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         }
 230         unlock_kernel();
 231         return 0;
 232 }
 233
 234 /*
 235  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 236  * As it's called within an interrupt, it may NOT sync: the only choice
 237  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 238  */
 239 void ctrl_alt_del(void)
 240 {
 241         if (C_A_D) {
 242                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 243                 machine_restart(NULL);
 244         } else
 245                 kill_proc(1, SIGINT, 1);
 246 }
 247
 248
 249 /*
 250  * Unprivileged users may change the real gid to the effective gid
 251  * or vice versa.  (BSD-style)
 252  *
 253  * If you set the real gid at all, or set the effective gid to a value not
 254  * equal to the real gid, then the saved gid is set to the new effective gid.
 255  *
 256  * This makes it possible for a setgid program to completely drop its
 257  * privileges, which is often a useful assertion to make when you are doing
 258  * a security audit over a program.
 259  *
 260  * The general idea is that a program which uses just setregid() will be
 261  * 100% compatible with BSD.  A program which uses just setgid() will be
 262  * 100% compatible with POSIX with saved IDs.
 263  *
 264  * SMP: There are not races, the GIDs are checked only by filesystem
 265  *      operations (as far as semantic preservation is concerned).
 266  */
 267 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 268 {
 269         int old_rgid = current->gid;
 270         int old_egid = current->egid;
 271
 272         if (rgid != (gid_t) -1) {
 273                 if ((old_rgid == rgid) ||
 274                     (current->egid==rgid) ||
 275                     capable(CAP_SETGID))
 276                         current->gid = rgid;
 277                 else
 278                         return -EPERM;
 279         }
 280         if (egid != (gid_t) -1) {
 281                 if ((old_rgid == egid) ||
 282                     (current->egid == egid) ||
 283                     (current->sgid == egid) ||
 284                     capable(CAP_SETGID))
 285                         current->fsgid = current->egid = egid;
 286                 else {
 287                         current->gid = old_rgid;
 288                         return -EPERM;
 289                 }
 290         }
 291         if (rgid != (gid_t) -1 ||
 292             (egid != (gid_t) -1 && egid != old_rgid))
 293                 current->sgid = current->egid;
 294         current->fsgid = current->egid;
 295         if (current->egid != old_egid)
 296                 current->dumpable = 0;
 297         return 0;
 298 }
 299
 300 /*
 301  * setgid() is implemented like SysV w/ SAVED_IDS
 302  *
 303  * SMP: Same implicit races as above.
 304  */
 305 asmlinkage long sys_setgid(gid_t gid)
 306 {
 307         int old_egid = current->egid;
 308
 309         if (capable(CAP_SETGID))
 310                 current->gid = current->egid = current->sgid = current->fsgid = gid;
 311         else if ((gid == current->gid) || (gid == current->sgid))
 312                 current->egid = current->fsgid = gid;
 313         else
 314                 return -EPERM;
 315
 316         if (current->egid != old_egid)
 317                 current->dumpable = 0;
 318         return 0;
 319 }
 320
 321 /*
 322  * cap_emulate_setxuid() fixes the effective / permitted capabilities of
 323  * a process after a call to setuid, setreuid, or setresuid.
 324  *
 325  *  1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of
 326  *  {r,e,s}uid != 0, the permitted and effective capabilities are
 327  *  cleared.
 328  *
 329  *  2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective
 330  *  capabilities of the process are cleared.
 331  *
 332  *  3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective
 333  *  capabilities are set to the permitted capabilities.
 334  *
 335  *  fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should
 336  *  never happen.
 337  *
 338  *  -astor
 339  */
 340 extern inline void cap_emulate_setxuid(int old_ruid, int old_euid,
 341                                        int old_suid)
 342 {
 343         if ((old_ruid == 0 || old_euid == 0 || old_suid == 0) &&
 344             (current->uid != 0 && current->euid != 0 && current->suid != 0)) {
 345                 cap_clear(current->cap_permitted);
 346                 cap_clear(current->cap_effective);
 347         }
 348         if (old_euid == 0 && current->euid != 0) {
 349                 cap_clear(current->cap_effective);
 350         }
 351         if (old_euid != 0 && current->euid == 0) {
 352                 current->cap_effective = current->cap_permitted;
 353         }
 354 }
 355
 356 /*
 357  * Unprivileged users may change the real uid to the effective uid
 358  * or vice versa.  (BSD-style)
 359  *
 360  * If you set the real uid at all, or set the effective uid to a value not
 361  * equal to the real uid, then the saved uid is set to the new effective uid.
