linux/kernel/ptrace.c
<<
>>
Prefs
   1/*
   2 * linux/kernel/ptrace.c
   3 *
   4 * (C) Copyright 1999 Linus Torvalds
   5 *
   6 * Common interfaces for "ptrace()" which we do not want
   7 * to continually duplicate across every architecture.
   8 */
   9
  10#include <linux/capability.h>
  11#include <linux/export.h>
  12#include <linux/sched.h>
  13#include <linux/errno.h>
  14#include <linux/mm.h>
  15#include <linux/highmem.h>
  16#include <linux/pagemap.h>
  17#include <linux/ptrace.h>
  18#include <linux/security.h>
  19#include <linux/signal.h>
  20#include <linux/uio.h>
  21#include <linux/audit.h>
  22#include <linux/pid_namespace.h>
  23#include <linux/syscalls.h>
  24#include <linux/uaccess.h>
  25#include <linux/regset.h>
  26#include <linux/hw_breakpoint.h>
  27#include <linux/cn_proc.h>
  28#include <linux/compat.h>
  29
  30
  31static int ptrace_trapping_sleep_fn(void *flags)
  32{
  33        schedule();
  34        return 0;
  35}
  36
  37/*
  38 * ptrace a task: make the debugger its new parent and
  39 * move it to the ptrace list.
  40 *
  41 * Must be called with the tasklist lock write-held.
  42 */
  43void __ptrace_link(struct task_struct *child, struct task_struct *new_parent)
  44{
  45        BUG_ON(!list_empty(&child->ptrace_entry));
  46        list_add(&child->ptrace_entry, &new_parent->ptraced);
  47        child->parent = new_parent;
  48}
  49
  50/**
  51 * __ptrace_unlink - unlink ptracee and restore its execution state
  52 * @child: ptracee to be unlinked
  53 *
  54 * Remove @child from the ptrace list, move it back to the original parent,
  55 * and restore the execution state so that it conforms to the group stop
  56 * state.
  57 *
  58 * Unlinking can happen via two paths - explicit PTRACE_DETACH or ptracer
  59 * exiting.  For PTRACE_DETACH, unless the ptracee has been killed between
  60 * ptrace_check_attach() and here, it's guaranteed to be in TASK_TRACED.
  61 * If the ptracer is exiting, the ptracee can be in any state.
  62 *
  63 * After detach, the ptracee should be in a state which conforms to the
  64 * group stop.  If the group is stopped or in the process of stopping, the
  65 * ptracee should be put into TASK_STOPPED; otherwise, it should be woken
  66 * up from TASK_TRACED.
  67 *
  68 * If the ptracee is in TASK_TRACED and needs to be moved to TASK_STOPPED,
  69 * it goes through TRACED -> RUNNING -> STOPPED transition which is similar
  70 * to but in the opposite direction of what happens while attaching to a
  71 * stopped task.  However, in this direction, the intermediate RUNNING
  72 * state is not hidden even from the current ptracer and if it immediately
  73 * re-attaches and performs a WNOHANG wait(2), it may fail.
  74 *
  75 * CONTEXT:
  76 * write_lock_irq(tasklist_lock)
  77 */
  78void __ptrace_unlink(struct task_struct *child)
  79{
  80        BUG_ON(!child->ptrace);
  81
  82        child->ptrace = 0;
  83        child->parent = child->real_parent;
  84        list_del_init(&child->ptrace_entry);
  85
  86        spin_lock(&child->sighand->siglock);
  87
  88        /*
  89         * Clear all pending traps and TRAPPING.  TRAPPING should be
  90         * cleared regardless of JOBCTL_STOP_PENDING.  Do it explicitly.
  91         */
  92        task_clear_jobctl_pending(child, JOBCTL_TRAP_MASK);
  93        task_clear_jobctl_trapping(child);
  94
  95        /*
  96         * Reinstate JOBCTL_STOP_PENDING if group stop is in effect and
  97         * @child isn't dead.
  98         */
  99        if (!(child->flags & PF_EXITING) &&
 100            (child->signal->flags & SIGNAL_STOP_STOPPED ||
 101             child->signal->group_stop_count)) {
 102                child->jobctl |= JOBCTL_STOP_PENDING;
 103
 104                /*
 105                 * This is only possible if this thread was cloned by the
 106                 * traced task running in the stopped group, set the signal
 107                 * for the future reports.
 108                 * FIXME: we should change ptrace_init_task() to handle this
 109                 * case.
