linux/kernel/sys.c
<<
>>
Prefs
   1/*
   2 *  linux/kernel/sys.c
   3 *
   4 *  Copyright (C) 1991, 1992  Linus Torvalds
   5 */
   6
   7#include <linux/export.h>
   8#include <linux/mm.h>
   9#include <linux/utsname.h>
  10#include <linux/mman.h>
  11#include <linux/reboot.h>
  12#include <linux/prctl.h>
  13#include <linux/highuid.h>
  14#include <linux/fs.h>
  15#include <linux/kmod.h>
  16#include <linux/perf_event.h>
  17#include <linux/resource.h>
  18#include <linux/kernel.h>
  19#include <linux/kexec.h>
  20#include <linux/workqueue.h>
  21#include <linux/capability.h>
  22#include <linux/device.h>
  23#include <linux/key.h>
  24#include <linux/times.h>
  25#include <linux/posix-timers.h>
  26#include <linux/security.h>
  27#include <linux/dcookies.h>
  28#include <linux/suspend.h>
  29#include <linux/tty.h>
  30#include <linux/signal.h>
  31#include <linux/cn_proc.h>
  32#include <linux/getcpu.h>
  33#include <linux/task_io_accounting_ops.h>
  34#include <linux/seccomp.h>
  35#include <linux/cpu.h>
  36#include <linux/personality.h>
  37#include <linux/ptrace.h>
  38#include <linux/fs_struct.h>
  39#include <linux/file.h>
  40#include <linux/mount.h>
  41#include <linux/gfp.h>
  42#include <linux/syscore_ops.h>
  43#include <linux/version.h>
  44#include <linux/ctype.h>
  45
  46#include <linux/compat.h>
  47#include <linux/syscalls.h>
  48#include <linux/kprobes.h>
  49#include <linux/user_namespace.h>
  50#include <linux/binfmts.h>
  51
  52#include <linux/sched.h>
  53#include <linux/rcupdate.h>
  54#include <linux/uidgid.h>
  55#include <linux/cred.h>
  56
  57#include <linux/kmsg_dump.h>
  58/* Move somewhere else to avoid recompiling? */
  59#include <generated/utsrelease.h>
  60
  61#include <asm/uaccess.h>
  62#include <asm/io.h>
  63#include <asm/unistd.h>
  64
  65#ifndef SET_UNALIGN_CTL
  66# define SET_UNALIGN_CTL(a,b)   (-EINVAL)
  67#endif
  68#ifndef GET_UNALIGN_CTL
  69# define GET_UNALIGN_CTL(a,b)   (-EINVAL)
  70#endif
  71#ifndef SET_FPEMU_CTL
  72# define SET_FPEMU_CTL(a,b)     (-EINVAL)
  73#endif
  74#ifndef GET_FPEMU_CTL
  75# define GET_FPEMU_CTL(a,b)     (-EINVAL)
  76#endif
  77#ifndef SET_FPEXC_CTL
  78# define SET_FPEXC_CTL(a,b)     (-EINVAL)
  79#endif
  80#ifndef GET_FPEXC_CTL
  81# define GET_FPEXC_CTL(a,b)     (-EINVAL)
  82#endif
  83#ifndef GET_ENDIAN
  84# define GET_ENDIAN(a,b)        (-EINVAL)
  85#endif
  86#ifndef SET_ENDIAN
  87# define SET_ENDIAN(a,b)        (-EINVAL)
  88#endif
  89#ifndef GET_TSC_CTL
  90# define GET_TSC_CTL(a)         (-EINVAL)
  91#endif
  92#ifndef SET_TSC_CTL
  93# define SET_TSC_CTL(a)         (-EINVAL)
  94#endif
  95
  96/*
  97 * this is where the system-wide overflow UID and GID are defined, for
  98 * architectures that now have 32-bit UID/GID but didn't in the past
  99 */
 100
 101int overflowuid = DEFAULT_OVERFLOWUID;
 102int overflowgid = DEFAULT_OVERFLOWGID;
 103
 104EXPORT_SYMBOL(overflowuid);
 105EXPORT_SYMBOL(overflowgid);
 106
 107/*
 108 * the same as above, but for filesystems which can only store a 16-bit
 109 * UID and GID. as such, this is needed on all architectures
 110 */
 111
 112int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
 113int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
 114
 115EXPORT_SYMBOL(fs_overflowuid);
 116EXPORT_SYMBOL(fs_overflowgid);
 117
 118/*
 119 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
 120 */
 121
 122int C_A_D = 1;
 123struct pid *cad_pid;
 124EXPORT_SYMBOL(cad_pid);
 125
 126/*
 127 * If set, this is used for preparing the system to power off.
 128 */
 129
 130void (*pm_power_off_prepare)(void);
 131
 132/*
 133 * Returns true if current's euid is same as p's uid or euid,
 134 * or has CAP_SYS_NICE to p's user_ns.
 135 *
 136 * Called with rcu_read_lock, creds are safe
 137 */
 138static bool set_one_prio_perm(struct task_struct *p)
 139{
 140        const struct cred *cred = current_cred(), *pcred = __task_cred(p);
 141
 142        if (uid_eq(pcred->uid,  cred->euid) ||
 143            uid_eq(pcred->euid, cred->euid))
 144                return true;
 145        if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
 146                return true;
 147        return false;
 148}
 149
 150/*
 151 * set the priority of a task
 152 * - the caller must hold the RCU read lock
 153 */
 154static int set_one_prio(struct task_struct *p, int niceval, int error)
 155{
 156        int no_nice;
 157
 158        if (!set_one_prio_perm(p)) {
 159                error = -EPERM;
 160                goto out;
 161        }
 162        if (niceval < task_nice(p) && !can_nice(p, niceval)) {
 163                error = -EACCES;
 164                goto out;
 165        }
 166        no_nice = security_task_setnice(p, niceval);
 167        if (no_nice) {
 168                error = no_nice;
 169                goto out;
 170        }
 171        if (error == -ESRCH)
 172                error = 0;
 173        set_user_nice(p, niceval);
 174out:
 175        return error;
 176}
 177
 178SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
 179{
 180        struct task_struct *g, *p;
 181        struct user_struct *user;
 182        const struct cred *cred = current_cred();
 183        int error = -EINVAL;
 184        struct pid *pgrp;
 185        kuid_t uid;
 186
 187        if (which > PRIO_USER || which < PRIO_PROCESS)
 188                goto out;
 189
 190        /* normalize: avoid signed division (rounding problems) */
 191        error = -ESRCH;
 192        if (niceval < -20)
 193                niceval = -20;
 194        if (niceval > 19)
 195                niceval = 19;
 196
 197        rcu_read_lock();
 198        read_lock(&tasklist_lock);
 199        switch (which) {
 200                case PRIO_PROCESS:
 201                        if (who)
 202                                p = find_task_by_vpid(who);
 203                        else
 204                                p = current;
 205                        if (p)
 206                                error = set_one_prio(p, niceval, error);
 207                        break;
 208                case PRIO_PGRP:
 209                        if (who)
 210                                pgrp = find_vpid(who);
 211                        else
 212                                pgrp = task_pgrp(current);
 213                        do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
 214                                error = set_one_prio(p, niceval, error);
 215                        } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
 216                        break;
 217                case PRIO_USER:
 218                        uid = make_kuid(cred->user_ns, who);
 219                        user = cred->user;
 220                        if (!who)
 221                                uid = cred->uid;
 222                        else if (!uid_eq(uid, cred->uid) &&
 223                                 !(user = find_user(uid)))
 224                                goto out_unlock;        /* No processes for this user */
 225
 226                        do_each_thread(g, p) {
 227                                if (uid_eq(task_uid(p), uid))
 228                                        error = set_one_prio(p, niceval, error);
 229                        } while_each_thread(g, p);
 230                        if (!uid_eq(uid, cred->uid))
 231                                free_uid(user);         /* For find_user() */
 232                        break;
 233        }
 234out_unlock:
 235        read_unlock(&tasklist_lock);
 236        rcu_read_unlock();
 237out:
 238        return error;
 239}
 240
 241/*
 242 * Ugh. To avoid negative return values, "getpriority()" will
 243 * not return the normal nice-value, but a negated value that
 244 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 245 * to stay compatible.
 246 */
 247SYSCALL_DEFINE2(getpriority, int, which, int, who)
 248{
 249        struct task_struct *g, *p;
 250        struct user_struct *user;
 251        const struct cred *cred = current_cred();
 252        long niceval, retval = -ESRCH;
 253        struct pid *pgrp;
 254        kuid_t uid;
 255
 256        if (which > PRIO_USER || which < PRIO_PROCESS)
 257                return -EINVAL;
 258
 259        rcu_read_lock();
 260        read_lock(&tasklist_lock);
 261        switch (which) {
 262                case PRIO_PROCESS:
 263                        if (who)
 264                                p = find_task_by_vpid(who);
 265                        else
 266                                p = current;
 267                        if (p) {
 268                                niceval = 20 - task_nice(p);
 269                                if (niceval > retval)
 270                                        retval = niceval;
 271                        }
 272                        break;
 273                case PRIO_PGRP:
 274                        if (who)
 275                                pgrp = find_vpid(who);
 276                        else
 277                                pgrp = task_pgrp(current);
 278                        do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
 279                                niceval = 20 - task_nice(p);
 280                                if (niceval > retval)
 281                                        retval = niceval;
 282                        } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
 283                        break;
 284                case PRIO_USER:
 285                        uid = make_kuid(cred->user_ns, who);
 286                        user = cred->user;
 287                        if (!who)
 288                                uid = cred->uid;
 289                        else if (!uid_eq(uid, cred->uid) &&
 290                                 !(user = find_user(uid)))
 291                                goto out_unlock;        /* No processes for this user */
 292
 293                        do_each_thread(g, p) {
 294                                if (uid_eq(task_uid(p), uid)) {
 295                                        niceval = 20 - task_nice(p);
 296                                        if (niceval > retval)
 297                                                retval = niceval;
 298                                }
 299                        } while_each_thread(g, p);
 300                        if (!uid_eq(uid, cred->uid))
 301                                free_uid(user);         /* for find_user() */
 302                        break;
 303        }
 304out_unlock:
 305        read_unlock(&tasklist_lock);
 306        rcu_read_unlock();
 307
 308        return retval;
 309}
 310
 311/**
 312 *      emergency_restart - reboot the system
 313 *
 314 *      Without shutting down any hardware or taking any locks
 315 *      reboot the system.  This is called when we know we are in
 316 *      trouble so this is our best effort to reboot.  This is
 317 *      safe to call in interrupt context.
