kernel/sys.c

   1 /*
   2  *  linux/kernel/sys.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6
   7 #include <linux/config.h>
   8 #include <linux/module.h>
   9 #include <linux/mm.h>
  10 #include <linux/utsname.h>
  11 #include <linux/mman.h>
  12 #include <linux/smp_lock.h>
  13 #include <linux/notifier.h>
  14 #include <linux/kmod.h>
  15 #include <linux/reboot.h>
  16 #include <linux/prctl.h>
  17 #include <linux/init.h>
  18 #include <linux/highuid.h>
  19 #include <linux/fs.h>
  20 #include <linux/kernel.h>
  21 #include <linux/kexec.h>
  22 #include <linux/workqueue.h>
  23 #include <linux/device.h>
  24 #include <linux/key.h>
  25 #include <linux/times.h>
  26 #include <linux/posix-timers.h>
  27 #include <linux/security.h>
  28 #include <linux/dcookies.h>
  29 #include <linux/suspend.h>
  30 #include <linux/tty.h>
  31 #include <linux/signal.h>
  32
  33 #include <linux/compat.h>
  34 #include <linux/syscalls.h>
  35 #include <linux/vs_cvirt.h>
  36
  37 #include <asm/uaccess.h>
  38 #include <asm/io.h>
  39 #include <asm/unistd.h>
  40
  41 #ifndef SET_UNALIGN_CTL
  42 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  43 #endif
  44 #ifndef GET_UNALIGN_CTL
  45 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  46 #endif
  47 #ifndef SET_FPEMU_CTL
  48 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
  49 #endif
  50 #ifndef GET_FPEMU_CTL
  51 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
  52 #endif
  53 #ifndef SET_FPEXC_CTL
  54 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
  55 #endif
  56 #ifndef GET_FPEXC_CTL
  57 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
  58 #endif
  59
  60 /*
  61  * this is where the system-wide overflow UID and GID are defined, for
  62  * architectures that now have 32-bit UID/GID but didn't in the past
  63  */
  64
  65 int overflowuid = DEFAULT_OVERFLOWUID;
  66 int overflowgid = DEFAULT_OVERFLOWGID;
  67
  68 #ifdef CONFIG_UID16
  69 EXPORT_SYMBOL(overflowuid);
  70 EXPORT_SYMBOL(overflowgid);
  71 #endif
  72
  73 /*
  74  * the same as above, but for filesystems which can only store a 16-bit
  75  * UID and GID. as such, this is needed on all architectures
  76  */
  77
  78 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
  79 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
  80
  81 EXPORT_SYMBOL(fs_overflowuid);
  82 EXPORT_SYMBOL(fs_overflowgid);
  83
  84 /*
  85  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
  86  */
  87
  88 int C_A_D = 1;
  89 int cad_pid = 1;
  90
  91 /*
  92  *      Notifier list for kernel code which wants to be called
  93  *      at shutdown. This is used to stop any idling DMA operations
  94  *      and the like.
  95  */
  96
  97 static struct notifier_block *reboot_notifier_list;
  98 static DEFINE_RWLOCK(notifier_lock);
  99
 100 /**
 101  *      notifier_chain_register - Add notifier to a notifier chain
 102  *      @list: Pointer to root list pointer
 103  *      @n: New entry in notifier chain
 104  *
 105  *      Adds a notifier to a notifier chain.
 106  *
 107  *      Currently always returns zero.
 108  */
 109
 110 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
 111 {
 112         write_lock(&notifier_lock);
 113         while(*list)
 114         {
 115                 if(n->priority > (*list)->priority)
 116                         break;
 117                 list= &((*list)->next);
 118         }
 119         n->next = *list;
 120         *list=n;
 121         write_unlock(&notifier_lock);
 122         return 0;
 123 }
 124
 125 EXPORT_SYMBOL(notifier_chain_register);
 126
 127 /**
 128  *      notifier_chain_unregister - Remove notifier from a notifier chain
 129  *      @nl: Pointer to root list pointer
 130  *      @n: New entry in notifier chain
 131  *
 132  *      Removes a notifier from a notifier chain.
 133  *
 134  *      Returns zero on success, or %-ENOENT on failure.
 135  */
 136
 137 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
 138 {
 139         write_lock(&notifier_lock);
 140         while((*nl)!=NULL)
 141         {
 142                 if((*nl)==n)
 143                 {
 144                         *nl=n->next;
 145                         write_unlock(&notifier_lock);
 146                         return 0;
 147                 }
 148                 nl=&((*nl)->next);
 149         }
 150         write_unlock(&notifier_lock);
 151         return -ENOENT;
 152 }
 153
 154 EXPORT_SYMBOL(notifier_chain_unregister);
 155
 156 /**
 157  *      notifier_call_chain - Call functions in a notifier chain
 158  *      @n: Pointer to root pointer of notifier chain
 159  *      @val: Value passed unmodified to notifier function
 160  *      @v: Pointer passed unmodified to notifier function
 161  *
 162  *      Calls each function in a notifier chain in turn.
 163  *
 164  *      If the return value of the notifier can be and'd
 165  *      with %NOTIFY_STOP_MASK, then notifier_call_chain
 166  *      will return immediately, with the return value of
 167  *      the notifier function which halted execution.
 168  *      Otherwise, the return value is the return value
 169  *      of the last notifier function called.
 170  */
 171
 172 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
 173 {
 174         int ret=NOTIFY_DONE;
 175         struct notifier_block *nb = *n;
 176
 177         while(nb)
 178         {
 179                 ret=nb->notifier_call(nb,val,v);
 180                 if(ret&NOTIFY_STOP_MASK)
 181                 {
 182                         return ret;
 183                 }
 184                 nb=nb->next;
 185         }
 186         return ret;
 187 }
 188
 189 EXPORT_SYMBOL(notifier_call_chain);
 190
 191 /**
 192  *      register_reboot_notifier - Register function to be called at reboot time
 193  *      @nb: Info about notifier function to be called
 194  *
 195  *      Registers a function with the list of functions
 196  *      to be called at reboot time.
 197  *
 198  *      Currently always returns zero, as notifier_chain_register
 199  *      always returns zero.
 200  */
 201
 202 int register_reboot_notifier(struct notifier_block * nb)
 203 {
 204         return notifier_chain_register(&reboot_notifier_list, nb);
 205 }
 206
 207 EXPORT_SYMBOL(register_reboot_notifier);
 208
 209 /**
 210  *      unregister_reboot_notifier - Unregister previously registered reboot notifier
 211  *      @nb: Hook to be unregistered
 212  *
 213  *      Unregisters a previously registered reboot
 214  *      notifier function.
 215  *
 216  *      Returns zero on success, or %-ENOENT on failure.
