kernel/sys.c

   1 /*
   2  *  linux/kernel/sys.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6
   7 #include <linux/errno.h>
   8 #include <linux/sched.h>
   9 #include <linux/kernel.h>
  10 #include <linux/times.h>
  11 #include <linux/utsname.h>
  12 #include <linux/param.h>
  13 #include <linux/resource.h>
  14 #include <linux/signal.h>
  15 #include <linux/string.h>
  16 #include <linux/ptrace.h>
  17 #include <linux/stat.h>
  18 #include <linux/mman.h>
  19 #include <linux/mm.h>
  20 #include <linux/fcntl.h>
  21 #include <linux/tty.h>
  22 #include <linux/smp.h>
  23 #include <linux/smp_lock.h>
  24 #include <linux/notifier.h>
  25 #include <linux/reboot.h>
  26 #include <linux/prctl.h>
  27
  28 #include <asm/uaccess.h>
  29 #include <asm/io.h>
  30
  31 /*
  32  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
  33  */
  34
  35 int C_A_D = 1;
  36
  37
  38 /*
  39  *      Notifier list for kernel code which wants to be called
  40  *      at shutdown. This is used to stop any idling DMA operations
  41  *      and the like.
  42  */
  43
  44 struct notifier_block *reboot_notifier_list = NULL;
  45
  46 int register_reboot_notifier(struct notifier_block * nb)
  47 {
  48         return notifier_chain_register(&reboot_notifier_list, nb);
  49 }
  50
  51 int unregister_reboot_notifier(struct notifier_block * nb)
  52 {
  53         return notifier_chain_unregister(&reboot_notifier_list, nb);
  54 }
  55
  56
  57
  58 extern void adjust_clock(void);
  59
  60 asmlinkage int sys_ni_syscall(void)
  61 {
  62         return -ENOSYS;
  63 }
  64
  65 static int proc_sel(struct task_struct *p, int which, int who)
  66 {
  67         if(p->pid)
  68         {
  69                 switch (which) {
  70                         case PRIO_PROCESS:
  71                                 if (!who && p == current)
  72                                         return 1;
  73                                 return(p->pid == who);
  74                         case PRIO_PGRP:
  75                                 if (!who)
  76                                         who = current->pgrp;
  77                                 return(p->pgrp == who);
  78                         case PRIO_USER:
  79                                 if (!who)
  80                                         who = current->uid;
  81                                 return(p->uid == who);
  82                 }
  83         }
  84         return 0;
  85 }
  86
  87 asmlinkage int sys_setpriority(int which, int who, int niceval)
  88 {
  89         struct task_struct *p;
  90         unsigned int priority;
  91         int error;
  92
  93         if (which > 2 || which < 0)
  94                 return -EINVAL;
  95
  96         /* normalize: avoid signed division (rounding problems) */
  97         error = ESRCH;
  98         priority = niceval;
  99         if (niceval < 0)
 100                 priority = -niceval;
 101         if (priority > 20)
 102                 priority = 20;
 103         priority = (priority * DEF_PRIORITY + 10) / 20 + DEF_PRIORITY;
 104
 105         if (niceval >= 0) {
 106                 priority = 2*DEF_PRIORITY - priority;
 107                 if (!priority)
 108                         priority = 1;
 109         }
 110
 111         read_lock(&tasklist_lock);
 112         for_each_task(p) {
 113                 if (!proc_sel(p, which, who))
 114                         continue;
 115                 if (p->uid != current->euid &&
 116                         p->uid != current->uid && !capable(CAP_SYS_NICE)) {
 117                         error = EPERM;
 118                         continue;
 119                 }
 120                 if (error == ESRCH)
 121                         error = 0;
 122                 if (priority > p->priority && !capable(CAP_SYS_NICE))
 123                         error = EACCES;
 124                 else
 125                         p->priority = priority;
 126         }
 127         read_unlock(&tasklist_lock);
 128
 129         return -error;
 130 }
 131
 132 /*
 133  * Ugh. To avoid negative return values, "getpriority()" will
 134  * not return the normal nice-value, but a value that has been
 135  * offset by 20 (ie it returns 0..40 instead of -20..20)
 136  */
 137 asmlinkage int sys_getpriority(int which, int who)
 138 {
 139         struct task_struct *p;
 140         long max_prio = -ESRCH;
 141
 142         if (which > 2 || which < 0)
 143                 return -EINVAL;
 144
 145         read_lock(&tasklist_lock);
 146         for_each_task (p) {
 147                 if (!proc_sel(p, which, who))
 148                         continue;
 149                 if (p->priority > max_prio)
 150                         max_prio = p->priority;
 151         }
 152         read_unlock(&tasklist_lock);
 153
 154         /* scale the priority from timeslice to 0..40 */
 155         if (max_prio > 0)
 156                 max_prio = (max_prio * 20 + DEF_PRIORITY/2) / DEF_PRIORITY;
 157         return max_prio;
 158 }
 159
 160
 161 /*
 162  * Reboot system call: for obvious reasons only root may call it,
 163  * and even root needs to set up some magic numbers in the registers
 164  * so that some mistake won't make this reboot the whole machine.