 318 */
 319void emergency_restart(void)
 320{
 321        kmsg_dump(KMSG_DUMP_EMERG);
 322        machine_emergency_restart();
 323}
 324EXPORT_SYMBOL_GPL(emergency_restart);
 325
 326void kernel_restart_prepare(char *cmd)
 327{
 328        blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
 329        system_state = SYSTEM_RESTART;
 330        usermodehelper_disable();
 331        device_shutdown();
 332}
 333
 334/**
 335 *      register_reboot_notifier - Register function to be called at reboot time
 336 *      @nb: Info about notifier function to be called
 337 *
 338 *      Registers a function with the list of functions
 339 *      to be called at reboot time.
 340 *
 341 *      Currently always returns zero, as blocking_notifier_chain_register()
 342 *      always returns zero.
 343 */
 344int register_reboot_notifier(struct notifier_block *nb)
 345{
 346        return blocking_notifier_chain_register(&reboot_notifier_list, nb);
 347}
 348EXPORT_SYMBOL(register_reboot_notifier);
 349
 350/**
 351 *      unregister_reboot_notifier - Unregister previously registered reboot notifier
 352 *      @nb: Hook to be unregistered
 353 *
 354 *      Unregisters a previously registered reboot
 355 *      notifier function.
 356 *
 357 *      Returns zero on success, or %-ENOENT on failure.
 358 */
 359int unregister_reboot_notifier(struct notifier_block *nb)
 360{
 361        return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
 362}
 363EXPORT_SYMBOL(unregister_reboot_notifier);
 364
 365/* Add backwards compatibility for stable trees. */
 366#ifndef PF_NO_SETAFFINITY
 367#define PF_NO_SETAFFINITY               PF_THREAD_BOUND
 368#endif
 369
 370static void migrate_to_reboot_cpu(void)
 371{
 372        /* The boot cpu is always logical cpu 0 */
 373        int cpu = 0;
 374
 375        cpu_hotplug_disable();
 376
 377        /* Make certain the cpu I'm about to reboot on is online */
 378        if (!cpu_online(cpu))
 379                cpu = cpumask_first(cpu_online_mask);
 380
 381        /* Prevent races with other tasks migrating this task */
 382        current->flags |= PF_NO_SETAFFINITY;
 383
 384        /* Make certain I only run on the appropriate processor */
 385        set_cpus_allowed_ptr(current, cpumask_of(cpu));
 386}
 387
 388/**
 389 *      kernel_restart - reboot the system
 390 *      @cmd: pointer to buffer containing command to execute for restart
 391 *              or %NULL
 392 *
 393 *      Shutdown everything and perform a clean reboot.
 394 *      This is not safe to call in interrupt context.
 395 */
 396void kernel_restart(char *cmd)
 397{
 398        kernel_restart_prepare(cmd);
 399        migrate_to_reboot_cpu();
 400        syscore_shutdown();
 401        if (!cmd)
 402                printk(KERN_EMERG "Restarting system.\n");
 403        else
 404                printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
 405        kmsg_dump(KMSG_DUMP_RESTART);
 406        machine_restart(cmd);
 407}
 408EXPORT_SYMBOL_GPL(kernel_restart);
 409
 410static void kernel_shutdown_prepare(enum system_states state)
 411{
 412        blocking_notifier_call_chain(&reboot_notifier_list,
 413                (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
 414        system_state = state;
 415        usermodehelper_disable();
 416        device_shutdown();
 417}
 418/**
 419 *      kernel_halt - halt the system
 420 *
 421 *      Shutdown everything and perform a clean system halt.
 422 */
 423void kernel_halt(void)
 424{
 425        kernel_shutdown_prepare(SYSTEM_HALT);
 426        migrate_to_reboot_cpu();
 427        syscore_shutdown();
 428        printk(KERN_EMERG "System halted.\n");
 429        kmsg_dump(KMSG_DUMP_HALT);
 430        machine_halt();
 431}
 432
 433EXPORT_SYMBOL_GPL(kernel_halt);
 434
 435/**
 436 *      kernel_power_off - power_off the system
 437 *
 438 *      Shutdown everything and perform a clean system power_off.
 439 */
 440void kernel_power_off(void)
 441{
 442        kernel_shutdown_prepare(SYSTEM_POWER_OFF);
 443        if (pm_power_off_prepare)
 444                pm_power_off_prepare();
 445        migrate_to_reboot_cpu();
 446        syscore_shutdown();
 447        printk(KERN_EMERG "Power down.\n");
 448        kmsg_dump(KMSG_DUMP_POWEROFF);
 449        machine_power_off();
 450}
 451EXPORT_SYMBOL_GPL(kernel_power_off);
 452
 453static DEFINE_MUTEX(reboot_mutex);
 454
 455/*
 456 * Reboot system call: for obvious reasons only root may call it,
 457 * and even root needs to set up some magic numbers in the registers
 458 * so that some mistake won't make this reboot the whole machine.
 459 * You can also set the meaning of the ctrl-alt-del-key here.
 460 *
 461 * reboot doesn't sync: do that yourself before calling this.
 462 */
 463SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
 464                void __user *, arg)
 465{
 466        struct pid_namespace *pid_ns = task_active_pid_ns(current);
 467        char buffer[256];
 468        int ret = 0;
 469
 470        /* We only trust the superuser with rebooting the system. */
 471        if (!ns_capable(pid_ns->user_ns, CAP_SYS_BOOT))
 472                return -EPERM;
 473
 474        /* For safety, we require "magic" arguments. */
 475        if (magic1 != LINUX_REBOOT_MAGIC1 ||
 476            (magic2 != LINUX_REBOOT_MAGIC2 &&
 477                        magic2 != LINUX_REBOOT_MAGIC2A &&
 478                        magic2 != LINUX_REBOOT_MAGIC2B &&
 479                        magic2 != LINUX_REBOOT_MAGIC2C))
 480                return -EINVAL;
 481
 482        /*
 483         * If pid namespaces are enabled and the current task is in a child
 484         * pid_namespace, the command is handled by reboot_pid_ns() which will
 485         * call do_exit().
 486         */
 487        ret = reboot_pid_ns(pid_ns, cmd);
 488        if (ret)
 489                return ret;
 490
 491        /* Instead of trying to make the power_off code look like
 492         * halt when pm_power_off is not set do it the easy way.
 493         */
 494        if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
 495                cmd = LINUX_REBOOT_CMD_HALT;
 496
 497        mutex_lock(&reboot_mutex);
 498        switch (cmd) {
 499        case LINUX_REBOOT_CMD_RESTART:
 500                kernel_restart(NULL);
 501                break;
 502
 503        case LINUX_REBOOT_CMD_CAD_ON:
 504                C_A_D = 1;
 505                break;
 506
 507        case LINUX_REBOOT_CMD_CAD_OFF:
 508                C_A_D = 0;
 509                break;
 510
 511        case LINUX_REBOOT_CMD_HALT:
 512                kernel_halt();
 513                do_exit(0);
 514                panic("cannot halt");
 515
 516        case LINUX_REBOOT_CMD_POWER_OFF:
 517                kernel_power_off();
 518                do_exit(0);
 519                break;
 520
 521        case LINUX_REBOOT_CMD_RESTART2:
 522                if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
 523                        ret = -EFAULT;
 524                        break;
 525                }
 526                buffer[sizeof(buffer) - 1] = '\0';
 527
 528                kernel_restart(buffer);
 529                break;
 530
 531#ifdef CONFIG_KEXEC
 532        case LINUX_REBOOT_CMD_KEXEC:
 533                ret = kernel_kexec();
 534                break;
 535#endif
 536
 537#ifdef CONFIG_HIBERNATION
 538        case LINUX_REBOOT_CMD_SW_SUSPEND:
 539                ret = hibernate();
 540                break;
 541#endif
 542
 543        default:
 544                ret = -EINVAL;
 545                break;
 546        }
 547        mutex_unlock(&reboot_mutex);
 548        return ret;
 549}
 550
 551static void deferred_cad(struct work_struct *dummy)
 552{
 553        kernel_restart(NULL);
 554}
 555
 556/*
 557 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
 558 * As it's called within an interrupt, it may NOT sync: the only choice
 559 * is whether to reboot at once, or just ignore the ctrl-alt-del.
 560 */
 561void ctrl_alt_del(void)
 562{
 563        static DECLARE_WORK(cad_work, deferred_cad);
 564
 565        if (C_A_D)
 566                schedule_work(&cad_work);
 567        else
 568                kill_cad_pid(SIGINT, 1);
 569}
 570        
 571/*
 572 * Unprivileged users may change the real gid to the effective gid
 573 * or vice versa.  (BSD-style)
 574 *
 575 * If you set the real gid at all, or set the effective gid to a value not
 576 * equal to the real gid, then the saved gid is set to the new effective gid.
 577 *
 578 * This makes it possible for a setgid program to completely drop its
 579 * privileges, which is often a useful assertion to make when you are doing
 580 * a security audit over a program.
 581 *
 582 * The general idea is that a program which uses just setregid() will be
 583 * 100% compatible with BSD.  A program which uses just setgid() will be
 584 * 100% compatible with POSIX with saved IDs. 
 585 *
 586 * SMP: There are not races, the GIDs are checked only by filesystem
 587 *      operations (as far as semantic preservation is concerned).