 217  */
 218
 219 int unregister_reboot_notifier(struct notifier_block * nb)
 220 {
 221         return notifier_chain_unregister(&reboot_notifier_list, nb);
 222 }
 223
 224 EXPORT_SYMBOL(unregister_reboot_notifier);
 225
 226 static int set_one_prio(struct task_struct *p, int niceval, int error)
 227 {
 228         int no_nice;
 229
 230         if (p->uid != current->euid &&
 231                 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
 232                 error = -EPERM;
 233                 goto out;
 234         }
 235         if (niceval < task_nice(p) && !can_nice(p, niceval)) {
 236                 if (vx_flags(VXF_IGNEG_NICE, 0))
 237                         error = 0;
 238                 else
 239                         error = -EACCES;
 240                 goto out;
 241         }
 242         no_nice = security_task_setnice(p, niceval);
 243         if (no_nice) {
 244                 error = no_nice;
 245                 goto out;
 246         }
 247         if (error == -ESRCH)
 248                 error = 0;
 249         set_user_nice(p, niceval);
 250 out:
 251         return error;
 252 }
 253
 254 asmlinkage long sys_setpriority(int which, int who, int niceval)
 255 {
 256         struct task_struct *g, *p;
 257         struct user_struct *user;
 258         int error = -EINVAL;
 259
 260         if (which > 2 || which < 0)
 261                 goto out;
 262
 263         /* normalize: avoid signed division (rounding problems) */
 264         error = -ESRCH;
 265         if (niceval < -20)
 266                 niceval = -20;
 267         if (niceval > 19)
 268                 niceval = 19;
 269
 270         read_lock(&tasklist_lock);
 271         switch (which) {
 272                 case PRIO_PROCESS:
 273                         if (!who)
 274                                 who = current->pid;
 275                         p = find_task_by_pid(who);
 276                         if (p)
 277                                 error = set_one_prio(p, niceval, error);
 278                         break;
 279                 case PRIO_PGRP:
 280                         if (!who)
 281                                 who = process_group(current);
 282                         do_each_task_pid(who, PIDTYPE_PGID, p) {
 283                                 error = set_one_prio(p, niceval, error);
 284                         } while_each_task_pid(who, PIDTYPE_PGID, p);
 285                         break;
 286                 case PRIO_USER:
 287                         user = current->user;
 288                         if (!who)
 289                                 who = current->uid;
 290                         else
 291                                 if ((who != current->uid) &&
 292                                         !(user = find_user(vx_current_xid(), who)))
 293                                         goto out_unlock;        /* No processes for this user */
 294
 295                         do_each_thread(g, p)
 296                                 if (p->uid == who)
 297                                         error = set_one_prio(p, niceval, error);
 298                         while_each_thread(g, p);
 299                         if (who != current->uid)
 300                                 free_uid(user);         /* For find_user() */
 301                         break;
 302         }
 303 out_unlock:
 304         read_unlock(&tasklist_lock);
 305 out:
 306         return error;
 307 }
 308
 309 /*
 310  * Ugh. To avoid negative return values, "getpriority()" will
 311  * not return the normal nice-value, but a negated value that
 312  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 313  * to stay compatible.
 314  */
 315 asmlinkage long sys_getpriority(int which, int who)
 316 {
 317         struct task_struct *g, *p;
 318         struct user_struct *user;
 319         long niceval, retval = -ESRCH;
 320
 321         if (which > 2 || which < 0)
 322                 return -EINVAL;
 323
 324         read_lock(&tasklist_lock);
 325         switch (which) {
 326                 case PRIO_PROCESS:
 327                         if (!who)
 328                                 who = current->pid;
 329                         p = find_task_by_pid(who);
 330                         if (p) {
 331                                 niceval = 20 - task_nice(p);
 332                                 if (niceval > retval)
 333                                         retval = niceval;
 334                         }
 335                         break;
 336                 case PRIO_PGRP:
 337                         if (!who)
 338                                 who = process_group(current);
 339                         do_each_task_pid(who, PIDTYPE_PGID, p) {
 340                                 niceval = 20 - task_nice(p);
 341                                 if (niceval > retval)
 342                                         retval = niceval;
 343                         } while_each_task_pid(who, PIDTYPE_PGID, p);
 344                         break;
 345                 case PRIO_USER:
 346                         user = current->user;
 347                         if (!who)
 348                                 who = current->uid;
 349                         else
 350                                 if ((who != current->uid) &&
 351                                         !(user = find_user(vx_current_xid(), who)))
 352                                         goto out_unlock;        /* No processes for this user */
 353
 354                         do_each_thread(g, p)
 355                                 if (p->uid == who) {
 356                                         niceval = 20 - task_nice(p);
 357                                         if (niceval > retval)
 358                                                 retval = niceval;
 359                                 }
 360                         while_each_thread(g, p);
 361                         if (who != current->uid)
 362                                 free_uid(user);         /* for find_user() */
 363                         break;
 364         }
 365 out_unlock:
 366         read_unlock(&tasklist_lock);
 367
 368         key_fsgid_changed(current);
 369         return 0;
 370 }
 371
 372 long vs_reboot(unsigned int, void *);
 373
 374 /*
 375  * Reboot system call: for obvious reasons only root may call it,
 376  * and even root needs to set up some magic numbers in the registers
 377  * so that some mistake won't make this reboot the whole machine.
 378  * You can also set the meaning of the ctrl-alt-del-key here.
 379  *
 380  * reboot doesn't sync: do that yourself before calling this.
