kernel/sys.c

   1 /*
   2  *  linux/kernel/sys.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6
   7 #include <linux/mm.h>
   8 #include <linux/utsname.h>
   9 #include <linux/mman.h>
  10 #include <linux/smp_lock.h>
  11 #include <linux/notifier.h>
  12 #include <linux/reboot.h>
  13 #include <linux/prctl.h>
  14 #include <linux/init.h>
  15 #include <linux/highuid.h>
  16
  17 #include <asm/uaccess.h>
  18 #include <asm/io.h>
  19
  20 /*
  21  * this is where the system-wide overflow UID and GID are defined, for
  22  * architectures that now have 32-bit UID/GID but didn't in the past
  23  */
  24
  25 int overflowuid = DEFAULT_OVERFLOWUID;
  26 int overflowgid = DEFAULT_OVERFLOWGID;
  27
  28 /*
  29  * the same as above, but for filesystems which can only store a 16-bit
  30  * UID and GID. as such, this is needed on all architectures
  31  */
  32
  33 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
  34 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
  35
  36 /*
  37  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
  38  */
  39
  40 int C_A_D = 1;
  41
  42
  43 /*
  44  *      Notifier list for kernel code which wants to be called
  45  *      at shutdown. This is used to stop any idling DMA operations
  46  *      and the like.
  47  */
  48
  49 struct notifier_block *reboot_notifier_list = NULL;
  50
  51 int register_reboot_notifier(struct notifier_block * nb)
  52 {
  53         return notifier_chain_register(&reboot_notifier_list, nb);
  54 }
  55
  56 int unregister_reboot_notifier(struct notifier_block * nb)
  57 {
  58         return notifier_chain_unregister(&reboot_notifier_list, nb);
  59 }
  60
  61 asmlinkage long sys_ni_syscall(void)
  62 {
  63         return -ENOSYS;
  64 }
  65
  66 static int proc_sel(struct task_struct *p, int which, int who)
  67 {
  68         if(p->pid)
  69         {
  70                 switch (which) {
  71                         case PRIO_PROCESS:
  72                                 if (!who && p == current)
  73                                         return 1;
  74                                 return(p->pid == who);
  75                         case PRIO_PGRP:
  76                                 if (!who)
  77                                         who = current->pgrp;
  78                                 return(p->pgrp == who);
  79                         case PRIO_USER:
  80                                 if (!who)
  81                                         who = current->uid;
  82                                 return(p->uid == who);
  83                 }
  84         }
  85         return 0;
  86 }
  87
  88 asmlinkage long sys_setpriority(int which, int who, int niceval)
  89 {
  90         struct task_struct *p;
  91         unsigned int priority;
  92         int error;
  93
  94         if (which > 2 || which < 0)
  95                 return -EINVAL;
  96
  97         /* normalize: avoid signed division (rounding problems) */
  98         error = ESRCH;
  99         priority = niceval;
 100         if (niceval < 0)
 101                 priority = -niceval;
 102         if (priority > 20)
 103                 priority = 20;
 104         priority = (priority * DEF_PRIORITY + 10) / 20 + DEF_PRIORITY;
 105
 106         if (niceval >= 0) {
 107                 priority = 2*DEF_PRIORITY - priority;
 108                 if (!priority)
 109                         priority = 1;
 110         }
 111
 112         read_lock(&tasklist_lock);
 113         for_each_task(p) {
 114                 if (!proc_sel(p, which, who))
 115                         continue;
 116                 if (p->uid != current->euid &&
 117                         p->uid != current->uid && !capable(CAP_SYS_NICE)) {
 118                         error = EPERM;
 119                         continue;
 120                 }
 121                 if (error == ESRCH)
 122                         error = 0;
 123                 if (priority > p->priority && !capable(CAP_SYS_NICE))
 124                         error = EACCES;
 125                 else
 126                         p->priority = priority;
 127         }
 128         read_unlock(&tasklist_lock);
 129
 130         return -error;
 131 }
 132
 133 /*
 134  * Ugh. To avoid negative return values, "getpriority()" will
 135  * not return the normal nice-value, but a value that has been
 136  * offset by 20 (ie it returns 0..40 instead of -20..20)
 137  */
 138 asmlinkage long sys_getpriority(int which, int who)
 139 {
 140         struct task_struct *p;
 141         long max_prio = -ESRCH;
 142
 143         if (which > 2 || which < 0)
 144                 return -EINVAL;
 145
 146         read_lock(&tasklist_lock);
 147         for_each_task (p) {
 148                 if (!proc_sel(p, which, who))
 149                         continue;
 150                 if (p->priority > max_prio)
 151                         max_prio = p->priority;
 152         }
 153         read_unlock(&tasklist_lock);
 154
 155         /* scale the priority from timeslice to 0..40 */
 156         if (max_prio > 0)
 157                 max_prio = (max_prio * 20 + DEF_PRIORITY/2) / DEF_PRIORITY;
 158         return max_prio;
 159 }
 160
 161
 162 /*
 163  * Reboot system call: for obvious reasons only root may call it,
 164  * and even root needs to set up some magic numbers in the registers
 165  * so that some mistake won't make this reboot the whole machine.
