_dumpable;
 592                         wmb();
 593                 }
 594                 current->gid = current->egid = current->sgid = current->fsgid = gid;
 595         }
 596         else if ((gid == current->gid) || (gid == current->sgid))
 597         {
 598                 if(old_egid != gid)
 599                 {
 600                         current->mm->dumpable = suid_dumpable;
 601                         wmb();
 602                 }
 603                 current->egid = current->fsgid = gid;
 604         }
 605         else
 606                 return -EPERM;
 607
 608         key_fsgid_changed(current);
 609         ckrm_cb_gid();
 610
 611         return 0;
 612 }
 613
 614 static int set_user(uid_t new_ruid, int dumpclear)
 615 {
 616         struct user_struct *new_user;
 617
 618         new_user = alloc_uid(vx_current_xid(), new_ruid);
 619         if (!new_user)
 620                 return -EAGAIN;
 621
 622         if (atomic_read(&new_user->processes) >=
 623                                 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
 624                         new_user != &root_user) {
 625                 free_uid(new_user);
 626                 return -EAGAIN;
 627         }
 628
 629         switch_uid(new_user);
 630
 631         if(dumpclear)
 632         {
 633                 current->mm->dumpable = suid_dumpable;
 634                 wmb();
 635         }
 636         current->uid = new_ruid;
 637         return 0;
 638 }
 639
 640 /*
 641  * Unprivileged users may change the real uid to the effective uid
 642  * or vice versa.  (BSD-style)
 643  *
 644  * If you set the real uid at all, or set the effective uid to a value not
 645  * equal to the real uid, then the saved uid is set to the new effective uid.
 646  *
 647  * This makes it possible for a setuid program to completely drop its
 648  * privileges, which is often a useful assertion to make when you are doing
 649  * a security audit over a program.
 650  *
 651  * The general idea is that a program which uses just setreuid() will be
 652  * 100% compatible with BSD.  A program which uses just setuid() will be
 653  * 100% compatible with POSIX with saved IDs.
 654  */
 655 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 656 {
 657         int old_ruid, old_euid, old_suid, new_ruid, new_euid;
 658         int retval;
 659
 660         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 661         if (retval)
 662                 return retval;
 663
 664         new_ruid = old_ruid = current->uid;
 665         new_euid = old_euid = current->euid;
 666         old_suid = current->suid;
 667
 668         if (ruid != (uid_t) -1) {
 669                 new_ruid = ruid;
 670                 if ((old_ruid != ruid) &&
 671                     (current->euid != ruid) &&
 672                     !capable(CAP_SETUID))
 673                         return -EPERM;
 674         }
 675
 676         if (euid != (uid_t) -1) {
 677                 new_euid = euid;
 678                 if ((old_ruid != euid) &&
 679                     (current->euid != euid) &&
 680                     (current->suid != euid) &&
 681                     !capable(CAP_SETUID))
 682                         return -EPERM;
 683         }
 684
 685         if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
 686                 return -EAGAIN;
 687
 688         if (new_euid != old_euid)
 689         {
 690                 current->mm->dumpable = suid_dumpable;
 691                 wmb();
 692         }
 693         current->fsuid = current->euid = new_euid;
 694         if (ruid != (uid_t) -1 ||
 695             (euid != (uid_t) -1 && euid != old_ruid))
 696                 current->suid = current->euid;
 697         current->fsuid = current->euid;
 698
 699         key_fsuid_changed(current);
 700         ckrm_cb_uid();
 701
 702         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
 703 }
 704
 705
 706
 707 /*
 708  * setuid() is implemented like SysV with SAVED_IDS
 709  *
 710  * Note that SAVED_ID's is deficient in that a setuid root program
 711  * like sendmail, for example, cannot set its uid to be a normal
 712  * user and then switch back, because if you're root, setuid() sets
 713  * the saved uid too.  If you don't like this, blame the bright people
 714  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 715  * will allow a root program to temporarily drop privileges and be able to
 716  * regain them by swapping the real and effective uid.