 110                 */
 111                if (!(child->jobctl & JOBCTL_STOP_SIGMASK))
 112                        child->jobctl |= SIGSTOP;
 113        }
 114
 115        /*
 116         * If transition to TASK_STOPPED is pending or in TASK_TRACED, kick
 117         * @child in the butt.  Note that @resume should be used iff @child
 118         * is in TASK_TRACED; otherwise, we might unduly disrupt
 119         * TASK_KILLABLE sleeps.
 120         */
 121        if (child->jobctl & JOBCTL_STOP_PENDING || task_is_traced(child))
 122                ptrace_signal_wake_up(child, true);
 123
 124        spin_unlock(&child->sighand->siglock);
 125}
 126
 127/* Ensure that nothing can wake it up, even SIGKILL */
 128static bool ptrace_freeze_traced(struct task_struct *task)
 129{
 130        bool ret = false;
 131
 132        /* Lockless, nobody but us can set this flag */
 133        if (task->jobctl & JOBCTL_LISTENING)
 134                return ret;
 135
 136        spin_lock_irq(&task->sighand->siglock);
 137        if (task_is_traced(task) && !__fatal_signal_pending(task)) {
 138                task->state = __TASK_TRACED;
 139                ret = true;
 140        }
 141        spin_unlock_irq(&task->sighand->siglock);
 142
 143        return ret;
 144}
 145
 146static void ptrace_unfreeze_traced(struct task_struct *task)
 147{
 148        if (task->state != __TASK_TRACED)
 149                return;
 150
 151        WARN_ON(!task->ptrace || task->parent != current);
 152
 153        spin_lock_irq(&task->sighand->siglock);
 154        if (__fatal_signal_pending(task))
 155                wake_up_state(task, __TASK_TRACED);
 156        else
 157                task->state = TASK_TRACED;
 158        spin_unlock_irq(&task->sighand->siglock);
 159}
 160
 161/**
 162 * ptrace_check_attach - check whether ptracee is ready for ptrace operation
 163 * @child: ptracee to check for
 164 * @ignore_state: don't check whether @child is currently %TASK_TRACED
 165 *
 166 * Check whether @child is being ptraced by %current and ready for further
 167 * ptrace operations.  If @ignore_state is %false, @child also should be in
 168 * %TASK_TRACED state and on return the child is guaranteed to be traced
 169 * and not executing.  If @ignore_state is %true, @child can be in any
 170 * state.
 171 *
 172 * CONTEXT:
 173 * Grabs and releases tasklist_lock and @child->sighand->siglock.
 174 *
 175 * RETURNS:
 176 * 0 on success, -ESRCH if %child is not ready.
 177 */
 178static int ptrace_check_attach(struct task_struct *child, bool ignore_state)
 179{
 180        int ret = -ESRCH;
 181
 182        /*
 183         * We take the read lock around doing both checks to close a
 184         * possible race where someone else was tracing our child and
 185         * detached between these two checks.  After this locked check,
 186         * we are sure that this is our traced child and that can only
 187         * be changed by us so it's not changing right after this.
 188         */
 189        read_lock(&tasklist_lock);
 190        if (child->ptrace && child->parent == current) {
 191                WARN_ON(child->state == __TASK_TRACED);
 192                /*
 193                 * child->sighand can't be NULL, release_task()
 194                 * does ptrace_unlink() before __exit_signal().
 195                 */
 196                if (ignore_state || ptrace_freeze_traced(child))
 197                        ret = 0;
 198        }
 199        read_unlock(&tasklist_lock);
 200
 201        if (!ret && !ignore_state) {
 202                if (!wait_task_inactive(child, __TASK_TRACED)) {
 203                        /*
 204                         * This can only happen if may_ptrace_stop() fails and
 205                         * ptrace_stop() changes ->state back to TASK_RUNNING,
 206                         * so we should not worry about leaking __TASK_TRACED.
 207                         */
 208                        WARN_ON(child->state == __TASK_TRACED);
 209                        ret = -ESRCH;
 210                }
 211        }
 212
 213        return ret;
 214}
 215
 216static int ptrace_has_cap(struct user_namespace *ns, unsigned int mode)
 217{
 218        if (mode & PTRACE_MODE_NOAUDIT)
 219                return has_ns_capability_noaudit(current, ns, CAP_SYS_PTRACE);
 220        else
 221                return has_ns_capability(current, ns, CAP_SYS_PTRACE);
 222}
 223
 224/* Returns 0 on success, -errno on denial. */
 225static int __ptrace_may_access(struct task_struct *task, unsigned int mode)
 226{
 227        const struct cred *cred = current_cred(), *tcred;
 228
 229        /* May we inspect the given task?
 230         * This check is used both for attaching with ptrace
 231         * and for allowing access to sensitive information in /proc.
 232         *
 233         * ptrace_attach denies several cases that /proc allows
 234         * because setting up the necessary parent/child relationship
 235         * or halting the specified task is impossible.
