linux/kernel/sys.c
<<
>>
Prefs
   1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 *  linux/kernel/sys.c
   4 *
   5 *  Copyright (C) 1991, 1992  Linus Torvalds
   6 */
   7
   8#include <linux/export.h>
   9#include <linux/mm.h>
  10#include <linux/utsname.h>
  11#include <linux/mman.h>
  12#include <linux/reboot.h>
  13#include <linux/prctl.h>
  14#include <linux/highuid.h>
  15#include <linux/fs.h>
  16#include <linux/kmod.h>
  17#include <linux/perf_event.h>
  18#include <linux/resource.h>
  19#include <linux/kernel.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/suspend.h>
  28#include <linux/tty.h>
  29#include <linux/signal.h>
  30#include <linux/cn_proc.h>
  31#include <linux/getcpu.h>
  32#include <linux/task_io_accounting_ops.h>
  33#include <linux/seccomp.h>
  34#include <linux/cpu.h>
  35#include <linux/personality.h>
  36#include <linux/ptrace.h>
  37#include <linux/fs_struct.h>
  38#include <linux/file.h>
  39#include <linux/mount.h>
  40#include <linux/gfp.h>
  41#include <linux/syscore_ops.h>
  42#include <linux/version.h>
  43#include <linux/ctype.h>
  44#include <linux/syscall_user_dispatch.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/time_namespace.h>
  51#include <linux/binfmts.h>
  52
  53#include <linux/sched.h>
  54#include <linux/sched/autogroup.h>
  55#include <linux/sched/loadavg.h>
  56#include <linux/sched/stat.h>
  57#include <linux/sched/mm.h>
  58#include <linux/sched/coredump.h>
  59#include <linux/sched/task.h>
  60#include <linux/sched/cputime.h>
  61#include <linux/rcupdate.h>
  62#include <linux/uidgid.h>
  63#include <linux/cred.h>
  64
  65#include <linux/nospec.h>
  66
  67#include <linux/kmsg_dump.h>
  68/* Move somewhere else to avoid recompiling? */
  69#include <generated/utsrelease.h>
  70
  71#include <linux/uaccess.h>
  72#include <asm/io.h>
  73#include <asm/unistd.h>
  74
  75#include "uid16.h"
  76
  77#ifndef SET_UNALIGN_CTL
  78# define SET_UNALIGN_CTL(a, b)  (-EINVAL)
  79#endif
  80#ifndef GET_UNALIGN_CTL
  81# define GET_UNALIGN_CTL(a, b)  (-EINVAL)
  82#endif
  83#ifndef SET_FPEMU_CTL
  84# define SET_FPEMU_CTL(a, b)    (-EINVAL)
  85#endif
  86#ifndef GET_FPEMU_CTL
  87# define GET_FPEMU_CTL(a, b)    (-EINVAL)
  88#endif
  89#ifndef SET_FPEXC_CTL
  90# define SET_FPEXC_CTL(a, b)    (-EINVAL)
  91#endif
  92#ifndef GET_FPEXC_CTL
  93# define GET_FPEXC_CTL(a, b)    (-EINVAL)
  94#endif
  95#ifndef GET_ENDIAN
  96# define GET_ENDIAN(a, b)       (-EINVAL)
  97#endif
  98#ifndef SET_ENDIAN
  99# define SET_ENDIAN(a, b)       (-EINVAL)
 100#endif
 101#ifndef GET_TSC_CTL
 102# define GET_TSC_CTL(a)         (-EINVAL)
 103#endif
 104#ifndef SET_TSC_CTL
 105# define SET_TSC_CTL(a)         (-EINVAL)
 106#endif
 107#ifndef GET_FP_MODE
 108# define GET_FP_MODE(a)         (-EINVAL)
 109#endif
 110#ifndef SET_FP_MODE
 111# define SET_FP_MODE(a,b)       (-EINVAL)
 112#endif
 113#ifndef SVE_SET_VL
 114# define SVE_SET_VL(a)          (-EINVAL)
 115#endif
 116#ifndef SVE_GET_VL
 117# define SVE_GET_VL()           (-EINVAL)
 118#endif
 119#ifndef PAC_RESET_KEYS
 120# define PAC_RESET_KEYS(a, b)   (-EINVAL)
 121#endif
 122#ifndef SET_TAGGED_ADDR_CTRL
 123# define SET_TAGGED_ADDR_CTRL(a)        (-EINVAL)
 124#endif
 125#ifndef GET_TAGGED_ADDR_CTRL
 126# define GET_TAGGED_ADDR_CTRL()         (-EINVAL)
 127#endif
 128
 129/*
 130 * this is where the system-wide overflow UID and GID are defined, for
 131 * architectures that now have 32-bit UID/GID but didn't in the past
 132 */
 133
 134int overflowuid = DEFAULT_OVERFLOWUID;
 135int overflowgid = DEFAULT_OVERFLOWGID;
 136
 137EXPORT_SYMBOL(overflowuid);
 138EXPORT_SYMBOL(overflowgid);
 139
 140/*
 141 * the same as above, but for filesystems which can only store a 16-bit
 142 * UID and GID. as such, this is needed on all architectures
 143 */
 144
 145int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
 146int fs_overflowgid = DEFAULT_FS_OVERFLOWGID;
 147
 148EXPORT_SYMBOL(fs_overflowuid);
 149EXPORT_SYMBOL(fs_overflowgid);
 150
 151/*
 152 * Returns true if current's euid is same as p's uid or euid,
 153 * or has CAP_SYS_NICE to p's user_ns.
 154 *
 155 * Called with rcu_read_lock, creds are safe
 156 */
 157static bool set_one_prio_perm(struct task_struct *p)
 158{
 159        const struct cred *cred = current_cred(), *pcred = __task_cred(p);
 160
 161        if (uid_eq(pcred->uid,  cred->euid) ||
 162            uid_eq(pcred->euid, cred->euid))
 163                return true;
 164        if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
 165                return true;
 166        return false;
 167}
 168
 169/*
 170 * set the priority of a task
 171 * - the caller must hold the RCU read lock
 172 */
 173static int set_one_prio(struct task_struct *p, int niceval, int error)
 174{
 175        int no_nice;
 176
 177        if (!set_one_prio_perm(p)) {
 178                error = -EPERM;
 179                goto out;
 180        }
 181        if (niceval < task_nice(p) && !can_nice(p, niceval)) {
 182                error = -EACCES;
 183                goto out;
 184        }
 185        no_nice = security_task_setnice(p, niceval);
 186        if (no_nice) {
 187                error = no_nice;
 188                goto out;
 189        }
 190        if (error == -ESRCH)
 191                error = 0;
 192        set_user_nice(p, niceval);
 193out:
 194        return error;
 195}
 196
 197SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
 198{
 199        struct task_struct *g, *p;
 200        struct user_struct *user;
 201        const struct cred *cred = current_cred();
 202        int error = -EINVAL;
 203        struct pid *pgrp;
 204        kuid_t uid;
 205
 206        if (which > PRIO_USER || which < PRIO_PROCESS)
 207                goto out;
 208
 209        /* normalize: avoid signed division (rounding problems) */
 210        error = -ESRCH;
 211        if (niceval < MIN_NICE)
 212                niceval = MIN_NICE;
 213        if (niceval > MAX_NICE)
 214                niceval = MAX_NICE;
 215
 216        rcu_read_lock();
 217        read_lock(&tasklist_lock);
 218        switch (which) {
 219        case PRIO_PROCESS:
 220                if (who)
 221                        p = find_task_by_vpid(who);
 222                else
 223                        p = current;
 224                if (p)
 225                        error = set_one_prio(p, niceval, error);
 226                break;
 227        case PRIO_PGRP:
 228                if (who)
 229                        pgrp = find_vpid(who);
 230                else
 231                        pgrp = task_pgrp(current);
 232                do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
 233                        error = set_one_prio(p, niceval, error);
 234                } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
 235                break;
 236        case PRIO_USER:
 237                uid = make_kuid(cred->user_ns, who);
 238                user = cred->user;
 239                if (!who)
 240                        uid = cred->uid;
 241                else if (!uid_eq(uid, cred->uid)) {
 242                        user = find_user(uid);
 243                        if (!user)
 244                                goto out_unlock;        /* No processes for this user */
 245                }
 246                do_each_thread(g, p) {
 247                        if (uid_eq(task_uid(p), uid) && task_pid_vnr(p))
 248                                error = set_one_prio(p, niceval, error);
 249                } while_each_thread(g, p);
 250                if (!uid_eq(uid, cred->uid))
 251                        free_uid(user);         /* For find_user() */
 252                break;
 253        }
 254out_unlock:
 255        read_unlock(&tasklist_lock);
 256        rcu_read_unlock();
 257out:
 258        return error;
 259}
 260
 261/*
 262 * Ugh. To avoid negative return values, "getpriority()" will
 263 * not return the normal nice-value, but a negated value that
 264 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
 265 * to stay compatible.
 266 */
 267SYSCALL_DEFINE2(getpriority, int, which, int, who)
 268{
 269        struct task_struct *g, *p;
 270        struct user_struct *user;
 271        const struct cred *cred = current_cred();
 272        long niceval, retval = -ESRCH;
 273        struct pid *pgrp;
 274        kuid_t uid;
 275
 276        if (which > PRIO_USER || which < PRIO_PROCESS)
 277                return -EINVAL;
 278
 279        rcu_read_lock();
 280        read_lock(&tasklist_lock);
 281        switch (which) {
 282        case PRIO_PROCESS:
 283                if (who)
 284                        p = find_task_by_vpid(who);
 285                else
 286                        p = current;
 287                if (p) {
 288                        niceval = nice_to_rlimit(task_nice(p));
 289                        if (niceval > retval)
 290                                retval = niceval;
 291                }
 292                break;
 293        case PRIO_PGRP:
 294                if (who)
 295                        pgrp = find_vpid(who);
 296                else
 297                        pgrp = task_pgrp(current);
 298                do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
 299                        niceval = nice_to_rlimit(task_nice(p));
 300                        if (niceval > retval)
 301                                retval = niceval;
 302                } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
 303                break;
 304        case PRIO_USER:
 305                uid = make_kuid(cred->user_ns, who);
 306                user = cred->user;
 307                if (!who)
 308                        uid = cred->uid;
 309                else if (!uid_eq(uid, cred->uid)) {
 310                        user = find_user(uid);
 311                        if (!user)
 312                                goto out_unlock;        /* No processes for this user */
 313                }
 314                do_each_thread(g, p) {
 315                        if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) {
 316                                niceval = nice_to_rlimit(task_nice(p));
 317                                if (niceval > retval)
 318                                        retval = niceval;
 319                        }
 320                } while_each_thread(g, p);
 321                if (!uid_eq(uid, cred->uid))
 322                        free_uid(user);         /* for find_user() */
 323                break;
 324        }
 325out_unlock:
 326        read_unlock(&tasklist_lock);
 327        rcu_read_unlock();
 328
 329        return retval;
 330}
 331
 332/*
 333 * Unprivileged users may change the real gid to the effective gid
 334 * or vice versa.  (BSD-style)
 335 *
 336 * If you set the real gid at all, or set the effective gid to a value not
 337 * equal to the real gid, then the saved gid is set to the new effective gid.