 588 */
 589SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
 590{
 591        struct user_namespace *ns = current_user_ns();
 592        const struct cred *old;
 593        struct cred *new;
 594        int retval;
 595        kgid_t krgid, kegid;
 596
 597        krgid = make_kgid(ns, rgid);
 598        kegid = make_kgid(ns, egid);
 599
 600        if ((rgid != (gid_t) -1) && !gid_valid(krgid))
 601                return -EINVAL;
 602        if ((egid != (gid_t) -1) && !gid_valid(kegid))
 603                return -EINVAL;
 604
 605        new = prepare_creds();
 606        if (!new)
 607                return -ENOMEM;
 608        old = current_cred();
 609
 610        retval = -EPERM;
 611        if (rgid != (gid_t) -1) {
 612                if (gid_eq(old->gid, krgid) ||
 613                    gid_eq(old->egid, krgid) ||
 614                    nsown_capable(CAP_SETGID))
 615                        new->gid = krgid;
 616                else
 617                        goto error;
 618        }
 619        if (egid != (gid_t) -1) {
 620                if (gid_eq(old->gid, kegid) ||
 621                    gid_eq(old->egid, kegid) ||
 622                    gid_eq(old->sgid, kegid) ||
 623                    nsown_capable(CAP_SETGID))
 624                        new->egid = kegid;
 625                else
 626                        goto error;
 627        }
 628
 629        if (rgid != (gid_t) -1 ||
 630            (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
 631                new->sgid = new->egid;
 632        new->fsgid = new->egid;
 633
 634        return commit_creds(new);
 635
 636error:
 637        abort_creds(new);
 638        return retval;
 639}
 640
 641/*
 642 * setgid() is implemented like SysV w/ SAVED_IDS 
 643 *
 644 * SMP: Same implicit races as above.
 645 */
 646SYSCALL_DEFINE1(setgid, gid_t, gid)
 647{
 648        struct user_namespace *ns = current_user_ns();
 649        const struct cred *old;
 650        struct cred *new;
 651        int retval;
 652        kgid_t kgid;
 653
 654        kgid = make_kgid(ns, gid);
 655        if (!gid_valid(kgid))
 656                return -EINVAL;
 657
 658        new = prepare_creds();
 659        if (!new)
 660                return -ENOMEM;
 661        old = current_cred();
 662
 663        retval = -EPERM;
 664        if (nsown_capable(CAP_SETGID))
 665                new->gid = new->egid = new->sgid = new->fsgid = kgid;
 666        else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
 667                new->egid = new->fsgid = kgid;
 668        else
 669                goto error;
 670
 671        return commit_creds(new);
 672
 673error:
 674        abort_creds(new);
 675        return retval;
 676}
 677
 678/*
 679 * change the user struct in a credentials set to match the new UID
 680 */
 681static int set_user(struct cred *new)
 682{
 683        struct user_struct *new_user;
 684
 685        new_user = alloc_uid(new->uid);
 686        if (!new_user)
 687                return -EAGAIN;
 688
 689        /*
 690         * We don't fail in case of NPROC limit excess here because too many
 691         * poorly written programs don't check set*uid() return code, assuming
 692         * it never fails if called by root.  We may still enforce NPROC limit
 693         * for programs doing set*uid()+execve() by harmlessly deferring the
 694         * failure to the execve() stage.
 695         */
 696        if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
 697                        new_user != INIT_USER)
 698                current->flags |= PF_NPROC_EXCEEDED;
 699        else
 700                current->flags &= ~PF_NPROC_EXCEEDED;
 701
 702        free_uid(new->user);
 703        new->user = new_user;
 704        return 0;
 705}
 706
 707/*
 708 * Unprivileged users may change the real uid to the effective uid
 709 * or vice versa.  (BSD-style)
 710 *
 711 * If you set the real uid at all, or set the effective uid to a value not
 712 * equal to the real uid, then the saved uid is set to the new effective uid.
 713 *
 714 * This makes it possible for a setuid program to completely drop its
 715 * privileges, which is often a useful assertion to make when you are doing
 716 * a security audit over a program.
 717 *
 718 * The general idea is that a program which uses just setreuid() will be
 719 * 100% compatible with BSD.  A program which uses just setuid() will be
 720 * 100% compatible with POSIX with saved IDs. 
 721 */
 722SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
 723{
 724        struct user_namespace *ns = current_user_ns();
 725        const struct cred *old;
 726        struct cred *new;
 727        int retval;
 728        kuid_t kruid, keuid;
 729
 730        kruid = make_kuid(ns, ruid);
 731        keuid = make_kuid(ns, euid);
 732
 733        if ((ruid != (uid_t) -1) && !uid_valid(kruid))
 734                return -EINVAL;
 735        if ((euid != (uid_t) -1) && !uid_valid(keuid))
 736                return -EINVAL;
 737
 738        new = prepare_creds();
 739        if (!new)
 740                return -ENOMEM;
 741        old = current_cred();
 742
 743        retval = -EPERM;
 744        if (ruid != (uid_t) -1) {
 745                new->uid = kruid;
 746                if (!uid_eq(old->uid, kruid) &&
 747                    !uid_eq(old->euid, kruid) &&
 748                    !nsown_capable(CAP_SETUID))
 749                        goto error;
 750        }
 751
 752        if (euid != (uid_t) -1) {
 753                new->euid = keuid;
 754                if (!uid_eq(old->uid, keuid) &&
 755                    !uid_eq(old->euid, keuid) &&
 756                    !uid_eq(old->suid, keuid) &&
 757                    !nsown_capable(CAP_SETUID))
 758                        goto error;
 759        }
 760
 761        if (!uid_eq(new->uid, old->uid)) {
 762                retval = set_user(new);
 763                if (retval < 0)
 764                        goto error;
 765        }
 766        if (ruid != (uid_t) -1 ||
 767            (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
 768                new->suid = new->euid;
 769        new->fsuid = new->euid;
 770
 771        retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
 772        if (retval < 0)
 773                goto error;
 774
 775        return commit_creds(new);
 776
 777error:
 778        abort_creds(new);
 779        return retval;
 780}
 781                
 782/*
 783 * setuid() is implemented like SysV with SAVED_IDS 
 784 * 
 785 * Note that SAVED_ID's is deficient in that a setuid root program
 786 * like sendmail, for example, cannot set its uid to be a normal 
 787 * user and then switch back, because if you're root, setuid() sets
 788 * the saved uid too.  If you don't like this, blame the bright people
 789 * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 790 * will allow a root program to temporarily drop privileges and be able to
 791 * regain them by swapping the real and effective uid.  
 792 */
 793SYSCALL_DEFINE1(setuid, uid_t, uid)
 794{
 795        struct user_namespace *ns = current_user_ns();
 796        const struct cred *old;
 797        struct cred *new;
 798        int retval;
 799        kuid_t kuid;
 800
 801        kuid = make_kuid(ns, uid);
 802        if (!uid_valid(kuid))
 803                return -EINVAL;
 804
 805        new = prepare_creds();
 806        if (!new)
 807                return -ENOMEM;
 808        old = current_cred();
 809
 810        retval = -EPERM;
 811        if (nsown_capable(CAP_SETUID)) {
 812                new->suid = new->uid = kuid;
 813                if (!uid_eq(kuid, old->uid)) {
 814                        retval = set_user(new);
 815                        if (retval < 0)
 816                                goto error;
 817                }
 818        } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
 819                goto error;
 820        }
 821
 822        new->fsuid = new->euid = kuid;
 823
 824        retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
 825        if (retval < 0)
 826                goto error;
 827
 828        return commit_creds(new);
 829
 830error:
 831        abort_creds(new);
 832        return retval;
 833}
 834
 835
 836/*
 837 * This function implements a generic ability to update ruid, euid,
 838 * and suid.  This allows you to implement the 4.4 compatible seteuid().
 839 */
 840SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
 841{
 842        struct user_namespace *ns = current_user_ns();
 843        const struct cred *old;
 844        struct cred *new;
 845        int retval;
 846        kuid_t kruid, keuid, ksuid;
 847
 848        kruid = make_kuid(ns, ruid);
 849        keuid = make_kuid(ns, euid);
 850        ksuid = make_kuid(ns, suid);
 851
 852        if ((ruid != (uid_t) -1) && !uid_valid(kruid))
 853                return -EINVAL;
 854
 855        if ((euid != (uid_t) -1) && !uid_valid(keuid))
 856                return -EINVAL;
 857
 858        if ((suid != (uid_t) -1) && !uid_valid(ksuid))
 859                return -EINVAL;
 860
 861        new = prepare_creds();
 862        if (!new)
 863                return -ENOMEM;
 864
 865        old = current_cred();
 866
 867        retval = -EPERM;
 868        if (!nsown_capable(CAP_SETUID)) {
 869                if (ruid != (uid_t) -1        && !uid_eq(kruid, old->uid) &&
 870                    !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
 871                        goto error;
 872                if (euid != (uid_t) -1        && !uid_eq(keuid, old->uid) &&
 873                    !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
 874                        goto error;
 875                if (suid != (uid_t) -1        && !uid_eq(ksuid, old->uid) &&
 876                    !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
 877                        goto error;
 878        }
 879
 880        if (ruid != (uid_t) -1) {
 881                new->uid = kruid;
 882                if (!uid_eq(kruid, old->uid)) {
 883                        retval = set_user(new);
 884                        if (retval < 0)
 885                                goto error;
 886                }
 887        }
 888        if (euid != (uid_t) -1)
 889                new->euid = keuid;
 890        if (suid != (uid_t) -1)
 891                new->suid = ksuid;
 892        new->fsuid = new->euid;
 893
 894        retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
 895        if (retval < 0)
 896                goto error;
 897
 898        return commit_creds(new);
 899
 900error:
 901        abort_creds(new);
 902        return retval;
 903}
 904
 905SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
 906{
 907        const struct cred *cred = current_cred();
 908        int retval;
 909        uid_t ruid, euid, suid;
 910
 911        ruid = from_kuid_munged(cred->user_ns, cred->uid);
 912        euid = from_kuid_munged(cred->user_ns, cred->euid);
 913        suid = from_kuid_munged(cred->user_ns, cred->suid);
 914
 915        if (!(retval   = put_user(ruid, ruidp)) &&
 916            !(retval   = put_user(euid, euidp)))
 917                retval = put_user(suid, suidp);
 918
 919        return retval;
 920}
 921
 922/*
 923 * Same as above, but for rgid, egid, sgid.