 717  */
 718 asmlinkage long sys_setuid(uid_t uid)
 719 {
 720         int old_euid = current->euid;
 721         int old_ruid, old_suid, new_ruid, new_suid;
 722         int retval;
 723
 724         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 725         if (retval)
 726                 return retval;
 727
 728         old_ruid = new_ruid = current->uid;
 729         old_suid = current->suid;
 730         new_suid = old_suid;
 731
 732         if (capable(CAP_SETUID)) {
 733                 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
 734                         return -EAGAIN;
 735                 new_suid = uid;
 736         } else if ((uid != current->uid) && (uid != new_suid))
 737                 return -EPERM;
 738
 739         if (old_euid != uid)
 740         {
 741                 current->mm->dumpable = suid_dumpable;
 742                 wmb();
 743         }
 744         current->fsuid = current->euid = uid;
 745         current->suid = new_suid;
 746
 747         key_fsuid_changed(current);
 748         ckrm_cb_uid();
 749
 750         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
 751 }
 752
 753
 754 /*
 755  * This function implements a generic ability to update ruid, euid,
 756  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 757  */
 758 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 759 {
 760         int old_ruid = current->uid;
 761         int old_euid = current->euid;
 762         int old_suid = current->suid;
 763         int retval;
 764
 765         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 766         if (retval)
 767                 return retval;
 768
 769         if (!capable(CAP_SETUID)) {
 770                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
 771                     (ruid != current->euid) && (ruid != current->suid))
 772                         return -EPERM;
 773                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
 774                     (euid != current->euid) && (euid != current->suid))
 775                         return -EPERM;
 776                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
 777                     (suid != current->euid) && (suid != current->suid))
 778                         return -EPERM;
 779         }
 780         if (ruid != (uid_t) -1) {
 781                 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
 782                         return -EAGAIN;
 783         }
 784         if (euid != (uid_t) -1) {
 785                 if (euid != current->euid)
 786                 {
 787                         current->mm->dumpable = suid_dumpable;
 788                         wmb();
 789                 }
 790                 current->euid = euid;
 791         }
 792         current->fsuid = current->euid;
 793         if (suid != (uid_t) -1)
 794                 current->suid = suid;
 795
 796         key_fsuid_changed(current);
 797         ckrm_cb_uid();
 798
 799         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
 800 }
 801
 802 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
 803 {
 804         int retval;
 805
 806         if (!(retval = put_user(current->uid, ruid)) &&
 807             !(retval = put_user(current->euid, euid)))
 808                 retval = put_user(current->suid, suid);
 809
 810         return retval;
 811 }
 812
 813 /*
 814  * Same as above, but for rgid, egid, sgid.
 815  */
 816 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 817 {
 818         int retval;
 819
 820         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 821         if (retval)
 822                 return retval;
 823
 824         if (!capable(CAP_SETGID)) {
 825                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
 826                     (rgid != current->egid) && (rgid != current->sgid))
 827                         return -EPERM;
 828                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
 829                     (egid != current->egid) && (egid != current->sgid))
 830                         return -EPERM;
 831                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
 832                     (sgid != current->egid) && (sgid != current->sgid))
 833                         return -EPERM;
 834         }
 835         if (egid != (gid_t) -1) {
 836                 if (egid != current->egid)
 837                 {
 838                         current->mm->dumpable = suid_dumpable;
 839                         wmb();
 840                 }
 841                 current->egid = egid;
 842         }
 843         current->fsgid = current->egid;
 844         if (rgid != (gid_t) -1)
 845                 current->gid = rgid;
 846         if (sgid != (gid_t) -1)
 847                 current->sgid = sgid;
 848
 849         key_fsgid_changed(current);
 850         ckrm_cb_gid();
 851         return 0;
 852 }
 853
 854 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
 855 {
 856         int retval;
 857
 858         if (!(retval = put_user(current->gid, rgid)) &&
 859             !(retval = put_user(current->egid, egid)))
 860                 retval = put_user(current->sgid, sgid);
 861
 862         return retval;
 863 }
 864
 865
 866 /*
 867  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 868  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 869  * whatever uid it wants to). It normally shadows "euid", except when
 870  * explicitly set by setfsuid() or for access..
 871  */
 872 asmlinkage long sys_setfsuid(uid_t uid)
 873 {
 874         int old_fsuid;
 875
 876         old_fsuid = current->fsuid;
 877         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
 878                 return old_fsuid;
 879
 880         if (uid == current->uid || uid == current->euid ||
 881             uid == current->suid || uid == current->fsuid ||
 882             capable(CAP_SETUID))
 883         {
 884                 if (uid != old_fsuid)
 885                 {
 886                         current->mm->dumpable = suid_dumpable;
 887                         wmb();
 888                 }
 889                 current->fsuid = uid;
 890         }
 891
 892         key_fsuid_changed(current);
 893
 894         security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
 895
 896         return old_fsuid;
 897 }
 898
 899 /*
 900  * Samma på svenska..