 381  */
 382 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
 383 {
 384         char buffer[256];
 385
 386         /* We only trust the superuser with rebooting the system. */
 387         if (!capable(CAP_SYS_BOOT))
 388                 return -EPERM;
 389
 390         /* For safety, we require "magic" arguments. */
 391         if (magic1 != LINUX_REBOOT_MAGIC1 ||
 392             (magic2 != LINUX_REBOOT_MAGIC2 &&
 393                         magic2 != LINUX_REBOOT_MAGIC2A &&
 394                         magic2 != LINUX_REBOOT_MAGIC2B &&
 395                         magic2 != LINUX_REBOOT_MAGIC2C))
 396                 return -EINVAL;
 397
 398         if (!vx_check(0, VX_ADMIN|VX_WATCH))
 399                 return vs_reboot(cmd, arg);
 400
 401         lock_kernel();
 402         switch (cmd) {
 403         case LINUX_REBOOT_CMD_RESTART:
 404                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 405                 system_state = SYSTEM_RESTART;
 406                 device_shutdown();
 407                 printk(KERN_EMERG "Restarting system.\n");
 408                 machine_restart(NULL);
 409                 break;
 410
 411         case LINUX_REBOOT_CMD_CAD_ON:
 412                 C_A_D = 1;
 413                 break;
 414
 415         case LINUX_REBOOT_CMD_CAD_OFF:
 416                 C_A_D = 0;
 417                 break;
 418
 419         case LINUX_REBOOT_CMD_HALT:
 420                 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
 421                 system_state = SYSTEM_HALT;
 422                 device_shutdown();
 423                 printk(KERN_EMERG "System halted.\n");
 424                 machine_halt();
 425                 unlock_kernel();
 426                 do_exit(0);
 427                 break;
 428
 429         case LINUX_REBOOT_CMD_POWER_OFF:
 430                 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
 431                 system_state = SYSTEM_POWER_OFF;
 432                 device_shutdown();
 433                 printk(KERN_EMERG "Power down.\n");
 434                 machine_power_off();
 435                 unlock_kernel();
 436                 do_exit(0);
 437                 break;
 438
 439         case LINUX_REBOOT_CMD_RESTART2:
 440                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 441                         unlock_kernel();
 442                         return -EFAULT;
 443                 }
 444                 buffer[sizeof(buffer) - 1] = '\0';
 445
 446                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
 447                 system_state = SYSTEM_RESTART;
 448                 device_shutdown();
 449                 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
 450                 machine_restart(buffer);
 451                 break;
 452
 453 #ifdef CONFIG_KEXEC
 454         case LINUX_REBOOT_CMD_KEXEC:
 455         {
 456                 struct kimage *image;
 457                 image = xchg(&kexec_image, 0);
 458                 if (!image) {
 459                         unlock_kernel();
 460                         return -EINVAL;
 461                 }
 462                 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 463                 system_state = SYSTEM_RESTART;
 464                 device_shutdown();
 465                 system_state = SYSTEM_BOOTING;
 466                 printk(KERN_EMERG "Starting new kernel\n");
 467                 machine_shutdown();
 468                 machine_kexec(image);
 469                 break;
 470         }
 471 #endif
 472
 473 #ifdef CONFIG_SOFTWARE_SUSPEND
 474         case LINUX_REBOOT_CMD_SW_SUSPEND:
 475                 {
 476                         int ret = software_suspend();
 477                         unlock_kernel();
 478                         return ret;
 479                 }
 480 #endif
 481
 482         default:
 483                 unlock_kernel();
 484                 return -EINVAL;
 485         }
 486         unlock_kernel();
 487         return 0;
 488 }
 489
 490 static void deferred_cad(void *dummy)
 491 {
 492         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
 493         machine_restart(NULL);
 494 }
 495
 496 /*
 497  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 498  * As it's called within an interrupt, it may NOT sync: the only choice
 499  * is whether to reboot at once, or just ignore the ctrl-alt-del.
 500  */
 501 void ctrl_alt_del(void)
 502 {
 503         static DECLARE_WORK(cad_work, deferred_cad, NULL);
 504
 505         if (C_A_D)
 506                 schedule_work(&cad_work);
 507         else
 508                 kill_proc(cad_pid, SIGINT, 1);
 509 }
 510
 511
 512
 513 /*
 514  * Unprivileged users may change the real gid to the effective gid
 515  * or vice versa.  (BSD-style)
 516  *
 517  * If you set the real gid at all, or set the effective gid to a value not
 518  * equal to the real gid, then the saved gid is set to the new effective gid.
 519  *
 520  * This makes it possible for a setgid program to completely drop its
 521  * privileges, which is often a useful assertion to make when you are doing
 522  * a security audit over a program.
 523  *
 524  * The general idea is that a program which uses just setregid() will be
 525  * 100% compatible with BSD.  A program which uses just setgid() will be
 526  * 100% compatible with POSIX with saved IDs.
 527  *
 528  * SMP: There are not races, the GIDs are checked only by filesystem
 529  *      operations (as far as semantic preservation is concerned).
 530  */
 531 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
 532 {
 533         int old_rgid = current->gid;
 534         int old_egid = current->egid;
 535         int new_rgid = old_rgid;
 536         int new_egid = old_egid;
 537         int retval;
 538
 539         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
 540         if (retval)
 541                 return retval;
 542
 543         if (rgid != (gid_t) -1) {
 544                 if ((old_rgid == rgid) ||
 545                     (current->egid==rgid) ||
 546                     capable(CAP_SETGID))
 547                         new_rgid = rgid;
 548                 else
 549                         return -EPERM;
 550         }
 551         if (egid != (gid_t) -1) {
 552                 if ((old_rgid == egid) ||
 553                     (current->egid == egid) ||
 554                     (current->sgid == egid) ||
 555                     capable(CAP_SETGID))
 556                         new_egid = egid;
 557                 else {
 558                         return -EPERM;
 559                 }
 560         }
 561         if (new_egid != old_egid)
 562         {
 563                 current->mm->dumpable = 0;
 564                 smp_wmb();
 565         }
 566         if (rgid != (gid_t) -1 ||
 567             (egid != (gid_t) -1 && egid != old_rgid))
 568                 current->sgid = new_egid;
 569         current->fsgid = new_egid;
 570         current->egid = new_egid;
 571         current->gid = new_rgid;
 572
 573         key_fsgid_changed(current);
 574         return 0;
 575 }
 576
 577 /*
 578  * setgid() is implemented like SysV w/ SAVED_IDS
 579  *
 580  * SMP: Same implicit races as above.
 581  */
 582 asmlinkage long sys_setgid(gid_t gid)
 583 {
 584         int old_egid = current->egid;
 585         int retval;
 586
 587         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
 588         if (retval)
 589                 return retval;
 590
 591         if (capable(CAP_SETGID))
 592         {
 593                 if(old_egid != gid)
 594                 {
 595                         current->mm->dumpable=0;
 596                         smp_wmb();
 597                 }
 598                 current->gid = current->egid = current->sgid = current->fsgid = gid;
 599         }
 600         else if ((gid == current->gid) || (gid == current->sgid))
 601         {
 602                 if(old_egid != gid)
 603                 {
 604                         current->mm->dumpable=0;
 605                         smp_wmb();
 606                 }
 607                 current->egid = current->fsgid = gid;
 608         }
 609         else
 610                 return -EPERM;
 611
 612         key_fsgid_changed(current);
 613
 614         return 0;
 615 }
 616
 617 static int set_user(uid_t new_ruid, int dumpclear)
 618 {
 619         struct user_struct *new_user;
 620
 621         new_user = alloc_uid(vx_current_xid(), new_ruid);
 622         if (!new_user)
 623                 return -EAGAIN;
 624
 625         if (atomic_read(&new_user->processes) >=
 626                                 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
 627                         new_user != &root_user) {
 628                 free_uid(new_user);
 629                 return -EAGAIN;
 630         }
 631
 632         switch_uid(new_user);
 633
 634         if(dumpclear)
 635         {
 636                 current->mm->dumpable = 0;
 637                 smp_wmb();
 638         }
 639         current->uid = new_ruid;
 640         return 0;
 641 }
 642
 643 /*
 644  * Unprivileged users may change the real uid to the effective uid
 645  * or vice versa.  (BSD-style)
 646  *
 647  * If you set the real uid at all, or set the effective uid to a value not
 648  * equal to the real uid, then the saved uid is set to the new effective uid.
 649  *
 650  * This makes it possible for a setuid program to completely drop its
 651  * privileges, which is often a useful assertion to make when you are doing
 652  * a security audit over a program.
 653  *
 654  * The general idea is that a program which uses just setreuid() will be
 655  * 100% compatible with BSD.  A program which uses just setuid() will be
 656  * 100% compatible with POSIX with saved IDs.