 924 */
 925SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
 926{
 927        struct user_namespace *ns = current_user_ns();
 928        const struct cred *old;
 929        struct cred *new;
 930        int retval;
 931        kgid_t krgid, kegid, ksgid;
 932
 933        krgid = make_kgid(ns, rgid);
 934        kegid = make_kgid(ns, egid);
 935        ksgid = make_kgid(ns, sgid);
 936
 937        if ((rgid != (gid_t) -1) && !gid_valid(krgid))
 938                return -EINVAL;
 939        if ((egid != (gid_t) -1) && !gid_valid(kegid))
 940                return -EINVAL;
 941        if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
 942                return -EINVAL;
 943
 944        new = prepare_creds();
 945        if (!new)
 946                return -ENOMEM;
 947        old = current_cred();
 948
 949        retval = -EPERM;
 950        if (!nsown_capable(CAP_SETGID)) {
 951                if (rgid != (gid_t) -1        && !gid_eq(krgid, old->gid) &&
 952                    !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
 953                        goto error;
 954                if (egid != (gid_t) -1        && !gid_eq(kegid, old->gid) &&
 955                    !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
 956                        goto error;
 957                if (sgid != (gid_t) -1        && !gid_eq(ksgid, old->gid) &&
 958                    !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
 959                        goto error;
 960        }
 961
 962        if (rgid != (gid_t) -1)
 963                new->gid = krgid;
 964        if (egid != (gid_t) -1)
 965                new->egid = kegid;
 966        if (sgid != (gid_t) -1)
 967                new->sgid = ksgid;
 968        new->fsgid = new->egid;
 969
 970        return commit_creds(new);
 971
 972error:
 973        abort_creds(new);
 974        return retval;
 975}
 976
 977SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
 978{
 979        const struct cred *cred = current_cred();
 980        int retval;
 981        gid_t rgid, egid, sgid;
 982
 983        rgid = from_kgid_munged(cred->user_ns, cred->gid);
 984        egid = from_kgid_munged(cred->user_ns, cred->egid);
 985        sgid = from_kgid_munged(cred->user_ns, cred->sgid);
 986
 987        if (!(retval   = put_user(rgid, rgidp)) &&
 988            !(retval   = put_user(egid, egidp)))
 989                retval = put_user(sgid, sgidp);
 990
 991        return retval;
 992}
 993
 994
 995/*
 996 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 997 * is used for "access()" and for the NFS daemon (letting nfsd stay at
 998 * whatever uid it wants to). It normally shadows "euid", except when
 999 * explicitly set by setfsuid() or for access..
1000 */
1001SYSCALL_DEFINE1(setfsuid, uid_t, uid)
1002{
1003        const struct cred *old;
1004        struct cred *new;
1005        uid_t old_fsuid;
1006        kuid_t kuid;
1007
1008        old = current_cred();
1009        old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
1010
1011        kuid = make_kuid(old->user_ns, uid);
1012        if (!uid_valid(kuid))
1013                return old_fsuid;
1014
1015        new = prepare_creds();
1016        if (!new)
1017                return old_fsuid;
1018
1019        if (uid_eq(kuid, old->uid)  || uid_eq(kuid, old->euid)  ||
1020            uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
1021            nsown_capable(CAP_SETUID)) {
1022                if (!uid_eq(kuid, old->fsuid)) {
1023                        new->fsuid = kuid;
1024                        if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
1025                                goto change_okay;
1026                }
1027        }
1028
1029        abort_creds(new);
1030        return old_fsuid;
1031
1032change_okay:
1033        commit_creds(new);
1034        return old_fsuid;
1035}
1036
1037/*
1038 * Samma på svenska..
1039 */
1040SYSCALL_DEFINE1(setfsgid, gid_t, gid)
1041{
1042        const struct cred *old;
1043        struct cred *new;
1044        gid_t old_fsgid;
1045        kgid_t kgid;
1046
1047        old = current_cred();
1048        old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
1049
1050        kgid = make_kgid(old->user_ns, gid);
1051        if (!gid_valid(kgid))
1052                return old_fsgid;
1053
1054        new = prepare_creds();
1055        if (!new)
1056                return old_fsgid;
1057
1058        if (gid_eq(kgid, old->gid)  || gid_eq(kgid, old->egid)  ||
1059            gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
1060            nsown_capable(CAP_SETGID)) {
1061                if (!gid_eq(kgid, old->fsgid)) {
1062                        new->fsgid = kgid;
1063                        goto change_okay;
1064                }
1065        }
1066
1067        abort_creds(new);
1068        return old_fsgid;
1069
1070change_okay:
1071        commit_creds(new);
1072        return old_fsgid;
1073}
1074
1075/**
1076 * sys_getpid - return the thread group id of the current process
1077 *
1078 * Note, despite the name, this returns the tgid not the pid.  The tgid and
1079 * the pid are identical unless CLONE_THREAD was specified on clone() in
1080 * which case the tgid is the same in all threads of the same group.
1081 *
1082 * This is SMP safe as current->tgid does not change.
1083 */
1084SYSCALL_DEFINE0(getpid)
1085{
1086        return task_tgid_vnr(current);
1087}
1088
1089/* Thread ID - the internal kernel "pid" */
1090SYSCALL_DEFINE0(gettid)
1091{
1092        return task_pid_vnr(current);
1093}
1094
1095/*
1096 * Accessing ->real_parent is not SMP-safe, it could
1097 * change from under us. However, we can use a stale
1098 * value of ->real_parent under rcu_read_lock(), see
1099 * release_task()->call_rcu(delayed_put_task_struct).
1100 */
1101SYSCALL_DEFINE0(getppid)
1102{
1103        int pid;
1104
1105        rcu_read_lock();
1106        pid = task_tgid_vnr(rcu_dereference(current->real_parent));
1107        rcu_read_unlock();
1108
1109        return pid;
1110}
1111
1112SYSCALL_DEFINE0(getuid)
1113{
1114        /* Only we change this so SMP safe */
1115        return from_kuid_munged(current_user_ns(), current_uid());
1116}
1117
1118SYSCALL_DEFINE0(geteuid)
1119{
1120        /* Only we change this so SMP safe */
1121        return from_kuid_munged(current_user_ns(), current_euid());
1122}
1123
1124SYSCALL_DEFINE0(getgid)
1125{
1126        /* Only we change this so SMP safe */
1127        return from_kgid_munged(current_user_ns(), current_gid());
1128}
1129
1130SYSCALL_DEFINE0(getegid)
1131{
1132        /* Only we change this so SMP safe */
1133        return from_kgid_munged(current_user_ns(), current_egid());
1134}
1135
1136void do_sys_times(struct tms *tms)
1137{
1138        cputime_t tgutime, tgstime, cutime, cstime;
1139
1140        spin_lock_irq(&current->sighand->siglock);
1141        thread_group_cputime_adjusted(current, &tgutime, &tgstime);
1142        cutime = current->signal->cutime;
1143        cstime = current->signal->cstime;
1144        spin_unlock_irq(&current->sighand->siglock);
1145        tms->tms_utime = cputime_to_clock_t(tgutime);
1146        tms->tms_stime = cputime_to_clock_t(tgstime);
1147        tms->tms_cutime = cputime_to_clock_t(cutime);
1148        tms->tms_cstime = cputime_to_clock_t(cstime);
1149}
1150
1151SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
1152{
1153        if (tbuf) {
1154                struct tms tmp;
1155
1156                do_sys_times(&tmp);
1157                if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1158                        return -EFAULT;
1159        }
1160        force_successful_syscall_return();
1161        return (long) jiffies_64_to_clock_t(get_jiffies_64());
1162}
1163
1164/*
1165 * This needs some heavy checking ...
1166 * I just haven't the stomach for it. I also don't fully
1167 * understand sessions/pgrp etc. Let somebody who does explain it.