 338 *
 339 * This makes it possible for a setgid program to completely drop its
 340 * privileges, which is often a useful assertion to make when you are doing
 341 * a security audit over a program.
 342 *
 343 * The general idea is that a program which uses just setregid() will be
 344 * 100% compatible with BSD.  A program which uses just setgid() will be
 345 * 100% compatible with POSIX with saved IDs.
 346 *
 347 * SMP: There are not races, the GIDs are checked only by filesystem
 348 *      operations (as far as semantic preservation is concerned).
 349 */
 350#ifdef CONFIG_MULTIUSER
 351long __sys_setregid(gid_t rgid, gid_t egid)
 352{
 353        struct user_namespace *ns = current_user_ns();
 354        const struct cred *old;
 355        struct cred *new;
 356        int retval;
 357        kgid_t krgid, kegid;
 358
 359        krgid = make_kgid(ns, rgid);
 360        kegid = make_kgid(ns, egid);
 361
 362        if ((rgid != (gid_t) -1) && !gid_valid(krgid))
 363                return -EINVAL;
 364        if ((egid != (gid_t) -1) && !gid_valid(kegid))
 365                return -EINVAL;
 366
 367        new = prepare_creds();
 368        if (!new)
 369                return -ENOMEM;
 370        old = current_cred();
 371
 372        retval = -EPERM;
 373        if (rgid != (gid_t) -1) {
 374                if (gid_eq(old->gid, krgid) ||
 375                    gid_eq(old->egid, krgid) ||
 376                    ns_capable_setid(old->user_ns, CAP_SETGID))
 377                        new->gid = krgid;
 378                else
 379                        goto error;
 380        }
 381        if (egid != (gid_t) -1) {
 382                if (gid_eq(old->gid, kegid) ||
 383                    gid_eq(old->egid, kegid) ||
 384                    gid_eq(old->sgid, kegid) ||
 385                    ns_capable_setid(old->user_ns, CAP_SETGID))
 386                        new->egid = kegid;
 387                else
 388                        goto error;
 389        }
 390
 391        if (rgid != (gid_t) -1 ||
 392            (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
 393                new->sgid = new->egid;
 394        new->fsgid = new->egid;
 395
 396        retval = security_task_fix_setgid(new, old, LSM_SETID_RE);
 397        if (retval < 0)
 398                goto error;
 399
 400        return commit_creds(new);
 401
 402error:
 403        abort_creds(new);
 404        return retval;
 405}
 406
 407SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
 408{
 409        return __sys_setregid(rgid, egid);
 410}
 411
 412/*
 413 * setgid() is implemented like SysV w/ SAVED_IDS
 414 *
 415 * SMP: Same implicit races as above.
 416 */
 417long __sys_setgid(gid_t gid)
 418{
 419        struct user_namespace *ns = current_user_ns();
 420        const struct cred *old;
 421        struct cred *new;
 422        int retval;
 423        kgid_t kgid;
 424
 425        kgid = make_kgid(ns, gid);
 426        if (!gid_valid(kgid))
 427                return -EINVAL;
 428
 429        new = prepare_creds();
 430        if (!new)
 431                return -ENOMEM;
 432        old = current_cred();
 433
 434        retval = -EPERM;
 435        if (ns_capable_setid(old->user_ns, CAP_SETGID))
 436                new->gid = new->egid = new->sgid = new->fsgid = kgid;
 437        else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
 438                new->egid = new->fsgid = kgid;
 439        else
 440                goto error;
 441
 442        retval = security_task_fix_setgid(new, old, LSM_SETID_ID);
 443        if (retval < 0)
 444                goto error;
 445
 446        return commit_creds(new);
 447
 448error:
 449        abort_creds(new);
 450        return retval;
 451}
 452
 453SYSCALL_DEFINE1(setgid, gid_t, gid)
 454{
 455        return __sys_setgid(gid);
 456}
 457
 458/*
 459 * change the user struct in a credentials set to match the new UID
 460 */
 461static int set_user(struct cred *new)
 462{
 463        struct user_struct *new_user;
 464
 465        new_user = alloc_uid(new->uid);
 466        if (!new_user)
 467                return -EAGAIN;
 468
 469        /*
 470         * We don't fail in case of NPROC limit excess here because too many
 471         * poorly written programs don't check set*uid() return code, assuming
 472         * it never fails if called by root.  We may still enforce NPROC limit
 473         * for programs doing set*uid()+execve() by harmlessly deferring the
 474         * failure to the execve() stage.
 475         */
 476        if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
 477                        new_user != INIT_USER)
 478                current->flags |= PF_NPROC_EXCEEDED;
 479        else
 480                current->flags &= ~PF_NPROC_EXCEEDED;
 481
 482        free_uid(new->user);
 483        new->user = new_user;
 484        return 0;
 485}
 486
 487/*
 488 * Unprivileged users may change the real uid to the effective uid
 489 * or vice versa.  (BSD-style)
 490 *
 491 * If you set the real uid at all, or set the effective uid to a value not
 492 * equal to the real uid, then the saved uid is set to the new effective uid.
 493 *
 494 * This makes it possible for a setuid program to completely drop its
 495 * privileges, which is often a useful assertion to make when you are doing
 496 * a security audit over a program.
 497 *
 498 * The general idea is that a program which uses just setreuid() will be
 499 * 100% compatible with BSD.  A program which uses just setuid() will be
 500 * 100% compatible with POSIX with saved IDs.
 501 */
 502long __sys_setreuid(uid_t ruid, uid_t euid)
 503{
 504        struct user_namespace *ns = current_user_ns();
 505        const struct cred *old;
 506        struct cred *new;
 507        int retval;
 508        kuid_t kruid, keuid;
 509
 510        kruid = make_kuid(ns, ruid);
 511        keuid = make_kuid(ns, euid);
 512
 513        if ((ruid != (uid_t) -1) && !uid_valid(kruid))
 514                return -EINVAL;
 515        if ((euid != (uid_t) -1) && !uid_valid(keuid))
 516                return -EINVAL;
 517
 518        new = prepare_creds();
 519        if (!new)
 520                return -ENOMEM;
 521        old = current_cred();
 522
 523        retval = -EPERM;
 524        if (ruid != (uid_t) -1) {
 525                new->uid = kruid;
 526                if (!uid_eq(old->uid, kruid) &&
 527                    !uid_eq(old->euid, kruid) &&
 528                    !ns_capable_setid(old->user_ns, CAP_SETUID))
 529                        goto error;
 530        }
 531
 532        if (euid != (uid_t) -1) {
 533                new->euid = keuid;
 534                if (!uid_eq(old->uid, keuid) &&
 535                    !uid_eq(old->euid, keuid) &&
 536                    !uid_eq(old->suid, keuid) &&
 537                    !ns_capable_setid(old->user_ns, CAP_SETUID))
 538                        goto error;
 539        }
 540
 541        if (!uid_eq(new->uid, old->uid)) {
 542                retval = set_user(new);
 543                if (retval < 0)
 544                        goto error;
 545        }
 546        if (ruid != (uid_t) -1 ||
 547            (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
 548                new->suid = new->euid;
 549        new->fsuid = new->euid;
 550
 551        retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
 552        if (retval < 0)
 553                goto error;
 554
 555        return commit_creds(new);
 556
 557error:
 558        abort_creds(new);
 559        return retval;
 560}
 561
 562SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
 563{
 564        return __sys_setreuid(ruid, euid);
 565}
 566
 567/*
 568 * setuid() is implemented like SysV with SAVED_IDS
 569 *
 570 * Note that SAVED_ID's is deficient in that a setuid root program
 571 * like sendmail, for example, cannot set its uid to be a normal
 572 * user and then switch back, because if you're root, setuid() sets
 573 * the saved uid too.  If you don't like this, blame the bright people
 574 * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
 575 * will allow a root program to temporarily drop privileges and be able to
 576 * regain them by swapping the real and effective uid.
 577 */
 578long __sys_setuid(uid_t uid)
 579{
 580        struct user_namespace *ns = current_user_ns();
 581        const struct cred *old;
 582        struct cred *new;
 583        int retval;
 584        kuid_t kuid;
 585
 586        kuid = make_kuid(ns, uid);
 587        if (!uid_valid(kuid))
 588                return -EINVAL;
 589
 590        new = prepare_creds();
 591        if (!new)
 592                return -ENOMEM;
 593        old = current_cred();
 594
 595        retval = -EPERM;
 596        if (ns_capable_setid(old->user_ns, CAP_SETUID)) {
 597                new->suid = new->uid = kuid;
 598                if (!uid_eq(kuid, old->uid)) {
 599                        retval = set_user(new);
 600                        if (retval < 0)
 601                                goto error;
 602                }
 603        } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
 604                goto error;
 605        }
 606
 607        new->fsuid = new->euid = kuid;
 608
 609        retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
 610        if (retval < 0)
 611                goto error;
 612
 613        return commit_creds(new);
 614
 615error:
 616        abort_creds(new);
 617        return retval;
 618}
 619
 620SYSCALL_DEFINE1(setuid, uid_t, uid)
 621{
 622        return __sys_setuid(uid);
 623}
 624
 625
 626/*
 627 * This function implements a generic ability to update ruid, euid,
 628 * and suid.  This allows you to implement the 4.4 compatible seteuid().
 629 */
 630long __sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
 631{
 632        struct user_namespace *ns = current_user_ns();
 633        const struct cred *old;
 634        struct cred *new;
 635        int retval;
 636        kuid_t kruid, keuid, ksuid;
 637
 638        kruid = make_kuid(ns, ruid);
 639        keuid = make_kuid(ns, euid);
 640        ksuid = make_kuid(ns, suid);
 641
 642        if ((ruid != (uid_t) -1) && !uid_valid(kruid))
 643                return -EINVAL;
 644
 645        if ((euid != (uid_t) -1) && !uid_valid(keuid))
 646                return -EINVAL;
 647
 648        if ((suid != (uid_t) -1) && !uid_valid(ksuid))
 649                return -EINVAL;
 650
 651        new = prepare_creds();
 652        if (!new)
 653                return -ENOMEM;
 654
 655        old = current_cred();
 656
 657        retval = -EPERM;
 658        if (!ns_capable_setid(old->user_ns, CAP_SETUID)) {
 659                if (ruid != (uid_t) -1        && !uid_eq(kruid, old->uid) &&
 660                    !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
 661                        goto error;
 662                if (euid != (uid_t) -1        && !uid_eq(keuid, old->uid) &&
 663                    !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
 664                        goto error;
 665                if (suid != (uid_t) -1        && !uid_eq(ksuid, old->uid) &&
 666                    !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
 667                        goto error;
 668        }
 669
 670        if (ruid != (uid_t) -1) {
 671                new->uid = kruid;
 672                if (!uid_eq(kruid, old->uid)) {
 673                        retval = set_user(new);
 674                        if (retval < 0)
 675                                goto error;
 676                }
 677        }
 678        if (euid != (uid_t) -1)
 679                new->euid = keuid;
 680        if (suid != (uid_t) -1)
 681                new->suid = ksuid;
 682        new->fsuid = new->euid;
 683
 684        retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
 685        if (retval < 0)
 686                goto error;
 687
 688        return commit_creds(new);
 689
 690error:
 691        abort_creds(new);
 692        return retval;
 693}
 694
 695SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
 696{
 697        return __sys_setresuid(ruid, euid, suid);
 698}
 699
 700SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
 701{
 702        const struct cred *cred = current_cred();
 703        int retval;
 704        uid_t ruid, euid, suid;
 705
 706        ruid = from_kuid_munged(cred->user_ns, cred->uid);
 707        euid = from_kuid_munged(cred->user_ns, cred->euid);
 708        suid = from_kuid_munged(cred->user_ns, cred->suid);
 709
 710        retval = put_user(ruid, ruidp);
 711        if (!retval) {
 712                retval = put_user(euid, euidp);
 713                if (!retval)
 714                        return put_user(suid, suidp);
 715        }
 716        return retval;
 717}
 718
 719/*
 720 * Same as above, but for rgid, egid, sgid.