 362  *
 363  * This makes it possible for a setuid program to completely drop its
 364  * privileges, which is often a useful assertion to make when you are doing
 365  * a security audit over a program.
 366  *
 367  * The general idea is that a program which uses just setreuid() will be
 368  * 100% compatible with BSD.  A program which uses just setuid() will be
 369  * 100% compatible with POSIX with saved IDs.
 370  */
 371 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 372 {
 373         int old_ruid, old_euid, old_suid, new_ruid;
 374
 375         new_ruid = old_ruid = current->uid;
 376         old_euid = current->euid;
 377         old_suid = current->suid;
 378         if (ruid != (uid_t) -1) {
 379                 if ((old_ruid == ruid) ||
 380                     (current->euid==ruid) ||
 381                     capable(CAP_SETUID))
 382                         new_ruid = ruid;
 383                 else
 384                         return -EPERM;
 385         }
 386         if (euid != (uid_t) -1) {
 387                 if ((old_ruid == euid) ||
 388                     (current->euid == euid) ||
 389                     (current->suid == euid) ||
 390                     capable(CAP_SETUID))
 391                         current->fsuid = current->euid = euid;
 392                 else
 393                         return -EPERM;
 394         }
 395         if (ruid != (uid_t) -1 ||
 396             (euid != (uid_t) -1 && euid != old_ruid))
 397                 current->suid = current->euid;
 398         current->fsuid = current->euid;
 399         if (current->euid != old_euid)
 400                 current->dumpable = 0;
 401
 402         if(new_ruid != old_ruid) {
 403                 /* What if a process setreuid()'s and this brings the
 404                  * new uid over his NPROC rlimit?  We can check this now
 405                  * cheaply with the new uid cache, so if it matters
 406                  * we should be checking for it.  -DaveM
 407                  */
 408                 free_uid(current);
 409                 current->uid = new_ruid;
 410                 alloc_uid(current);
 411         }
 412
 413         if (!issecure(SECURE_NO_SETUID_FIXUP)) {
 414                 cap_emulate_setxuid(old_ruid, old_euid, old_suid);
 415         }
 416
 417         return 0;
 418 }
 419
 420
 421
 422 /*
 423  * setuid() is implemented like SysV with SAVED_IDS
 424  *
 425  * Note that SAVED_ID's is deficient in that a setuid root program
 426  * like sendmail, for example, cannot set its uid to be a normal
 427  * user and then switch back, because if you're root, setuid() sets
 428  * the saved uid too.  If you don't like this, blame the bright people
 429  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 430  * will allow a root program to temporarily drop privileges and be able to
 431  * regain them by swapping the real and effective uid.