 166  * You can also set the meaning of the ctrl-alt-del-key here.
 167  *
 168  * reboot doesn't sync: do that yourself before calling this.
 169  */
 170 asmlinkage long sys_reboot(int magic1, int magic2, int cmd, void * arg)
 171 {
 172         char buffer[256];
 173
 174         /* We only trust the superuser with rebooting the system. */
 175         if (!capable(CAP_SYS_BOOT))
 176                 return -EPERM;
 177
 178         /* For safety, we require "magic" arguments. */
 179         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 180             (magic2 != LINUX_REBOOT_MAGIC2 && magic2 != LINUX_REBOOT_MAGIC2A &&
 181                         magic2 != LINUX_REBOOT_MAGIC2B))
 182                 return -EINVAL;
 183
 184         lock_kernel();
 185         switch (cmd) {
 186         case LINUX_REBOOT_CMD_RESTART:
 187                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 188                 printk(KERN_EMERG "Restarting system.\n");
 189                 machine_restart(NULL);
 190                 break;
 191
 192         case LINUX_REBOOT_CMD_CAD_ON:
 193                 C_A_D = 1;
 194                 break;
 195
 196         case LINUX_REBOOT_CMD_CAD_OFF:
 197                 C_A_D = 0;
 198                 break;
 199
 200         case LINUX_REBOOT_CMD_HALT:
 201                 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
 202                 printk(KERN_EMERG "System halted.\n");
 203                 machine_halt();
 204                 do_exit(0);
 205                 break;
 206
 207         case LINUX_REBOOT_CMD_POWER_OFF:
 208                 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
 209                 printk(KERN_EMERG "Power down.\n");
 210                 machine_power_off();
 211                 do_exit(0);
 212                 break;
 213
 214         case LINUX_REBOOT_CMD_RESTART2:
 215                 if (strncpy_from_user(&buffer[0], (char *)arg, sizeof(buffer) - 1) < 0) {
 216                         unlock_kernel();
 217                         return -EFAULT;
 218                 }
 219                 buffer[sizeof(buffer) - 1] = '\0';
 220
 221                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
 222                 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
 223                 machine_restart(buffer);
 224                 break;
 225
 226         default:
 227                 unlock_kernel();
 228                 return -EINVAL;
 229                 break;
 230         };
 231         unlock_kernel();
 232         return 0;
 233 }
 234
 235 /*
 236  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 237  * As it's called within an interrupt, it may NOT sync: the only choice
 238  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 239  */
 240 void ctrl_alt_del(void)
 241 {
 242         if (C_A_D) {
 243                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 244                 machine_restart(NULL);
 245         } else
 246                 kill_proc(1, SIGINT, 1);
 247 }
 248
 249
 250 /*
 251  * Unprivileged users may change the real gid to the effective gid
 252  * or vice versa.  (BSD-style)
 253  *
 254  * If you set the real gid at all, or set the effective gid to a value not
 255  * equal to the real gid, then the saved gid is set to the new effective gid.
 256  *
 257  * This makes it possible for a setgid program to completely drop its
 258  * privileges, which is often a useful assertion to make when you are doing
 259  * a security audit over a program.
 260  *
 261  * The general idea is that a program which uses just setregid() will be
 262  * 100% compatible with BSD.  A program which uses just setgid() will be
 263  * 100% compatible with POSIX with saved IDs.
 264  *
 265  * SMP: There are not races, the GIDs are checked only by filesystem
 266  *      operations (as far as semantic preservation is concerned).