 721 */
 722long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
 723{
 724        struct user_namespace *ns = current_user_ns();
 725        const struct cred *old;
 726        struct cred *new;
 727        int retval;
 728        kgid_t krgid, kegid, ksgid;
 729
 730        krgid = make_kgid(ns, rgid);
 731        kegid = make_kgid(ns, egid);
 732        ksgid = make_kgid(ns, sgid);
 733
 734        if ((rgid != (gid_t) -1) && !gid_valid(krgid))
 735                return -EINVAL;
 736        if ((egid != (gid_t) -1) && !gid_valid(kegid))
 737                return -EINVAL;
 738        if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
 739                return -EINVAL;
 740
 741        new = prepare_creds();
 742        if (!new)
 743                return -ENOMEM;
 744        old = current_cred();
 745
 746        retval = -EPERM;
 747        if (!ns_capable_setid(old->user_ns, CAP_SETGID)) {
 748                if (rgid != (gid_t) -1        && !gid_eq(krgid, old->gid) &&
 749                    !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
 750                        goto error;
 751                if (egid != (gid_t) -1        && !gid_eq(kegid, old->gid) &&
 752                    !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
 753                        goto error;
 754                if (sgid != (gid_t) -1        && !gid_eq(ksgid, old->gid) &&
 755                    !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
 756                        goto error;
 757        }
 758
 759        if (rgid != (gid_t) -1)
 760                new->gid = krgid;
 761        if (egid != (gid_t) -1)
 762                new->egid = kegid;
 763        if (sgid != (gid_t) -1)
 764                new->sgid = ksgid;
 765        new->fsgid = new->egid;
 766
 767        retval = security_task_fix_setgid(new, old, LSM_SETID_RES);
 768        if (retval < 0)
 769                goto error;
 770
 771        return commit_creds(new);
 772
 773error:
 774        abort_creds(new);
 775        return retval;
 776}
 777
 778SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
 779{
 780        return __sys_setresgid(rgid, egid, sgid);
 781}
 782
 783SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
 784{
 785        const struct cred *cred = current_cred();
 786        int retval;
 787        gid_t rgid, egid, sgid;
 788
 789        rgid = from_kgid_munged(cred->user_ns, cred->gid);
 790        egid = from_kgid_munged(cred->user_ns, cred->egid);
 791        sgid = from_kgid_munged(cred->user_ns, cred->sgid);
 792
 793        retval = put_user(rgid, rgidp);
 794        if (!retval) {
 795                retval = put_user(egid, egidp);
 796                if (!retval)
 797                        retval = put_user(sgid, sgidp);
 798        }
 799
 800        return retval;
 801}
 802
 803
 804/*
 805 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
 806 * is used for "access()" and for the NFS daemon (letting nfsd stay at
 807 * whatever uid it wants to). It normally shadows "euid", except when
 808 * explicitly set by setfsuid() or for access..
 809 */
 810long __sys_setfsuid(uid_t uid)
 811{
 812        const struct cred *old;
 813        struct cred *new;
 814        uid_t old_fsuid;
 815        kuid_t kuid;
 816
 817        old = current_cred();
 818        old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
 819
 820        kuid = make_kuid(old->user_ns, uid);
 821        if (!uid_valid(kuid))
 822                return old_fsuid;
 823
 824        new = prepare_creds();
 825        if (!new)
 826                return old_fsuid;
 827
 828        if (uid_eq(kuid, old->uid)  || uid_eq(kuid, old->euid)  ||
 829            uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
 830            ns_capable_setid(old->user_ns, CAP_SETUID)) {
 831                if (!uid_eq(kuid, old->fsuid)) {
 832                        new->fsuid = kuid;
 833                        if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
 834                                goto change_okay;
 835                }
 836        }
 837
 838        abort_creds(new);
 839        return old_fsuid;
 840
 841change_okay:
 842        commit_creds(new);
 843        return old_fsuid;
 844}
 845
 846SYSCALL_DEFINE1(setfsuid, uid_t, uid)
 847{
 848        return __sys_setfsuid(uid);
 849}
 850
 851/*
 852 * Samma på svenska..
 853 */
 854long __sys_setfsgid(gid_t gid)
 855{
 856        const struct cred *old;
 857        struct cred *new;
 858        gid_t old_fsgid;
 859        kgid_t kgid;
 860
 861        old = current_cred();
 862        old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
 863
 864        kgid = make_kgid(old->user_ns, gid);
 865        if (!gid_valid(kgid))
 866                return old_fsgid;
 867
 868        new = prepare_creds();
 869        if (!new)
 870                return old_fsgid;
 871
 872        if (gid_eq(kgid, old->gid)  || gid_eq(kgid, old->egid)  ||
 873            gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
 874            ns_capable_setid(old->user_ns, CAP_SETGID)) {
 875                if (!gid_eq(kgid, old->fsgid)) {
 876                        new->fsgid = kgid;
 877                        if (security_task_fix_setgid(new,old,LSM_SETID_FS) == 0)
 878                                goto change_okay;
 879                }
 880        }
 881
 882        abort_creds(new);
 883        return old_fsgid;
 884
 885change_okay:
 886        commit_creds(new);
 887        return old_fsgid;
 888}
 889
 890SYSCALL_DEFINE1(setfsgid, gid_t, gid)
 891{
 892        return __sys_setfsgid(gid);
 893}
 894#endif /* CONFIG_MULTIUSER */
 895
 896/**
 897 * sys_getpid - return the thread group id of the current process
 898 *
 899 * Note, despite the name, this returns the tgid not the pid.  The tgid and
 900 * the pid are identical unless CLONE_THREAD was specified on clone() in
 901 * which case the tgid is the same in all threads of the same group.
 902 *
 903 * This is SMP safe as current->tgid does not change.
 904 */
 905SYSCALL_DEFINE0(getpid)
 906{
 907        return task_tgid_vnr(current);
 908}
 909
 910/* Thread ID - the internal kernel "pid" */
 911SYSCALL_DEFINE0(gettid)
 912{
 913        return task_pid_vnr(current);
 914}
 915
 916/*
 917 * Accessing ->real_parent is not SMP-safe, it could
 918 * change from under us. However, we can use a stale
 919 * value of ->real_parent under rcu_read_lock(), see
 920 * release_task()->call_rcu(delayed_put_task_struct).
 921 */
 922SYSCALL_DEFINE0(getppid)
 923{
 924        int pid;
 925
 926        rcu_read_lock();
 927        pid = task_tgid_vnr(rcu_dereference(current->real_parent));
 928        rcu_read_unlock();
 929
 930        return pid;
 931}
 932
 933SYSCALL_DEFINE0(getuid)
 934{
 935        /* Only we change this so SMP safe */
 936        return from_kuid_munged(current_user_ns(), current_uid());
 937}
 938
 939SYSCALL_DEFINE0(geteuid)
 940{
 941        /* Only we change this so SMP safe */
 942        return from_kuid_munged(current_user_ns(), current_euid());
 943}
 944
 945SYSCALL_DEFINE0(getgid)
 946{
 947        /* Only we change this so SMP safe */
 948        return from_kgid_munged(current_user_ns(), current_gid());
 949}
 950
 951SYSCALL_DEFINE0(getegid)
 952{
 953        /* Only we change this so SMP safe */
 954        return from_kgid_munged(current_user_ns(), current_egid());
 955}
 956
 957static void do_sys_times(struct tms *tms)
 958{
 959        u64 tgutime, tgstime, cutime, cstime;
 960
 961        thread_group_cputime_adjusted(current, &tgutime, &tgstime);
 962        cutime = current->signal->cutime;
 963        cstime = current->signal->cstime;
 964        tms->tms_utime = nsec_to_clock_t(tgutime);
 965        tms->tms_stime = nsec_to_clock_t(tgstime);
 966        tms->tms_cutime = nsec_to_clock_t(cutime);
 967        tms->tms_cstime = nsec_to_clock_t(cstime);
 968}
 969
 970SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
 971{
 972        if (tbuf) {
 973                struct tms tmp;
 974
 975                do_sys_times(&tmp);
 976                if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
 977                        return -EFAULT;
 978        }
 979        force_successful_syscall_return();
 980        return (long) jiffies_64_to_clock_t(get_jiffies_64());
 981}
 982
 983#ifdef CONFIG_COMPAT
 984static compat_clock_t clock_t_to_compat_clock_t(clock_t x)
 985{
 986        return compat_jiffies_to_clock_t(clock_t_to_jiffies(x));
 987}
 988
 989COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
 990{
 991        if (tbuf) {
 992                struct tms tms;
 993                struct compat_tms tmp;
 994
 995                do_sys_times(&tms);
 996                /* Convert our struct tms to the compat version. */
 997                tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime);
 998                tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime);
 999                tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime);
1000                tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime);
1001                if (copy_to_user(tbuf, &tmp, sizeof(tmp)))
1002                        return -EFAULT;
1003        }
1004        force_successful_syscall_return();
1005        return compat_jiffies_to_clock_t(jiffies);
1006}
1007#endif
1008
1009/*
1010 * This needs some heavy checking ...
1011 * I just haven't the stomach for it. I also don't fully
1012 * understand sessions/pgrp etc. Let somebody who does explain it.
1013 *
1014 * OK, I think I have the protection semantics right.... this is really
1015 * only important on a multi-user system anyway, to make sure one user
1016 * can't send a signal to a process owned by another.  -TYT, 12/12/91
1017 *
1018 * !PF_FORKNOEXEC check to conform completely to POSIX.