 901  */
 902 asmlinkage long sys_setfsgid(gid_t gid)
 903 {
 904         int old_fsgid;
 905
 906         old_fsgid = current->fsgid;
 907         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 908                 return old_fsgid;
 909
 910         if (gid == current->gid || gid == current->egid ||
 911             gid == current->sgid || gid == current->fsgid ||
 912             capable(CAP_SETGID))
 913         {
 914                 if (gid != old_fsgid)
 915                 {
 916                         current->mm->dumpable = suid_dumpable;
 917                         wmb();
 918                 }
 919                 current->fsgid = gid;
 920                 key_fsgid_changed(current);
 921         }
 922         return old_fsgid;
 923 }
 924
 925 asmlinkage long sys_times(struct tms __user * tbuf)
 926 {
 927         /*
 928          *      In the SMP world we might just be unlucky and have one of
 929          *      the times increment as we use it. Since the value is an
 930          *      atomically safe type this is just fine. Conceptually its
 931          *      as if the syscall took an instant longer to occur.
 932          */
 933         if (tbuf) {
 934                 struct tms tmp;
 935                 struct task_struct *tsk = current;
 936                 struct task_struct *t;
 937                 unsigned long utime, stime, cutime, cstime;
 938
 939                 read_lock(&tasklist_lock);
 940                 utime = tsk->signal->utime;
 941                 stime = tsk->signal->stime;
 942                 t = tsk;
 943                 do {
 944                         utime += t->utime;
 945                         stime += t->stime;
 946                         t = next_thread(t);
 947                 } while (t != tsk);
 948
 949                 /*
 950                  * While we have tasklist_lock read-locked, no dying thread
 951                  * can be updating current->signal->[us]time.  Instead,
 952                  * we got their counts included in the live thread loop.
 953                  * However, another thread can come in right now and
 954                  * do a wait call that updates current->signal->c[us]time.
 955                  * To make sure we always see that pair updated atomically,
 956                  * we take the siglock around fetching them.
 957                  */
 958                 spin_lock_irq(&tsk->sighand->siglock);
 959                 cutime = tsk->signal->cutime;
 960                 cstime = tsk->signal->cstime;
 961                 spin_unlock_irq(&tsk->sighand->siglock);
 962                 read_unlock(&tasklist_lock);
 963
 964                 tmp.tms_utime = jiffies_to_clock_t(utime);
 965                 tmp.tms_stime = jiffies_to_clock_t(stime);
 966                 tmp.tms_cutime = jiffies_to_clock_t(cutime);
 967                 tmp.tms_cstime = jiffies_to_clock_t(cstime);
 968                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
 969                         return -EFAULT;
 970         }
 971         return (long) jiffies_64_to_clock_t(get_jiffies_64());
 972 }
 973
 974 /*
 975  * This needs some heavy checking ...
 976  * I just haven't the stomach for it. I also don't fully
 977  * understand sessions/pgrp etc. Let somebody who does explain it.
 978  *
 979  * OK, I think I have the protection semantics right.... this is really
 980  * only important on a multi-user system anyway, to make sure one user
 981  * can't send a signal to a process owned by another.  -TYT, 12/12/91
 982  *
 983  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
 984  * LBT 04.03.94
 985  */
 986
 987 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
 988 {
 989         struct task_struct *p;
 990         int err = -EINVAL;
 991         pid_t rpgid;
 992
 993         if (!pid)
 994                 pid = vx_map_pid(current->pid);
 995         if (!pgid)
 996                 pgid = pid;
 997         if (pgid < 0)
 998                 return -EINVAL;
 999
1000         rpgid = vx_rmap_pid(pgid);
1001
1002         /* From this point forward we keep holding onto the tasklist lock
1003          * so that our parent does not change from under us. -DaveM
1004          */
1005         write_lock_irq(&tasklist_lock);
1006
1007         err = -ESRCH;
1008         p = find_task_by_pid(pid);
1009         if (!p)
1010                 goto out;
1011
1012         err = -EINVAL;
1013         if (!thread_group_leader(p))
1014                 goto out;
1015
1016         if (p->parent == current || p->real_parent == current) {
1017                 err = -EPERM;
1018                 if (p->signal->session != current->signal->session)
1019                         goto out;
1020                 err = -EACCES;
1021                 if (p->did_exec)
1022                         goto out;
1023         } else {
1024                 err = -ESRCH;
1025                 if (p != current)
1026                         goto out;
1027         }
1028
1029         err = -EPERM;
1030         if (p->signal->leader)
1031                 goto out;
1032
1033         if (pgid != pid) {