 236         */
 237        int dumpable = 0;
 238        /* Don't let security modules deny introspection */
 239        if (task == current)
 240                return 0;
 241        rcu_read_lock();
 242        tcred = __task_cred(task);
 243        if (uid_eq(cred->uid, tcred->euid) &&
 244            uid_eq(cred->uid, tcred->suid) &&
 245            uid_eq(cred->uid, tcred->uid)  &&
 246            gid_eq(cred->gid, tcred->egid) &&
 247            gid_eq(cred->gid, tcred->sgid) &&
 248            gid_eq(cred->gid, tcred->gid))
 249                goto ok;
 250        if (ptrace_has_cap(tcred->user_ns, mode))
 251                goto ok;
 252        rcu_read_unlock();
 253        return -EPERM;
 254ok:
 255        rcu_read_unlock();
 256        smp_rmb();
 257        if (task->mm)
 258                dumpable = get_dumpable(task->mm);
 259        rcu_read_lock();
 260        if (!dumpable && !ptrace_has_cap(__task_cred(task)->user_ns, mode)) {
 261                rcu_read_unlock();
 262                return -EPERM;
 263        }
 264        rcu_read_unlock();
 265
 266        return security_ptrace_access_check(task, mode);
 267}
 268
 269bool ptrace_may_access(struct task_struct *task, unsigned int mode)
 270{
 271        int err;
 272        task_lock(task);
 273        err = __ptrace_may_access(task, mode);
 274        task_unlock(task);
 275        return !err;
 276}
 277
 278static int ptrace_attach(struct task_struct *task, long request,
 279                         unsigned long addr,
 280                         unsigned long flags)
 281{
 282        bool seize = (request == PTRACE_SEIZE);
 283        int retval;
 284
 285        retval = -EIO;
 286        if (seize) {
 287                if (addr != 0)
 288                        goto out;
 289                if (flags & ~(unsigned long)PTRACE_O_MASK)
 290                        goto out;
 291                flags = PT_PTRACED | PT_SEIZED | (flags << PT_OPT_FLAG_SHIFT);
 292        } else {
 293                flags = PT_PTRACED;
 294        }
 295
 296        audit_ptrace(task);
 297
 298        retval = -EPERM;
 299        if (unlikely(task->flags & PF_KTHREAD))
 300                goto out;
 301        if (same_thread_group(task, current))
 302                goto out;
 303
 304        /*
 305         * Protect exec's credential calculations against our interference;
 306         * SUID, SGID and LSM creds get determined differently
 307         * under ptrace.
 308         */
 309        retval = -ERESTARTNOINTR;
 310        if (mutex_lock_interruptible(&task->signal->cred_guard_mutex))
 311                goto out;
 312
 313        task_lock(task);
 314        retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH);
 315        task_unlock(task);
 316        if (retval)
 317                goto unlock_creds;
 318
 319        write_lock_irq(&tasklist_lock);
 320        retval = -EPERM;
 321        if (unlikely(task->exit_state))
 322                goto unlock_tasklist;
 323        if (task->ptrace)
 324                goto unlock_tasklist;
 325
 326        if (seize)
 327                flags |= PT_SEIZED;
 328        rcu_read_lock();
 329        if (ns_capable(__task_cred(task)->user_ns, CAP_SYS_PTRACE))
 330                flags |= PT_PTRACE_CAP;
 331        rcu_read_unlock();
 332        task->ptrace = flags;
 333
 334        __ptrace_link(task, current);
 335
 336        /* SEIZE doesn't trap tracee on attach */
 337        if (!seize)
 338                send_sig_info(SIGSTOP, SEND_SIG_FORCED, task);
 339
 340        spin_lock(&task->sighand->siglock);
 341
 342        /*
 343         * If the task is already STOPPED, set JOBCTL_TRAP_STOP and
 344         * TRAPPING, and kick it so that it transits to TRACED.  TRAPPING
 345         * will be cleared if the child completes the transition or any
 346         * event which clears the group stop states happens.  We'll wait
 347         * for the transition to complete before returning from this
 348         * function.
 349         *
 350         * This hides STOPPED -> RUNNING -> TRACED transition from the
 351         * attaching thread but a different thread in the same group can
 352         * still observe the transient RUNNING state.  IOW, if another
 353         * thread's WNOHANG wait(2) on the stopped tracee races against
 354         * ATTACH, the wait(2) may fail due to the transient RUNNING.
 355         *
 356         * The following task_is_stopped() test is safe as both transitions
 357         * in and out of STOPPED are protected by siglock.