 657  */
 658 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
 659 {
 660         int old_ruid, old_euid, old_suid, new_ruid, new_euid;
 661         int retval;
 662
 663         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
 664         if (retval)
 665                 return retval;
 666
 667         new_ruid = old_ruid = current->uid;
 668         new_euid = old_euid = current->euid;
 669         old_suid = current->suid;
 670
 671         if (ruid != (uid_t) -1) {
 672                 new_ruid = ruid;
 673                 if ((old_ruid != ruid) &&
 674                     (current->euid != ruid) &&
 675                     !capable(CAP_SETUID))
 676                         return -EPERM;
 677         }
 678
 679         if (euid != (uid_t) -1) {
 680                 new_euid = euid;
 681                 if ((old_ruid != euid) &&
 682                     (current->euid != euid) &&
 683                     (current->suid != euid) &&
 684                     !capable(CAP_SETUID))
 685                         return -EPERM;
 686         }
 687
 688         if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
 689                 return -EAGAIN;
 690
 691         if (new_euid != old_euid)
 692         {
 693                 current->mm->dumpable=0;
 694                 smp_wmb();
 695         }
 696         current->fsuid = current->euid = new_euid;
 697         if (ruid != (uid_t) -1 ||
 698             (euid != (uid_t) -1 && euid != old_ruid))
 699                 current->suid = current->euid;
 700         current->fsuid = current->euid;
 701
 702         key_fsuid_changed(current);
 703
 704         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
 705 }
 706
 707
 708
 709 /*
 710  * setuid() is implemented like SysV with SAVED_IDS
 711  *
 712  * Note that SAVED_ID's is deficient in that a setuid root program
 713  * like sendmail, for example, cannot set its uid to be a normal
 714  * user and then switch back, because if you're root, setuid() sets
 715  * the saved uid too.  If you don't like this, blame the bright people
 716  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 717  * will allow a root program to temporarily drop privileges and be able to
 718  * regain them by swapping the real and effective uid.
 719  */
 720 asmlinkage long sys_setuid(uid_t uid)
 721 {
 722         int old_euid = current->euid;
 723         int old_ruid, old_suid, new_ruid, new_suid;
 724         int retval;
 725
 726         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
 727         if (retval)
 728                 return retval;
 729
 730         old_ruid = new_ruid = current->uid;
 731         old_suid = current->suid;
 732         new_suid = old_suid;
 733
 734         if (capable(CAP_SETUID)) {
 735                 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
 736                         return -EAGAIN;
 737                 new_suid = uid;
 738         } else if ((uid != current->uid) && (uid != new_suid))
 739                 return -EPERM;
 740
 741         if (old_euid != uid)
 742         {
 743                 current->mm->dumpable = 0;
 744                 smp_wmb();
 745         }
 746         current->fsuid = current->euid = uid;
 747         current->suid = new_suid;
 748
 749         key_fsuid_changed(current);
 750
 751         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
 752 }
 753
 754
 755 /*
 756  * This function implements a generic ability to update ruid, euid,
 757  * and suid.  This allows you to implement the 4.4 compatible seteuid().
 758  */
 759 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 760 {
 761         int old_ruid = current->uid;
 762         int old_euid = current->euid;
 763         int old_suid = current->suid;
 764         int retval;
 765
 766         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
 767         if (retval)
 768                 return retval;
 769
 770         if (!capable(CAP_SETUID)) {
 771                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
 772                     (ruid != current->euid) && (ruid != current->suid))
 773                         return -EPERM;
 774                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
 775                     (euid != current->euid) && (euid != current->suid))
 776                         return -EPERM;
 777                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
 778                     (suid != current->euid) && (suid != current->suid))
 779                         return -EPERM;
 780         }
 781         if (ruid != (uid_t) -1) {
 782                 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
 783                         return -EAGAIN;
 784         }
 785         if (euid != (uid_t) -1) {
 786                 if (euid != current->euid)
 787                 {
 788                         current->mm->dumpable = 0;
 789                         smp_wmb();
 790                 }
 791                 current->euid = euid;
 792         }
 793         current->fsuid = current->euid;
 794         if (suid != (uid_t) -1)
 795                 current->suid = suid;
 796
 797         key_fsuid_changed(current);
 798
 799         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
 800 }
 801
 802 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
 803 {
 804         int retval;
 805
 806         if (!(retval = put_user(current->uid, ruid)) &&
 807             !(retval = put_user(current->euid, euid)))
 808                 retval = put_user(current->suid, suid);
 809
 810         return retval;
 811 }
 812
 813 /*
 814  * Same as above, but for rgid, egid, sgid.
 815  */
 816 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 817 {
 818         int retval;
 819
 820         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
 821         if (retval)
 822                 return retval;
 823
 824         if (!capable(CAP_SETGID)) {
 825                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
 826                     (rgid != current->egid) && (rgid != current->sgid))
 827                         return -EPERM;
 828                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
 829                     (egid != current->egid) && (egid != current->sgid))
 830                         return -EPERM;
 831                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
 832                     (sgid != current->egid) && (sgid != current->sgid))
 833                         return -EPERM;
 834         }
 835         if (egid != (gid_t) -1) {
 836                 if (egid != current->egid)
 837                 {
 838                         current->mm->dumpable = 0;
 839                         smp_wmb();
 840                 }
 841                 current->egid = egid;
 842         }
 843         current->fsgid = current->egid;
 844         if (rgid != (gid_t) -1)
 845                 current->gid = rgid;
 846         if (sgid != (gid_t) -1)
 847                 current->sgid = sgid;
 848
 849         key_fsgid_changed(current);
 850         return 0;
 851 }
 852
 853 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
 854 {
 855         int retval;
 856
 857         if (!(retval = put_user(current->gid, rgid)) &&
 858             !(retval = put_user(current->egid, egid)))
 859                 retval = put_user(current->sgid, sgid);
 860
 861         return retval;
 862 }
 863
 864
 865 /*
 866  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 867  * is used for "access()" and for the NFS daemon (letting nfsd stay at
 868  * whatever uid it wants to). It normally shadows "euid", except when
 869  * explicitly set by setfsuid() or for access..
 870  */
 871 asmlinkage long sys_setfsuid(uid_t uid)
 872 {
 873         int old_fsuid;
 874
 875         old_fsuid = current->fsuid;
 876         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
 877                 return old_fsuid;
 878
 879         if (uid == current->uid || uid == current->euid ||
 880             uid == current->suid || uid == current->fsuid ||
 881             capable(CAP_SETUID))
 882         {
 883                 if (uid != old_fsuid)
 884                 {
 885                         current->mm->dumpable = 0;
 886                         smp_wmb();
 887                 }
 888                 current->fsuid = uid;
 889         }
 890
 891         key_fsuid_changed(current);
 892
 893         security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
 894
 895         return old_fsuid;
 896 }
 897
 898 /*
 899  * Samma på svenska..