1168 *
1169 * OK, I think I have the protection semantics right.... this is really
1170 * only important on a multi-user system anyway, to make sure one user
1171 * can't send a signal to a process owned by another.  -TYT, 12/12/91
1172 *
1173 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1174 * LBT 04.03.94
1175 */
1176SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1177{
1178        struct task_struct *p;
1179        struct task_struct *group_leader = current->group_leader;
1180        struct pid *pgrp;
1181        int err;
1182
1183        if (!pid)
1184                pid = task_pid_vnr(group_leader);
1185        if (!pgid)
1186                pgid = pid;
1187        if (pgid < 0)
1188                return -EINVAL;
1189        rcu_read_lock();
1190
1191        /* From this point forward we keep holding onto the tasklist lock
1192         * so that our parent does not change from under us. -DaveM
1193         */
1194        write_lock_irq(&tasklist_lock);
1195
1196        err = -ESRCH;
1197        p = find_task_by_vpid(pid);
1198        if (!p)
1199                goto out;
1200
1201        err = -EINVAL;
1202        if (!thread_group_leader(p))
1203                goto out;
1204
1205        if (same_thread_group(p->real_parent, group_leader)) {
1206                err = -EPERM;
1207                if (task_session(p) != task_session(group_leader))
1208                        goto out;
1209                err = -EACCES;
1210                if (p->did_exec)
1211                        goto out;
1212        } else {
1213                err = -ESRCH;
1214                if (p != group_leader)
1215                        goto out;
1216        }
1217
1218        err = -EPERM;
1219        if (p->signal->leader)
1220                goto out;
1221
1222        pgrp = task_pid(p);
1223        if (pgid != pid) {
1224                struct task_struct *g;
1225
1226                pgrp = find_vpid(pgid);
1227                g = pid_task(pgrp, PIDTYPE_PGID);
1228                if (!g || task_session(g) != task_session(group_leader))
1229                        goto out;
1230        }
1231
1232        err = security_task_setpgid(p, pgid);
1233        if (err)
1234                goto out;
1235
1236        if (task_pgrp(p) != pgrp)
1237                change_pid(p, PIDTYPE_PGID, pgrp);
1238
1239        err = 0;
1240out:
1241        /* All paths lead to here, thus we are safe. -DaveM */
1242        write_unlock_irq(&tasklist_lock);
1243        rcu_read_unlock();
1244        return err;
1245}
1246
1247SYSCALL_DEFINE1(getpgid, pid_t, pid)
1248{
1249        struct task_struct *p;
1250        struct pid *grp;
1251        int retval;
1252
1253        rcu_read_lock();
1254        if (!pid)
1255                grp = task_pgrp(current);
1256        else {
1257                retval = -ESRCH;
1258                p = find_task_by_vpid(pid);
1259                if (!p)
1260                        goto out;
1261                grp = task_pgrp(p);
1262                if (!grp)
1263                        goto out;
1264
1265                retval = security_task_getpgid(p);
1266                if (retval)
1267                        goto out;
1268        }
1269        retval = pid_vnr(grp);
1270out:
1271        rcu_read_unlock();
1272        return retval;
1273}
1274
1275#ifdef __ARCH_WANT_SYS_GETPGRP
1276
1277SYSCALL_DEFINE0(getpgrp)
1278{
1279        return sys_getpgid(0);
1280}
1281
1282#endif
1283
1284SYSCALL_DEFINE1(getsid, pid_t, pid)
1285{
1286        struct task_struct *p;
1287        struct pid *sid;
1288        int retval;
1289
1290        rcu_read_lock();
1291        if (!pid)
1292                sid = task_session(current);
1293        else {
1294                retval = -ESRCH;
1295                p = find_task_by_vpid(pid);
1296                if (!p)
1297                        goto out;
1298                sid = task_session(p);
1299                if (!sid)
1300                        goto out;
1301
1302                retval = security_task_getsid(p);
1303                if (retval)
1304                        goto out;
1305        }
1306        retval = pid_vnr(sid);
1307out:
1308        rcu_read_unlock();
1309        return retval;
1310}
1311
1312SYSCALL_DEFINE0(setsid)
1313{
1314        struct task_struct *group_leader = current->group_leader;
1315        struct pid *sid = task_pid(group_leader);
1316        pid_t session = pid_vnr(sid);
1317        int err = -EPERM;
1318
1319        write_lock_irq(&tasklist_lock);
1320        /* Fail if I am already a session leader */
1321        if (group_leader->signal->leader)
1322                goto out;
1323
1324        /* Fail if a process group id already exists that equals the
1325         * proposed session id.
1326         */
1327        if (pid_task(sid, PIDTYPE_PGID))
1328                goto out;
1329
1330        group_leader->signal->leader = 1;
1331        __set_special_pids(sid);
1332
1333        proc_clear_tty(group_leader);
1334
1335        err = session;
1336out:
1337        write_unlock_irq(&tasklist_lock);
1338        if (err > 0) {
1339                proc_sid_connector(group_leader);
1340                sched_autogroup_create_attach(group_leader);
1341        }
1342        return err;
1343}
1344
1345DECLARE_RWSEM(uts_sem);
1346
1347#ifdef COMPAT_UTS_MACHINE
1348#define override_architecture(name) \
1349        (personality(current->personality) == PER_LINUX32 && \
1350         copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1351                      sizeof(COMPAT_UTS_MACHINE)))
1352#else
1353#define override_architecture(name)     0
1354#endif
1355
1356/*
1357 * Work around broken programs that cannot handle "Linux 3.0".
1358 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1359 */
1360static int override_release(char __user *release, size_t len)
1361{
1362        int ret = 0;
1363
1364        if (current->personality & UNAME26) {
1365                const char *rest = UTS_RELEASE;
1366                char buf[65] = { 0 };
1367                int ndots = 0;
1368                unsigned v;
1369                size_t copy;
1370
1371                while (*rest) {
1372                        if (*rest == '.' && ++ndots >= 3)
1373                                break;
1374                        if (!isdigit(*rest) && *rest != '.')
1375                                break;
1376                        rest++;
1377                }
1378                v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 40;
1379                copy = clamp_t(size_t, len, 1, sizeof(buf));
1380                copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1381                ret = copy_to_user(release, buf, copy + 1);
1382        }
1383        return ret;
1384}
1385
1386SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1387{
1388        int errno = 0;
1389
1390        down_read(&uts_sem);
1391        if (copy_to_user(name, utsname(), sizeof *name))
1392                errno = -EFAULT;
1393        up_read(&uts_sem);
1394
1395        if (!errno && override_release(name->release, sizeof(name->release)))
1396                errno = -EFAULT;
1397        if (!errno && override_architecture(name))
1398                errno = -EFAULT;
1399        return errno;
1400}
1401
1402#ifdef __ARCH_WANT_SYS_OLD_UNAME
1403/*
1404 * Old cruft
1405 */
1406SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1407{
1408        int error = 0;
1409
1410        if (!name)
1411                return -EFAULT;
1412
1413        down_read(&uts_sem);
1414        if (copy_to_user(name, utsname(), sizeof(*name)))
1415                error = -EFAULT;
1416        up_read(&uts_sem);
1417
1418        if (!error && override_release(name->release, sizeof(name->release)))
1419                error = -EFAULT;
1420        if (!error && override_architecture(name))
1421                error = -EFAULT;
1422        return error;
1423}
1424
1425SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1426{
1427        int error;
1428
1429        if (!name)
1430                return -EFAULT;
1431        if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1432                return -EFAULT;
1433
1434        down_read(&uts_sem);
1435        error = __copy_to_user(&name->sysname, &utsname()->sysname,
1436                               __OLD_UTS_LEN);
1437        error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1438        error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1439                                __OLD_UTS_LEN);
1440        error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1441        error |= __copy_to_user(&name->release, &utsname()->release,
1442                                __OLD_UTS_LEN);
1443        error |= __put_user(0, name->release + __OLD_UTS_LEN);
1444        error |= __copy_to_user(&name->version, &utsname()->version,
1445                                __OLD_UTS_LEN);
1446        error |= __put_user(0, name->version + __OLD_UTS_LEN);
1447        error |= __copy_to_user(&name->machine, &utsname()->machine,
1448                                __OLD_UTS_LEN);
1449        error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1450        up_read(&uts_sem);
1451
1452        if (!error && override_architecture(name))
1453                error = -EFAULT;
1454        if (!error && override_release(name->release, sizeof(name->release)))
1455                error = -EFAULT;
1456        return error ? -EFAULT : 0;
1457}
1458#endif
1459
1460SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1461{
1462        int errno;
1463        char tmp[__NEW_UTS_LEN];
1464
1465        if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1466                return -EPERM;
1467
1468        if (len < 0 || len > __NEW_UTS_LEN)
1469                return -EINVAL;
1470        down_write(&uts_sem);
1471        errno = -EFAULT;
1472        if (!copy_from_user(tmp, name, len)) {
1473                struct new_utsname *u = utsname();
1474
1475                memcpy(u->nodename, tmp, len);
1476                memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1477                errno = 0;
1478                uts_proc_notify(UTS_PROC_HOSTNAME);
1479        }
1480        up_write(&uts_sem);
1481        return errno;
1482}
1483
1484#ifdef __ARCH_WANT_SYS_GETHOSTNAME
1485
1486SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1487{
1488        int i, errno;
1489        struct new_utsname *u;
1490
1491        if (len < 0)
1492                return -EINVAL;
1493        down_read(&uts_sem);
1494        u = utsname();
1495        i = 1 + strlen(u->nodename);
1496        if (i > len)
1497                i = len;
1498        errno = 0;
1499        if (copy_to_user(name, u->nodename, i))
1500                errno = -EFAULT;
1501        up_read(&uts_sem);
1502        return errno;
1503}
1504
1505#endif
1506
1507/*
1508 * Only setdomainname; getdomainname can be implemented by calling
1509 * uname()
1510 */
1511SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1512{
1513        int errno;
1514        char tmp[__NEW_UTS_LEN];
1515
1516        if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1517                return -EPERM;
1518        if (len < 0 || len > __NEW_UTS_LEN)
1519                return -EINVAL;
1520
1521        down_write(&uts_sem);
1522        errno = -EFAULT;
1523        if (!copy_from_user(tmp, name, len)) {
1524                struct new_utsname *u = utsname();
1525
1526                memcpy(u->domainname, tmp, len);
1527                memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1528                errno = 0;
1529                uts_proc_notify(UTS_PROC_DOMAINNAME);
1530        }
1531        up_write(&uts_sem);
1532        return errno;
1533}
1534
1535SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1536{
1537        struct rlimit value;
1538        int ret;
1539
1540        ret = do_prlimit(current, resource, NULL, &value);
1541        if (!ret)
1542                ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1543
1544        return ret;
1545}
1546
1547#ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1548
1549/*
1550 *      Back compatibility for getrlimit. Needed for some apps.