 358         */
 359        if (task_is_stopped(task) &&
 360            task_set_jobctl_pending(task, JOBCTL_TRAP_STOP | JOBCTL_TRAPPING))
 361                signal_wake_up_state(task, __TASK_STOPPED);
 362
 363        spin_unlock(&task->sighand->siglock);
 364
 365        retval = 0;
 366unlock_tasklist:
 367        write_unlock_irq(&tasklist_lock);
 368unlock_creds:
 369        mutex_unlock(&task->signal->cred_guard_mutex);
 370out:
 371        if (!retval) {
 372                wait_on_bit(&task->jobctl, JOBCTL_TRAPPING_BIT,
 373                            ptrace_trapping_sleep_fn, TASK_UNINTERRUPTIBLE);
 374                proc_ptrace_connector(task, PTRACE_ATTACH);
 375        }
 376
 377        return retval;
 378}
 379
 380/**
 381 * ptrace_traceme  --  helper for PTRACE_TRACEME
 382 *
 383 * Performs checks and sets PT_PTRACED.
 384 * Should be used by all ptrace implementations for PTRACE_TRACEME.
 385 */
 386static int ptrace_traceme(void)
 387{
 388        int ret = -EPERM;
 389
 390        write_lock_irq(&tasklist_lock);
 391        /* Are we already being traced? */
 392        if (!current->ptrace) {
 393                ret = security_ptrace_traceme(current->parent);
 394                /*
 395                 * Check PF_EXITING to ensure ->real_parent has not passed
 396                 * exit_ptrace(). Otherwise we don't report the error but
 397                 * pretend ->real_parent untraces us right after return.
 398                 */
 399                if (!ret && !(current->real_parent->flags & PF_EXITING)) {
 400                        current->ptrace = PT_PTRACED;
 401                        __ptrace_link(current, current->real_parent);
 402                }
 403        }
 404        write_unlock_irq(&tasklist_lock);
 405
 406        return ret;
 407}
 408
 409/*
 410 * Called with irqs disabled, returns true if childs should reap themselves.
 411 */
 412static int ignoring_children(struct sighand_struct *sigh)
 413{
 414        int ret;
 415        spin_lock(&sigh->siglock);
 416        ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
 417              (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
 418        spin_unlock(&sigh->siglock);
 419        return ret;
 420}
 421
 422/*
 423 * Called with tasklist_lock held for writing.
 424 * Unlink a traced task, and clean it up if it was a traced zombie.
 425 * Return true if it needs to be reaped with release_task().
 426 * (We can't call release_task() here because we already hold tasklist_lock.)
 427 *
 428 * If it's a zombie, our attachedness prevented normal parent notification
 429 * or self-reaping.  Do notification now if it would have happened earlier.
 430 * If it should reap itself, return true.
 431 *
 432 * If it's our own child, there is no notification to do. But if our normal
 433 * children self-reap, then this child was prevented by ptrace and we must
 434 * reap it now, in that case we must also wake up sub-threads sleeping in
 435 * do_wait().
 436 */
 437static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
 438{
 439        bool dead;
 440
 441        __ptrace_unlink(p);
 442
 443        if (p->exit_state != EXIT_ZOMBIE)
 444                return false;
 445
 446        dead = !thread_group_leader(p);
 447
 448        if (!dead && thread_group_empty(p)) {
 449                if (!same_thread_group(p->real_parent, tracer))
 450                        dead = do_notify_parent(p, p->exit_signal);
 451                else if (ignoring_children(tracer->sighand)) {
 452                        __wake_up_parent(p, tracer);
 453                        dead = true;
 454                }
 455        }
 456        /* Mark it as in the process of being reaped. */
 457        if (dead)
 458                p->exit_state = EXIT_DEAD;
 459        return dead;
 460}
 461
 462static int ptrace_detach(struct task_struct *child, unsigned int data)
 463{
 464        bool dead = false;
 465
 466        if (!valid_signal(data))
 467                return -EIO;
 468
 469        /* Architecture-specific hardware disable .. */
 470        ptrace_disable(child);
 471        clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
 472
 473        write_lock_irq(&tasklist_lock);
 474        /*
 475         * This child can be already killed. Make sure de_thread() or
 476         * our sub-thread doing do_wait() didn't do release_task() yet.
 477         */
 478        if (child->ptrace) {
 479                child->exit_code = data;
 480                dead = __ptrace_detach(current, child);
 481        }
 482        write_unlock_irq(&tasklist_lock);
 483
 484        proc_ptrace_connector(child, PTRACE_DETACH);
 485        if (unlikely(dead))
 486                release_task(child);
 487
 488        return 0;
 489}
 490
 491/*
 492 * Detach all tasks we were using ptrace on. Called with tasklist held
 493 * for writing, and returns with it held too. But note it can release
 494 * and reacquire the lock.