 900  */
 901 asmlinkage long sys_setfsgid(gid_t gid)
 902 {
 903         int old_fsgid;
 904
 905         old_fsgid = current->fsgid;
 906         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
 907                 return old_fsgid;
 908
 909         if (gid == current->gid || gid == current->egid ||
 910             gid == current->sgid || gid == current->fsgid ||
 911             capable(CAP_SETGID))
 912         {
 913                 if (gid != old_fsgid)
 914                 {
 915                         current->mm->dumpable = 0;
 916                         smp_wmb();
 917                 }
 918                 current->fsgid = gid;
 919                 key_fsgid_changed(current);
 920         }
 921         return old_fsgid;
 922 }
 923
 924 asmlinkage long sys_times(struct tms __user * tbuf)
 925 {
 926         /*
 927          *      In the SMP world we might just be unlucky and have one of
 928          *      the times increment as we use it. Since the value is an
 929          *      atomically safe type this is just fine. Conceptually its
 930          *      as if the syscall took an instant longer to occur.
 931          */
 932         if (tbuf) {
 933                 struct tms tmp;
 934                 struct task_struct *tsk = current;
 935                 struct task_struct *t;
 936                 cputime_t utime, stime, cutime, cstime;
 937
 938                 read_lock(&tasklist_lock);
 939                 utime = tsk->signal->utime;
 940                 stime = tsk->signal->stime;
 941                 t = tsk;
 942                 do {
 943                         utime = cputime_add(utime, t->utime);
 944                         stime = cputime_add(stime, t->stime);
 945                         t = next_thread(t);
 946                 } while (t != tsk);
 947
 948                 /*
 949                  * While we have tasklist_lock read-locked, no dying thread
 950                  * can be updating current->signal->[us]time.  Instead,
 951                  * we got their counts included in the live thread loop.
 952                  * However, another thread can come in right now and
 953                  * do a wait call that updates current->signal->c[us]time.
 954                  * To make sure we always see that pair updated atomically,
 955                  * we take the siglock around fetching them.
 956                  */
 957                 spin_lock_irq(&tsk->sighand->siglock);
 958                 cutime = tsk->signal->cutime;
 959                 cstime = tsk->signal->cstime;
 960                 spin_unlock_irq(&tsk->sighand->siglock);
 961                 read_unlock(&tasklist_lock);
 962
 963                 tmp.tms_utime = cputime_to_clock_t(utime);
 964                 tmp.tms_stime = cputime_to_clock_t(stime);
 965                 tmp.tms_cutime = cputime_to_clock_t(cutime);
 966                 tmp.tms_cstime = cputime_to_clock_t(cstime);
 967                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
 968                         return -EFAULT;
 969         }
 970         return (long) jiffies_64_to_clock_t(get_jiffies_64());
 971 }
 972
 973 /*
 974  * This needs some heavy checking ...
 975  * I just haven't the stomach for it. I also don't fully
 976  * understand sessions/pgrp etc. Let somebody who does explain it.
 977  *
 978  * OK, I think I have the protection semantics right.... this is really
 979  * only important on a multi-user system anyway, to make sure one user
 980  * can't send a signal to a process owned by another.  -TYT, 12/12/91
 981  *
 982  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
 983  * LBT 04.03.94
 984  */
 985
 986 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
 987 {
 988         struct task_struct *p;
 989         int err = -EINVAL;
 990         pid_t rpgid;
 991
 992         if (!pid)
 993                 pid = vx_map_pid(current->pid);
 994         if (!pgid)
 995                 pgid = pid;
 996         if (pgid < 0)
 997                 return -EINVAL;
 998
 999         rpgid = vx_rmap_pid(pgid);
1000
1001         /* From this point forward we keep holding onto the tasklist lock
1002          * so that our parent does not change from under us. -DaveM
1003          */
1004         write_lock_irq(&tasklist_lock);
1005
1006         err = -ESRCH;
1007         p = find_task_by_pid(pid);
1008         if (!p)
1009                 goto out;
1010
1011         err = -EINVAL;
1012         if (!thread_group_leader(p))
1013                 goto out;
1014
1015         if (p->parent == current || p->real_parent == current) {
1016                 err = -EPERM;
1017                 if (p->signal->session != current->signal->session)
1018                         goto out;
1019                 err = -EACCES;
1020                 if (p->did_exec)
1021                         goto out;
1022         } else {
1023                 err = -ESRCH;
1024                 if (p != current)
1025                         goto out;
1026         }
1027
1028         err = -EPERM;
1029         if (p->signal->leader)
1030                 goto out;
1031
1032         if (pgid != pid) {
1033                 struct task_struct *p;
1034
1035                 do_each_task_pid(rpgid, PIDTYPE_PGID, p) {
1036                         if (p->signal->session == current->signal->session)
1037                                 goto ok_pgid;
1038                 } while_each_task_pid(rpgid, PIDTYPE_PGID, p);
1039                 goto out;
1040         }
1041
1042 ok_pgid:
1043         err = security_task_setpgid(p, rpgid);
1044         if (err)
1045                 goto out;
1046
1047         if (process_group(p) != rpgid) {
1048                 detach_pid(p, PIDTYPE_PGID);
1049                 p->signal->pgrp = rpgid;
1050                 attach_pid(p, PIDTYPE_PGID, rpgid);
1051         }
1052
1053         err = 0;
1054 out:
1055         /* All paths lead to here, thus we are safe. -DaveM */
1056         write_unlock_irq(&tasklist_lock);
1057         return err;
1058 }
1059
1060 asmlinkage long sys_getpgid(pid_t pid)
1061 {
1062         if (!pid) {
1063                 return vx_rmap_pid(process_group(current));
1064         } else {
1065                 int retval;
1066                 struct task_struct *p;
1067
1068                 read_lock(&tasklist_lock);
1069                 p = find_task_by_pid(pid);
1070
1071                 retval = -ESRCH;
1072                 if (p) {
1073                         retval = security_task_getpgid(p);
1074                         if (!retval)
1075                                 retval = vx_rmap_pid(process_group(p));
1076                 }
1077                 read_unlock(&tasklist_lock);
1078                 return retval;
1079         }
1080 }
1081
1082 #ifdef __ARCH_WANT_SYS_GETPGRP
1083
1084 asmlinkage long sys_getpgrp(void)
1085 {
1086         /* SMP - assuming writes are word atomic this is fine */
1087         return process_group(current);
1088 }
1089
1090 #endif
1091
1092 asmlinkage long sys_getsid(pid_t pid)
1093 {
1094         if (!pid) {
1095                 return current->signal->session;
1096         } else {
1097                 int retval;
1098                 struct task_struct *p;
1099
1100                 read_lock(&tasklist_lock);
1101                 p = find_task_by_pid(pid);
1102
1103                 retval = -ESRCH;
1104                 if(p) {
1105                         retval = security_task_getsid(p);
1106                         if (!retval)
1107                                 retval = p->signal->session;
1108                 }
1109                 read_unlock(&tasklist_lock);
1110                 return retval;
1111         }
1112 }
1113