 267  */
 268 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 269 {
 270         int old_rgid = current->gid;
 271         int old_egid = current->egid;
 272
 273         if (rgid != (gid_t) -1) {
 274                 if ((old_rgid == rgid) ||
 275                     (current->egid==rgid) ||
 276                     capable(CAP_SETGID))
 277                         current->gid = rgid;
 278                 else
 279                         return -EPERM;
 280         }
 281         if (egid != (gid_t) -1) {
 282                 if ((old_rgid == egid) ||
 283                     (current->egid == egid) ||
 284                     (current->sgid == egid) ||
 285                     capable(CAP_SETGID))
 286                         current->fsgid = current->egid = egid;
 287                 else {
 288                         current->gid = old_rgid;
 289                         return -EPERM;
 290                 }
 291         }
 292         if (rgid != (gid_t) -1 ||
 293             (egid != (gid_t) -1 && egid != old_rgid))
 294                 current->sgid = current->egid;
 295         current->fsgid = current->egid;
 296         if (current->egid != old_egid)
 297                 current->dumpable = 0;
 298         return 0;
 299 }
 300
 301 /*
 302  * setgid() is implemented like SysV w/ SAVED_IDS
 303  *
 304  * SMP: Same implicit races as above.
 305  */
 306 asmlinkage long sys_setgid(gid_t gid)
 307 {
 308         int old_egid = current->egid;
 309
 310         if (capable(CAP_SETGID))
 311                 current->gid = current->egid = current->sgid = current->fsgid = gid;
 312         else if ((gid == current->gid) || (gid == current->sgid))
 313                 current->egid = current->fsgid = gid;
 314         else
 315                 return -EPERM;
 316
 317         if (current->egid != old_egid)
 318                 current->dumpable = 0;
 319         return 0;
 320 }
 321
 322 /*
 323  * cap_emulate_setxuid() fixes the effective / permitted capabilities of
 324  * a process after a call to setuid, setreuid, or setresuid.
 325  *
 326  *  1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of
 327  *  {r,e,s}uid != 0, the permitted and effective capabilities are
 328  *  cleared.
 329  *
 330  *  2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective
 331  *  capabilities of the process are cleared.
 332  *
 333  *  3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective
 334  *  capabilities are set to the permitted capabilities.
 335  *
 336  *  fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should
 337  *  never happen.
 338  *
 339  *  -astor
 340  */
 341 extern inline void cap_emulate_setxuid(int old_ruid, int old_euid,
 342                                        int old_suid)
 343 {
 344         if ((old_ruid == 0 || old_euid == 0 || old_suid == 0) &&
 345             (current->uid != 0 && current->euid != 0 && current->suid != 0)) {
 346                 cap_clear(current->cap_permitted);
 347                 cap_clear(current->cap_effective);
 348         }
 349         if (old_euid == 0 && current->euid != 0) {
 350                 cap_clear(current->cap_effective);
 351         }
 352         if (old_euid != 0 && current->euid == 0) {
 353                 current->cap_effective = current->cap_permitted;
 354         }
 355 }
 356
 357 /*
 358  * Unprivileged users may change the real uid to the effective uid
 359  * or vice versa.  (BSD-style)
 360  *
 361  * If you set the real uid at all, or set the effective uid to a value not
 362  * equal to the real uid, then the saved uid is set to the new effective uid.
 363  *
 364  * This makes it possible for a setuid program to completely drop its
 365  * privileges, which is often a useful assertion to make when you are doing
 366  * a security audit over a program.
 367  *
 368  * The general idea is that a program which uses just setreuid() will be
 369  * 100% compatible with BSD.  A program which uses just setuid() will be
 370  * 100% compatible with POSIX with saved IDs.
 371  */
 372 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 373 {
 374         int old_ruid, old_euid, old_suid, new_ruid;
 375
 376         new_ruid = old_ruid = current->uid;
 377         old_euid = current->euid;
 378         old_suid = current->suid;
 379         if (ruid != (uid_t) -1) {
 380                 if ((old_ruid == ruid) ||
 381                     (current->euid==ruid) ||
 382                     capable(CAP_SETUID))
 383                         new_ruid = ruid;
 384                 else
 385                         return -EPERM;
 386         }
 387         if (euid != (uid_t) -1) {
 388                 if ((old_ruid == euid) ||
 389                     (current->euid == euid) ||
 390                     (current->suid == euid) ||
 391                     capable(CAP_SETUID))
 392                         current->fsuid = current->euid = euid;
 393                 else
 394                         return -EPERM;
 395         }
 396         if (ruid != (uid_t) -1 ||
 397             (euid != (uid_t) -1 && euid != old_ruid))
 398                 current->suid = current->euid;
 399         current->fsuid = current->euid;
 400         if (current->euid != old_euid)
 401                 current->dumpable = 0;
 402
 403         if(new_ruid != old_ruid) {
 404                 /* What if a process setreuid()'s and this brings the
 405                  * new uid over his NPROC rlimit?  We can check this now
 406                  * cheaply with the new uid cache, so if it matters
 407                  * we should be checking for it.  -DaveM
 408                  */
 409                 free_uid(current);
 410                 current->uid = new_ruid;
 411                 alloc_uid(current);
 412         }
 413
 414         if (!issecure(SECURE_NO_SETUID_FIXUP)) {
 415                 cap_emulate_setxuid(old_ruid, old_euid, old_suid);
 416         }
 417
 418         return 0;
 419 }
 420
 421
 422
 423 /*
 424  * setuid() is implemented like SysV with SAVED_IDS
 425  *
 426  * Note that SAVED_ID's is deficient in that a setuid root program
 427  * like sendmail, for example, cannot set its uid to be a normal
 428  * user and then switch back, because if you're root, setuid() sets
 429  * the saved uid too.  If you don't like this, blame the bright people
 430  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 431  * will allow a root program to temporarily drop privileges and be able to
 432  * regain them by swapping the real and effective uid.