 432  */
 433 asmlinkage long sys_setuid(uid_t uid)
 434 {
 435         int old_euid = current->euid;
 436         int old_ruid, old_suid, new_ruid;
 437
 438         old_ruid = new_ruid = current->uid;
 439         old_suid = current->suid;
 440         if (capable(CAP_SETUID))
 441                 new_ruid = current->euid = current->suid = current->fsuid = uid;
 442         else if ((uid == current->uid) || (uid == current->suid))
 443                 current->fsuid = current->euid = uid;
 444         else
 445                 return -EPERM;
 446
 447         if (current->euid != old_euid)
 448                 current->dumpable = 0;
 449
 450        if (new_ruid != old_ruid) {
 451                 /* See comment above about NPROC rlimit issues... */
 452                 free_uid(current);
 453                 current->uid = new_ruid;
 454                 alloc_uid(current);
 455         }
 456
 457         if (!issecure(SECURE_NO_SETUID_FIXUP)) {
 458                 cap_emulate_setxuid(old_ruid, old_euid, old_suid);
 459         }
 460
 461         return 0;
 462 }
 463
 464
 465 /*
 466  * This function implements a generic ability to update ruid, euid,
 467  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 468  */
 469 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 470 {
 471         int old_ruid = current->uid;
 472         int old_euid = current->euid;
 473         int old_suid = current->suid;
 474
 475         if (!capable(CAP_SETUID)) {
 476                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
 477                     (ruid != current->euid) && (ruid != current->suid))
 478                         return -EPERM;
 479                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
 480                     (euid != current->euid) && (euid != current->suid))
 481                         return -EPERM;
 482                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
 483                     (suid != current->euid) && (suid != current->suid))
 484                         return -EPERM;
 485         }
 486         if (ruid != (uid_t) -1) {
 487                 /* See above commentary about NPROC rlimit issues here. */
 488                 free_uid(current);
 489                 current->uid = ruid;
 490                 alloc_uid(current);
 491         }
 492         if (euid != (uid_t) -1) {
 493                 if (euid != current->euid)
 494                         current->dumpable = 0;
 495                 current->euid = euid;
 496                 current->fsuid = euid;
 497         }
 498         if (suid != (uid_t) -1)
 499                 current->suid = suid;
 500
 501         if (!issecure(SECURE_NO_SETUID_FIXUP)) {
 502                 cap_emulate_setxuid(old_ruid, old_euid, old_suid);
 503         }
 504
 505         return 0;
 506 }
 507
 508 asmlinkage long sys_getresuid(uid_t *ruid, uid_t *euid, uid_t *suid)
 509 {
 510         int retval;
 511
 512         if (!(retval = put_user(current->uid, ruid)) &&
 513             !(retval = put_user(current->euid, euid)))
 514                 retval = put_user(current->suid, suid);
 515
 516         return retval;
 517 }
 518
 519 /*
 520  * Same as above, but for rgid, egid, sgid.
 521  */
 522 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 523 {
 524        if (!capable(CAP_SETGID)) {
 525                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
 526                     (rgid != current->egid) && (rgid != current->sgid))
 527                         return -EPERM;
 528                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
 529                     (egid != current->egid) && (egid != current->sgid))
 530                         return -EPERM;
 531                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
 532                     (sgid != current->egid) && (sgid != current->sgid))
 533                         return -EPERM;
 534         }
 535         if (rgid != (gid_t) -1)
 536                 current->gid = rgid;
 537         if (egid != (gid_t) -1) {
 538                 if (egid != current->egid)
 539                         current->dumpable = 0;
 540                 current->egid = egid;
 541                 current->fsgid = egid;
 542         }
 543         if (sgid != (gid_t) -1)
 544                 current->sgid = sgid;
 545         return 0;
 546 }
 547
 548 asmlinkage long sys_getresgid(gid_t *rgid, gid_t *egid, gid_t *sgid)
 549 {
 550         int retval;
 551
 552         if (!(retval = put_user(current->gid, rgid)) &&
 553             !(retval = put_user(current->egid, egid)))
 554                 retval = put_user(current->sgid, sgid);
 555
 556         return retval;
 557 }
 558
 559
 560 /*
 561  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 562  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 563  * whatever uid it wants to). It normally shadows "euid", except when
 564  * explicitly set by setfsuid() or for access..
 565  */
 566 asmlinkage long sys_setfsuid(uid_t uid)
 567 {
 568         int old_fsuid;
 569
 570         old_fsuid = current->fsuid;
 571         if (uid == current->uid || uid == current->euid ||
 572             uid == current->suid || uid == current->fsuid ||
 573             capable(CAP_SETUID))
 574                 current->fsuid = uid;
 575         if (current->fsuid != old_fsuid)
 576                 current->dumpable = 0;
 577
 578         /* We emulate fsuid by essentially doing a scaled-down version
 579          * of what we did in setresuid and friends. However, we only
 580          * operate on the fs-specific bits of the process' effective
 581          * capabilities
 582          *
 583          * FIXME - is fsuser used for all CAP_FS_MASK capabilities?
 584          *          if not, we might be a bit too harsh here.