1019 */
1020SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1021{
1022        struct task_struct *p;
1023        struct task_struct *group_leader = current->group_leader;
1024        struct pid *pgrp;
1025        int err;
1026
1027        if (!pid)
1028                pid = task_pid_vnr(group_leader);
1029        if (!pgid)
1030                pgid = pid;
1031        if (pgid < 0)
1032                return -EINVAL;
1033        rcu_read_lock();
1034
1035        /* From this point forward we keep holding onto the tasklist lock
1036         * so that our parent does not change from under us. -DaveM
1037         */
1038        write_lock_irq(&tasklist_lock);
1039
1040        err = -ESRCH;
1041        p = find_task_by_vpid(pid);
1042        if (!p)
1043                goto out;
1044
1045        err = -EINVAL;
1046        if (!thread_group_leader(p))
1047                goto out;
1048
1049        if (same_thread_group(p->real_parent, group_leader)) {
1050                err = -EPERM;
1051                if (task_session(p) != task_session(group_leader))
1052                        goto out;
1053                err = -EACCES;
1054                if (!(p->flags & PF_FORKNOEXEC))
1055                        goto out;
1056        } else {
1057                err = -ESRCH;
1058                if (p != group_leader)
1059                        goto out;
1060        }
1061
1062        err = -EPERM;
1063        if (p->signal->leader)
1064                goto out;
1065
1066        pgrp = task_pid(p);
1067        if (pgid != pid) {
1068                struct task_struct *g;
1069
1070                pgrp = find_vpid(pgid);
1071                g = pid_task(pgrp, PIDTYPE_PGID);
1072                if (!g || task_session(g) != task_session(group_leader))
1073                        goto out;
1074        }
1075
1076        err = security_task_setpgid(p, pgid);
1077        if (err)
1078                goto out;
1079
1080        if (task_pgrp(p) != pgrp)
1081                change_pid(p, PIDTYPE_PGID, pgrp);
1082
1083        err = 0;
1084out:
1085        /* All paths lead to here, thus we are safe. -DaveM */
1086        write_unlock_irq(&tasklist_lock);
1087        rcu_read_unlock();
1088        return err;
1089}
1090
1091static int do_getpgid(pid_t pid)
1092{
1093        struct task_struct *p;
1094        struct pid *grp;
1095        int retval;
1096
1097        rcu_read_lock();
1098        if (!pid)
1099                grp = task_pgrp(current);
1100        else {
1101                retval = -ESRCH;
1102                p = find_task_by_vpid(pid);
1103                if (!p)
1104                        goto out;
1105                grp = task_pgrp(p);
1106                if (!grp)
1107                        goto out;
1108
1109                retval = security_task_getpgid(p);
1110                if (retval)
1111                        goto out;
1112        }
1113        retval = pid_vnr(grp);
1114out:
1115        rcu_read_unlock();
1116        return retval;
1117}
1118
1119SYSCALL_DEFINE1(getpgid, pid_t, pid)
1120{
1121        return do_getpgid(pid);
1122}
1123
1124#ifdef __ARCH_WANT_SYS_GETPGRP
1125
1126SYSCALL_DEFINE0(getpgrp)
1127{
1128        return do_getpgid(0);
1129}
1130
1131#endif
1132
1133SYSCALL_DEFINE1(getsid, pid_t, pid)
1134{
1135        struct task_struct *p;
1136        struct pid *sid;
1137        int retval;
1138
1139        rcu_read_lock();
1140        if (!pid)
1141                sid = task_session(current);
1142        else {
1143                retval = -ESRCH;
1144                p = find_task_by_vpid(pid);
1145                if (!p)
1146                        goto out;
1147                sid = task_session(p);
1148                if (!sid)
1149                        goto out;
1150
1151                retval = security_task_getsid(p);
1152                if (retval)
1153                        goto out;
1154        }
1155        retval = pid_vnr(sid);
1156out:
1157        rcu_read_unlock();
1158        return retval;
1159}
1160
1161static void set_special_pids(struct pid *pid)
1162{
1163        struct task_struct *curr = current->group_leader;
1164
1165        if (task_session(curr) != pid)
1166                change_pid(curr, PIDTYPE_SID, pid);
1167
1168        if (task_pgrp(curr) != pid)
1169                change_pid(curr, PIDTYPE_PGID, pid);
1170}
1171
1172int ksys_setsid(void)
1173{
1174        struct task_struct *group_leader = current->group_leader;
1175        struct pid *sid = task_pid(group_leader);
1176        pid_t session = pid_vnr(sid);
1177        int err = -EPERM;
1178
1179        write_lock_irq(&tasklist_lock);
1180        /* Fail if I am already a session leader */
1181        if (group_leader->signal->leader)
1182                goto out;
1183
1184        /* Fail if a process group id already exists that equals the
1185         * proposed session id.
1186         */
1187        if (pid_task(sid, PIDTYPE_PGID))
1188                goto out;
1189
1190        group_leader->signal->leader = 1;
1191        set_special_pids(sid);
1192
1193        proc_clear_tty(group_leader);
1194
1195        err = session;
1196out:
1197        write_unlock_irq(&tasklist_lock);
1198        if (err > 0) {
1199                proc_sid_connector(group_leader);
1200                sched_autogroup_create_attach(group_leader);
1201        }
1202        return err;
1203}
1204
1205SYSCALL_DEFINE0(setsid)
1206{
1207        return ksys_setsid();
1208}
1209
1210DECLARE_RWSEM(uts_sem);
1211
1212#ifdef COMPAT_UTS_MACHINE
1213#define override_architecture(name) \
1214        (personality(current->personality) == PER_LINUX32 && \
1215         copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1216                      sizeof(COMPAT_UTS_MACHINE)))
1217#else
1218#define override_architecture(name)     0
1219#endif
1220
1221/*
1222 * Work around broken programs that cannot handle "Linux 3.0".
1223 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1224 * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1225 * 2.6.60.
1226 */
1227static int override_release(char __user *release, size_t len)
1228{
1229        int ret = 0;
1230
1231        if (current->personality & UNAME26) {
1232                const char *rest = UTS_RELEASE;
1233                char buf[65] = { 0 };
1234                int ndots = 0;
1235                unsigned v;
1236                size_t copy;
1237
1238                while (*rest) {
1239                        if (*rest == '.' && ++ndots >= 3)
1240                                break;
1241                        if (!isdigit(*rest) && *rest != '.')
1242                                break;
1243                        rest++;
1244                }
1245                v = LINUX_VERSION_PATCHLEVEL + 60;
1246                copy = clamp_t(size_t, len, 1, sizeof(buf));
1247                copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1248                ret = copy_to_user(release, buf, copy + 1);
1249        }
1250        return ret;
1251}
1252
1253SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1254{
1255        struct new_utsname tmp;
1256
1257        down_read(&uts_sem);
1258        memcpy(&tmp, utsname(), sizeof(tmp));
1259        up_read(&uts_sem);
1260        if (copy_to_user(name, &tmp, sizeof(tmp)))
1261                return -EFAULT;
1262
1263        if (override_release(name->release, sizeof(name->release)))
1264                return -EFAULT;
1265        if (override_architecture(name))
1266                return -EFAULT;
1267        return 0;
1268}
1269
1270#ifdef __ARCH_WANT_SYS_OLD_UNAME
1271/*
1272 * Old cruft
1273 */
1274SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1275{
1276        struct old_utsname tmp;
1277
1278        if (!name)
1279                return -EFAULT;
1280
1281        down_read(&uts_sem);
1282        memcpy(&tmp, utsname(), sizeof(tmp));
1283        up_read(&uts_sem);
1284        if (copy_to_user(name, &tmp, sizeof(tmp)))
1285                return -EFAULT;
1286
1287        if (override_release(name->release, sizeof(name->release)))
1288                return -EFAULT;
1289        if (override_architecture(name))
1290                return -EFAULT;
1291        return 0;
1292}
1293
1294SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1295{
1296        struct oldold_utsname tmp;
1297
1298        if (!name)
1299                return -EFAULT;
1300
1301        memset(&tmp, 0, sizeof(tmp));
1302
1303        down_read(&uts_sem);
1304        memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN);
1305        memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN);
1306        memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN);
1307        memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN);
1308        memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN);
1309        up_read(&uts_sem);
1310        if (copy_to_user(name, &tmp, sizeof(tmp)))
1311                return -EFAULT;
1312
1313        if (override_architecture(name))
1314                return -EFAULT;
1315        if (override_release(name->release, sizeof(name->release)))
1316                return -EFAULT;
1317        return 0;
1318}
1319#endif
1320
1321SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1322{
1323        int errno;
1324        char tmp[__NEW_UTS_LEN];
1325
1326        if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1327                return -EPERM;
1328
1329        if (len < 0 || len > __NEW_UTS_LEN)
1330                return -EINVAL;
1331        errno = -EFAULT;
1332        if (!copy_from_user(tmp, name, len)) {
1333                struct new_utsname *u;
1334
1335                down_write(&uts_sem);
1336                u = utsname();
1337                memcpy(u->nodename, tmp, len);
1338                memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1339                errno = 0;
1340                uts_proc_notify(UTS_PROC_HOSTNAME);
1341                up_write(&uts_sem);
1342        }
1343        return errno;
1344}
1345
1346#ifdef __ARCH_WANT_SYS_GETHOSTNAME
1347
1348SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1349{
1350        int i;
1351        struct new_utsname *u;
1352        char tmp[__NEW_UTS_LEN + 1];
1353
1354        if (len < 0)
1355                return -EINVAL;
1356        down_read(&uts_sem);
1357        u = utsname();
1358        i = 1 + strlen(u->nodename);
1359        if (i > len)
1360                i = len;
1361        memcpy(tmp, u->nodename, i);
1362        up_read(&uts_sem);
1363        if (copy_to_user(name, tmp, i))
1364                return -EFAULT;
1365        return 0;
1366}
1367
1368#endif
1369
1370/*
1371 * Only setdomainname; getdomainname can be implemented by calling
1372 * uname()
1373 */
1374SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1375{
1376        int errno;
1377        char tmp[__NEW_UTS_LEN];
1378
1379        if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1380                return -EPERM;
1381        if (len < 0 || len > __NEW_UTS_LEN)
1382                return -EINVAL;
1383
1384        errno = -EFAULT;
1385        if (!copy_from_user(tmp, name, len)) {
1386                struct new_utsname *u;
1387
1388                down_write(&uts_sem);
1389                u = utsname();
1390                memcpy(u->domainname, tmp, len);
1391                memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1392                errno = 0;
1393                uts_proc_notify(UTS_PROC_DOMAINNAME);
1394                up_write(&uts_sem);
1395        }
1396        return errno;
1397}
1398
1399SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1400{
1401        struct rlimit value;
1402        int ret;
1403
1404        ret = do_prlimit(current, resource, NULL, &value);
1405        if (!ret)
1406                ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1407
1408        return ret;
1409}
1410
1411#ifdef CONFIG_COMPAT
1412
1413COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource,
1414                       struct compat_rlimit __user *, rlim)
1415{
1416        struct rlimit r;
1417        struct compat_rlimit r32;
1418
1419        if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit)))
1420                return -EFAULT;
1421
1422        if (r32.rlim_cur == COMPAT_RLIM_INFINITY)
1423                r.rlim_cur = RLIM_INFINITY;
1424        else
1425                r.rlim_cur = r32.rlim_cur;
1426        if (r32.rlim_max == COMPAT_RLIM_INFINITY)
1427                r.rlim_max = RLIM_INFINITY;
1428        else
1429                r.rlim_max = r32.rlim_max;
1430        return do_prlimit(current, resource, &r, NULL);
1431}
1432
1433COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource,
1434                       struct compat_rlimit __user *, rlim)
1435{
1436        struct rlimit r;
1437        int ret;
1438
1439        ret = do_prlimit(current, resource, NULL, &r);
1440        if (!ret) {
1441                struct compat_rlimit r32;
1442                if (r.rlim_cur > COMPAT_RLIM_INFINITY)
1443                        r32.rlim_cur = COMPAT_RLIM_INFINITY;
1444                else
1445                        r32.rlim_cur = r.rlim_cur;
1446                if (r.rlim_max > COMPAT_RLIM_INFINITY)
1447                        r32.rlim_max = COMPAT_RLIM_INFINITY;
1448                else
1449                        r32.rlim_max = r.rlim_max;
1450
1451                if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit)))
1452                        return -EFAULT;
1453        }
1454        return ret;
1455}
1456
1457#endif
1458
1459#ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1460
1461/*
1462 *      Back compatibility for getrlimit. Needed for some apps.