 433  */
 434 asmlinkage long sys_setuid(uid_t uid)
 435 {
 436         int old_euid = current->euid;
 437         int old_ruid, old_suid, new_ruid;
 438
 439         old_ruid = new_ruid = current->uid;
 440         old_suid = current->suid;
 441         if (capable(CAP_SETUID))
 442                 new_ruid = current->euid = current->suid = current->fsuid = uid;
 443         else if ((uid == current->uid) || (uid == current->suid))
 444                 current->fsuid = current->euid = uid;
 445         else
 446                 return -EPERM;
 447
 448         if (current->euid != old_euid)
 449                 current->dumpable = 0;
 450
 451        if (new_ruid != old_ruid) {
 452                 /* See comment above about NPROC rlimit issues... */
 453                 free_uid(current);
 454                 current->uid = new_ruid;
 455                 alloc_uid(current);
 456         }
 457
 458         if (!issecure(SECURE_NO_SETUID_FIXUP)) {
 459                 cap_emulate_setxuid(old_ruid, old_euid, old_suid);
 460         }
 461
 462         return 0;
 463 }
 464
 465
 466 /*
 467  * This function implements a generic ability to update ruid, euid,
 468  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 469  */
 470 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 471 {
 472         int old_ruid = current->uid;
 473         int old_euid = current->euid;
 474         int old_suid = current->suid;
 475
 476         if (!capable(CAP_SETUID)) {
 477                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
 478                     (ruid != current->euid) && (ruid != current->suid))
 479                         return -EPERM;
 480                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
 481                     (euid != current->euid) && (euid != current->suid))
 482                         return -EPERM;
 483                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
 484                     (suid != current->euid) && (suid != current->suid))
 485                         return -EPERM;
 486         }
 487         if (ruid != (uid_t) -1) {
 488                 /* See above commentary about NPROC rlimit issues here. */
 489                 free_uid(current);
 490                 current->uid = ruid;
 491                 alloc_uid(current);
 492         }
 493         if (euid != (uid_t) -1) {
 494                 if (euid != current->euid)
 495                         current->dumpable = 0;
 496                 current->euid = euid;
 497                 current->fsuid = euid;
 498         }
 499         if (suid != (uid_t) -1)
 500                 current->suid = suid;
 501
 502         if (!issecure(SECURE_NO_SETUID_FIXUP)) {
 503                 cap_emulate_setxuid(old_ruid, old_euid, old_suid);
 504         }
 505
 506         return 0;
 507 }
 508
 509 asmlinkage long sys_getresuid(uid_t *ruid, uid_t *euid, uid_t *suid)
 510 {
 511         int retval;
 512
 513         if (!(retval = put_user(current->uid, ruid)) &&
 514             !(retval = put_user(current->euid, euid)))
 515                 retval = put_user(current->suid, suid);
 516
 517         return retval;
 518 }
 519
 520 /*
 521  * Same as above, but for rgid, egid, sgid.