 585          */
 586
 587         if (!issecure(SECURE_NO_SETUID_FIXUP)) {
 588                 if (old_fsuid == 0 && current->fsuid != 0) {
 589                         cap_t(current->cap_effective) &= ~CAP_FS_MASK;
 590                 }
 591                 if (old_fsuid != 0 && current->fsuid == 0) {
 592                         cap_t(current->cap_effective) |=
 593                                 (cap_t(current->cap_permitted) & CAP_FS_MASK);
 594                 }
 595         }
 596
 597         return old_fsuid;
 598 }
 599
 600 /*
 601  * Samma på svenska..
 602  */
 603 asmlinkage long sys_setfsgid(gid_t gid)
 604 {
 605         int old_fsgid;
 606
 607         old_fsgid = current->fsgid;
 608         if (gid == current->gid || gid == current->egid ||
 609             gid == current->sgid || gid == current->fsgid ||
 610             capable(CAP_SETGID))
 611                 current->fsgid = gid;
 612         if (current->fsgid != old_fsgid)
 613                 current->dumpable = 0;
 614
 615         return old_fsgid;
 616 }
 617
 618 asmlinkage long sys_times(struct tms * tbuf)
 619 {
 620         struct tms temp;
 621
 622         /*
 623          *      In the SMP world we might just be unlucky and have one of
 624          *      the times increment as we use it. Since the value is an
 625          *      atomically safe type this is just fine. Conceptually its
 626          *      as if the syscall took an instant longer to occur.
 627          */
 628         if (tbuf) {
 629                 temp.tms_utime = HZ_TO_STD(current->times.tms_utime);
 630                 temp.tms_stime = HZ_TO_STD(current->times.tms_stime);
 631                 temp.tms_cutime = HZ_TO_STD(current->times.tms_cutime);
 632                 temp.tms_cstime = HZ_TO_STD(current->times.tms_cstime);
 633                 if (copy_to_user(tbuf, &temp, sizeof(struct tms)))
 634                         return -EFAULT;
 635         }
 636         return HZ_TO_STD(jiffies);
 637 }
 638
 639 /*
 640  * This needs some heavy checking ...
 641  * I just haven't the stomach for it. I also don't fully
 642  * understand sessions/pgrp etc. Let somebody who does explain it.
 643  *
 644  * OK, I think I have the protection semantics right.... this is really
 645  * only important on a multi-user system anyway, to make sure one user
 646  * can't send a signal to a process owned by another.  -TYT, 12/12/91
 647  *
 648  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
 649  * LBT 04.03.94
 650  */
 651
 652 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
 653 {
 654         struct task_struct * p;
 655         int err = -EINVAL;
 656
 657         if (!pid)
 658                 pid = current->pid;
 659         if (!pgid)
 660                 pgid = pid;
 661         if (pgid < 0)
 662                 return -EINVAL;
 663
 664         /* From this point forward we keep holding onto the tasklist lock
 665          * so that our parent does not change from under us. -DaveM
 666          */
 667         read_lock(&tasklist_lock);
 668
 669         err = -ESRCH;
 670         p = find_task_by_pid(pid);
 671         if (!p)
 672                 goto out;
 673
 674         if (p->p_pptr == current || p->p_opptr == current) {
 675                 err = -EPERM;
 676                 if (p->session != current->session)
 677                         goto out;
 678                 err = -EACCES;
 679                 if (p->did_exec)
 680                         goto out;
 681         } else if (p != current)
 682                 goto out;
 683         err = -EPERM;
 684         if (p->leader)
 685                 goto out;
 686         if (pgid != pid) {
 687                 struct task_struct * tmp;
 688                 for_each_task (tmp) {
 689                         if (tmp->pgrp == pgid &&
 690                             tmp->session == current->session)
 691                                 goto ok_pgid;
 692                 }
 693                 goto out;
 694         }
 695
 696 ok_pgid:
 697         p->pgrp = pgid;
 698         err = 0;
 699 out:
 700         /* All paths lead to here, thus we are safe. -DaveM */
 701         read_unlock(&tasklist_lock);
 702         return err;
 703 }
 704
 705 asmlinkage long sys_getpgid(pid_t pid)
 706 {