1463 */
1464SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1465                struct rlimit __user *, rlim)
1466{
1467        struct rlimit x;
1468        if (resource >= RLIM_NLIMITS)
1469                return -EINVAL;
1470
1471        resource = array_index_nospec(resource, RLIM_NLIMITS);
1472        task_lock(current->group_leader);
1473        x = current->signal->rlim[resource];
1474        task_unlock(current->group_leader);
1475        if (x.rlim_cur > 0x7FFFFFFF)
1476                x.rlim_cur = 0x7FFFFFFF;
1477        if (x.rlim_max > 0x7FFFFFFF)
1478                x.rlim_max = 0x7FFFFFFF;
1479        return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0;
1480}
1481
1482#ifdef CONFIG_COMPAT
1483COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1484                       struct compat_rlimit __user *, rlim)
1485{
1486        struct rlimit r;
1487
1488        if (resource >= RLIM_NLIMITS)
1489                return -EINVAL;
1490
1491        resource = array_index_nospec(resource, RLIM_NLIMITS);
1492        task_lock(current->group_leader);
1493        r = current->signal->rlim[resource];
1494        task_unlock(current->group_leader);
1495        if (r.rlim_cur > 0x7FFFFFFF)
1496                r.rlim_cur = 0x7FFFFFFF;
1497        if (r.rlim_max > 0x7FFFFFFF)
1498                r.rlim_max = 0x7FFFFFFF;
1499
1500        if (put_user(r.rlim_cur, &rlim->rlim_cur) ||
1501            put_user(r.rlim_max, &rlim->rlim_max))
1502                return -EFAULT;
1503        return 0;
1504}
1505#endif
1506
1507#endif
1508
1509static inline bool rlim64_is_infinity(__u64 rlim64)
1510{
1511#if BITS_PER_LONG < 64
1512        return rlim64 >= ULONG_MAX;
1513#else
1514        return rlim64 == RLIM64_INFINITY;
1515#endif
1516}
1517
1518static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1519{
1520        if (rlim->rlim_cur == RLIM_INFINITY)
1521                rlim64->rlim_cur = RLIM64_INFINITY;
1522        else
1523                rlim64->rlim_cur = rlim->rlim_cur;
1524        if (rlim->rlim_max == RLIM_INFINITY)
1525                rlim64->rlim_max = RLIM64_INFINITY;
1526        else
1527                rlim64->rlim_max = rlim->rlim_max;
1528}
1529
1530static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1531{
1532        if (rlim64_is_infinity(rlim64->rlim_cur))
1533                rlim->rlim_cur = RLIM_INFINITY;
1534        else
1535                rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1536        if (rlim64_is_infinity(rlim64->rlim_max))
1537                rlim->rlim_max = RLIM_INFINITY;
1538        else
1539                rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1540}
1541
1542/* make sure you are allowed to change @tsk limits before calling this */
1543int do_prlimit(struct task_struct *tsk, unsigned int resource,
1544                struct rlimit *new_rlim, struct rlimit *old_rlim)
1545{
1546        struct rlimit *rlim;
1547        int retval = 0;
1548
1549        if (resource >= RLIM_NLIMITS)
1550                return -EINVAL;
1551        if (new_rlim) {
1552                if (new_rlim->rlim_cur > new_rlim->rlim_max)
1553                        return -EINVAL;
1554                if (resource == RLIMIT_NOFILE &&
1555                                new_rlim->rlim_max > sysctl_nr_open)
1556                        return -EPERM;
1557        }
1558
1559        /* protect tsk->signal and tsk->sighand from disappearing */
1560        read_lock(&tasklist_lock);
1561        if (!tsk->sighand) {
1562                retval = -ESRCH;
1563                goto out;
1564        }
1565
1566        rlim = tsk->signal->rlim + resource;
1567        task_lock(tsk->group_leader);
1568        if (new_rlim) {
1569                /* Keep the capable check against init_user_ns until
1570                   cgroups can contain all limits */
1571                if (new_rlim->rlim_max > rlim->rlim_max &&
1572                                !capable(CAP_SYS_RESOURCE))
1573                        retval = -EPERM;
1574                if (!retval)
1575                        retval = security_task_setrlimit(tsk, resource, new_rlim);
1576        }
1577        if (!retval) {
1578                if (old_rlim)
1579                        *old_rlim = *rlim;
1580                if (new_rlim)
1581                        *rlim = *new_rlim;
1582        }
1583        task_unlock(tsk->group_leader);
1584
1585        /*
1586         * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1587         * infite. In case of RLIM_INFINITY the posix CPU timer code
1588         * ignores the rlimit.
1589         */
1590         if (!retval && new_rlim && resource == RLIMIT_CPU &&
1591             new_rlim->rlim_cur != RLIM_INFINITY &&
1592             IS_ENABLED(CONFIG_POSIX_TIMERS))
1593                update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1594out:
1595        read_unlock(&tasklist_lock);
1596        return retval;
1597}
1598
1599/* rcu lock must be held */
1600static int check_prlimit_permission(struct task_struct *task,
1601                                    unsigned int flags)
1602{
1603        const struct cred *cred = current_cred(), *tcred;
1604        bool id_match;
1605
1606        if (current == task)
1607                return 0;
1608
1609        tcred = __task_cred(task);
1610        id_match = (uid_eq(cred->uid, tcred->euid) &&
1611                    uid_eq(cred->uid, tcred->suid) &&
1612                    uid_eq(cred->uid, tcred->uid)  &&
1613                    gid_eq(cred->gid, tcred->egid) &&
1614                    gid_eq(cred->gid, tcred->sgid) &&
1615                    gid_eq(cred->gid, tcred->gid));
1616        if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1617                return -EPERM;
1618
1619        return security_task_prlimit(cred, tcred, flags);
1620}
1621
1622SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1623                const struct rlimit64 __user *, new_rlim,
1624                struct rlimit64 __user *, old_rlim)
1625{
1626        struct rlimit64 old64, new64;
1627        struct rlimit old, new;
1628        struct task_struct *tsk;
1629        unsigned int checkflags = 0;
1630        int ret;
1631
1632        if (old_rlim)
1633                checkflags |= LSM_PRLIMIT_READ;
1634
1635        if (new_rlim) {
1636                if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1637                        return -EFAULT;
1638                rlim64_to_rlim(&new64, &new);
1639                checkflags |= LSM_PRLIMIT_WRITE;
1640        }
1641
1642        rcu_read_lock();
1643        tsk = pid ? find_task_by_vpid(pid) : current;
1644        if (!tsk) {
1645                rcu_read_unlock();
1646                return -ESRCH;
1647        }
1648        ret = check_prlimit_permission(tsk, checkflags);
1649        if (ret) {
1650                rcu_read_unlock();
1651                return ret;
1652        }
1653        get_task_struct(tsk);
1654        rcu_read_unlock();
1655
1656        ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1657                        old_rlim ? &old : NULL);
1658
1659        if (!ret && old_rlim) {
1660                rlim_to_rlim64(&old, &old64);
1661                if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1662                        ret = -EFAULT;
1663        }
1664
1665        put_task_struct(tsk);
1666        return ret;
1667}
1668
1669SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1670{
1671        struct rlimit new_rlim;
1672
1673        if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1674                return -EFAULT;
1675        return do_prlimit(current, resource, &new_rlim, NULL);
1676}
1677
1678/*
1679 * It would make sense to put struct rusage in the task_struct,
1680 * except that would make the task_struct be *really big*.  After
1681 * task_struct gets moved into malloc'ed memory, it would
1682 * make sense to do this.  It will make moving the rest of the information
1683 * a lot simpler!  (Which we're not doing right now because we're not
1684 * measuring them yet).
1685 *
1686 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1687 * races with threads incrementing their own counters.  But since word
1688 * reads are atomic, we either get new values or old values and we don't
1689 * care which for the sums.  We always take the siglock to protect reading
1690 * the c* fields from p->signal from races with exit.c updating those
1691 * fields when reaping, so a sample either gets all the additions of a
1692 * given child after it's reaped, or none so this sample is before reaping.
1693 *
1694 * Locking:
1695 * We need to take the siglock for CHILDEREN, SELF and BOTH
1696 * for  the cases current multithreaded, non-current single threaded
1697 * non-current multithreaded.  Thread traversal is now safe with
1698 * the siglock held.
1699 * Strictly speaking, we donot need to take the siglock if we are current and
1700 * single threaded,  as no one else can take our signal_struct away, no one
1701 * else can  reap the  children to update signal->c* counters, and no one else
1702 * can race with the signal-> fields. If we do not take any lock, the
1703 * signal-> fields could be read out of order while another thread was just
1704 * exiting. So we should  place a read memory barrier when we avoid the lock.
1705 * On the writer side,  write memory barrier is implied in  __exit_signal
1706 * as __exit_signal releases  the siglock spinlock after updating the signal->
1707 * fields. But we don't do this yet to keep things simple.