 522  */
 523 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 524 {
 525        if (!capable(CAP_SETGID)) {
 526                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
 527                     (rgid != current->egid) && (rgid != current->sgid))
 528                         return -EPERM;
 529                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
 530                     (egid != current->egid) && (egid != current->sgid))
 531                         return -EPERM;
 532                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
 533                     (sgid != current->egid) && (sgid != current->sgid))
 534                         return -EPERM;
 535         }
 536         if (rgid != (gid_t) -1)
 537                 current->gid = rgid;
 538         if (egid != (gid_t) -1) {
 539                 if (egid != current->egid)
 540                         current->dumpable = 0;
 541                 current->egid = egid;
 542                 current->fsgid = egid;
 543         }
 544         if (sgid != (gid_t) -1)
 545                 current->sgid = sgid;
 546         return 0;
 547 }
 548
 549 asmlinkage long sys_getresgid(gid_t *rgid, gid_t *egid, gid_t *sgid)
 550 {
 551         int retval;
 552
 553         if (!(retval = put_user(current->gid, rgid)) &&
 554             !(retval = put_user(current->egid, egid)))
 555                 retval = put_user(current->sgid, sgid);
 556
 557         return retval;
 558 }
 559
 560
 561 /*
 562  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 563  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 564  * whatever uid it wants to). It normally shadows "euid", except when
 565  * explicitly set by setfsuid() or for access..
 566  */
 567 asmlinkage long sys_setfsuid(uid_t uid)
 568 {
 569         int old_fsuid;
 570
 571         old_fsuid = current->fsuid;
 572         if (uid == current->uid || uid == current->euid ||
 573             uid == current->suid || uid == current->fsuid ||
 574             capable(CAP_SETUID))
 575                 current->fsuid = uid;
 576         if (current->fsuid != old_fsuid)
 577                 current->dumpable = 0;
 578
 579         /* We emulate fsuid by essentially doing a scaled-down version
 580          * of what we did in setresuid and friends. However, we only
 581          * operate on the fs-specific bits of the process' effective
 582          * capabilities
 583          *
 584          * FIXME - is fsuser used for all CAP_FS_MASK capabilities?
 585          *          if not, we might be a bit too harsh here.
 586          */
 587
 588         if (!issecure(SECURE_NO_SETUID_FIXUP)) {
 589                 if (old_fsuid == 0 && current->fsuid != 0) {
 590                         cap_t(current->cap_effective) &= ~CAP_FS_MASK;
 591                 }
 592                 if (old_fsuid != 0 && current->fsuid == 0) {
 593                         cap_t(current->cap_effective) |=
 594                                 (cap_t(current->cap_permitted) & CAP_FS_MASK);
 595                 }
 596         }
 597
 598         return old_fsuid;
 599 }
 600
 601 /*
 602  * Samma på svenska..
 603  */
 604 asmlinkage long sys_setfsgid(gid_t gid)
 605 {
 606         int old_fsgid;
 607
 608         old_fsgid = current->fsgid;
 609         if (gid == current->gid || gid == current->egid ||
 610             gid == current->sgid || gid == current->fsgid ||
 611             capable(CAP_SETGID))
 612                 current->fsgid = gid;
 613         if (current->fsgid != old_fsgid)
 614                 current->dumpable = 0;
 615
 616         return old_fsgid;
 617 }
 618
 619 asmlinkage long sys_times(struct tms * tbuf)
 620 {
 621         struct tms temp;
 622
 623         /*
 624          *      In the SMP world we might just be unlucky and have one of
 625          *      the times increment as we use it. Since the value is an
 626          *      atomically safe type this is just fine. Conceptually its
 627          *      as if the syscall took an instant longer to occur.
 628          */
 629         if (tbuf) {
 630                 temp.tms_utime = HZ_TO_STD(current->times.tms_utime);
 631                 temp.tms_stime = HZ_TO_STD(current->times.tms_stime);
 632                 temp.tms_cutime = HZ_TO_STD(current->times.tms_cutime);
 633                 temp.tms_cstime = HZ_TO_STD(current->times.tms_cstime);
 634                 if (copy_to_user(tbuf, &temp, sizeof(struct tms)))
 635                         return -EFAULT;
 636         }
 637         return HZ_TO_STD(jiffies);
 638 }
 639
 640 /*
 641  * This needs some heavy checking ...
 642  * I just haven't the stomach for it. I also don't fully
 643  * understand sessions/pgrp etc. Let somebody who does explain it.