 707         if (!pid) {
 708                 return current->pgrp;
 709         } else {
 710                 int retval;
 711                 struct task_struct *p;
 712
 713                 read_lock(&tasklist_lock);
 714                 p = find_task_by_pid(pid);
 715
 716                 retval = -ESRCH;
 717                 if (p)
 718                         retval = p->pgrp;
 719                 read_unlock(&tasklist_lock);
 720                 return retval;
 721         }
 722 }
 723
 724 asmlinkage long sys_getpgrp(void)
 725 {
 726         /* SMP - assuming writes are word atomic this is fine */
 727         return current->pgrp;
 728 }
 729
 730 asmlinkage long sys_getsid(pid_t pid)
 731 {
 732         if (!pid) {
 733                 return current->session;
 734         } else {
 735                 int retval;
 736                 struct task_struct *p;
 737
 738                 read_lock(&tasklist_lock);
 739                 p = find_task_by_pid(pid);
 740
 741                 retval = -ESRCH;
 742                 if(p)
 743                         retval = p->session;
 744                 read_unlock(&tasklist_lock);
 745                 return retval;
 746         }
 747 }
 748
 749 asmlinkage long sys_setsid(void)
 750 {
 751         struct task_struct * p;
 752         int err = -EPERM;
 753
 754         read_lock(&tasklist_lock);
 755         for_each_task(p) {
 756                 if (p->pgrp == current->pid)
 757                         goto out;
 758         }
 759
 760         current->leader = 1;
 761         current->session = current->pgrp = current->pid;
 762         current->tty = NULL;
 763         current->tty_old_pgrp = 0;
 764         err = current->pgrp;
 765 out:
 766         read_unlock(&tasklist_lock);
 767         return err;
 768 }
 769
 770 /*
 771  * Supplementary group IDs
 772  */
 773 asmlinkage long sys_getgroups(int gidsetsize, gid_t *grouplist)
 774 {
 775         int i;
 776
 777         /*
 778          *      SMP: Nobody else can change our grouplist. Thus we are
 779          *      safe.
 780          */
 781
 782         if (gidsetsize < 0)
 783                 return -EINVAL;
 784         i = current->ngroups;
 785         if (gidsetsize) {
 786                 if (i > gidsetsize)
 787                         return -EINVAL;
 788                 if (copy_to_user(grouplist, current->groups, sizeof(gid_t)*i))
 789                         return -EFAULT;
 790         }
 791         return i;
 792 }
 793
 794 /*
 795  *      SMP: Our groups are not shared. We can copy to/from them safely
 796  *      without another task interfering.
 797  */
 798
 799 asmlinkage long sys_setgroups(int gidsetsize, gid_t *grouplist)
 800 {
 801         if (!capable(CAP_SETGID))
 802                 return -EPERM;
 803         if ((unsigned) gidsetsize > NGROUPS)
 804                 return -EINVAL;
 805         if(copy_from_user(current->groups, grouplist, gidsetsize * sizeof(gid_t)))
 806                 return -EFAULT;
 807         current->ngroups = gidsetsize;
 808         return 0;
 809 }
 810
 811 int in_group_p(gid_t grp)
 812 {
 813         if (grp != current->fsgid) {
 814                 int i = current->ngroups;
 815                 if (i) {
 816                         gid_t *groups = current->groups;
 817                         do {
 818                                 if (*groups == grp)
 819                                         goto out;
 820                                 groups++;
 821                                 i--;
 822                         } while (i);
 823                 }
 824                 return 0;
 825         }
 826 out:
 827         return 1;
 828 }
 829
 830 DECLARE_RWSEM(uts_sem);
 831
 832 asmlinkage long sys_newuname(struct new_utsname * name)
 833 {
 834         int errno = 0;
 835
 836         down_read(&uts_sem);
 837         if (copy_to_user(name,&system_utsname,sizeof *name))
 838                 errno = -EFAULT;
 839         up_read(&uts_sem);
 840         return errno;
 841 }
 842
 843 asmlinkage long sys_sethostname(char *name, int len)
 844 {
 845         int errno;
 846
 847         if (!capable(CAP_SYS_ADMIN))
 848                 return -EPERM;
 849         if (len < 0 || len > __NEW_UTS_LEN)
 850                 return -EINVAL;