1551 */
1552 
1553SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1554                struct rlimit __user *, rlim)
1555{
1556        struct rlimit x;
1557        if (resource >= RLIM_NLIMITS)
1558                return -EINVAL;
1559
1560        task_lock(current->group_leader);
1561        x = current->signal->rlim[resource];
1562        task_unlock(current->group_leader);
1563        if (x.rlim_cur > 0x7FFFFFFF)
1564                x.rlim_cur = 0x7FFFFFFF;
1565        if (x.rlim_max > 0x7FFFFFFF)
1566                x.rlim_max = 0x7FFFFFFF;
1567        return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1568}
1569
1570#endif
1571
1572static inline bool rlim64_is_infinity(__u64 rlim64)
1573{
1574#if BITS_PER_LONG < 64
1575        return rlim64 >= ULONG_MAX;
1576#else
1577        return rlim64 == RLIM64_INFINITY;
1578#endif
1579}
1580
1581static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1582{
1583        if (rlim->rlim_cur == RLIM_INFINITY)
1584                rlim64->rlim_cur = RLIM64_INFINITY;
1585        else
1586                rlim64->rlim_cur = rlim->rlim_cur;
1587        if (rlim->rlim_max == RLIM_INFINITY)
1588                rlim64->rlim_max = RLIM64_INFINITY;
1589        else
1590                rlim64->rlim_max = rlim->rlim_max;
1591}
1592
1593static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1594{
1595        if (rlim64_is_infinity(rlim64->rlim_cur))
1596                rlim->rlim_cur = RLIM_INFINITY;
1597        else
1598                rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1599        if (rlim64_is_infinity(rlim64->rlim_max))
1600                rlim->rlim_max = RLIM_INFINITY;
1601        else
1602                rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1603}
1604
1605/* make sure you are allowed to change @tsk limits before calling this */
1606int do_prlimit(struct task_struct *tsk, unsigned int resource,
1607                struct rlimit *new_rlim, struct rlimit *old_rlim)
1608{
1609        struct rlimit *rlim;
1610        int retval = 0;
1611
1612        if (resource >= RLIM_NLIMITS)
1613                return -EINVAL;
1614        if (new_rlim) {
1615                if (new_rlim->rlim_cur > new_rlim->rlim_max)
1616                        return -EINVAL;
1617                if (resource == RLIMIT_NOFILE &&
1618                                new_rlim->rlim_max > sysctl_nr_open)
1619                        return -EPERM;
1620        }
1621
1622        /* protect tsk->signal and tsk->sighand from disappearing */
1623        read_lock(&tasklist_lock);
1624        if (!tsk->sighand) {
1625                retval = -ESRCH;
1626                goto out;
1627        }
1628
1629        rlim = tsk->signal->rlim + resource;
1630        task_lock(tsk->group_leader);
1631        if (new_rlim) {
1632                /* Keep the capable check against init_user_ns until
1633                   cgroups can contain all limits */
1634                if (new_rlim->rlim_max > rlim->rlim_max &&
1635                                !capable(CAP_SYS_RESOURCE))
1636                        retval = -EPERM;
1637                if (!retval)
1638                        retval = security_task_setrlimit(tsk->group_leader,
1639                                        resource, new_rlim);
1640                if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1641                        /*
1642                         * The caller is asking for an immediate RLIMIT_CPU
1643                         * expiry.  But we use the zero value to mean "it was
1644                         * never set".  So let's cheat and make it one second
1645                         * instead
1646                         */
1647                        new_rlim->rlim_cur = 1;
1648                }
1649        }
1650        if (!retval) {
1651                if (old_rlim)
1652                        *old_rlim = *rlim;
1653                if (new_rlim)
1654                        *rlim = *new_rlim;
1655        }
1656        task_unlock(tsk->group_leader);
1657
1658        /*
1659         * RLIMIT_CPU handling.   Note that the kernel fails to return an error
1660         * code if it rejected the user's attempt to set RLIMIT_CPU.  This is a
1661         * very long-standing error, and fixing it now risks breakage of
1662         * applications, so we live with it
1663         */
1664         if (!retval && new_rlim && resource == RLIMIT_CPU &&
1665                         new_rlim->rlim_cur != RLIM_INFINITY)
1666                update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1667out:
1668        read_unlock(&tasklist_lock);
1669        return retval;
1670}
1671
1672/* rcu lock must be held */
1673static int check_prlimit_permission(struct task_struct *task)
1674{
1675        const struct cred *cred = current_cred(), *tcred;
1676
1677        if (current == task)
1678                return 0;
1679
1680        tcred = __task_cred(task);
1681        if (uid_eq(cred->uid, tcred->euid) &&
1682            uid_eq(cred->uid, tcred->suid) &&
1683            uid_eq(cred->uid, tcred->uid)  &&
1684            gid_eq(cred->gid, tcred->egid) &&
1685            gid_eq(cred->gid, tcred->sgid) &&
1686            gid_eq(cred->gid, tcred->gid))
1687                return 0;
1688        if (ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1689                return 0;
1690
1691        return -EPERM;
1692}
1693
1694SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1695                const struct rlimit64 __user *, new_rlim,
1696                struct rlimit64 __user *, old_rlim)
1697{
1698        struct rlimit64 old64, new64;
1699        struct rlimit old, new;
1700        struct task_struct *tsk;
1701        int ret;
1702
1703        if (new_rlim) {
1704                if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1705                        return -EFAULT;
1706                rlim64_to_rlim(&new64, &new);
1707        }
1708
1709        rcu_read_lock();
1710        tsk = pid ? find_task_by_vpid(pid) : current;
1711        if (!tsk) {
1712                rcu_read_unlock();
1713                return -ESRCH;
1714        }
1715        ret = check_prlimit_permission(tsk);
1716        if (ret) {
1717                rcu_read_unlock();
1718                return ret;
1719        }
1720        get_task_struct(tsk);
1721        rcu_read_unlock();
1722
1723        ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1724                        old_rlim ? &old : NULL);
1725
1726        if (!ret && old_rlim) {
1727                rlim_to_rlim64(&old, &old64);
1728                if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1729                        ret = -EFAULT;
1730        }
1731
1732        put_task_struct(tsk);
1733        return ret;
1734}
1735
1736SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1737{
1738        struct rlimit new_rlim;
1739
1740        if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1741                return -EFAULT;
1742        return do_prlimit(current, resource, &new_rlim, NULL);
1743}
1744
1745/*
1746 * It would make sense to put struct rusage in the task_struct,
1747 * except that would make the task_struct be *really big*.  After
1748 * task_struct gets moved into malloc'ed memory, it would
1749 * make sense to do this.  It will make moving the rest of the information
1750 * a lot simpler!  (Which we're not doing right now because we're not
1751 * measuring them yet).
1752 *
1753 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1754 * races with threads incrementing their own counters.  But since word
1755 * reads are atomic, we either get new values or old values and we don't
1756 * care which for the sums.  We always take the siglock to protect reading
1757 * the c* fields from p->signal from races with exit.c updating those
1758 * fields when reaping, so a sample either gets all the additions of a
1759 * given child after it's reaped, or none so this sample is before reaping.
1760 *
1761 * Locking:
1762 * We need to take the siglock for CHILDEREN, SELF and BOTH
1763 * for  the cases current multithreaded, non-current single threaded
1764 * non-current multithreaded.  Thread traversal is now safe with
1765 * the siglock held.
1766 * Strictly speaking, we donot need to take the siglock if we are current and
1767 * single threaded,  as no one else can take our signal_struct away, no one
1768 * else can  reap the  children to update signal->c* counters, and no one else
1769 * can race with the signal-> fields. If we do not take any lock, the
1770 * signal-> fields could be read out of order while another thread was just
1771 * exiting. So we should  place a read memory barrier when we avoid the lock.
1772 * On the writer side,  write memory barrier is implied in  __exit_signal
1773 * as __exit_signal releases  the siglock spinlock after updating the signal->
1774 * fields. But we don't do this yet to keep things simple.
1775 *
1776 */
1777
1778static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1779{
1780        r->ru_nvcsw += t->nvcsw;
1781        r->ru_nivcsw += t->nivcsw;
1782        r->ru_minflt += t->min_flt;
1783        r->ru_majflt += t->maj_flt;
1784        r->ru_inblock += task_io_get_inblock(t);
1785        r->ru_oublock += task_io_get_oublock(t);
1786}
1787
1788static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1789{
1790        struct task_struct *t;
1791        unsigned long flags;
1792        cputime_t tgutime, tgstime, utime, stime;
1793        unsigned long maxrss = 0;
1794
1795        memset((char *) r, 0, sizeof *r);
1796        utime = stime = 0;
1797
1798        if (who == RUSAGE_THREAD) {
1799                task_cputime_adjusted(current, &utime, &stime);
1800                accumulate_thread_rusage(p, r);
1801                maxrss = p->signal->maxrss;
1802                goto out;
1803        }
1804
1805        if (!lock_task_sighand(p, &flags))
1806                return;
1807
1808        switch (who) {
1809                case RUSAGE_BOTH:
1810                case RUSAGE_CHILDREN:
1811                        utime = p->signal->cutime;
1812                        stime = p->signal->cstime;
1813                        r->ru_nvcsw = p->signal->cnvcsw;
1814                        r->ru_nivcsw = p->signal->cnivcsw;
1815                        r->ru_minflt = p->signal->cmin_flt;
1816                        r->ru_majflt = p->signal->cmaj_flt;
1817                        r->ru_inblock = p->signal->cinblock;
1818                        r->ru_oublock = p->signal->coublock;
1819                        maxrss = p->signal->cmaxrss;
1820
1821                        if (who == RUSAGE_CHILDREN)
1822                                break;
1823
1824                case RUSAGE_SELF:
1825                        thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1826                        utime += tgutime;
1827                        stime += tgstime;
1828                        r->ru_nvcsw += p->signal->nvcsw;
1829                        r->ru_nivcsw += p->signal->nivcsw;
1830                        r->ru_minflt += p->signal->min_flt;
1831                        r->ru_majflt += p->signal->maj_flt;
1832                        r->ru_inblock += p->signal->inblock;
1833                        r->ru_oublock += p->signal->oublock;
1834                        if (maxrss < p->signal->maxrss)
1835                                maxrss = p->signal->maxrss;
1836                        t = p;
1837                        do {
1838                                accumulate_thread_rusage(t, r);
1839                                t = next_thread(t);
1840                        } while (t != p);
1841                        break;
1842
1843                default:
1844                        BUG();
1845        }
1846        unlock_task_sighand(p, &flags);
1847
1848out:
1849        cputime_to_timeval(utime, &r->ru_utime);
1850        cputime_to_timeval(stime, &r->ru_stime);
1851
1852        if (who != RUSAGE_CHILDREN) {
1853                struct mm_struct *mm = get_task_mm(p);
1854                if (mm) {
1855                        setmax_mm_hiwater_rss(&maxrss, mm);
1856                        mmput(mm);
1857                }
1858        }
1859        r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1860}
1861
1862int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1863{
1864        struct rusage r;
1865        k_getrusage(p, who, &r);
1866        return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1867}
1868
1869SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1870{
1871        if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1872            who != RUSAGE_THREAD)
1873                return -EINVAL;
1874        return getrusage(current, who, ru);
1875}
1876
1877#ifdef CONFIG_COMPAT
1878COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1879{
1880        struct rusage r;
1881
1882        if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1883            who != RUSAGE_THREAD)
1884                return -EINVAL;
1885
1886        k_getrusage(current, who, &r);
1887        return put_compat_rusage(&r, ru);
1888}
1889#endif
1890
1891SYSCALL_DEFINE1(umask, int, mask)
1892{
1893        mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1894        return mask;
1895}
1896
1897static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1898{
1899        struct fd exe;
1900        struct inode *inode;
1901        int err;
1902
1903        exe = fdget(fd);
1904        if (!exe.file)
1905                return -EBADF;
1906
1907        inode = file_inode(exe.file);
1908
1909        /*
1910         * Because the original mm->exe_file points to executable file, make
1911         * sure that this one is executable as well, to avoid breaking an
1912         * overall picture.