 644  *
 645  * OK, I think I have the protection semantics right.... this is really
 646  * only important on a multi-user system anyway, to make sure one user
 647  * can't send a signal to a process owned by another.  -TYT, 12/12/91
 648  *
 649  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
 650  * LBT 04.03.94
 651  */
 652
 653 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
 654 {
 655         struct task_struct * p;
 656         int err = -EINVAL;
 657
 658         if (!pid)
 659                 pid = current->pid;
 660         if (!pgid)
 661                 pgid = pid;
 662         if (pgid < 0)
 663                 return -EINVAL;
 664
 665         /* From this point forward we keep holding onto the tasklist lock
 666          * so that our parent does not change from under us. -DaveM
 667          */
 668         read_lock(&tasklist_lock);
 669
 670         err = -ESRCH;
 671         p = find_task_by_pid(pid);
 672         if (!p)
 673                 goto out;
 674
 675         if (p->p_pptr == current || p->p_opptr == current) {
 676                 err = -EPERM;
 677                 if (p->session != current->session)
 678                         goto out;
 679                 err = -EACCES;
 680                 if (p->did_exec)
 681                         goto out;
 682         } else if (p != current)
 683                 goto out;
 684         err = -EPERM;
 685         if (p->leader)
 686                 goto out;
 687         if (pgid != pid) {
 688                 struct task_struct * tmp;
 689                 for_each_task (tmp) {
 690                         if (tmp->pgrp == pgid &&
 691                             tmp->session == current->session)
 692                                 goto ok_pgid;
 693                 }
 694                 goto out;
 695         }
 696
 697 ok_pgid:
 698         p->pgrp = pgid;
 699         err = 0;
 700 out:
 701         /* All paths lead to here, thus we are safe. -DaveM */
 702         read_unlock(&tasklist_lock);
 703         return err;
 704 }
 705
 706 asmlinkage long sys_getpgid(pid_t pid)
 707 {
 708         if (!pid) {
 709                 return current->pgrp;
 710         } else {
 711                 int retval;
 712                 struct task_struct *p;
 713
 714                 read_lock(&tasklist_lock);
 715                 p = find_task_by_pid(pid);
 716
 717                 retval = -ESRCH;
 718                 if (p)
 719                         retval = p->pgrp;
 720                 read_unlock(&tasklist_lock);
 721                 return retval;
 722         }
 723 }
 724
 725 asmlinkage long sys_getpgrp(void)
 726 {
 727         /* SMP - assuming writes are word atomic this is fine */
 728         return current->pgrp;
 729 }
 730
 731 asmlinkage long sys_getsid(pid_t pid)
 732 {
 733         if (!pid) {
 734                 return current->session;
 735         } else {
 736                 int retval;
 737                 struct task_struct *p;
 738
 739                 read_lock(&tasklist_lock);
 740                 p = find_task_by_pid(pid);
 741
 742                 retval = -ESRCH;
 743                 if(p)
 744                         retval = p->session;
 745                 read_unlock(&tasklist_lock);
 746                 return retval;
 747         }
 748 }
 749
 750 asmlinkage long sys_setsid(void)
 751 {
 752         struct task_struct * p;
 753         int err = -EPERM;
 754
 755         read_lock(&tasklist_lock);
 756         for_each_task(p) {
 757                 if (p->pgrp == current->pid)
 758                         goto out;
 759         }
 760
 761         current->leader = 1;
 762         current->session = current->pgrp = current->pid;
 763         current->tty = NULL;
 764         current->tty_old_pgrp = 0;
 765         err = current->pgrp;
 766 out:
 767         read_unlock(&tasklist_lock);
 768         return err;
 769 }
 770
 771 /*
 772  * Supplementary group IDs
 773  */
 774 asmlinkage long sys_getgroups(int gidsetsize, gid_t *grouplist)
 775 {
 776         int i;
 777
 778         /*
 779          *      SMP: Nobody else can change our grouplist. Thus we are
 780          *      safe.
 781          */
 782
 783         if (gidsetsize < 0)
 784                 return -EINVAL;
 785         i = current->ngroups;
 786         if (gidsetsize) {
 787                 if (i > gidsetsize)
 788                         return -EINVAL;
 789                 if (copy_to_user(grouplist, current->groups, sizeof(gid_t)*i))
 790                         return -EFAULT;
 791         }
 792         return i;
 793 }
 794
 795 /*
 796  *      SMP: Our groups are not shared. We can copy to/from them safely
 797  *      without another task interfering.