 851         down_write(&uts_sem);
 852         errno = -EFAULT;
 853         if (!copy_from_user(system_utsname.nodename, name, len)) {
 854                 system_utsname.nodename[len] = 0;
 855                 errno = 0;
 856         }
 857         up_write(&uts_sem);
 858         return errno;
 859 }
 860
 861 asmlinkage long sys_gethostname(char *name, int len)
 862 {
 863         int i, errno;
 864
 865         if (len < 0)
 866                 return -EINVAL;
 867         down_read(&uts_sem);
 868         i = 1 + strlen(system_utsname.nodename);
 869         if (i > len)
 870                 i = len;
 871         errno = 0;
 872         if (copy_to_user(name, system_utsname.nodename, i))
 873                 errno = -EFAULT;
 874         up_read(&uts_sem);
 875         return errno;
 876 }
 877
 878 /*
 879  * Only setdomainname; getdomainname can be implemented by calling
 880  * uname()
 881  */
 882 asmlinkage long sys_setdomainname(char *name, int len)
 883 {
 884         int errno;
 885
 886         if (!capable(CAP_SYS_ADMIN))
 887                 return -EPERM;
 888         if (len < 0 || len > __NEW_UTS_LEN)
 889                 return -EINVAL;
 890
 891         down_write(&uts_sem);
 892         errno = -EFAULT;
 893         if (!copy_from_user(system_utsname.domainname, name, len)) {
 894                 errno = 0;
 895                 system_utsname.domainname[len] = 0;
 896         }
 897         up_write(&uts_sem);
 898         return errno;
 899 }
 900
 901 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit *rlim)
 902 {
 903         if (resource >= RLIM_NLIMITS)
 904                 return -EINVAL;
 905         else
 906                 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
 907                         ? -EFAULT : 0;
 908 }
 909
 910 #if !defined(__ia64__)
 911
 912 /*
 913  *      Back compatibility for getrlimit. Needed for some apps.
 914  */
 915
 916 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit *rlim)
 917 {
 918         struct rlimit x;
 919         if (resource >= RLIM_NLIMITS)
 920                 return -EINVAL;
 921
 922         memcpy(&x, current->rlim + resource, sizeof(*rlim));
 923         if(x.rlim_cur > 0x7FFFFFFF)
 924                 x.rlim_cur = 0x7FFFFFFF;
 925         if(x.rlim_max > 0x7FFFFFFF)
 926                 x.rlim_max = 0x7FFFFFFF;
 927         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
 928 }
 929
 930 #endif
 931
 932 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit *rlim)
 933 {
 934         struct rlimit new_rlim, *old_rlim;
 935
 936         if (resource >= RLIM_NLIMITS)
 937                 return -EINVAL;
 938         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
 939                 return -EFAULT;
 940         if (new_rlim.rlim_cur < 0 || new_rlim.rlim_max < 0)
 941                 return -EINVAL;
 942         old_rlim = current->rlim + resource;
 943         if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
 944              (new_rlim.rlim_max > old_rlim->rlim_max)) &&
 945             !capable(CAP_SYS_RESOURCE))
 946                 return -EPERM;
 947         if (resource == RLIMIT_NOFILE) {
 948                 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
 949                         return -EPERM;
 950         }
 951         *old_rlim = new_rlim;
 952         return 0;
 953 }
 954
 955 /*
 956  * It would make sense to put struct rusage in the task_struct,
 957  * except that would make the task_struct be *really big*.  After
 958  * task_struct gets moved into malloc'ed memory, it would
 959  * make sense to do this.  It will make moving the rest of the information
 960  * a lot simpler!  (Which we're not doing right now because we're not
 961  * measuring them yet).
 962  *
 963  * This is SMP safe.  Either we are called from sys_getrusage on ourselves
 964  * below (we know we aren't going to exit/disappear and only we change our
 965  * rusage counters), or we are called from wait4() on a process which is
 966  * either stopped or zombied.  In the zombied case the task won't get
 967  * reaped till shortly after the call to getrusage(), in both cases the
 968  * task being examined is in a frozen state so the counters won't change.
 969  *
 970  * FIXME! Get the fault counts properly!