 165  * You can also set the meaning of the ctrl-alt-del-key here.
 166  *
 167  * reboot doesn't sync: do that yourself before calling this.
 168  */
 169 asmlinkage int sys_reboot(int magic1, int magic2, int cmd, void * arg)
 170 {
 171         char buffer[256];
 172
 173         /* We only trust the superuser with rebooting the system. */
 174         if (!capable(CAP_SYS_BOOT))
 175                 return -EPERM;
 176
 177         /* For safety, we require "magic" arguments. */
 178         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 179             (magic2 != LINUX_REBOOT_MAGIC2 && magic2 != LINUX_REBOOT_MAGIC2A &&
 180                         magic2 != LINUX_REBOOT_MAGIC2B))
 181                 return -EINVAL;
 182
 183         lock_kernel();
 184         switch (cmd) {
 185         case LINUX_REBOOT_CMD_RESTART:
 186                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 187                 printk(KERN_EMERG "Restarting system.\n");
 188                 machine_restart(NULL);
 189                 break;
 190
 191         case LINUX_REBOOT_CMD_CAD_ON:
 192                 C_A_D = 1;
 193                 break;
 194
 195         case LINUX_REBOOT_CMD_CAD_OFF:
 196                 C_A_D = 0;
 197                 break;
 198
 199         case LINUX_REBOOT_CMD_HALT:
 200                 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
 201                 printk(KERN_EMERG "System halted.\n");
 202                 machine_halt();
 203                 do_exit(0);
 204                 break;
 205
 206         case LINUX_REBOOT_CMD_POWER_OFF:
 207                 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
 208                 printk(KERN_EMERG "Power down.\n");
 209                 machine_power_off();
 210                 do_exit(0);
 211                 break;
 212
 213         case LINUX_REBOOT_CMD_RESTART2:
 214                 if (strncpy_from_user(&buffer[0], (char *)arg, sizeof(buffer) - 1) < 0) {
 215                         unlock_kernel();
 216                         return -EFAULT;
 217                 }
 218                 buffer[sizeof(buffer) - 1] = '\0';
 219
 220                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
 221                 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
 222                 machine_restart(buffer);
 223                 break;
 224
 225         default:
 226                 unlock_kernel();
 227                 return -EINVAL;
 228                 break;
 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 int 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 int 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 int 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 int 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 int 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 int 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 int 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 int 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 int 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                         current->cap_effective.cap &= ~CAP_FS_MASK;
 590                 }
 591                 if (old_fsuid != 0 && current->fsuid == 0) {
 592                         current->cap_effective.cap |=
 593                                 (current->cap_permitted.cap & CAP_FS_MASK);
 594                 }
 595         }
 596
 597         return old_fsuid;
 598 }
 599
 600 /*
 601  * Samma på svenska..
 602  */
 603 asmlinkage int 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         /*
 621          *      In the SMP world we might just be unlucky and have one of
 622          *      the times increment as we use it. Since the value is an
 623          *      atomically safe type this is just fine. Conceptually its
 624          *      as if the syscall took an instant longer to occur.
 625          */
 626         if (tbuf)
 627                 if (copy_to_user(tbuf, &current->times, sizeof(struct tms)))
 628                         return -EFAULT;
 629         return jiffies;
 630 }
 631
 632 /*
 633  * This needs some heavy checking ...
 634  * I just haven't the stomach for it. I also don't fully
 635  * understand sessions/pgrp etc. Let somebody who does explain it.
