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