 971  */
 972 int getrusage(struct task_struct *p, int who, struct rusage *ru)
 973 {
 974         struct rusage r;
 975
 976         memset((char *) &r, 0, sizeof(r));
 977         switch (who) {
 978                 case RUSAGE_SELF:
 979                         r.ru_utime.tv_sec = CT_TO_SECS(p->times.tms_utime);
 980                         r.ru_utime.tv_usec = CT_TO_USECS(p->times.tms_utime);
 981                         r.ru_stime.tv_sec = CT_TO_SECS(p->times.tms_stime);
 982                         r.ru_stime.tv_usec = CT_TO_USECS(p->times.tms_stime);
 983                         r.ru_minflt = p->min_flt;
 984                         r.ru_majflt = p->maj_flt;
 985                         r.ru_nswap = p->nswap;
 986                         break;
 987                 case RUSAGE_CHILDREN:
 988                         r.ru_utime.tv_sec = CT_TO_SECS(p->times.tms_cutime);
 989                         r.ru_utime.tv_usec = CT_TO_USECS(p->times.tms_cutime);
 990                         r.ru_stime.tv_sec = CT_TO_SECS(p->times.tms_cstime);
 991                         r.ru_stime.tv_usec = CT_TO_USECS(p->times.tms_cstime);
 992                         r.ru_minflt = p->cmin_flt;
 993                         r.ru_majflt = p->cmaj_flt;
 994                         r.ru_nswap = p->cnswap;
 995                         break;
 996                 default:
 997                         r.ru_utime.tv_sec = CT_TO_SECS(p->times.tms_utime + p->times.tms_cutime);
 998                         r.ru_utime.tv_usec = CT_TO_USECS(p->times.tms_utime + p->times.tms_cutime);
 999                         r.ru_stime.tv_sec = CT_TO_SECS(p->times.tms_stime + p->times.tms_cstime);
1000                         r.ru_stime.tv_usec = CT_TO_USECS(p->times.tms_stime + p->times.tms_cstime);
1001                         r.ru_minflt = p->min_flt + p->cmin_flt;
1002                         r.ru_majflt = p->maj_flt + p->cmaj_flt;
1003                         r.ru_nswap = p->nswap + p->cnswap;
1004                         break;
1005         }
1006         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1007 }
1008
1009 asmlinkage long sys_getrusage(int who, struct rusage *ru)
1010 {
1011         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1012                 return -EINVAL;
1013         return getrusage(current, who, ru);
1014 }
1015
1016 asmlinkage long sys_umask(int mask)
1017 {
1018         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1019         return mask;
1020 }
1021
1022 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1023                           unsigned long arg4, unsigned long arg5)
1024 {
1025         int error = 0;
1026         int sig;
1027
1028         switch (option) {
1029                 case PR_SET_PDEATHSIG:
1030                         sig = arg2;
1031                         if (sig > _NSIG) {
1032                                 error = -EINVAL;
1033                                 break;
1034                         }
1035                         current->pdeath_signal = sig;
1036                         break;
1037                 case PR_GET_PDEATHSIG:
1038                         error = put_user(current->pdeath_signal, (int *)arg2);
1039                         break;
1040                 case PR_GET_DUMPABLE:
1041                         if (current->dumpable)
1042                                 error = 1;
1043                         break;
1044                 case PR_SET_DUMPABLE:
1045                         if (arg2 != 0 && arg2 != 1) {
1046                                 error = -EINVAL;
1047                                 break;
1048                         }
1049                         current->dumpable = arg2;
1050                         break;
1051                 case PR_SET_UNALIGN:
1052 #ifdef SET_UNALIGN_CTL
1053                         error = SET_UNALIGN_CTL(current, arg2);
1054 #else
1055                         error = -EINVAL;
1056 #endif
1057                         break;
1058
1059                 case PR_GET_UNALIGN:
1060 #ifdef GET_UNALIGN_CTL
1061                         error = GET_UNALIGN_CTL(current, arg2);
1062 #else
1063                         error = -EINVAL;
1064 #endif
1065                         break;
1066
1067                 default:
1068                         error = -EINVAL;
1069                         break;
1070         }
1071         return error;
1072 }
1073