1034                 struct task_struct *p;
1035
1036                 do_each_task_pid(rpgid, PIDTYPE_PGID, p) {
1037                         if (p->signal->session == current->signal->session)
1038                                 goto ok_pgid;
1039                 } while_each_task_pid(rpgid, PIDTYPE_PGID, p);
1040                 goto out;
1041         }
1042
1043 ok_pgid:
1044         err = security_task_setpgid(p, rpgid);
1045         if (err)
1046                 goto out;
1047
1048         if (process_group(p) != rpgid) {
1049                 detach_pid(p, PIDTYPE_PGID);
1050                 p->signal->pgrp = rpgid;
1051                 attach_pid(p, PIDTYPE_PGID, rpgid);
1052         }
1053
1054         err = 0;
1055 out:
1056         /* All paths lead to here, thus we are safe. -DaveM */
1057         write_unlock_irq(&tasklist_lock);
1058         return err;
1059 }
1060
1061 asmlinkage long sys_getpgid(pid_t pid)
1062 {
1063         if (!pid) {
1064                 return vx_rmap_pid(process_group(current));
1065         } else {
1066                 int retval;
1067                 struct task_struct *p;
1068
1069                 read_lock(&tasklist_lock);
1070                 p = find_task_by_pid(pid);
1071
1072                 retval = -ESRCH;
1073                 if (p) {
1074                         retval = security_task_getpgid(p);
1075                         if (!retval)
1076                                 retval = vx_rmap_pid(process_group(p));
1077                 }
1078                 read_unlock(&tasklist_lock);
1079                 return retval;
1080         }
1081 }
1082
1083 #ifdef __ARCH_WANT_SYS_GETPGRP
1084
1085 asmlinkage long sys_getpgrp(void)
1086 {
1087         /* SMP - assuming writes are word atomic this is fine */
1088         return process_group(current);
1089 }
1090
1091 #endif
1092
1093 asmlinkage long sys_getsid(pid_t pid)
1094 {
1095         if (!pid) {
1096                 return current->signal->session;
1097         } else {
1098                 int retval;
1099                 struct task_struct *p;
1100
1101                 read_lock(&tasklist_lock);
1102                 p = find_task_by_pid(pid);
1103
1104                 retval = -ESRCH;
1105                 if(p) {
1106                         retval = security_task_getsid(p);
1107                         if (!retval)
1108                                 retval = p->signal->session;
1109                 }
1110                 read_unlock(&tasklist_lock);
1111                 return retval;
1112         }
1113 }
1114
1115 asmlinkage long sys_setsid(void)
1116 {
1117         struct pid *pid;
1118         int err = -EPERM;
1119
1120         if (!thread_group_leader(current))
1121                 return -EINVAL;
1122
1123         down(&tty_sem);
1124         write_lock_irq(&tasklist_lock);
1125
1126         pid = find_pid(PIDTYPE_PGID, current->pid);
1127         if (pid)
1128                 goto out;
1129
1130         current->signal->leader = 1;
1131         __set_special_pids(current->pid, current->pid);
1132         current->signal->tty = NULL;
1133         current->signal->tty_old_pgrp = 0;
1134         err = process_group(current);
1135 out:
1136         write_unlock_irq(&tasklist_lock);
1137         up(&tty_sem);
1138         return err;
1139 }
1140
1141 /*
1142  * Supplementary group IDs
1143  */
1144
1145 /* init to 2 - one for init_task, one to ensure it is never freed */
1146 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1147
1148 struct group_info *groups_alloc(int gidsetsize)
1149 {
1150         struct group_info *group_info;
1151         int nblocks;
1152         int i;
1153
1154         nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1155         /* Make sure we always allocate at least one indirect block pointer */
1156         nblocks = nblocks ? : 1;
1157         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1158         if (!group_info)
1159                 return NULL;
1160         group_info->ngroups = gidsetsize;
1161         group_info->nblocks = nblocks;
1162         atomic_set(&group_info->usage, 1);
1163
1164         if (gidsetsize <= NGROUPS_SMALL) {
1165                 group_info->blocks[0] = group_info->small_block;
1166         } else {
1167                 for (i = 0; i < nblocks; i++) {
1168                         gid_t *b;
1169                         b = (void *)__get_free_page(GFP_USER);
1170                         if (!b)
1171                                 goto out_undo_partial_alloc;
1172                         group_info->blocks[i] = b;
1173                 }
1174         }
1175         return group_info;
1176
1177 out_undo_partial_alloc:
1178         while (--i >= 0) {
1179                 free_page((unsigned long)group_info->blocks[i]);
1180         }
1181         kfree(group_info);
1182         return NULL;
1183 }
1184
1185 EXPORT_SYMBOL(groups_alloc);
1186
1187 void groups_free(struct group_info *group_info)
1188 {
1189         if (group_info->blocks[0] != group_info->small_block) {
1190                 int i;
1191                 for (i = 0; i < group_info->nblocks; i++)