1913         */
1914        err = -EACCES;
1915        if (!S_ISREG(inode->i_mode)     ||
1916            exe.file->f_path.mnt->mnt_flags & MNT_NOEXEC)
1917                goto exit;
1918
1919        err = inode_permission(inode, MAY_EXEC);
1920        if (err)
1921                goto exit;
1922
1923        down_write(&mm->mmap_sem);
1924
1925        /*
1926         * Forbid mm->exe_file change if old file still mapped.
1927         */
1928        err = -EBUSY;
1929        if (mm->exe_file) {
1930                struct vm_area_struct *vma;
1931
1932                for (vma = mm->mmap; vma; vma = vma->vm_next)
1933                        if (vma->vm_file &&
1934                            path_equal(&vma->vm_file->f_path,
1935                                       &mm->exe_file->f_path))
1936                                goto exit_unlock;
1937        }
1938
1939        /*
1940         * The symlink can be changed only once, just to disallow arbitrary
1941         * transitions malicious software might bring in. This means one
1942         * could make a snapshot over all processes running and monitor
1943         * /proc/pid/exe changes to notice unusual activity if needed.
1944         */
1945        err = -EPERM;
1946        if (test_and_set_bit(MMF_EXE_FILE_CHANGED, &mm->flags))
1947                goto exit_unlock;
1948
1949        err = 0;
1950        set_mm_exe_file(mm, exe.file);  /* this grabs a reference to exe.file */
1951exit_unlock:
1952        up_write(&mm->mmap_sem);
1953
1954exit:
1955        fdput(exe);
1956        return err;
1957}
1958
1959static int prctl_set_mm(int opt, unsigned long addr,
1960                        unsigned long arg4, unsigned long arg5)
1961{
1962        unsigned long rlim = rlimit(RLIMIT_DATA);
1963        struct mm_struct *mm = current->mm;
1964        struct vm_area_struct *vma;
1965        int error;
1966
1967        if (arg5 || (arg4 && opt != PR_SET_MM_AUXV))
1968                return -EINVAL;
1969
1970        if (!capable(CAP_SYS_RESOURCE))
1971                return -EPERM;
1972
1973        if (opt == PR_SET_MM_EXE_FILE)
1974                return prctl_set_mm_exe_file(mm, (unsigned int)addr);
1975
1976        if (addr >= TASK_SIZE || addr < mmap_min_addr)
1977                return -EINVAL;
1978
1979        error = -EINVAL;
1980
1981        down_read(&mm->mmap_sem);
1982        vma = find_vma(mm, addr);
1983
1984        switch (opt) {
1985        case PR_SET_MM_START_CODE:
1986                mm->start_code = addr;
1987                break;
1988        case PR_SET_MM_END_CODE:
1989                mm->end_code = addr;
1990                break;
1991        case PR_SET_MM_START_DATA:
1992                mm->start_data = addr;
1993                break;
1994        case PR_SET_MM_END_DATA:
1995                mm->end_data = addr;
1996                break;
1997
1998        case PR_SET_MM_START_BRK:
1999                if (addr <= mm->end_data)
2000                        goto out;
2001
2002                if (rlim < RLIM_INFINITY &&
2003                    (mm->brk - addr) +
2004                    (mm->end_data - mm->start_data) > rlim)
2005                        goto out;
2006
2007                mm->start_brk = addr;
2008                break;
2009
2010        case PR_SET_MM_BRK:
2011                if (addr <= mm->end_data)
2012                        goto out;
2013
2014                if (rlim < RLIM_INFINITY &&
2015                    (addr - mm->start_brk) +
2016                    (mm->end_data - mm->start_data) > rlim)
2017                        goto out;
2018
2019                mm->brk = addr;
2020                break;
2021
2022        /*
2023         * If command line arguments and environment
2024         * are placed somewhere else on stack, we can
2025         * set them up here, ARG_START/END to setup
2026         * command line argumets and ENV_START/END
2027         * for environment.
2028         */
2029        case PR_SET_MM_START_STACK:
2030        case PR_SET_MM_ARG_START:
2031        case PR_SET_MM_ARG_END:
2032        case PR_SET_MM_ENV_START:
2033        case PR_SET_MM_ENV_END:
2034                if (!vma) {
2035                        error = -EFAULT;
2036                        goto out;
2037                }
2038                if (opt == PR_SET_MM_START_STACK)
2039                        mm->start_stack = addr;
2040                else if (opt == PR_SET_MM_ARG_START)
2041                        mm->arg_start = addr;
2042                else if (opt == PR_SET_MM_ARG_END)
2043                        mm->arg_end = addr;
2044                else if (opt == PR_SET_MM_ENV_START)
2045                        mm->env_start = addr;
2046                else if (opt == PR_SET_MM_ENV_END)
2047                        mm->env_end = addr;
2048                break;
2049
2050        /*
2051         * This doesn't move auxiliary vector itself
2052         * since it's pinned to mm_struct, but allow
2053         * to fill vector with new values. It's up
2054         * to a caller to provide sane values here
2055         * otherwise user space tools which use this
2056         * vector might be unhappy.
2057         */
2058        case PR_SET_MM_AUXV: {
2059                unsigned long user_auxv[AT_VECTOR_SIZE];
2060
2061                if (arg4 > sizeof(user_auxv))
2062                        goto out;
2063                up_read(&mm->mmap_sem);
2064
2065                if (copy_from_user(user_auxv, (const void __user *)addr, arg4))
2066                        return -EFAULT;
2067
2068                /* Make sure the last entry is always AT_NULL */
2069                user_auxv[AT_VECTOR_SIZE - 2] = 0;
2070                user_auxv[AT_VECTOR_SIZE - 1] = 0;
2071
2072                BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2073
2074                task_lock(current);
2075                memcpy(mm->saved_auxv, user_auxv, arg4);
2076                task_unlock(current);
2077
2078                return 0;
2079        }
2080        default:
2081                goto out;
2082        }
2083
2084        error = 0;
2085out:
2086        up_read(&mm->mmap_sem);
2087        return error;
2088}
2089
2090#ifdef CONFIG_CHECKPOINT_RESTORE
2091static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2092{
2093        return put_user(me->clear_child_tid, tid_addr);
2094}
2095#else
2096static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2097{
2098        return -EINVAL;
2099}
2100#endif
2101
2102SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2103                unsigned long, arg4, unsigned long, arg5)
2104{
2105        struct task_struct *me = current;
2106        unsigned char comm[sizeof(me->comm)];
2107        long error;
2108
2109        error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2110        if (error != -ENOSYS)
2111                return error;
2112
2113        error = 0;
2114        switch (option) {
2115        case PR_SET_PDEATHSIG:
2116                if (!valid_signal(arg2)) {
2117                        error = -EINVAL;
2118                        break;
2119                }
2120                me->pdeath_signal = arg2;
2121                break;
2122        case PR_GET_PDEATHSIG:
2123                error = put_user(me->pdeath_signal, (int __user *)arg2);
2124                break;
2125        case PR_GET_DUMPABLE:
2126                error = get_dumpable(me->mm);
2127                break;
2128        case PR_SET_DUMPABLE:
2129                if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2130                        error = -EINVAL;
2131                        break;
2132                }
2133                set_dumpable(me->mm, arg2);
2134                break;
2135
2136        case PR_SET_UNALIGN:
2137                error = SET_UNALIGN_CTL(me, arg2);
2138                break;
2139        case PR_GET_UNALIGN:
2140                error = GET_UNALIGN_CTL(me, arg2);
2141                break;
2142        case PR_SET_FPEMU:
2143                error = SET_FPEMU_CTL(me, arg2);
2144                break;
2145        case PR_GET_FPEMU:
2146                error = GET_FPEMU_CTL(me, arg2);
2147                break;
2148        case PR_SET_FPEXC:
2149                error = SET_FPEXC_CTL(me, arg2);
2150                break;
2151        case PR_GET_FPEXC:
2152                error = GET_FPEXC_CTL(me, arg2);
2153                break;
2154        case PR_GET_TIMING:
2155                error = PR_TIMING_STATISTICAL;
2156                break;
2157        case PR_SET_TIMING:
2158                if (arg2 != PR_TIMING_STATISTICAL)
2159                        error = -EINVAL;
2160                break;
2161        case PR_SET_NAME:
2162                comm[sizeof(me->comm) - 1] = 0;
2163                if (strncpy_from_user(comm, (char __user *)arg2,
2164                                      sizeof(me->comm) - 1) < 0)
2165                        return -EFAULT;
2166                set_task_comm(me, comm);
2167                proc_comm_connector(me);
2168                break;
2169        case PR_GET_NAME:
2170                get_task_comm(comm, me);
2171                if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2172                        return -EFAULT;
2173                break;
2174        case PR_GET_ENDIAN:
2175                error = GET_ENDIAN(me, arg2);
2176                break;
2177        case PR_SET_ENDIAN:
2178                error = SET_ENDIAN(me, arg2);
2179                break;
2180        case PR_GET_SECCOMP:
2181                error = prctl_get_seccomp();
2182                break;
2183        case PR_SET_SECCOMP:
2184                error = prctl_set_seccomp(arg2, (char __user *)arg3);
2185                break;
2186        case PR_GET_TSC:
2187                error = GET_TSC_CTL(arg2);
2188                break;
2189        case PR_SET_TSC:
2190                error = SET_TSC_CTL(arg2);
2191                break;
2192        case PR_TASK_PERF_EVENTS_DISABLE:
2193                error = perf_event_task_disable();
2194                break;
2195        case PR_TASK_PERF_EVENTS_ENABLE:
2196                error = perf_event_task_enable();
2197                break;
2198        case PR_GET_TIMERSLACK:
2199                error = current->timer_slack_ns;
2200                break;
2201        case PR_SET_TIMERSLACK:
2202                if (arg2 <= 0)
2203                        current->timer_slack_ns =
2204                                        current->default_timer_slack_ns;
2205                else
2206                        current->timer_slack_ns = arg2;
2207                break;
2208        case PR_MCE_KILL:
2209                if (arg4 | arg5)
2210                        return -EINVAL;
2211                switch (arg2) {
2212                case PR_MCE_KILL_CLEAR:
2213                        if (arg3 != 0)
2214                                return -EINVAL;
2215                        current->flags &= ~PF_MCE_PROCESS;
2216                        break;
2217                case PR_MCE_KILL_SET:
2218                        current->flags |= PF_MCE_PROCESS;
2219                        if (arg3 == PR_MCE_KILL_EARLY)
2220                                current->flags |= PF_MCE_EARLY;
2221                        else if (arg3 == PR_MCE_KILL_LATE)
2222                                current->flags &= ~PF_MCE_EARLY;
2223                        else if (arg3 == PR_MCE_KILL_DEFAULT)
2224                                current->flags &=
2225                                                ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2226                        else
2227                                return -EINVAL;
2228                        break;
2229                default:
2230                        return -EINVAL;
2231                }
2232                break;
2233        case PR_MCE_KILL_GET:
2234                if (arg2 | arg3 | arg4 | arg5)
2235                        return -EINVAL;
2236                if (current->flags & PF_MCE_PROCESS)
2237                        error = (current->flags & PF_MCE_EARLY) ?