 636  *
 637  * OK, I think I have the protection semantics right.... this is really
 638  * only important on a multi-user system anyway, to make sure one user
 639  * can't send a signal to a process owned by another.  -TYT, 12/12/91
 640  *
 641  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
 642  * LBT 04.03.94
 643  */
 644
 645 asmlinkage int sys_setpgid(pid_t pid, pid_t pgid)
 646 {
 647         struct task_struct * p;
 648         int err = -EINVAL;
 649
 650         if (!pid)
 651                 pid = current->pid;
 652         if (!pgid)
 653                 pgid = pid;
 654         if (pgid < 0)
 655                 return -EINVAL;
 656
 657         /* From this point forward we keep holding onto the tasklist lock
 658          * so that our parent does not change from under us. -DaveM
 659          */
 660         read_lock(&tasklist_lock);
 661
 662         err = -ESRCH;
 663         p = find_task_by_pid(pid);
 664         if (!p)
 665                 goto out;
 666
 667         if (p->p_pptr == current || p->p_opptr == current) {
 668                 err = -EPERM;
 669                 if (p->session != current->session)
 670                         goto out;
 671                 err = -EACCES;
 672                 if (p->did_exec)
 673                         goto out;
 674         } else if (p != current)
 675                 goto out;
 676         err = -EPERM;
 677         if (p->leader)
 678                 goto out;
 679         if (pgid != pid) {
 680                 struct task_struct * tmp;
 681                 for_each_task (tmp) {
 682                         if (tmp->pgrp == pgid &&
 683                             tmp->session == current->session)
 684                                 goto ok_pgid;
 685                 }
 686                 goto out;
 687         }
 688
 689 ok_pgid:
 690         p->pgrp = pgid;
 691         err = 0;
 692 out:
 693         /* All paths lead to here, thus we are safe. -DaveM */
 694         read_unlock(&tasklist_lock);
 695         return err;
 696 }
 697
 698 asmlinkage int sys_getpgid(pid_t pid)
 699 {
 700         if (!pid) {
 701                 return current->pgrp;
 702         } else {
 703                 int retval;
 704                 struct task_struct *p;
 705
 706                 read_lock(&tasklist_lock);
 707                 p = find_task_by_pid(pid);
 708
 709                 retval = -ESRCH;
 710                 if (p)
 711                         retval = p->pgrp;
 712                 read_unlock(&tasklist_lock);
 713                 return retval;
 714         }
 715 }
 716
 717 asmlinkage int sys_getpgrp(void)
 718 {
 719         /* SMP - assuming writes are word atomic this is fine */
 720         return current->pgrp;
 721 }
 722
 723 asmlinkage int sys_getsid(pid_t pid)
 724 {
 725         if (!pid) {
 726                 return current->session;
 727         } else {
 728                 int retval;
 729                 struct task_struct *p;
 730
 731                 read_lock(&tasklist_lock);
 732                 p = find_task_by_pid(pid);
 733
 734                 retval = -ESRCH;
 735                 if(p)
 736                         retval = p->session;
 737                 read_unlock(&tasklist_lock);
 738                 return retval;
 739         }
 740 }
 741
 742 asmlinkage int sys_setsid(void)
 743 {
 744         struct task_struct * p;
 745         int err = -EPERM;
 746
 747         read_lock(&tasklist_lock);
 748         for_each_task(p) {
 749                 if (p->pgrp == current->pid)
 750                         goto out;
 751         }
 752
 753         current->leader = 1;
 754         current->session = current->pgrp = current->pid;
 755         current->tty = NULL;
 756         current->tty_old_pgrp = 0;
 757         err = current->pgrp;
 758 out:
 759         read_unlock(&tasklist_lock);
 760         return err;
 761 }
 762
 763 /*
 764  * Supplementary group IDs
 765  */
 766 asmlinkage int sys_getgroups(int gidsetsize, gid_t *grouplist)
 767 {
 768         int i;
 769
 770         /*
 771          *      SMP: Nobody else can change our grouplist. Thus we are
 772          *      safe.
 773          */
 774
 775         if (gidsetsize < 0)
 776                 return -EINVAL;
 777         i = current->ngroups;
 778         if (gidsetsize) {
 779                 if (i > gidsetsize)
 780                         return -EINVAL;
 781                 if (copy_to_user(grouplist, current->groups, sizeof(gid_t)*i))
 782                         return -EFAULT;
 783         }
 784         return i;
 785 }
 786
 787 /*
 788  *      SMP: Our groups are not shared. We can copy to/from them safely
 789  *      without another task interfering.