1708 *
1709 */
1710
1711static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1712{
1713        r->ru_nvcsw += t->nvcsw;
1714        r->ru_nivcsw += t->nivcsw;
1715        r->ru_minflt += t->min_flt;
1716        r->ru_majflt += t->maj_flt;
1717        r->ru_inblock += task_io_get_inblock(t);
1718        r->ru_oublock += task_io_get_oublock(t);
1719}
1720
1721void getrusage(struct task_struct *p, int who, struct rusage *r)
1722{
1723        struct task_struct *t;
1724        unsigned long flags;
1725        u64 tgutime, tgstime, utime, stime;
1726        unsigned long maxrss = 0;
1727
1728        memset((char *)r, 0, sizeof (*r));
1729        utime = stime = 0;
1730
1731        if (who == RUSAGE_THREAD) {
1732                task_cputime_adjusted(current, &utime, &stime);
1733                accumulate_thread_rusage(p, r);
1734                maxrss = p->signal->maxrss;
1735                goto out;
1736        }
1737
1738        if (!lock_task_sighand(p, &flags))
1739                return;
1740
1741        switch (who) {
1742        case RUSAGE_BOTH:
1743        case RUSAGE_CHILDREN:
1744                utime = p->signal->cutime;
1745                stime = p->signal->cstime;
1746                r->ru_nvcsw = p->signal->cnvcsw;
1747                r->ru_nivcsw = p->signal->cnivcsw;
1748                r->ru_minflt = p->signal->cmin_flt;
1749                r->ru_majflt = p->signal->cmaj_flt;
1750                r->ru_inblock = p->signal->cinblock;
1751                r->ru_oublock = p->signal->coublock;
1752                maxrss = p->signal->cmaxrss;
1753
1754                if (who == RUSAGE_CHILDREN)
1755                        break;
1756                fallthrough;
1757
1758        case RUSAGE_SELF:
1759                thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1760                utime += tgutime;
1761                stime += tgstime;
1762                r->ru_nvcsw += p->signal->nvcsw;
1763                r->ru_nivcsw += p->signal->nivcsw;
1764                r->ru_minflt += p->signal->min_flt;
1765                r->ru_majflt += p->signal->maj_flt;
1766                r->ru_inblock += p->signal->inblock;
1767                r->ru_oublock += p->signal->oublock;
1768                if (maxrss < p->signal->maxrss)
1769                        maxrss = p->signal->maxrss;
1770                t = p;
1771                do {
1772                        accumulate_thread_rusage(t, r);
1773                } while_each_thread(p, t);
1774                break;
1775
1776        default:
1777                BUG();
1778        }
1779        unlock_task_sighand(p, &flags);
1780
1781out:
1782        r->ru_utime = ns_to_kernel_old_timeval(utime);
1783        r->ru_stime = ns_to_kernel_old_timeval(stime);
1784
1785        if (who != RUSAGE_CHILDREN) {
1786                struct mm_struct *mm = get_task_mm(p);
1787
1788                if (mm) {
1789                        setmax_mm_hiwater_rss(&maxrss, mm);
1790                        mmput(mm);
1791                }
1792        }
1793        r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1794}
1795
1796SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1797{
1798        struct rusage r;
1799
1800        if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1801            who != RUSAGE_THREAD)
1802                return -EINVAL;
1803
1804        getrusage(current, who, &r);
1805        return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1806}
1807
1808#ifdef CONFIG_COMPAT
1809COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1810{
1811        struct rusage r;
1812
1813        if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1814            who != RUSAGE_THREAD)
1815                return -EINVAL;
1816
1817        getrusage(current, who, &r);
1818        return put_compat_rusage(&r, ru);
1819}
1820#endif
1821
1822SYSCALL_DEFINE1(umask, int, mask)
1823{
1824        mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1825        return mask;
1826}
1827
1828static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1829{
1830        struct fd exe;
1831        struct file *old_exe, *exe_file;
1832        struct inode *inode;
1833        int err;
1834
1835        exe = fdget(fd);
1836        if (!exe.file)
1837                return -EBADF;
1838
1839        inode = file_inode(exe.file);
1840
1841        /*
1842         * Because the original mm->exe_file points to executable file, make
1843         * sure that this one is executable as well, to avoid breaking an
1844         * overall picture.
1845         */
1846        err = -EACCES;
1847        if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path))
1848                goto exit;
1849
1850        err = file_permission(exe.file, MAY_EXEC);
1851        if (err)
1852                goto exit;
1853
1854        /*
1855         * Forbid mm->exe_file change if old file still mapped.
1856         */
1857        exe_file = get_mm_exe_file(mm);
1858        err = -EBUSY;
1859        if (exe_file) {
1860                struct vm_area_struct *vma;
1861
1862                mmap_read_lock(mm);
1863                for (vma = mm->mmap; vma; vma = vma->vm_next) {
1864                        if (!vma->vm_file)
1865                                continue;
1866                        if (path_equal(&vma->vm_file->f_path,
1867                                       &exe_file->f_path))
1868                                goto exit_err;
1869                }
1870
1871                mmap_read_unlock(mm);
1872                fput(exe_file);
1873        }
1874
1875        err = 0;
1876        /* set the new file, lockless */
1877        get_file(exe.file);
1878        old_exe = xchg(&mm->exe_file, exe.file);
1879        if (old_exe)
1880                fput(old_exe);
1881exit:
1882        fdput(exe);
1883        return err;
1884exit_err:
1885        mmap_read_unlock(mm);
1886        fput(exe_file);
1887        goto exit;
1888}
1889
1890/*
1891 * Check arithmetic relations of passed addresses.
1892 *
1893 * WARNING: we don't require any capability here so be very careful
1894 * in what is allowed for modification from userspace.
1895 */
1896static int validate_prctl_map_addr(struct prctl_mm_map *prctl_map)
1897{
1898        unsigned long mmap_max_addr = TASK_SIZE;
1899        int error = -EINVAL, i;
1900
1901        static const unsigned char offsets[] = {
1902                offsetof(struct prctl_mm_map, start_code),
1903                offsetof(struct prctl_mm_map, end_code),
1904                offsetof(struct prctl_mm_map, start_data),
1905                offsetof(struct prctl_mm_map, end_data),
1906                offsetof(struct prctl_mm_map, start_brk),
1907                offsetof(struct prctl_mm_map, brk),
1908                offsetof(struct prctl_mm_map, start_stack),
1909                offsetof(struct prctl_mm_map, arg_start),
1910                offsetof(struct prctl_mm_map, arg_end),
1911                offsetof(struct prctl_mm_map, env_start),
1912                offsetof(struct prctl_mm_map, env_end),
1913        };
1914
1915        /*
1916         * Make sure the members are not somewhere outside
1917         * of allowed address space.
1918         */
1919        for (i = 0; i < ARRAY_SIZE(offsets); i++) {
1920                u64 val = *(u64 *)((char *)prctl_map + offsets[i]);
1921
1922                if ((unsigned long)val >= mmap_max_addr ||
1923                    (unsigned long)val < mmap_min_addr)
1924                        goto out;
1925        }
1926
1927        /*
1928         * Make sure the pairs are ordered.
1929         */
1930#define __prctl_check_order(__m1, __op, __m2)                           \
1931        ((unsigned long)prctl_map->__m1 __op                            \
1932         (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1933        error  = __prctl_check_order(start_code, <, end_code);
1934        error |= __prctl_check_order(start_data,<=, end_data);
1935        error |= __prctl_check_order(start_brk, <=, brk);
1936        error |= __prctl_check_order(arg_start, <=, arg_end);
1937        error |= __prctl_check_order(env_start, <=, env_end);
1938        if (error)
1939                goto out;
1940#undef __prctl_check_order
1941
1942        error = -EINVAL;
1943
1944        /*
1945         * @brk should be after @end_data in traditional maps.
1946         */
1947        if (prctl_map->start_brk <= prctl_map->end_data ||
1948            prctl_map->brk <= prctl_map->end_data)
1949                goto out;
1950
1951        /*
1952         * Neither we should allow to override limits if they set.
1953         */
1954        if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk,
1955                              prctl_map->start_brk, prctl_map->end_data,
1956                              prctl_map->start_data))
1957                        goto out;
1958
1959        error = 0;
1960out:
1961        return error;
1962}
1963
1964#ifdef CONFIG_CHECKPOINT_RESTORE
1965static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size)
1966{
1967        struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, };
1968        unsigned long user_auxv[AT_VECTOR_SIZE];
1969        struct mm_struct *mm = current->mm;
1970        int error;
1971
1972        BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1973        BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256);
1974
1975        if (opt == PR_SET_MM_MAP_SIZE)
1976                return put_user((unsigned int)sizeof(prctl_map),
1977                                (unsigned int __user *)addr);
1978
1979        if (data_size != sizeof(prctl_map))
1980                return -EINVAL;
1981
1982        if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
1983                return -EFAULT;
1984
1985        error = validate_prctl_map_addr(&prctl_map);
1986        if (error)
1987                return error;
1988
1989        if (prctl_map.auxv_size) {
1990                /*
1991                 * Someone is trying to cheat the auxv vector.
1992                 */
1993                if (!prctl_map.auxv ||
1994                                prctl_map.auxv_size > sizeof(mm->saved_auxv))
1995                        return -EINVAL;
1996
1997                memset(user_auxv, 0, sizeof(user_auxv));
1998                if (copy_from_user(user_auxv,
1999                                   (const void __user *)prctl_map.auxv,
2000                                   prctl_map.auxv_size))
2001                        return -EFAULT;
2002
2003                /* Last entry must be AT_NULL as specification requires */
2004                user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL;
2005                user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL;
2006        }
2007
2008        if (prctl_map.exe_fd != (u32)-1) {
2009                /*
2010                 * Check if the current user is checkpoint/restore capable.
2011                 * At the time of this writing, it checks for CAP_SYS_ADMIN
2012                 * or CAP_CHECKPOINT_RESTORE.
2013                 * Note that a user with access to ptrace can masquerade an
2014                 * arbitrary program as any executable, even setuid ones.
2015                 * This may have implications in the tomoyo subsystem.
2016                 */
2017                if (!checkpoint_restore_ns_capable(current_user_ns()))
2018                        return -EPERM;
2019
2020                error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd);
2021                if (error)
2022                        return error;
2023        }
2024
2025        /*
2026         * arg_lock protects concurent updates but we still need mmap_lock for
2027         * read to exclude races with sys_brk.
2028         */
2029        mmap_read_lock(mm);
2030
2031        /*
2032         * We don't validate if these members are pointing to
2033         * real present VMAs because application may have correspond
2034         * VMAs already unmapped and kernel uses these members for statistics
2035         * output in procfs mostly, except
2036         *
2037         *  - @start_brk/@brk which are used in do_brk_flags but kernel lookups
2038         *    for VMAs when updating these memvers so anything wrong written
2039         *    here cause kernel to swear at userspace program but won't lead
2040         *    to any problem in kernel itself
2041         */
2042
2043        spin_lock(&mm->arg_lock);
2044        mm->start_code  = prctl_map.start_code;
2045        mm->end_code    = prctl_map.end_code;
2046        mm->start_data  = prctl_map.start_data;
2047        mm->end_data    = prctl_map.end_data;
2048        mm->start_brk   = prctl_map.start_brk;
2049        mm->brk         = prctl_map.brk;
2050        mm->start_stack = prctl_map.start_stack;
2051        mm->arg_start   = prctl_map.arg_start;
2052        mm->arg_end     = prctl_map.arg_end;
2053        mm->env_start   = prctl_map.env_start;
2054        mm->env_end     = prctl_map.env_end;
2055        spin_unlock(&mm->arg_lock);
2056
2057        /*
2058         * Note this update of @saved_auxv is lockless thus
2059         * if someone reads this member in procfs while we're
2060         * updating -- it may get partly updated results. It's
2061         * known and acceptable trade off: we leave it as is to
2062         * not introduce additional locks here making the kernel
2063         * more complex.
