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