1114 asmlinkage long sys_setsid(void)
1115 {
1116         struct pid *pid;
1117         int err = -EPERM;
1118
1119         if (!thread_group_leader(current))
1120                 return -EINVAL;
1121
1122         down(&tty_sem);
1123         write_lock_irq(&tasklist_lock);
1124
1125         pid = find_pid(PIDTYPE_PGID, current->pid);
1126         if (pid)
1127                 goto out;
1128
1129         current->signal->leader = 1;
1130         __set_special_pids(current->pid, current->pid);
1131         current->signal->tty = NULL;
1132         current->signal->tty_old_pgrp = 0;
1133         err = process_group(current);
1134 out:
1135         write_unlock_irq(&tasklist_lock);
1136         up(&tty_sem);
1137         return err;
1138 }
1139
1140 /*
1141  * Supplementary group IDs
1142  */
1143
1144 /* init to 2 - one for init_task, one to ensure it is never freed */
1145 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1146
1147 struct group_info *groups_alloc(int gidsetsize)
1148 {
1149         struct group_info *group_info;
1150         int nblocks;
1151         int i;
1152
1153         nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1154         /* Make sure we always allocate at least one indirect block pointer */
1155         nblocks = nblocks ? : 1;
1156         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1157         if (!group_info)
1158                 return NULL;
1159         group_info->ngroups = gidsetsize;
1160         group_info->nblocks = nblocks;
1161         atomic_set(&group_info->usage, 1);
1162
1163         if (gidsetsize <= NGROUPS_SMALL) {
1164                 group_info->blocks[0] = group_info->small_block;
1165         } else {
1166                 for (i = 0; i < nblocks; i++) {
1167                         gid_t *b;
1168                         b = (void *)__get_free_page(GFP_USER);
1169                         if (!b)
1170                                 goto out_undo_partial_alloc;
1171                         group_info->blocks[i] = b;
1172                 }
1173         }
1174         return group_info;
1175
1176 out_undo_partial_alloc:
1177         while (--i >= 0) {
1178                 free_page((unsigned long)group_info->blocks[i]);
1179         }
1180         kfree(group_info);
1181         return NULL;
1182 }
1183
1184 EXPORT_SYMBOL(groups_alloc);
1185
1186 void groups_free(struct group_info *group_info)
1187 {
1188         if (group_info->blocks[0] != group_info->small_block) {
1189                 int i;
1190                 for (i = 0; i < group_info->nblocks; i++)
1191                         free_page((unsigned long)group_info->blocks[i]);
1192         }
1193         kfree(group_info);
1194 }
1195
1196 EXPORT_SYMBOL(groups_free);
1197
1198 /* export the group_info to a user-space array */
1199 static int groups_to_user(gid_t __user *grouplist,
1200     struct group_info *group_info)
1201 {
1202         int i;
1203         int count = group_info->ngroups;
1204
1205         for (i = 0; i < group_info->nblocks; i++) {
1206                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1207                 int off = i * NGROUPS_PER_BLOCK;
1208                 int len = cp_count * sizeof(*grouplist);
1209
1210                 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1211                         return -EFAULT;
1212
1213                 count -= cp_count;
1214         }
1215         return 0;
1216 }
1217
1218 /* fill a group_info from a user-space array - it must be allocated already */
1219 static int groups_from_user(struct group_info *group_info,
1220     gid_t __user *grouplist)
1221  {
1222         int i;
1223         int count = group_info->ngroups;
1224
1225         for (i = 0; i < group_info->nblocks; i++) {
1226                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1227                 int off = i * NGROUPS_PER_BLOCK;
1228                 int len = cp_count * sizeof(*grouplist);
1229
1230                 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1231                         return -EFAULT;
1232
1233                 count -= cp_count;
1234         }
1235         return 0;
1236 }
1237
1238 /* a simple Shell sort */
1239 static void groups_sort(struct group_info *group_info)
1240 {
1241         int base, max, stride;
1242         int gidsetsize = group_info->ngroups;
1243
1244         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1245                 ; /* nothing */
1246         stride /= 3;
1247
1248         while (stride) {
1249                 max = gidsetsize - stride;
1250                 for (base = 0; base < max; base++) {
1251                         int left = base;
1252                         int right = left + stride;
1253                         gid_t tmp = GROUP_AT(group_info, right);
1254
1255                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1256                                 GROUP_AT(group_info, right) =
1257                                     GROUP_AT(group_info, left);
1258                                 right = left;
1259                                 left -= stride;
1260                         }
1261                         GROUP_AT(group_info, right) = tmp;
1262                 }
1263                 stride /= 3;
1264         }
1265 }
1266
1267 /* a simple bsearch */
1268 static int groups_search(struct group_info *group_info, gid_t grp)
1269 {
1270         int left, right;
1271
1272         if (!group_info)
1273                 return 0;
1274
1275         left = 0;
1276         right = group_info->ngroups;
1277         while (left < right) {
1278                 int mid = (left+right)/2;
1279                 int cmp = grp - GROUP_AT(group_info, mid);
1280                 if (cmp > 0)
1281                         left = mid + 1;
1282                 else if (cmp < 0)
1283                         right = mid;
1284                 else
1285                         return 1;
1286         }
1287         return 0;
1288 }
1289
1290 /* validate and set current->group_info */
1291 int set_current_groups(struct group_info *group_info)
1292 {
1293         int retval;
1294         struct group_info *old_info;
1295
1296         retval = security_task_setgroups(group_info);
1297         if (retval)
1298                 return retval;
1299
1300         groups_sort(group_info);
1301         get_group_info(group_info);
1302
1303         task_lock(current);
1304         old_info = current->group_info;
1305         current->group_info = group_info;
1306         task_unlock(current);
1307
1308         put_group_info(old_info);
1309
1310         return 0;
1311 }
1312
1313 EXPORT_SYMBOL(set_current_groups);
1314
1315 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1316 {
1317         int i = 0;
1318
1319         /*
1320          *      SMP: Nobody else can change our grouplist. Thus we are
1321          *      safe.
1322          */
1323
1324         if (gidsetsize < 0)
1325                 return -EINVAL;
1326
1327         /* no need to grab task_lock here; it cannot change */
1328         get_group_info(current->group_info);
1329         i = current->group_info->ngroups;
1330         if (gidsetsize) {
1331                 if (i > gidsetsize) {
1332                         i = -EINVAL;
1333                         goto out;
1334                 }
1335                 if (groups_to_user(grouplist, current->group_info)) {
1336                         i = -EFAULT;
1337                         goto out;
1338                 }
1339         }
1340 out:
1341         put_group_info(current->group_info);
1342         return i;
1343 }
1344
1345 /*
1346  *      SMP: Our groups are copy-on-write. We can set them safely
1347  *      without another task interfering.