2238                                PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2239                else
2240                        error = PR_MCE_KILL_DEFAULT;
2241                break;
2242        case PR_SET_MM:
2243                error = prctl_set_mm(arg2, arg3, arg4, arg5);
2244                break;
2245        case PR_GET_TID_ADDRESS:
2246                error = prctl_get_tid_address(me, (int __user **)arg2);
2247                break;
2248        case PR_SET_CHILD_SUBREAPER:
2249                me->signal->is_child_subreaper = !!arg2;
2250                break;
2251        case PR_GET_CHILD_SUBREAPER:
2252                error = put_user(me->signal->is_child_subreaper,
2253                                 (int __user *)arg2);
2254                break;
2255        case PR_SET_NO_NEW_PRIVS:
2256                if (arg2 != 1 || arg3 || arg4 || arg5)
2257                        return -EINVAL;
2258
2259                current->no_new_privs = 1;
2260                break;
2261        case PR_GET_NO_NEW_PRIVS:
2262                if (arg2 || arg3 || arg4 || arg5)
2263                        return -EINVAL;
2264                return current->no_new_privs ? 1 : 0;
2265        default:
2266                error = -EINVAL;
2267                break;
2268        }
2269        return error;
2270}
2271
2272SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2273                struct getcpu_cache __user *, unused)
2274{
2275        int err = 0;
2276        int cpu = raw_smp_processor_id();
2277        if (cpup)
2278                err |= put_user(cpu, cpup);
2279        if (nodep)
2280                err |= put_user(cpu_to_node(cpu), nodep);
2281        return err ? -EFAULT : 0;
2282}
2283
2284char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
2285
2286static int __orderly_poweroff(bool force)
2287{
2288        char **argv;
2289        static char *envp[] = {
2290                "HOME=/",
2291                "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
2292                NULL
2293        };
2294        int ret;
2295
2296        argv = argv_split(GFP_KERNEL, poweroff_cmd, NULL);
2297        if (argv) {
2298                ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_EXEC);
2299                argv_free(argv);
2300        } else {
2301                printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
2302                                         __func__, poweroff_cmd);
2303                ret = -ENOMEM;
2304        }
2305
2306        if (ret && force) {
2307                printk(KERN_WARNING "Failed to start orderly shutdown: "
2308                                        "forcing the issue\n");
2309                /*
2310                 * I guess this should try to kick off some daemon to sync and
2311                 * poweroff asap.  Or not even bother syncing if we're doing an
2312                 * emergency shutdown?
2313                 */
2314                emergency_sync();
2315                kernel_power_off();
2316        }
2317
2318        return ret;
2319}
2320
2321static bool poweroff_force;
2322
2323static void poweroff_work_func(struct work_struct *work)
2324{
2325        __orderly_poweroff(poweroff_force);
2326}
2327
2328static DECLARE_WORK(poweroff_work, poweroff_work_func);
2329
2330/**
2331 * orderly_poweroff - Trigger an orderly system poweroff
2332 * @force: force poweroff if command execution fails
2333 *
2334 * This may be called from any context to trigger a system shutdown.
2335 * If the orderly shutdown fails, it will force an immediate shutdown.
2336 */
2337int orderly_poweroff(bool force)
2338{
2339        if (force) /* do not override the pending "true" */
2340                poweroff_force = true;
2341        schedule_work(&poweroff_work);
2342        return 0;
2343}
2344EXPORT_SYMBOL_GPL(orderly_poweroff);
2345
2346/**
2347 * do_sysinfo - fill in sysinfo struct
2348 * @info: pointer to buffer to fill
2349 */
2350static int do_sysinfo(struct sysinfo *info)
2351{
2352        unsigned long mem_total, sav_total;
2353        unsigned int mem_unit, bitcount;
2354        struct timespec tp;
2355
2356        memset(info, 0, sizeof(struct sysinfo));
2357
2358        ktime_get_ts(&tp);
2359        monotonic_to_bootbased(&tp);
2360        info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2361
2362        get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2363
2364        info->procs = nr_threads;
2365
2366        si_meminfo(info);
2367        si_swapinfo(info);
2368
2369        /*
2370         * If the sum of all the available memory (i.e. ram + swap)
2371         * is less than can be stored in a 32 bit unsigned long then
2372         * we can be binary compatible with 2.2.x kernels.  If not,
2373         * well, in that case 2.2.x was broken anyways...
2374         *
2375         *  -Erik Andersen <andersee@debian.org>
2376         */
2377
2378        mem_total = info->totalram + info->totalswap;
2379        if (mem_total < info->totalram || mem_total < info->totalswap)
2380                goto out;
2381        bitcount = 0;
2382        mem_unit = info->mem_unit;
2383        while (mem_unit > 1) {
2384                bitcount++;
2385                mem_unit >>= 1;
2386                sav_total = mem_total;
2387                mem_total <<= 1;
2388                if (mem_total < sav_total)
2389                        goto out;
2390        }
2391
2392        /*
2393         * If mem_total did not overflow, multiply all memory values by
2394         * info->mem_unit and set it to 1.  This leaves things compatible
2395         * with 2.2.x, and also retains compatibility with earlier 2.4.x
2396         * kernels...
2397         */
2398
2399        info->mem_unit = 1;
2400        info->totalram <<= bitcount;
2401        info->freeram <<= bitcount;
2402        info->sharedram <<= bitcount;
2403        info->bufferram <<= bitcount;
2404        info->totalswap <<= bitcount;
2405        info->freeswap <<= bitcount;
2406        info->totalhigh <<= bitcount;
2407        info->freehigh <<= bitcount;
2408
2409out:
2410        return 0;
2411}
2412
2413SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2414{
2415        struct sysinfo val;
2416
2417        do_sysinfo(&val);
2418
2419        if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2420                return -EFAULT;
2421
2422        return 0;
2423}
2424
2425#ifdef CONFIG_COMPAT
2426struct compat_sysinfo {
2427        s32 uptime;
2428        u32 loads[3];
2429        u32 totalram;
2430        u32 freeram;
2431        u32 sharedram;
2432        u32 bufferram;
2433        u32 totalswap;
2434        u32 freeswap;
2435        u16 procs;
2436        u16 pad;
2437        u32 totalhigh;
2438        u32 freehigh;
2439        u32 mem_unit;
2440        char _f[20-2*sizeof(u32)-sizeof(int)];
2441};
2442
2443COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2444{
2445        struct sysinfo s;
2446
2447        do_sysinfo(&s);
2448
2449        /* Check to see if any memory value is too large for 32-bit and scale
2450         *  down if needed
2451         */
2452        if ((s.totalram >> 32) || (s.totalswap >> 32)) {
2453                int bitcount = 0;
2454
2455                while (s.mem_unit < PAGE_SIZE) {
2456                        s.mem_unit <<= 1;
2457                        bitcount++;
2458                }
2459
2460                s.totalram >>= bitcount;
2461                s.freeram >>= bitcount;
2462                s.sharedram >>= bitcount;
2463                s.bufferram >>= bitcount;
2464                s.totalswap >>= bitcount;
2465                s.freeswap >>= bitcount;
2466                s.totalhigh >>= bitcount;
2467                s.freehigh >>= bitcount;
2468        }
2469
2470        if (!access_ok(VERIFY_WRITE, info, sizeof(struct compat_sysinfo)) ||
2471            __put_user(s.uptime, &info->uptime) ||
2472            __put_user(s.loads[0], &info->loads[0]) ||
2473            __put_user(s.loads[1], &info->loads[1]) ||
2474            __put_user(s.loads[2], &info->loads[2]) ||
2475            __put_user(s.totalram, &info->totalram) ||
2476            __put_user(s.freeram, &info->freeram) ||
2477            __put_user(s.sharedram, &info->sharedram) ||
2478            __put_user(s.bufferram, &info->bufferram) ||
2479            __put_user(s.totalswap, &info->totalswap) ||
2480            __put_user(s.freeswap, &info->freeswap) ||
2481            __put_user(s.procs, &info->procs) ||
2482            __put_user(s.totalhigh, &info->totalhigh) ||
2483            __put_user(s.freehigh, &info->freehigh) ||
2484            __put_user(s.mem_unit, &info->mem_unit))
2485                return -EFAULT;
2486
2487        return 0;
2488}
2489#endif /* CONFIG_COMPAT */
2490
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.