 495 */
 496void exit_ptrace(struct task_struct *tracer)
 497        __releases(&tasklist_lock)
 498        __acquires(&tasklist_lock)
 499{
 500        struct task_struct *p, *n;
 501        LIST_HEAD(ptrace_dead);
 502
 503        if (likely(list_empty(&tracer->ptraced)))
 504                return;
 505
 506        list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
 507                if (unlikely(p->ptrace & PT_EXITKILL))
 508                        send_sig_info(SIGKILL, SEND_SIG_FORCED, p);
 509
 510                if (__ptrace_detach(tracer, p))
 511                        list_add(&p->ptrace_entry, &ptrace_dead);
 512        }
 513
 514        write_unlock_irq(&tasklist_lock);
 515        BUG_ON(!list_empty(&tracer->ptraced));
 516
 517        list_for_each_entry_safe(p, n, &ptrace_dead, ptrace_entry) {
 518                list_del_init(&p->ptrace_entry);
 519                release_task(p);
 520        }
 521
 522        write_lock_irq(&tasklist_lock);
 523}
 524
 525int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
 526{
 527        int copied = 0;
 528
 529        while (len > 0) {
 530                char buf[128];
 531                int this_len, retval;
 532
 533                this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 534                retval = access_process_vm(tsk, src, buf, this_len, 0);
 535                if (!retval) {
 536                        if (copied)
 537                                break;
 538                        return -EIO;
 539                }
 540                if (copy_to_user(dst, buf, retval))
 541                        return -EFAULT;
 542                copied += retval;
 543                src += retval;
 544                dst += retval;
 545                len -= retval;
 546        }
 547        return copied;
 548}
 549
 550int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
 551{
 552        int copied = 0;
 553
 554        while (len > 0) {
 555                char buf[128];
 556                int this_len, retval;
 557
 558                this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 559                if (copy_from_user(buf, src, this_len))
 560                        return -EFAULT;
 561                retval = access_process_vm(tsk, dst, buf, this_len, 1);
 562                if (!retval) {
 563                        if (copied)
 564                                break;
 565                        return -EIO;
 566                }
 567                copied += retval;
 568                src += retval;
 569                dst += retval;
 570                len -= retval;
 571        }
 572        return copied;
 573}
 574
 575static int ptrace_setoptions(struct task_struct *child, unsigned long data)
 576{
 577        unsigned flags;
 578
 579        if (data & ~(unsigned long)PTRACE_O_MASK)
 580                return -EINVAL;
 581
 582        /* Avoid intermediate state when all opts are cleared */
 583        flags = child->ptrace;
 584        flags &= ~(PTRACE_O_MASK << PT_OPT_FLAG_SHIFT);
 585        flags |= (data << PT_OPT_FLAG_SHIFT);
 586        child->ptrace = flags;
 587
 588        return 0;
 589}
 590
 591static int ptrace_getsiginfo(struct task_struct *child, siginfo_t *info)
 592{
 593        unsigned long flags;
 594        int error = -ESRCH;
 595
 596        if (lock_task_sighand(child, &flags)) {
 597                error = -EINVAL;
 598                if (likely(child->last_siginfo != NULL)) {
 599                        *info = *child->last_siginfo;
 600                        error = 0;
 601                }
 602                unlock_task_sighand(child, &flags);
 603        }
 604        return error;
 605}
 606
 607static int ptrace_setsiginfo(struct task_struct *child, const siginfo_t *info)
 608{
 609        unsigned long flags;
 610        int error = -ESRCH;
 611
 612        if (lock_task_sighand(child, &flags)) {
 613                error = -EINVAL;
 614                if (likely(child->last_siginfo != NULL)) {
 615                        *child->last_siginfo = *info;
 616                        error = 0;
 617                }
 618                unlock_task_sighand(child, &flags);
 619        }
 620        return error;
 621}
 622
 623static int ptrace_peek_siginfo(struct task_struct *child,
 624                                unsigned long addr,
 625                                unsigned long data)
 626{
 627        struct ptrace_peeksiginfo_args arg;
 628        struct sigpending *pending;
 629        struct sigqueue *q;
 630        int ret, i;
 631
 632        ret = copy_from_user(&arg, (void __user *) addr,
 633                                sizeof(struct ptrace_peeksiginfo_args));
 634        if (ret)
 635                return -EFAULT;
 636
 637        if (arg.flags & ~PTRACE_PEEKSIGINFO_SHARED)
 638                return -EINVAL; /* unknown flags */
 639
 640        if (arg.nr < 0)
 641                return -EINVAL;
 642
 643        if (arg.flags & PTRACE_PEEKSIGINFO_SHARED)
 644                pending = &child->signal->shared_pending;
 645        else
 646                pending = &child->pending;
 647
 648        for (i = 0; i < arg.nr; ) {
 649                siginfo_t info;
 650                s32 off = arg.off + i;