2064         */
2065        if (prctl_map.auxv_size)
2066                memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
2067
2068        mmap_read_unlock(mm);
2069        return 0;
2070}
2071#endif /* CONFIG_CHECKPOINT_RESTORE */
2072
2073static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr,
2074                          unsigned long len)
2075{
2076        /*
2077         * This doesn't move the auxiliary vector itself since it's pinned to
2078         * mm_struct, but it permits filling the vector with new values.  It's
2079         * up to the caller to provide sane values here, otherwise userspace
2080         * tools which use this vector might be unhappy.
2081         */
2082        unsigned long user_auxv[AT_VECTOR_SIZE] = {};
2083
2084        if (len > sizeof(user_auxv))
2085                return -EINVAL;
2086
2087        if (copy_from_user(user_auxv, (const void __user *)addr, len))
2088                return -EFAULT;
2089
2090        /* Make sure the last entry is always AT_NULL */
2091        user_auxv[AT_VECTOR_SIZE - 2] = 0;
2092        user_auxv[AT_VECTOR_SIZE - 1] = 0;
2093
2094        BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2095
2096        task_lock(current);
2097        memcpy(mm->saved_auxv, user_auxv, len);
2098        task_unlock(current);
2099
2100        return 0;
2101}
2102
2103static int prctl_set_mm(int opt, unsigned long addr,
2104                        unsigned long arg4, unsigned long arg5)
2105{
2106        struct mm_struct *mm = current->mm;
2107        struct prctl_mm_map prctl_map = {
2108                .auxv = NULL,
2109                .auxv_size = 0,
2110                .exe_fd = -1,
2111        };
2112        struct vm_area_struct *vma;
2113        int error;
2114
2115        if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV &&
2116                              opt != PR_SET_MM_MAP &&
2117                              opt != PR_SET_MM_MAP_SIZE)))
2118                return -EINVAL;
2119
2120#ifdef CONFIG_CHECKPOINT_RESTORE
2121        if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE)
2122                return prctl_set_mm_map(opt, (const void __user *)addr, arg4);
2123#endif
2124
2125        if (!capable(CAP_SYS_RESOURCE))
2126                return -EPERM;
2127
2128        if (opt == PR_SET_MM_EXE_FILE)
2129                return prctl_set_mm_exe_file(mm, (unsigned int)addr);
2130
2131        if (opt == PR_SET_MM_AUXV)
2132                return prctl_set_auxv(mm, addr, arg4);
2133
2134        if (addr >= TASK_SIZE || addr < mmap_min_addr)
2135                return -EINVAL;
2136
2137        error = -EINVAL;
2138
2139        /*
2140         * arg_lock protects concurent updates of arg boundaries, we need
2141         * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr
2142         * validation.
2143         */
2144        mmap_read_lock(mm);
2145        vma = find_vma(mm, addr);
2146
2147        spin_lock(&mm->arg_lock);
2148        prctl_map.start_code    = mm->start_code;
2149        prctl_map.end_code      = mm->end_code;
2150        prctl_map.start_data    = mm->start_data;
2151        prctl_map.end_data      = mm->end_data;
2152        prctl_map.start_brk     = mm->start_brk;
2153        prctl_map.brk           = mm->brk;
2154        prctl_map.start_stack   = mm->start_stack;
2155        prctl_map.arg_start     = mm->arg_start;
2156        prctl_map.arg_end       = mm->arg_end;
2157        prctl_map.env_start     = mm->env_start;
2158        prctl_map.env_end       = mm->env_end;
2159
2160        switch (opt) {
2161        case PR_SET_MM_START_CODE:
2162                prctl_map.start_code = addr;
2163                break;
2164        case PR_SET_MM_END_CODE:
2165                prctl_map.end_code = addr;
2166                break;
2167        case PR_SET_MM_START_DATA:
2168                prctl_map.start_data = addr;
2169                break;
2170        case PR_SET_MM_END_DATA:
2171                prctl_map.end_data = addr;
2172                break;
2173        case PR_SET_MM_START_STACK:
2174                prctl_map.start_stack = addr;
2175                break;
2176        case PR_SET_MM_START_BRK:
2177                prctl_map.start_brk = addr;
2178                break;
2179        case PR_SET_MM_BRK:
2180                prctl_map.brk = addr;
2181                break;
2182        case PR_SET_MM_ARG_START:
2183                prctl_map.arg_start = addr;
2184                break;
2185        case PR_SET_MM_ARG_END:
2186                prctl_map.arg_end = addr;
2187                break;
2188        case PR_SET_MM_ENV_START:
2189                prctl_map.env_start = addr;
2190                break;
2191        case PR_SET_MM_ENV_END:
2192                prctl_map.env_end = addr;
2193                break;
2194        default:
2195                goto out;
2196        }
2197
2198        error = validate_prctl_map_addr(&prctl_map);
2199        if (error)
2200                goto out;
2201
2202        switch (opt) {
2203        /*
2204         * If command line arguments and environment
2205         * are placed somewhere else on stack, we can
2206         * set them up here, ARG_START/END to setup
2207         * command line argumets and ENV_START/END
2208         * for environment.
2209         */
2210        case PR_SET_MM_START_STACK:
2211        case PR_SET_MM_ARG_START:
2212        case PR_SET_MM_ARG_END:
2213        case PR_SET_MM_ENV_START:
2214        case PR_SET_MM_ENV_END:
2215                if (!vma) {
2216                        error = -EFAULT;
2217                        goto out;
2218                }
2219        }
2220
2221        mm->start_code  = prctl_map.start_code;
2222        mm->end_code    = prctl_map.end_code;
2223        mm->start_data  = prctl_map.start_data;
2224        mm->end_data    = prctl_map.end_data;
2225        mm->start_brk   = prctl_map.start_brk;
2226        mm->brk         = prctl_map.brk;
2227        mm->start_stack = prctl_map.start_stack;
2228        mm->arg_start   = prctl_map.arg_start;
2229        mm->arg_end     = prctl_map.arg_end;
2230        mm->env_start   = prctl_map.env_start;
2231        mm->env_end     = prctl_map.env_end;
2232
2233        error = 0;
2234out:
2235        spin_unlock(&mm->arg_lock);
2236        mmap_read_unlock(mm);
2237        return error;
2238}
2239
2240#ifdef CONFIG_CHECKPOINT_RESTORE
2241static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2242{
2243        return put_user(me->clear_child_tid, tid_addr);
2244}
2245#else
2246static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2247{
2248        return -EINVAL;
2249}
2250#endif
2251
2252static int propagate_has_child_subreaper(struct task_struct *p, void *data)
2253{
2254        /*
2255         * If task has has_child_subreaper - all its decendants
2256         * already have these flag too and new decendants will
2257         * inherit it on fork, skip them.
2258         *
2259         * If we've found child_reaper - skip descendants in
2260         * it's subtree as they will never get out pidns.
2261         */
2262        if (p->signal->has_child_subreaper ||
2263            is_child_reaper(task_pid(p)))
2264                return 0;
2265
2266        p->signal->has_child_subreaper = 1;
2267        return 1;
2268}
2269
2270int __weak arch_prctl_spec_ctrl_get(struct task_struct *t, unsigned long which)
2271{
2272        return -EINVAL;
2273}
2274
2275int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which,
2276                                    unsigned long ctrl)
2277{
2278        return -EINVAL;
2279}
2280
2281#define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE)
2282
2283SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2284                unsigned long, arg4, unsigned long, arg5)
2285{
2286        struct task_struct *me = current;
2287        unsigned char comm[sizeof(me->comm)];
2288        long error;
2289
2290        error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2291        if (error != -ENOSYS)
2292                return error;
2293
2294        error = 0;
2295        switch (option) {
2296        case PR_SET_PDEATHSIG:
2297                if (!valid_signal(arg2)) {
2298                        error = -EINVAL;
2299                        break;
2300                }
2301                me->pdeath_signal = arg2;
2302                break;
2303        case PR_GET_PDEATHSIG:
2304                error = put_user(me->pdeath_signal, (int __user *)arg2);
2305                break;
2306        case PR_GET_DUMPABLE:
2307                error = get_dumpable(me->mm);
2308                break;
2309        case PR_SET_DUMPABLE:
2310                if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2311                        error = -EINVAL;
2312                        break;
2313                }
2314                set_dumpable(me->mm, arg2);
2315                break;
2316
2317        case PR_SET_UNALIGN:
2318                error = SET_UNALIGN_CTL(me, arg2);
2319                break;
2320        case PR_GET_UNALIGN:
2321                error = GET_UNALIGN_CTL(me, arg2);
2322                break;
2323        case PR_SET_FPEMU:
2324                error = SET_FPEMU_CTL(me, arg2);
2325                break;
2326        case PR_GET_FPEMU:
2327                error = GET_FPEMU_CTL(me, arg2);
2328                break;
2329        case PR_SET_FPEXC:
2330                error = SET_FPEXC_CTL(me, arg2);
2331                break;
2332        case PR_GET_FPEXC:
2333                error = GET_FPEXC_CTL(me, arg2);
2334                break;
2335        case PR_GET_TIMING:
2336                error = PR_TIMING_STATISTICAL;
2337                break;
2338        case PR_SET_TIMING:
2339                if (arg2 != PR_TIMING_STATISTICAL)
2340                        error = -EINVAL;
2341                break;
2342        case PR_SET_NAME:
2343                comm[sizeof(me->comm) - 1] = 0;
2344                if (strncpy_from_user(comm, (char __user *)arg2,
2345                                      sizeof(me->comm) - 1) < 0)
2346                        return -EFAULT;
2347                set_task_comm(me, comm);
2348                proc_comm_connector(me);
2349                break;
2350        case PR_GET_NAME:
2351                get_task_comm(comm, me);
2352                if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2353                        return -EFAULT;
2354                break;
2355        case PR_GET_ENDIAN:
2356                error = GET_ENDIAN(me, arg2);
2357                break;
2358        case PR_SET_ENDIAN:
2359                error = SET_ENDIAN(me, arg2);
2360                break;
2361        case PR_GET_SECCOMP:
2362                error = prctl_get_seccomp();
2363                break;
2364        case PR_SET_SECCOMP:
2365                error = prctl_set_seccomp(arg2, (char __user *)arg3);
2366                break;
2367        case PR_GET_TSC:
2368                error = GET_TSC_CTL(arg2);
2369                break;
2370        case PR_SET_TSC:
2371                error = SET_TSC_CTL(arg2);
2372                break;
2373        case PR_TASK_PERF_EVENTS_DISABLE:
2374                error = perf_event_task_disable();
2375                break;
2376        case PR_TASK_PERF_EVENTS_ENABLE:
2377                error = perf_event_task_enable();
2378                break;
2379        case PR_GET_TIMERSLACK:
2380                if (current->timer_slack_ns > ULONG_MAX)
2381                        error = ULONG_MAX;
2382                else
2383                        error = current->timer_slack_ns;
2384                break;
2385        case PR_SET_TIMERSLACK:
2386                if (arg2 <= 0)
2387                        current->timer_slack_ns =
2388                                        current->default_timer_slack_ns;
2389                else
2390                        current->timer_slack_ns = arg2;
2391                break;
2392        case PR_MCE_KILL:
2393                if (arg4 | arg5)
2394                        return -EINVAL;
2395                switch (arg2) {
2396                case PR_MCE_KILL_CLEAR:
2397                        if (arg3 != 0)
2398                                return -EINVAL;
2399                        current->flags &= ~PF_MCE_PROCESS;
2400                        break;
2401                case PR_MCE_KILL_SET:
2402                        current->flags |= PF_MCE_PROCESS;
2403                        if (arg3 == PR_MCE_KILL_EARLY)
2404                                current->flags |= PF_MCE_EARLY;
2405                        else if (arg3 == PR_MCE_KILL_LATE)
2406                                current->flags &= ~PF_MCE_EARLY;
2407                        else if (arg3 == PR_MCE_KILL_DEFAULT)
2408                                current->flags &=
2409                                                ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2410                        else
2411                                return -EINVAL;
2412                        break;
2413                default:
2414                        return -EINVAL;
2415                }
2416                break;
2417        case PR_MCE_KILL_GET:
2418                if (arg2 | arg3 | arg4 | arg5)
2419                        return -EINVAL;
2420                if (current->flags & PF_MCE_PROCESS)
2421                        error = (current->flags & PF_MCE_EARLY) ?