1348  */
1349
1350 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1351 {
1352         struct group_info *group_info;
1353         int retval;
1354
1355         if (!capable(CAP_SETGID))
1356                 return -EPERM;
1357         if ((unsigned)gidsetsize > NGROUPS_MAX)
1358                 return -EINVAL;
1359
1360         group_info = groups_alloc(gidsetsize);
1361         if (!group_info)
1362                 return -ENOMEM;
1363         retval = groups_from_user(group_info, grouplist);
1364         if (retval) {
1365                 put_group_info(group_info);
1366                 return retval;
1367         }
1368
1369         retval = set_current_groups(group_info);
1370         put_group_info(group_info);
1371
1372         return retval;
1373 }
1374
1375 /*
1376  * Check whether we're fsgid/egid or in the supplemental group..
1377  */
1378 int in_group_p(gid_t grp)
1379 {
1380         int retval = 1;
1381         if (grp != current->fsgid) {
1382                 get_group_info(current->group_info);
1383                 retval = groups_search(current->group_info, grp);
1384                 put_group_info(current->group_info);
1385         }
1386         return retval;
1387 }
1388
1389 EXPORT_SYMBOL(in_group_p);
1390
1391 int in_egroup_p(gid_t grp)
1392 {
1393         int retval = 1;
1394         if (grp != current->egid) {
1395                 get_group_info(current->group_info);
1396                 retval = groups_search(current->group_info, grp);
1397                 put_group_info(current->group_info);
1398         }
1399         return retval;
1400 }
1401
1402 EXPORT_SYMBOL(in_egroup_p);
1403
1404 DECLARE_RWSEM(uts_sem);
1405
1406 EXPORT_SYMBOL(uts_sem);
1407
1408 asmlinkage long sys_newuname(struct new_utsname __user * name)
1409 {
1410         int errno = 0;
1411
1412         down_read(&uts_sem);
1413         if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1414                 errno = -EFAULT;
1415         up_read(&uts_sem);
1416         return errno;
1417 }
1418
1419 asmlinkage long sys_sethostname(char __user *name, int len)
1420 {
1421         int errno;
1422         char tmp[__NEW_UTS_LEN];
1423
1424         if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1425                 return -EPERM;
1426         if (len < 0 || len > __NEW_UTS_LEN)
1427                 return -EINVAL;
1428         down_write(&uts_sem);
1429         errno = -EFAULT;
1430         if (!copy_from_user(tmp, name, len)) {
1431                 char *ptr = vx_new_uts(nodename);
1432
1433                 memcpy(ptr, tmp, len);
1434                 ptr[len] = 0;
1435                 errno = 0;
1436         }
1437         up_write(&uts_sem);
1438         return errno;
1439 }
1440
1441 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1442
1443 asmlinkage long sys_gethostname(char __user *name, int len)
1444 {
1445         int i, errno;
1446         char *ptr;
1447
1448         if (len < 0)
1449                 return -EINVAL;
1450         down_read(&uts_sem);
1451         ptr = vx_new_uts(nodename);
1452         i = 1 + strlen(ptr);
1453         if (i > len)
1454                 i = len;
1455         errno = 0;
1456         if (copy_to_user(name, ptr, i))
1457                 errno = -EFAULT;
1458         up_read(&uts_sem);
1459         return errno;
1460 }
1461
1462 #endif
1463
1464 /*
1465  * Only setdomainname; getdomainname can be implemented by calling
1466  * uname()
1467  */
1468 asmlinkage long sys_setdomainname(char __user *name, int len)
1469 {
1470         int errno;
1471         char tmp[__NEW_UTS_LEN];
1472
1473         if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1474                 return -EPERM;
1475         if (len < 0 || len > __NEW_UTS_LEN)
1476                 return -EINVAL;
1477
1478         down_write(&uts_sem);
1479         errno = -EFAULT;
1480         if (!copy_from_user(tmp, name, len)) {
1481                 char *ptr = vx_new_uts(domainname);
1482
1483                 memcpy(ptr, tmp, len);
1484                 ptr[len] = 0;
1485                 errno = 0;
1486         }
1487         up_write(&uts_sem);
1488         return errno;
1489 }
1490
1491 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1492 {
1493         if (resource >= RLIM_NLIMITS)
1494                 return -EINVAL;
1495         else {
1496                 struct rlimit value;
1497                 task_lock(current->group_leader);
1498                 value = current->signal->rlim[resource];
1499                 task_unlock(current->group_leader);
1500                 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1501         }
1502 }
1503
1504 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1505
1506 /*
1507  *      Back compatibility for getrlimit. Needed for some apps.
1508  */
1509
1510 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1511 {
1512         struct rlimit x;
1513         if (resource >= RLIM_NLIMITS)
1514                 return -EINVAL;
1515
1516         task_lock(current->group_leader);
1517         x = current->signal->rlim[resource];
1518         task_unlock(current->group_leader);
1519         if(x.rlim_cur > 0x7FFFFFFF)
1520                 x.rlim_cur = 0x7FFFFFFF;
1521         if(x.rlim_max > 0x7FFFFFFF)
1522                 x.rlim_max = 0x7FFFFFFF;
1523         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1524 }
1525
1526 #endif
1527
1528 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1529 {
1530         struct rlimit new_rlim, *old_rlim;
1531         int retval;
1532
1533         if (resource >= RLIM_NLIMITS)
1534                 return -EINVAL;
1535         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1536                 return -EFAULT;
1537        if (new_rlim.rlim_cur > new_rlim.rlim_max)
1538                return -EINVAL;
1539         old_rlim = current->signal->rlim + resource;
1540         if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1541             !capable(CAP_SYS_RESOURCE) && !vx_ccaps(VXC_SET_RLIMIT))
1542                 return -EPERM;
1543         if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1544                         return -EPERM;
1545
1546         retval = security_task_setrlimit(resource, &new_rlim);
1547         if (retval)
1548                 return retval;
1549
1550         task_lock(current->group_leader);
1551         *old_rlim = new_rlim;
1552         task_unlock(current->group_leader);
1553
1554         if (resource == RLIMIT_CPU && new_rlim.rlim_cur != RLIM_INFINITY &&
1555             (cputime_eq(current->signal->it_prof_expires, cputime_zero) ||
1556              new_rlim.rlim_cur <= cputime_to_secs(
1557                      current->signal->it_prof_expires))) {
1558                 cputime_t cputime = secs_to_cputime(new_rlim.rlim_cur);
1559                 read_lock(&tasklist_lock);
1560                 spin_lock_irq(&current->sighand->siglock);
1561                 set_process_cpu_timer(current, CPUCLOCK_PROF,
1562                                       &cputime, NULL);
1563                 spin_unlock_irq(&current->sighand->siglock);
1564                 read_unlock(&tasklist_lock);
1565         }
1566
1567         return 0;
1568 }
1569
1570 /*
1571  * It would make sense to put struct rusage in the task_struct,
1572  * except that would make the task_struct be *really big*.  After
1573  * task_struct gets moved into malloc'ed memory, it would
1574  * make sense to do this.  It will make moving the rest of the information
1575  * a lot simpler!  (Which we're not doing right now because we're not
1576  * measuring them yet).
1577  *
1578  * This expects to be called with tasklist_lock read-locked or better,
1579  * and the siglock not locked.  It may momentarily take the siglock.