 651
 652                spin_lock_irq(&child->sighand->siglock);
 653                list_for_each_entry(q, &pending->list, list) {
 654                        if (!off--) {
 655                                copy_siginfo(&info, &q->info);
 656                                break;
 657                        }
 658                }
 659                spin_unlock_irq(&child->sighand->siglock);
 660
 661                if (off >= 0) /* beyond the end of the list */
 662                        break;
 663
 664#ifdef CONFIG_COMPAT
 665                if (unlikely(is_compat_task())) {
 666                        compat_siginfo_t __user *uinfo = compat_ptr(data);
 667
 668                        if (copy_siginfo_to_user32(uinfo, &info) ||
 669                            __put_user(info.si_code, &uinfo->si_code)) {
 670                                ret = -EFAULT;
 671                                break;
 672                        }
 673
 674                } else
 675#endif
 676                {
 677                        siginfo_t __user *uinfo = (siginfo_t __user *) data;
 678
 679                        if (copy_siginfo_to_user(uinfo, &info) ||
 680                            __put_user(info.si_code, &uinfo->si_code)) {
 681                                ret = -EFAULT;
 682                                break;
 683                        }
 684                }
 685
 686                data += sizeof(siginfo_t);
 687                i++;
 688
 689                if (signal_pending(current))
 690                        break;
 691
 692                cond_resched();
 693        }
 694
 695        if (i > 0)
 696                return i;
 697
 698        return ret;
 699}
 700
 701#ifdef PTRACE_SINGLESTEP
 702#define is_singlestep(request)          ((request) == PTRACE_SINGLESTEP)
 703#else
 704#define is_singlestep(request)          0
 705#endif
 706
 707#ifdef PTRACE_SINGLEBLOCK
 708#define is_singleblock(request)         ((request) == PTRACE_SINGLEBLOCK)
 709#else
 710#define is_singleblock(request)         0
 711#endif
 712
 713#ifdef PTRACE_SYSEMU
 714#define is_sysemu_singlestep(request)   ((request) == PTRACE_SYSEMU_SINGLESTEP)
 715#else
 716#define is_sysemu_singlestep(request)   0
 717#endif
 718
 719static int ptrace_resume(struct task_struct *child, long request,
 720                         unsigned long data)
 721{
 722        if (!valid_signal(data))
 723                return -EIO;
 724
 725        if (request == PTRACE_SYSCALL)
 726                set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
 727        else
 728                clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
 729
 730#ifdef TIF_SYSCALL_EMU
 731        if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
 732                set_tsk_thread_flag(child, TIF_SYSCALL_EMU);
 733        else
 734                clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
 735#endif
 736
 737        if (is_singleblock(request)) {
 738                if (unlikely(!arch_has_block_step()))
 739                        return -EIO;
 740                user_enable_block_step(child);
 741        } else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
 742                if (unlikely(!arch_has_single_step()))
 743                        return -EIO;
 744                user_enable_single_step(child);
 745        } else {
 746                user_disable_single_step(child);
 747        }
 748
 749        child->exit_code = data;
 750        wake_up_state(child, __TASK_TRACED);
 751
 752        return 0;
 753}
 754
 755#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
 756
 757static const struct user_regset *
 758find_regset(const struct user_regset_view *view, unsigned int type)
 759{
 760        const struct user_regset *regset;
 761        int n;
 762
 763        for (n = 0; n < view->n; ++n) {
 764                regset = view->regsets + n;
 765                if (regset->core_note_type == type)
 766                        return regset;
 767        }
 768
 769        return NULL;
 770}
 771
 772static int ptrace_regset(struct task_struct *task, int req, unsigned int type,
 773                         struct iovec *kiov)
 774{
 775        const struct user_regset_view *view = task_user_regset_view(task);
 776        const struct user_regset *regset = find_regset(view, type);
 777        int regset_no;
 778
 779        if (!regset || (kiov->iov_len % regset->size) != 0)
 780                return -EINVAL;
 781
 782        regset_no = regset - view->regsets;
 783        kiov->iov_len = min(kiov->iov_len,
 784                            (__kernel_size_t) (regset->n * regset->size));
 785
 786        if (req == PTRACE_GETREGSET)
 787                return copy_regset_to_user(task, view, regset_no, 0,
 788                                           kiov->iov_len, kiov->iov_base);
 789        else
 790                return copy_regset_from_user(task, view, regset_no, 0,
 791                                             kiov->iov_len, kiov->iov_base);
 792}
 793
 794/*
 795 * This is declared in linux/regset.h and defined in machine-dependent
 796 * code.  We put the export here, near the primary machine-neutral use,
 797 * to ensure no machine forgets it.