2422                                PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2423                else
2424                        error = PR_MCE_KILL_DEFAULT;
2425                break;
2426        case PR_SET_MM:
2427                error = prctl_set_mm(arg2, arg3, arg4, arg5);
2428                break;
2429        case PR_GET_TID_ADDRESS:
2430                error = prctl_get_tid_address(me, (int __user * __user *)arg2);
2431                break;
2432        case PR_SET_CHILD_SUBREAPER:
2433                me->signal->is_child_subreaper = !!arg2;
2434                if (!arg2)
2435                        break;
2436
2437                walk_process_tree(me, propagate_has_child_subreaper, NULL);
2438                break;
2439        case PR_GET_CHILD_SUBREAPER:
2440                error = put_user(me->signal->is_child_subreaper,
2441                                 (int __user *)arg2);
2442                break;
2443        case PR_SET_NO_NEW_PRIVS:
2444                if (arg2 != 1 || arg3 || arg4 || arg5)
2445                        return -EINVAL;
2446
2447                task_set_no_new_privs(current);
2448                break;
2449        case PR_GET_NO_NEW_PRIVS:
2450                if (arg2 || arg3 || arg4 || arg5)
2451                        return -EINVAL;
2452                return task_no_new_privs(current) ? 1 : 0;
2453        case PR_GET_THP_DISABLE:
2454                if (arg2 || arg3 || arg4 || arg5)
2455                        return -EINVAL;
2456                error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags);
2457                break;
2458        case PR_SET_THP_DISABLE:
2459                if (arg3 || arg4 || arg5)
2460                        return -EINVAL;
2461                if (mmap_write_lock_killable(me->mm))
2462                        return -EINTR;
2463                if (arg2)
2464                        set_bit(MMF_DISABLE_THP, &me->mm->flags);
2465                else
2466                        clear_bit(MMF_DISABLE_THP, &me->mm->flags);
2467                mmap_write_unlock(me->mm);
2468                break;
2469        case PR_MPX_ENABLE_MANAGEMENT:
2470        case PR_MPX_DISABLE_MANAGEMENT:
2471                /* No longer implemented: */
2472                return -EINVAL;
2473        case PR_SET_FP_MODE:
2474                error = SET_FP_MODE(me, arg2);
2475                break;
2476        case PR_GET_FP_MODE:
2477                error = GET_FP_MODE(me);
2478                break;
2479        case PR_SVE_SET_VL:
2480                error = SVE_SET_VL(arg2);
2481                break;
2482        case PR_SVE_GET_VL:
2483                error = SVE_GET_VL();
2484                break;
2485        case PR_GET_SPECULATION_CTRL:
2486                if (arg3 || arg4 || arg5)
2487                        return -EINVAL;
2488                error = arch_prctl_spec_ctrl_get(me, arg2);
2489                break;
2490        case PR_SET_SPECULATION_CTRL:
2491                if (arg4 || arg5)
2492                        return -EINVAL;
2493                error = arch_prctl_spec_ctrl_set(me, arg2, arg3);
2494                break;
2495        case PR_PAC_RESET_KEYS:
2496                if (arg3 || arg4 || arg5)
2497                        return -EINVAL;
2498                error = PAC_RESET_KEYS(me, arg2);
2499                break;
2500        case PR_SET_TAGGED_ADDR_CTRL:
2501                if (arg3 || arg4 || arg5)
2502                        return -EINVAL;
2503                error = SET_TAGGED_ADDR_CTRL(arg2);
2504                break;
2505        case PR_GET_TAGGED_ADDR_CTRL:
2506                if (arg2 || arg3 || arg4 || arg5)
2507                        return -EINVAL;
2508                error = GET_TAGGED_ADDR_CTRL();
2509                break;
2510        case PR_SET_IO_FLUSHER:
2511                if (!capable(CAP_SYS_RESOURCE))
2512                        return -EPERM;
2513
2514                if (arg3 || arg4 || arg5)
2515                        return -EINVAL;
2516
2517                if (arg2 == 1)
2518                        current->flags |= PR_IO_FLUSHER;
2519                else if (!arg2)
2520                        current->flags &= ~PR_IO_FLUSHER;
2521                else
2522                        return -EINVAL;
2523                break;
2524        case PR_GET_IO_FLUSHER:
2525                if (!capable(CAP_SYS_RESOURCE))
2526                        return -EPERM;
2527
2528                if (arg2 || arg3 || arg4 || arg5)
2529                        return -EINVAL;
2530
2531                error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER;
2532                break;
2533        case PR_SET_SYSCALL_USER_DISPATCH:
2534                error = set_syscall_user_dispatch(arg2, arg3, arg4,
2535                                                  (char __user *) arg5);
2536                break;
2537        default:
2538                error = -EINVAL;
2539                break;
2540        }
2541        return error;
2542}
2543
2544SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2545                struct getcpu_cache __user *, unused)
2546{
2547        int err = 0;
2548        int cpu = raw_smp_processor_id();
2549
2550        if (cpup)
2551                err |= put_user(cpu, cpup);
2552        if (nodep)
2553                err |= put_user(cpu_to_node(cpu), nodep);
2554        return err ? -EFAULT : 0;
2555}
2556
2557/**
2558 * do_sysinfo - fill in sysinfo struct
2559 * @info: pointer to buffer to fill
2560 */
2561static int do_sysinfo(struct sysinfo *info)
2562{
2563        unsigned long mem_total, sav_total;
2564        unsigned int mem_unit, bitcount;
2565        struct timespec64 tp;
2566
2567        memset(info, 0, sizeof(struct sysinfo));
2568
2569        ktime_get_boottime_ts64(&tp);
2570        timens_add_boottime(&tp);
2571        info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2572
2573        get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2574
2575        info->procs = nr_threads;
2576
2577        si_meminfo(info);
2578        si_swapinfo(info);
2579
2580        /*
2581         * If the sum of all the available memory (i.e. ram + swap)
2582         * is less than can be stored in a 32 bit unsigned long then
2583         * we can be binary compatible with 2.2.x kernels.  If not,
2584         * well, in that case 2.2.x was broken anyways...
2585         *
2586         *  -Erik Andersen <andersee@debian.org>
2587         */
2588
2589        mem_total = info->totalram + info->totalswap;
2590        if (mem_total < info->totalram || mem_total < info->totalswap)
2591                goto out;
2592        bitcount = 0;
2593        mem_unit = info->mem_unit;
2594        while (mem_unit > 1) {
2595                bitcount++;
2596                mem_unit >>= 1;
2597                sav_total = mem_total;
2598                mem_total <<= 1;
2599                if (mem_total < sav_total)
2600                        goto out;
2601        }
2602
2603        /*
2604         * If mem_total did not overflow, multiply all memory values by
2605         * info->mem_unit and set it to 1.  This leaves things compatible
2606         * with 2.2.x, and also retains compatibility with earlier 2.4.x
2607         * kernels...
2608         */
2609
2610        info->mem_unit = 1;
2611        info->totalram <<= bitcount;
2612        info->freeram <<= bitcount;
2613        info->sharedram <<= bitcount;
2614        info->bufferram <<= bitcount;
2615        info->totalswap <<= bitcount;
2616        info->freeswap <<= bitcount;
2617        info->totalhigh <<= bitcount;
2618        info->freehigh <<= bitcount;
2619
2620out:
2621        return 0;
2622}
2623
2624SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2625{
2626        struct sysinfo val;
2627
2628        do_sysinfo(&val);
2629
2630        if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2631                return -EFAULT;
2632
2633        return 0;
2634}
2635
2636#ifdef CONFIG_COMPAT
2637struct compat_sysinfo {
2638        s32 uptime;
2639        u32 loads[3];
2640        u32 totalram;
2641        u32 freeram;
2642        u32 sharedram;
2643        u32 bufferram;
2644        u32 totalswap;
2645        u32 freeswap;
2646        u16 procs;
2647        u16 pad;
2648        u32 totalhigh;
2649        u32 freehigh;
2650        u32 mem_unit;
2651        char _f[20-2*sizeof(u32)-sizeof(int)];
2652};
2653
2654COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2655{
2656        struct sysinfo s;
2657        struct compat_sysinfo s_32;
2658
2659        do_sysinfo(&s);
2660
2661        /* Check to see if any memory value is too large for 32-bit and scale
2662         *  down if needed
2663         */
2664        if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) {
2665                int bitcount = 0;
2666
2667                while (s.mem_unit < PAGE_SIZE) {
2668                        s.mem_unit <<= 1;
2669                        bitcount++;
2670                }
2671
2672                s.totalram >>= bitcount;
2673                s.freeram >>= bitcount;
2674                s.sharedram >>= bitcount;
2675                s.bufferram >>= bitcount;
2676                s.totalswap >>= bitcount;
2677                s.freeswap >>= bitcount;
2678                s.totalhigh >>= bitcount;
2679                s.freehigh >>= bitcount;
2680        }
2681
2682        memset(&s_32, 0, sizeof(s_32));
2683        s_32.uptime = s.uptime;
2684        s_32.loads[0] = s.loads[0];
2685        s_32.loads[1] = s.loads[1];
2686        s_32.loads[2] = s.loads[2];
2687        s_32.totalram = s.totalram;
2688        s_32.freeram = s.freeram;
2689        s_32.sharedram = s.sharedram;
2690        s_32.bufferram = s.bufferram;
2691        s_32.totalswap = s.totalswap;
2692        s_32.freeswap = s.freeswap;
2693        s_32.procs = s.procs;
2694        s_32.totalhigh = s.totalhigh;
2695        s_32.freehigh = s.freehigh;
2696        s_32.mem_unit = s.mem_unit;
2697        if (copy_to_user(info, &s_32, sizeof(s_32)))
2698                return -EFAULT;
2699        return 0;
2700}
2701#endif /* CONFIG_COMPAT */
2702