1580  *
1581  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1582  * races with threads incrementing their own counters.  But since word
1583  * reads are atomic, we either get new values or old values and we don't
1584  * care which for the sums.  We always take the siglock to protect reading
1585  * the c* fields from p->signal from races with exit.c updating those
1586  * fields when reaping, so a sample either gets all the additions of a
1587  * given child after it's reaped, or none so this sample is before reaping.
1588  */
1589
1590 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1591 {
1592         struct task_struct *t;
1593         unsigned long flags;
1594         cputime_t utime, stime;
1595
1596         memset((char *) r, 0, sizeof *r);
1597
1598         if (unlikely(!p->signal))
1599                 return;
1600
1601         switch (who) {
1602                 case RUSAGE_CHILDREN:
1603                         spin_lock_irqsave(&p->sighand->siglock, flags);
1604                         utime = p->signal->cutime;
1605                         stime = p->signal->cstime;
1606                         r->ru_nvcsw = p->signal->cnvcsw;
1607                         r->ru_nivcsw = p->signal->cnivcsw;
1608                         r->ru_minflt = p->signal->cmin_flt;
1609                         r->ru_majflt = p->signal->cmaj_flt;
1610                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1611                         cputime_to_timeval(utime, &r->ru_utime);
1612                         cputime_to_timeval(stime, &r->ru_stime);
1613                         break;
1614                 case RUSAGE_SELF:
1615                         spin_lock_irqsave(&p->sighand->siglock, flags);
1616                         utime = stime = cputime_zero;
1617                         goto sum_group;
1618                 case RUSAGE_BOTH:
1619                         spin_lock_irqsave(&p->sighand->siglock, flags);
1620                         utime = p->signal->cutime;
1621                         stime = p->signal->cstime;
1622                         r->ru_nvcsw = p->signal->cnvcsw;
1623                         r->ru_nivcsw = p->signal->cnivcsw;
1624                         r->ru_minflt = p->signal->cmin_flt;
1625                         r->ru_majflt = p->signal->cmaj_flt;
1626                 sum_group:
1627                         utime = cputime_add(utime, p->signal->utime);
1628                         stime = cputime_add(stime, p->signal->stime);
1629                         r->ru_nvcsw += p->signal->nvcsw;
1630                         r->ru_nivcsw += p->signal->nivcsw;
1631                         r->ru_minflt += p->signal->min_flt;
1632                         r->ru_majflt += p->signal->maj_flt;
1633                         t = p;
1634                         do {
1635                                 utime = cputime_add(utime, t->utime);
1636                                 stime = cputime_add(stime, t->stime);
1637                                 r->ru_nvcsw += t->nvcsw;
1638                                 r->ru_nivcsw += t->nivcsw;
1639                                 r->ru_minflt += t->min_flt;
1640                                 r->ru_majflt += t->maj_flt;
1641                                 t = next_thread(t);
1642                         } while (t != p);
1643                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1644                         cputime_to_timeval(utime, &r->ru_utime);
1645                         cputime_to_timeval(stime, &r->ru_stime);
1646                         break;
1647                 default:
1648                         BUG();
1649         }
1650 }
1651
1652 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1653 {
1654         struct rusage r;
1655         read_lock(&tasklist_lock);
1656         k_getrusage(p, who, &r);
1657         read_unlock(&tasklist_lock);
1658         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1659 }
1660
1661 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1662 {
1663         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1664                 return -EINVAL;
1665         return getrusage(current, who, ru);
1666 }
1667
1668 asmlinkage long sys_umask(int mask)
1669 {
1670         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1671         return mask;
1672 }
1673
1674 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1675                           unsigned long arg4, unsigned long arg5)
1676 {
1677         long error;
1678         int sig;
1679
1680         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1681         if (error)
1682                 return error;
1683
1684         switch (option) {
1685                 case PR_SET_PDEATHSIG:
1686                         sig = arg2;
1687                         if (!valid_signal(sig)) {
1688                                 error = -EINVAL;
1689                                 break;
1690                         }
1691                         current->pdeath_signal = sig;
1692                         break;
1693                 case PR_GET_PDEATHSIG:
1694                         error = put_user(current->pdeath_signal, (int __user *)arg2);
1695                         break;
1696                 case PR_GET_DUMPABLE:
1697                         if (current->mm->dumpable)
1698                                 error = 1;
1699                         break;
1700                 case PR_SET_DUMPABLE:
1701                         if (arg2 != 0 && arg2 != 1) {
1702                                 error = -EINVAL;
1703                                 break;
1704                         }
1705                         current->mm->dumpable = arg2;
1706                         break;
1707
1708                 case PR_SET_UNALIGN:
1709                         error = SET_UNALIGN_CTL(current, arg2);
1710                         break;
1711                 case PR_GET_UNALIGN:
1712                         error = GET_UNALIGN_CTL(current, arg2);
1713                         break;
1714                 case PR_SET_FPEMU:
1715                         error = SET_FPEMU_CTL(current, arg2);
1716                         break;
1717                 case PR_GET_FPEMU:
1718                         error = GET_FPEMU_CTL(current, arg2);
1719                         break;
1720                 case PR_SET_FPEXC:
1721                         error = SET_FPEXC_CTL(current, arg2);
1722                         break;
1723                 case PR_GET_FPEXC:
1724                         error = GET_FPEXC_CTL(current, arg2);
1725                         break;
1726                 case PR_GET_TIMING:
1727                         error = PR_TIMING_STATISTICAL;
1728                         break;
1729                 case PR_SET_TIMING:
1730                         if (arg2 == PR_TIMING_STATISTICAL)
1731                                 error = 0;
1732                         else
1733                                 error = -EINVAL;
1734                         break;
1735
1736                 case PR_GET_KEEPCAPS:
1737                         if (current->keep_capabilities)
1738                                 error = 1;
1739                         break;
1740                 case PR_SET_KEEPCAPS:
1741                         if (arg2 != 0 && arg2 != 1) {
1742                                 error = -EINVAL;
1743                                 break;
1744                         }
1745                         current->keep_capabilities = arg2;
1746                         break;
1747                 case PR_SET_NAME: {
1748                         struct task_struct *me = current;
1749                         unsigned char ncomm[sizeof(me->comm)];
1750
1751                         ncomm[sizeof(me->comm)-1] = 0;
1752                         if (strncpy_from_user(ncomm, (char __user *)arg2,
1753                                                 sizeof(me->comm)-1) < 0)
1754                                 return -EFAULT;
1755                         set_task_comm(me, ncomm);
1756                         return 0;
1757                 }
1758                 case PR_GET_NAME: {
1759                         struct task_struct *me = current;
1760                         unsigned char tcomm[sizeof(me->comm)];
1761
1762                         get_task_comm(tcomm, me);
1763                         if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
1764                                 return -EFAULT;
1765                         return 0;
1766                 }
1767                 default:
1768                         error = -EINVAL;
1769                         break;
1770         }
1771         return error;
1772 }