 798 */
 799EXPORT_SYMBOL_GPL(task_user_regset_view);
 800#endif
 801
 802int ptrace_request(struct task_struct *child, long request,
 803                   unsigned long addr, unsigned long data)
 804{
 805        bool seized = child->ptrace & PT_SEIZED;
 806        int ret = -EIO;
 807        siginfo_t siginfo, *si;
 808        void __user *datavp = (void __user *) data;
 809        unsigned long __user *datalp = datavp;
 810        unsigned long flags;
 811
 812        switch (request) {
 813        case PTRACE_PEEKTEXT:
 814        case PTRACE_PEEKDATA:
 815                return generic_ptrace_peekdata(child, addr, data);
 816        case PTRACE_POKETEXT:
 817        case PTRACE_POKEDATA:
 818                return generic_ptrace_pokedata(child, addr, data);
 819
 820#ifdef PTRACE_OLDSETOPTIONS
 821        case PTRACE_OLDSETOPTIONS:
 822#endif
 823        case PTRACE_SETOPTIONS:
 824                ret = ptrace_setoptions(child, data);
 825                break;
 826        case PTRACE_GETEVENTMSG:
 827                ret = put_user(child->ptrace_message, datalp);
 828                break;
 829
 830        case PTRACE_PEEKSIGINFO:
 831                ret = ptrace_peek_siginfo(child, addr, data);
 832                break;
 833
 834        case PTRACE_GETSIGINFO:
 835                ret = ptrace_getsiginfo(child, &siginfo);
 836                if (!ret)
 837                        ret = copy_siginfo_to_user(datavp, &siginfo);
 838                break;
 839
 840        case PTRACE_SETSIGINFO:
 841                if (copy_from_user(&siginfo, datavp, sizeof siginfo))
 842                        ret = -EFAULT;
 843                else
 844                        ret = ptrace_setsiginfo(child, &siginfo);
 845                break;
 846
 847        case PTRACE_INTERRUPT:
 848                /*
 849                 * Stop tracee without any side-effect on signal or job
 850                 * control.  At least one trap is guaranteed to happen
 851                 * after this request.  If @child is already trapped, the
 852                 * current trap is not disturbed and another trap will
 853                 * happen after the current trap is ended with PTRACE_CONT.
 854                 *
 855                 * The actual trap might not be PTRACE_EVENT_STOP trap but
 856                 * the pending condition is cleared regardless.
 857                 */
 858                if (unlikely(!seized || !lock_task_sighand(child, &flags)))
 859                        break;
 860
 861                /*
 862                 * INTERRUPT doesn't disturb existing trap sans one
 863                 * exception.  If ptracer issued LISTEN for the current
 864                 * STOP, this INTERRUPT should clear LISTEN and re-trap
 865                 * tracee into STOP.
 866                 */
 867                if (likely(task_set_jobctl_pending(child, JOBCTL_TRAP_STOP)))
 868                        ptrace_signal_wake_up(child, child->jobctl & JOBCTL_LISTENING);
 869
 870                unlock_task_sighand(child, &flags);
 871                ret = 0;
 872                break;
 873
 874        case PTRACE_LISTEN:
 875                /*
 876                 * Listen for events.  Tracee must be in STOP.  It's not
 877                 * resumed per-se but is not considered to be in TRACED by
 878                 * wait(2) or ptrace(2).  If an async event (e.g. group
 879                 * stop state change) happens, tracee will enter STOP trap
 880                 * again.  Alternatively, ptracer can issue INTERRUPT to
 881                 * finish listening and re-trap tracee into STOP.
 882                 */
 883                if (unlikely(!seized || !lock_task_sighand(child, &flags)))
 884                        break;
 885
 886                si = child->last_siginfo;
 887                if (likely(si && (si->si_code >> 8) == PTRACE_EVENT_STOP)) {
 888                        child->jobctl |= JOBCTL_LISTENING;
 889                        /*
 890                         * If NOTIFY is set, it means event happened between
 891                         * start of this trap and now.  Trigger re-trap.