1192                         free_page((unsigned long)group_info->blocks[i]);
1193         }
1194         kfree(group_info);
1195 }
1196
1197 EXPORT_SYMBOL(groups_free);
1198
1199 /* export the group_info to a user-space array */
1200 static int groups_to_user(gid_t __user *grouplist,
1201     struct group_info *group_info)
1202 {
1203         int i;
1204         int count = group_info->ngroups;
1205
1206         for (i = 0; i < group_info->nblocks; i++) {
1207                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1208                 int off = i * NGROUPS_PER_BLOCK;
1209                 int len = cp_count * sizeof(*grouplist);
1210
1211                 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1212                         return -EFAULT;
1213
1214                 count -= cp_count;
1215         }
1216         return 0;
1217 }
1218
1219 /* fill a group_info from a user-space array - it must be allocated already */
1220 static int groups_from_user(struct group_info *group_info,
1221     gid_t __user *grouplist)
1222  {
1223         int i;
1224         int count = group_info->ngroups;
1225
1226         for (i = 0; i < group_info->nblocks; i++) {
1227                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1228                 int off = i * NGROUPS_PER_BLOCK;
1229                 int len = cp_count * sizeof(*grouplist);
1230
1231                 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1232                         return -EFAULT;
1233
1234                 count -= cp_count;
1235         }
1236         return 0;
1237 }
1238
1239 /* a simple shell-metzner sort */
1240 static void groups_sort(struct group_info *group_info)
1241 {
1242         int base, max, stride;
1243         int gidsetsize = group_info->ngroups;
1244
1245         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1246                 ; /* nothing */
1247         stride /= 3;
1248
1249         while (stride) {
1250                 max = gidsetsize - stride;
1251                 for (base = 0; base < max; base++) {
1252                         int left = base;
1253                         int right = left + stride;
1254                         gid_t tmp = GROUP_AT(group_info, right);
1255
1256                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1257                                 GROUP_AT(group_info, right) =
1258                                     GROUP_AT(group_info, left);
1259                                 right = left;
1260                                 left -= stride;
1261                         }
1262                         GROUP_AT(group_info, right) = tmp;
1263                 }
1264                 stride /= 3;
1265         }
1266 }
1267
1268 /* a simple bsearch */
1269 static int groups_search(struct group_info *group_info, gid_t grp)
1270 {
1271         int left, right;
1272
1273         if (!group_info)
1274                 return 0;
1275
1276         left = 0;
1277         right = group_info->ngroups;
1278         while (left < right) {
1279                 int mid = (left+right)/2;
1280                 int cmp = grp - GROUP_AT(group_info, mid);
1281                 if (cmp > 0)
1282                         left = mid + 1;
1283                 else if (cmp < 0)
1284                         right = mid;
1285                 else
1286                         return 1;
1287         }
1288         return 0;
1289 }
1290
1291 /* validate and set current->group_info */
1292 int set_current_groups(struct group_info *group_info)
1293 {
1294         int retval;
1295         struct group_info *old_info;
1296
1297         retval = security_task_setgroups(group_info);
1298         if (retval)
1299                 return retval;
1300
1301         groups_sort(group_info);
1302         get_group_info(group_info);
1303
1304         task_lock(current);
1305         old_info = current->group_info;
1306         current->group_info = group_info;
1307         task_unlock(current);
1308
1309         put_group_info(old_info);
1310
1311         return 0;
1312 }
1313
1314 EXPORT_SYMBOL(set_current_groups);
1315
1316 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1317 {
1318         int i = 0;
1319
1320         /*
1321          *      SMP: Nobody else can change our grouplist. Thus we are
1322          *      safe.
1323          */
1324
1325         if (gidsetsize < 0)
1326                 return -EINVAL;
1327
1328         /* no need to grab task_lock here; it cannot change */
1329         get_group_info(current->group_info);
1330         i = current->group_info->ngroups;
1331         if (gidsetsize) {
1332                 if (i > gidsetsize) {
1333                         i = -EINVAL;
1334                         goto out;
1335                 }
1336                 if (groups_to_user(grouplist, current->group_info)) {
1337                         i = -EFAULT;
1338                         goto out;
1339                 }
1340         }
1341 out:
1342         put_group_info(current->group_info);
1343         return i;
1344 }
1345
1346 /*
1347  *      SMP: Our groups are copy-on-write. We can set them safely
1348  *      without another task interfering.