 790  */
 791
 792 asmlinkage int sys_setgroups(int gidsetsize, gid_t *grouplist)
 793 {
 794         if (!capable(CAP_SETGID))
 795                 return -EPERM;
 796         if ((unsigned) gidsetsize > NGROUPS)
 797                 return -EINVAL;
 798         if(copy_from_user(current->groups, grouplist, gidsetsize * sizeof(gid_t)))
 799                 return -EFAULT;
 800         current->ngroups = gidsetsize;
 801         return 0;
 802 }
 803
 804 int in_group_p(gid_t grp)
 805 {
 806         if (grp != current->fsgid) {
 807                 int i = current->ngroups;
 808                 if (i) {
 809                         gid_t *groups = current->groups;
 810                         do {
 811                                 if (*groups == grp)
 812                                         goto out;
 813                                 groups++;
 814                                 i--;
 815                         } while (i);
 816                 }
 817                 return 0;
 818         }
 819 out:
 820         return 1;
 821 }
 822
 823 /*
 824  * This should really be a blocking read-write lock
 825  * rather than a semaphore. Anybody want to implement
 826  * one?
 827  */
 828 struct semaphore uts_sem = MUTEX;
 829
 830 asmlinkage int sys_newuname(struct new_utsname * name)
 831 {
 832         int errno = 0;
 833
 834         down(&uts_sem);
 835         if (copy_to_user(name,&system_utsname,sizeof *name))
 836                 errno = -EFAULT;
 837         up(&uts_sem);
 838         return errno;
 839 }
 840
 841 asmlinkage int sys_sethostname(char *name, int len)
 842 {
 843         int errno;
 844
 845         if (!capable(CAP_SYS_ADMIN))
 846                 return -EPERM;
 847         if (len < 0 || len > __NEW_UTS_LEN)
 848                 return -EINVAL;
 849         down(&uts_sem);
 850         errno = -EFAULT;
 851         if (!copy_from_user(system_utsname.nodename, name, len)) {
 852                 system_utsname.nodename[len] = 0;
 853                 errno = 0;
 854         }
 855         up(&uts_sem);
 856         return errno;
 857 }
 858
 859 asmlinkage int sys_gethostname(char *name, int len)
 860 {
 861         int i, errno;
 862
 863         if (len < 0)
 864                 return -EINVAL;
 865         down(&uts_sem);
 866         i = 1 + strlen(system_utsname.nodename);
 867         if (i > len)
 868                 i = len;
 869         errno = 0;
 870         if (copy_to_user(name, system_utsname.nodename, i))
 871                 errno = -EFAULT;
 872         up(&uts_sem);
 873         return errno;
 874 }
 875
 876 /*
 877  * Only setdomainname; getdomainname can be implemented by calling
 878  * uname()
 879  */
 880 asmlinkage int sys_setdomainname(char *name, int len)
 881 {
 882         int errno;
 883
 884         if (!capable(CAP_SYS_ADMIN))
 885                 return -EPERM;
 886         if (len < 0 || len > __NEW_UTS_LEN)
 887                 return -EINVAL;
 888
 889         down(&uts_sem);
 890         errno = -EFAULT;
 891         if (!copy_from_user(system_utsname.domainname, name, len)) {
 892                 errno = 0;
 893                 system_utsname.domainname[len] = 0;
 894         }
 895         up(&uts_sem);
 896         return errno;
 897 }
 898
 899 asmlinkage int sys_getrlimit(unsigned int resource, struct rlimit *rlim)
 900 {
 901         if (resource >= RLIM_NLIMITS)
 902                 return -EINVAL;
 903         else
 904                 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
 905                         ? -EFAULT : 0;
 906 }
 907
 908 asmlinkage int sys_setrlimit(unsigned int resource, struct rlimit *rlim)
 909 {
 910         struct rlimit new_rlim, *old_rlim;
 911
 912         if (resource >= RLIM_NLIMITS)
 913                 return -EINVAL;
 914         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
 915                 return -EFAULT;
 916         old_rlim = current->rlim + resource;
 917         if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
 918              (new_rlim.rlim_max > old_rlim->rlim_max)) &&
 919             !capable(CAP_SYS_RESOURCE))
 920                 return -EPERM;
 921         if (resource == RLIMIT_NOFILE) {
 922                 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
 923                         return -EPERM;
 924         }
 925         *old_rlim = new_rlim;
 926         return 0;
 927 }
 928
 929 /*
 930  * It would make sense to put struct rusage in the task_struct,
 931  * except that would make the task_struct be *really big*.  After
 932  * task_struct gets moved into malloc'ed memory, it would
 933  * make sense to do this.  It will make moving the rest of the information
 934  * a lot simpler!  (Which we're not doing right now because we're not
 935  * measuring them yet).