 798  */
 799
 800 asmlinkage long sys_setgroups(int gidsetsize, gid_t *grouplist)
 801 {
 802         if (!capable(CAP_SETGID))
 803                 return -EPERM;
 804         if ((unsigned) gidsetsize > NGROUPS)
 805                 return -EINVAL;
 806         if(copy_from_user(current->groups, grouplist, gidsetsize * sizeof(gid_t)))
 807                 return -EFAULT;
 808         current->ngroups = gidsetsize;
 809         return 0;
 810 }
 811
 812 int in_group_p(gid_t grp)
 813 {
 814         if (grp != current->fsgid) {
 815                 int i = current->ngroups;
 816                 if (i) {
 817                         gid_t *groups = current->groups;
 818                         do {
 819                                 if (*groups == grp)
 820                                         goto out;
 821                                 groups++;
 822                                 i--;
 823                         } while (i);
 824                 }
 825                 return 0;
 826         }
 827 out:
 828         return 1;
 829 }
 830
 831 /*
 832  * This should really be a blocking read-write lock
 833  * rather than a semaphore. Anybody want to implement
 834  * one?
 835  */
 836 DECLARE_MUTEX(uts_sem);
 837
 838 asmlinkage long sys_newuname(struct new_utsname * name)
 839 {
 840         int errno = 0;
 841
 842         down(&uts_sem);
 843         if (copy_to_user(name,&system_utsname,sizeof *name))
 844                 errno = -EFAULT;
 845         up(&uts_sem);
 846         return errno;
 847 }
 848
 849 asmlinkage long sys_sethostname(char *name, int len)
 850 {
 851         int errno;
 852
 853         if (!capable(CAP_SYS_ADMIN))
 854                 return -EPERM;
 855         if (len < 0 || len > __NEW_UTS_LEN)
 856                 return -EINVAL;
 857         down(&uts_sem);
 858         errno = -EFAULT;
 859         if (!copy_from_user(system_utsname.nodename, name, len)) {
 860                 system_utsname.nodename[len] = 0;
 861                 errno = 0;
 862         }
 863         up(&uts_sem);
 864         return errno;
 865 }
 866
 867 asmlinkage long sys_gethostname(char *name, int len)
 868 {
 869         int i, errno;
 870
 871         if (len < 0)
 872                 return -EINVAL;
 873         down(&uts_sem);
 874         i = 1 + strlen(system_utsname.nodename);
 875         if (i > len)
 876                 i = len;
 877         errno = 0;
 878         if (copy_to_user(name, system_utsname.nodename, i))
 879                 errno = -EFAULT;
 880         up(&uts_sem);
 881         return errno;
 882 }
 883
 884 /*
 885  * Only setdomainname; getdomainname can be implemented by calling
 886  * uname()
 887  */
 888 asmlinkage long sys_setdomainname(char *name, int len)
 889 {
 890         int errno;
 891
 892         if (!capable(CAP_SYS_ADMIN))
 893                 return -EPERM;
 894         if (len < 0 || len > __NEW_UTS_LEN)
 895                 return -EINVAL;
 896
 897         down(&uts_sem);
 898         errno = -EFAULT;
 899         if (!copy_from_user(system_utsname.domainname, name, len)) {
 900                 errno = 0;
 901                 system_utsname.domainname[len] = 0;
 902         }
 903         up(&uts_sem);
 904         return errno;
 905 }
 906
 907 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit *rlim)
 908 {
 909         if (resource >= RLIM_NLIMITS)
 910                 return -EINVAL;
 911         else
 912                 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
 913                         ? -EFAULT : 0;
 914 }
 915
 916 /*
 917  *      Back compatibility for getrlimit. Needed for some apps.
 918  */
 919
 920 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit *rlim)
 921 {
 922         struct rlimit x;
 923         if (resource >= RLIM_NLIMITS)
 924                 return -EINVAL;
 925
 926         memcpy(&x, current->rlim + resource, sizeof(*rlim));
 927         if(x.rlim_cur > 0x7FFFFFFF)
 928                 x.rlim_cur = 0x7FFFFFFF;
 929         if(x.rlim_max > 0x7FFFFFFF)
 930                 x.rlim_max = 0x7FFFFFFF;
 931         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
 932 }
 933
 934
 935 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit *rlim)
 936 {
 937         struct rlimit new_rlim, *old_rlim;
 938
 939         if (resource >= RLIM_NLIMITS)
 940                 return -EINVAL;
 941         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
 942                 return -EFAULT;
 943         if (new_rlim.rlim_cur < 0 || new_rlim.rlim_max < 0)
 944                 return -EINVAL;
 945         old_rlim = current->rlim + resource;
 946         if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
 947              (new_rlim.rlim_max > old_rlim->rlim_max)) &&
 948             !capable(CAP_SYS_RESOURCE))
 949                 return -EPERM;
 950         if (resource == RLIMIT_NOFILE) {
 951                 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
 952                         return -EPERM;
 953         }
 954         *old_rlim = new_rlim;
 955         return 0;
 956 }
 957
 958 /*
 959  * It would make sense to put struct rusage in the task_struct,
 960  * except that would make the task_struct be *really big*.  After
 961  * task_struct gets moved into malloc'ed memory, it would
 962  * make sense to do this.  It will make moving the rest of the information
 963  * a lot simpler!  (Which we're not doing right now because we're not
 964  * measuring them yet).