1349  */
1350
1351 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1352 {
1353         struct group_info *group_info;
1354         int retval;
1355
1356         if (!capable(CAP_SETGID))
1357                 return -EPERM;
1358         if ((unsigned)gidsetsize > NGROUPS_MAX)
1359                 return -EINVAL;
1360
1361         group_info = groups_alloc(gidsetsize);
1362         if (!group_info)
1363                 return -ENOMEM;
1364         retval = groups_from_user(group_info, grouplist);
1365         if (retval) {
1366                 put_group_info(group_info);
1367                 return retval;
1368         }
1369
1370         retval = set_current_groups(group_info);
1371         put_group_info(group_info);
1372
1373         return retval;
1374 }
1375
1376 /*
1377  * Check whether we're fsgid/egid or in the supplemental group..
1378  */
1379 int in_group_p(gid_t grp)
1380 {
1381         int retval = 1;
1382         if (grp != current->fsgid) {
1383                 get_group_info(current->group_info);
1384                 retval = groups_search(current->group_info, grp);
1385                 put_group_info(current->group_info);
1386         }
1387         return retval;
1388 }
1389
1390 EXPORT_SYMBOL(in_group_p);
1391
1392 int in_egroup_p(gid_t grp)
1393 {
1394         int retval = 1;
1395         if (grp != current->egid) {
1396                 get_group_info(current->group_info);
1397                 retval = groups_search(current->group_info, grp);
1398                 put_group_info(current->group_info);
1399         }
1400         return retval;
1401 }
1402
1403 EXPORT_SYMBOL(in_egroup_p);
1404
1405 DECLARE_RWSEM(uts_sem);
1406
1407 EXPORT_SYMBOL(uts_sem);
1408
1409 asmlinkage long sys_newuname(struct new_utsname __user * name)
1410 {
1411         int errno = 0;
1412
1413         down_read(&uts_sem);
1414         if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1415                 errno = -EFAULT;
1416         up_read(&uts_sem);
1417         return errno;
1418 }
1419
1420 asmlinkage long sys_sethostname(char __user *name, int len)
1421 {
1422         int errno;
1423         char tmp[__NEW_UTS_LEN];
1424
1425         if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1426                 return -EPERM;
1427         if (len < 0 || len > __NEW_UTS_LEN)
1428                 return -EINVAL;
1429         down_write(&uts_sem);
1430         errno = -EFAULT;
1431         if (!copy_from_user(tmp, name, len)) {
1432                 char *ptr = vx_new_uts(nodename);
1433
1434                 memcpy(ptr, tmp, len);
1435                 ptr[len] = 0;
1436                 errno = 0;
1437         }
1438         up_write(&uts_sem);
1439         return errno;
1440 }
1441
1442 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1443
1444 asmlinkage long sys_gethostname(char __user *name, int len)
1445 {
1446         int i, errno;
1447         char *ptr;
1448
1449         if (len < 0)
1450                 return -EINVAL;
1451         down_read(&uts_sem);
1452         ptr = vx_new_uts(nodename);
1453         i = 1 + strlen(ptr);
1454         if (i > len)
1455                 i = len;
1456         errno = 0;
1457         if (copy_to_user(name, ptr, i))
1458                 errno = -EFAULT;
1459         up_read(&uts_sem);
1460         return errno;
1461 }
1462
1463 #endif
1464
1465 /*
1466  * Only setdomainname; getdomainname can be implemented by calling
1467  * uname()
1468  */
1469 asmlinkage long sys_setdomainname(char __user *name, int len)
1470 {
1471         int errno;
1472         char tmp[__NEW_UTS_LEN];
1473
1474         if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1475                 return -EPERM;
1476         if (len < 0 || len > __NEW_UTS_LEN)
1477                 return -EINVAL;
1478
1479         down_write(&uts_sem);
1480         errno = -EFAULT;
1481         if (!copy_from_user(tmp, name, len)) {
1482                 char *ptr = vx_new_uts(domainname);
1483
1484                 memcpy(ptr, tmp, len);
1485                 ptr[len] = 0;
1486                 errno = 0;
1487         }
1488         up_write(&uts_sem);
1489         return errno;
1490 }
1491
1492 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1493 {
1494         if (resource >= RLIM_NLIMITS)
1495                 return -EINVAL;
1496         else {
1497                 struct rlimit value;
1498                 task_lock(current->group_leader);
1499                 value = current->signal->rlim[resource];
1500                 task_unlock(current->group_leader);
1501                 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1502         }
1503 }
1504
1505 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1506
1507 /*
1508  *      Back compatibility for getrlimit. Needed for some apps.
