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