 965  *
 966  * This is SMP safe.  Either we are called from sys_getrusage on ourselves
 967  * below (we know we aren't going to exit/disappear and only we change our
 968  * rusage counters), or we are called from wait4() on a process which is
 969  * either stopped or zombied.  In the zombied case the task won't get
 970  * reaped till shortly after the call to getrusage(), in both cases the
 971  * task being examined is in a frozen state so the counters won't change.
 972  *
 973  * FIXME! Get the fault counts properly!
 974  */
 975 int getrusage(struct task_struct *p, int who, struct rusage *ru)
 976 {
 977         struct rusage r;
 978
 979         memset((char *) &r, 0, sizeof(r));
 980         switch (who) {
 981                 case RUSAGE_SELF:
 982                         r.ru_utime.tv_sec = CT_TO_SECS(p->times.tms_utime);
 983                         r.ru_utime.tv_usec = CT_TO_USECS(p->times.tms_utime);
 984                         r.ru_stime.tv_sec = CT_TO_SECS(p->times.tms_stime);
 985                         r.ru_stime.tv_usec = CT_TO_USECS(p->times.tms_stime);
 986                         r.ru_minflt = p->min_flt;
 987                         r.ru_majflt = p->maj_flt;
 988                         r.ru_nswap = p->nswap;
 989                         break;
 990                 case RUSAGE_CHILDREN:
 991                         r.ru_utime.tv_sec = CT_TO_SECS(p->times.tms_cutime);
 992                         r.ru_utime.tv_usec = CT_TO_USECS(p->times.tms_cutime);
 993                         r.ru_stime.tv_sec = CT_TO_SECS(p->times.tms_cstime);
 994                         r.ru_stime.tv_usec = CT_TO_USECS(p->times.tms_cstime);
 995                         r.ru_minflt = p->cmin_flt;
 996                         r.ru_majflt = p->cmaj_flt;
 997                         r.ru_nswap = p->cnswap;
 998                         break;
 999                 default:
1000                         r.ru_utime.tv_sec = CT_TO_SECS(p->times.tms_utime + p->times.tms_cutime);
1001                         r.ru_utime.tv_usec = CT_TO_USECS(p->times.tms_utime + p->times.tms_cutime);
1002                         r.ru_stime.tv_sec = CT_TO_SECS(p->times.tms_stime + p->times.tms_cstime);
1003                         r.ru_stime.tv_usec = CT_TO_USECS(p->times.tms_stime + p->times.tms_cstime);
1004                         r.ru_minflt = p->min_flt + p->cmin_flt;
1005                         r.ru_majflt = p->maj_flt + p->cmaj_flt;
1006                         r.ru_nswap = p->nswap + p->cnswap;
1007                         break;
1008         }
1009         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1010 }
1011
1012 asmlinkage long sys_getrusage(int who, struct rusage *ru)
1013 {
1014         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1015                 return -EINVAL;
1016         return getrusage(current, who, ru);
1017 }
1018
1019 asmlinkage long sys_umask(int mask)
1020 {
1021         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1022         return mask;
1023 }
1024
1025 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1026                           unsigned long arg4, unsigned long arg5)
1027 {
1028         int error = 0;
1029         int sig;
1030
1031         switch (option) {
1032                 case PR_SET_PDEATHSIG:
1033                         sig = arg2;
1034                         if (sig > _NSIG) {
1035                                 error = -EINVAL;
1036                                 break;
1037                         }
1038                         current->pdeath_signal = sig;
1039                         break;
1040                 case PR_GET_PDEATHSIG:
1041                         error = put_user(current->pdeath_signal, (int *)arg2);
1042                         break;
1043                 case PR_GET_DUMPABLE:
1044                         if (current->dumpable)
1045                                 error = 1;
1046                         break;
1047                 case PR_SET_DUMPABLE:
1048                         if (arg2 != 0 && arg2 != 1) {
1049                                 error = -EINVAL;
1050                                 break;
1051                         }
1052                         current->dumpable = arg2;
1053                         break;
1054                 default:
1055                         error = -EINVAL;
1056                         break;
1057         }
1058         return error;
1059 }
1060