1509  */
1510
1511 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1512 {
1513         struct rlimit x;
1514         if (resource >= RLIM_NLIMITS)
1515                 return -EINVAL;
1516
1517         task_lock(current->group_leader);
1518         x = current->signal->rlim[resource];
1519         task_unlock(current->group_leader);
1520         if(x.rlim_cur > 0x7FFFFFFF)
1521                 x.rlim_cur = 0x7FFFFFFF;
1522         if(x.rlim_max > 0x7FFFFFFF)
1523                 x.rlim_max = 0x7FFFFFFF;
1524         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1525 }
1526
1527 #endif
1528
1529 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1530 {
1531         struct rlimit new_rlim, *old_rlim;
1532         int retval;
1533
1534         if (resource >= RLIM_NLIMITS)
1535                 return -EINVAL;
1536         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1537                 return -EFAULT;
1538        if (new_rlim.rlim_cur > new_rlim.rlim_max)
1539                return -EINVAL;
1540         old_rlim = current->signal->rlim + resource;
1541         if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1542             !capable(CAP_SYS_RESOURCE) && !vx_ccaps(VXC_SET_RLIMIT))
1543                 return -EPERM;
1544         if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1545                         return -EPERM;
1546
1547         retval = security_task_setrlimit(resource, &new_rlim);
1548         if (retval)
1549                 return retval;
1550
1551         task_lock(current->group_leader);
1552         *old_rlim = new_rlim;
1553         task_unlock(current->group_leader);
1554         return 0;
1555 }
1556
1557 /*
1558  * It would make sense to put struct rusage in the task_struct,
1559  * except that would make the task_struct be *really big*.  After
1560  * task_struct gets moved into malloc'ed memory, it would
1561  * make sense to do this.  It will make moving the rest of the information
1562  * a lot simpler!  (Which we're not doing right now because we're not
1563  * measuring them yet).
1564  *
1565  * This expects to be called with tasklist_lock read-locked or better,
1566  * and the siglock not locked.  It may momentarily take the siglock.
1567  *
1568  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1569  * races with threads incrementing their own counters.  But since word
1570  * reads are atomic, we either get new values or old values and we don't
1571  * care which for the sums.  We always take the siglock to protect reading
1572  * the c* fields from p->signal from races with exit.c updating those
1573  * fields when reaping, so a sample either gets all the additions of a
1574  * given child after it's reaped, or none so this sample is before reaping.
1575  */
1576
1577 void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1578 {
1579         struct task_struct *t;
1580         unsigned long flags;
1581         unsigned long utime, stime;
1582
1583         memset((char *) r, 0, sizeof *r);
1584
1585         if (unlikely(!p->signal))
1586                 return;
1587
1588         switch (who) {
1589                 case RUSAGE_CHILDREN:
1590                         spin_lock_irqsave(&p->sighand->siglock, flags);
1591                         utime = p->signal->cutime;
1592                         stime = p->signal->cstime;
1593                         r->ru_nvcsw = p->signal->cnvcsw;
1594                         r->ru_nivcsw = p->signal->cnivcsw;
1595                         r->ru_minflt = p->signal->cmin_flt;
1596                         r->ru_majflt = p->signal->cmaj_flt;
1597                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1598                         jiffies_to_timeval(utime, &r->ru_utime);
1599                         jiffies_to_timeval(stime, &r->ru_stime);
1600                         break;
1601                 case RUSAGE_SELF:
1602                         spin_lock_irqsave(&p->sighand->siglock, flags);
1603                         utime = stime = 0;
1604                         goto sum_group;
1605                 case RUSAGE_BOTH:
1606                         spin_lock_irqsave(&p->sighand->siglock, flags);
1607                         utime = p->signal->cutime;
1608                         stime = p->signal->cstime;
1609                         r->ru_nvcsw = p->signal->cnvcsw;
1610                         r->ru_nivcsw = p->signal->cnivcsw;
1611                         r->ru_minflt = p->signal->cmin_flt;
1612                         r->ru_majflt = p->signal->cmaj_flt;
1613                 sum_group:
1614                         utime += p->signal->utime;
1615                         stime += p->signal->stime;
1616                         r->ru_nvcsw += p->signal->nvcsw;
1617                         r->ru_nivcsw += p->signal->nivcsw;
1618                         r->ru_minflt += p->signal->min_flt;
1619                         r->ru_majflt += p->signal->maj_flt;
1620                         t = p;
1621                         do {
1622                                 utime += t->utime;
1623                                 stime += t->stime;
1624                                 r->ru_nvcsw += t->nvcsw;