 936  *
 937  * This is SMP safe.  Either we are called from sys_getrusage on ourselves
 938  * below (we know we aren't going to exit/disappear and only we change our
 939  * rusage counters), or we are called from wait4() on a process which is
 940  * either stopped or zombied.  In the zombied case the task won't get
 941  * reaped till shortly after the call to getrusage(), in both cases the
 942  * task being examined is in a frozen state so the counters won't change.
 943  */
 944 int getrusage(struct task_struct *p, int who, struct rusage *ru)
 945 {
 946         struct rusage r;
 947
 948         memset((char *) &r, 0, sizeof(r));
 949         switch (who) {
 950                 case RUSAGE_SELF:
 951                         r.ru_utime.tv_sec = CT_TO_SECS(p->times.tms_utime);
 952                         r.ru_utime.tv_usec = CT_TO_USECS(p->times.tms_utime);
 953                         r.ru_stime.tv_sec = CT_TO_SECS(p->times.tms_stime);
 954                         r.ru_stime.tv_usec = CT_TO_USECS(p->times.tms_stime);
 955                         r.ru_minflt = p->min_flt;
 956                         r.ru_majflt = p->maj_flt;
 957                         r.ru_nswap = p->nswap;
 958                         break;
 959                 case RUSAGE_CHILDREN:
 960                         r.ru_utime.tv_sec = CT_TO_SECS(p->times.tms_cutime);
 961                         r.ru_utime.tv_usec = CT_TO_USECS(p->times.tms_cutime);
 962                         r.ru_stime.tv_sec = CT_TO_SECS(p->times.tms_cstime);
 963                         r.ru_stime.tv_usec = CT_TO_USECS(p->times.tms_cstime);
 964                         r.ru_minflt = p->cmin_flt;
 965                         r.ru_majflt = p->cmaj_flt;
 966                         r.ru_nswap = p->cnswap;
 967                         break;
 968                 default:
 969                         r.ru_utime.tv_sec = CT_TO_SECS(p->times.tms_utime + p->times.tms_cutime);
 970                         r.ru_utime.tv_usec = CT_TO_USECS(p->times.tms_utime + p->times.tms_cutime);
 971                         r.ru_stime.tv_sec = CT_TO_SECS(p->times.tms_stime + p->times.tms_cstime);
 972                         r.ru_stime.tv_usec = CT_TO_USECS(p->times.tms_stime + p->times.tms_cstime);
 973                         r.ru_minflt = p->min_flt + p->cmin_flt;
 974                         r.ru_majflt = p->maj_flt + p->cmaj_flt;
 975                         r.ru_nswap = p->nswap + p->cnswap;
 976                         break;
 977         }
 978         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
 979 }
 980
 981 asmlinkage int sys_getrusage(int who, struct rusage *ru)
 982 {
 983         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
 984                 return -EINVAL;
 985         return getrusage(current, who, ru);
 986 }
 987
 988 asmlinkage int sys_umask(int mask)
 989 {
 990         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
 991         return mask;
 992 }
 993
 994 asmlinkage int sys_prctl(int option, unsigned long arg2, unsigned long arg3,
 995                          unsigned long arg4, unsigned long arg5)
 996 {
 997         int error = 0;
 998         int sig;
 999
1000         switch (option) {
1001                 case PR_SET_PDEATHSIG:
1002                         sig = arg2;
1003                         if (sig > _NSIG) {
1004                                 error = -EINVAL;
1005                                 break;
1006                         }
1007                         current->pdeath_signal = sig;
1008                         break;
1009                 default:
1010                         error = -EINVAL;
1011                         break;
1012         }
1013         return error;
1014 }
1015