1625                                 r->ru_nivcsw += t->nivcsw;
1626                                 r->ru_minflt += t->min_flt;
1627                                 r->ru_majflt += t->maj_flt;
1628                                 t = next_thread(t);
1629                         } while (t != p);
1630                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1631                         jiffies_to_timeval(utime, &r->ru_utime);
1632                         jiffies_to_timeval(stime, &r->ru_stime);
1633                         break;
1634                 default:
1635                         BUG();
1636         }
1637 }
1638
1639 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1640 {
1641         struct rusage r;
1642         read_lock(&tasklist_lock);
1643         k_getrusage(p, who, &r);
1644         read_unlock(&tasklist_lock);
1645         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1646 }
1647
1648 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1649 {
1650         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1651                 return -EINVAL;
1652         return getrusage(current, who, ru);
1653 }
1654
1655 asmlinkage long sys_umask(int mask)
1656 {
1657         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1658         return mask;
1659 }
1660
1661 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1662                           unsigned long arg4, unsigned long arg5)
1663 {
1664         long error;
1665         int sig;
1666
1667         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1668         if (error)
1669                 return error;
1670
1671         switch (option) {
1672                 case PR_SET_PDEATHSIG:
1673                         sig = arg2;
1674                         if (sig < 0 || sig > _NSIG) {
1675                                 error = -EINVAL;
1676                                 break;
1677                         }
1678                         current->pdeath_signal = sig;
1679                         break;
1680                 case PR_GET_PDEATHSIG:
1681                         error = put_user(current->pdeath_signal, (int __user *)arg2);
1682                         break;
1683                 case PR_GET_DUMPABLE:
1684                         if (current->mm->dumpable)
1685                                 error = 1;
1686                         break;
1687                 case PR_SET_DUMPABLE:
1688                         if (arg2 < 0 && arg2 > 2) {
1689                                 error = -EINVAL;
1690                                 break;
1691                         }
1692                         current->mm->dumpable = arg2;
1693                         break;
1694
1695                 case PR_SET_UNALIGN:
1696                         error = SET_UNALIGN_CTL(current, arg2);
1697                         break;
1698                 case PR_GET_UNALIGN:
1699                         error = GET_UNALIGN_CTL(current, arg2);
1700                         break;
1701                 case PR_SET_FPEMU:
1702                         error = SET_FPEMU_CTL(current, arg2);
1703                         break;
1704                 case PR_GET_FPEMU:
1705                         error = GET_FPEMU_CTL(current, arg2);
1706                         break;
1707                 case PR_SET_FPEXC:
1708                         error = SET_FPEXC_CTL(current, arg2);
1709                         break;
1710                 case PR_GET_FPEXC:
1711                         error = GET_FPEXC_CTL(current, arg2);
1712                         break;
1713                 case PR_GET_TIMING:
1714                         error = PR_TIMING_STATISTICAL;
1715                         break;
1716                 case PR_SET_TIMING:
1717                         if (arg2 == PR_TIMING_STATISTICAL)
1718                                 error = 0;
1719                         else
1720                                 error = -EINVAL;
1721                         break;
1722
1723                 case PR_GET_KEEPCAPS:
1724                         if (current->keep_capabilities)
1725                                 error = 1;
1726                         break;
1727                 case PR_SET_KEEPCAPS:
1728                         if (arg2 != 0 && arg2 != 1) {
1729                                 error = -EINVAL;
1730                                 break;
1731                         }
1732                         current->keep_capabilities = arg2;
1733                         break;
1734                 case PR_SET_NAME: {
1735                         struct task_struct *me = current;
1736                         unsigned char ncomm[sizeof(me->comm)];
1737
1738                         ncomm[sizeof(me->comm)-1] = 0;
1739                         if (strncpy_from_user(ncomm, (char __user *)arg2,
1740                                                 sizeof(me->comm)-1) < 0)
1741                                 return -EFAULT;
1742                         set_task_comm(me, ncomm);
1743                         return 0;
1744                 }
1745                 default:
1746                         error = -EINVAL;
1747                         break;
1748         }
1749         return error;
1750 }