 892                         */
 893                        if (child->jobctl & JOBCTL_TRAP_NOTIFY)
 894                                ptrace_signal_wake_up(child, true);
 895                        ret = 0;
 896                }
 897                unlock_task_sighand(child, &flags);
 898                break;
 899
 900        case PTRACE_DETACH:      /* detach a process that was attached. */
 901                ret = ptrace_detach(child, data);
 902                break;
 903
 904#ifdef CONFIG_BINFMT_ELF_FDPIC
 905        case PTRACE_GETFDPIC: {
 906                struct mm_struct *mm = get_task_mm(child);
 907                unsigned long tmp = 0;
 908
 909                ret = -ESRCH;
 910                if (!mm)
 911                        break;
 912
 913                switch (addr) {
 914                case PTRACE_GETFDPIC_EXEC:
 915                        tmp = mm->context.exec_fdpic_loadmap;
 916                        break;
 917                case PTRACE_GETFDPIC_INTERP:
 918                        tmp = mm->context.interp_fdpic_loadmap;
 919                        break;
 920                default:
 921                        break;
 922                }
 923                mmput(mm);
 924
 925                ret = put_user(tmp, datalp);
 926                break;
 927        }
 928#endif
 929
 930#ifdef PTRACE_SINGLESTEP
 931        case PTRACE_SINGLESTEP:
 932#endif
 933#ifdef PTRACE_SINGLEBLOCK
 934        case PTRACE_SINGLEBLOCK:
 935#endif
 936#ifdef PTRACE_SYSEMU
 937        case PTRACE_SYSEMU:
 938        case PTRACE_SYSEMU_SINGLESTEP:
 939#endif
 940        case PTRACE_SYSCALL:
 941        case PTRACE_CONT:
 942                return ptrace_resume(child, request, data);
 943
 944        case PTRACE_KILL:
 945                if (child->exit_state)  /* already dead */
 946                        return 0;
 947                return ptrace_resume(child, request, SIGKILL);
 948
 949#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
 950        case PTRACE_GETREGSET:
 951        case PTRACE_SETREGSET:
 952        {
 953                struct iovec kiov;
 954                struct iovec __user *uiov = datavp;
 955
 956                if (!access_ok(VERIFY_WRITE, uiov, sizeof(*uiov)))
 957                        return -EFAULT;
 958
 959                if (__get_user(kiov.iov_base, &uiov->iov_base) ||
 960                    __get_user(kiov.iov_len, &uiov->iov_len))
 961                        return -EFAULT;
 962
 963                ret = ptrace_regset(child, request, addr, &kiov);
 964                if (!ret)
 965                        ret = __put_user(kiov.iov_len, &uiov->iov_len);
 966                break;
 967        }
 968#endif
 969        default:
 970                break;
 971        }
 972
 973        return ret;
 974}
 975
 976static struct task_struct *ptrace_get_task_struct(pid_t pid)
 977{
 978        struct task_struct *child;
 979
 980        rcu_read_lock();
 981        child = find_task_by_vpid(pid);
 982        if (child)
 983                get_task_struct(child);
 984        rcu_read_unlock();
 985
 986        if (!child)
 987                return ERR_PTR(-ESRCH);
 988        return child;
 989}
 990
 991#ifndef arch_ptrace_attach
 992#define arch_ptrace_attach(child)       do { } while (0)
 993#endif
 994
 995SYSCALL_DEFINE4(ptrace, long, request, long, pid, unsigned long, addr,
 996                unsigned long, data)
 997{
 998        struct task_struct *child;
 999        long ret;
1000
1001        if (request == PTRACE_TRACEME) {
1002                ret = ptrace_traceme();
1003                if (!ret)
1004                        arch_ptrace_attach(current);
1005                goto out;
1006        }
1007
1008        child = ptrace_get_task_struct(pid);
1009        if (IS_ERR(child)) {
1010                ret = PTR_ERR(child);
1011                goto out;
1012        }
1013
1014        if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1015                ret = ptrace_attach(child, request, addr, data);
1016                /*
1017                 * Some architectures need to do book-keeping after
1018                 * a ptrace attach.
1019                 */
1020                if (!ret)
1021                        arch_ptrace_attach(child);
1022                goto out_put_task_struct;
1023        }
1024
1025        ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1026                                  request == PTRACE_INTERRUPT);
1027        if (ret < 0)
1028                goto out_put_task_struct;
1029
1030        ret = arch_ptrace(child, request, addr, data);
1031        if (ret || request != PTRACE_DETACH)
1032                ptrace_unfreeze_traced(child);
1033
1034 out_put_task_struct:
1035        put_task_struct(child);
1036 out:
1037        return ret;
1038}
1039
1040int generic_ptrace_peekdata(struct task_struct *tsk, unsigned long addr,
1041                            unsigned long data)
1042{
1043        unsigned long tmp;
1044        int copied;
1045
1046        copied = access_process_vm(tsk, addr, &tmp, sizeof(tmp), 0);
1047        if (copied != sizeof(tmp))
1048                return -EIO;
1049        return put_user(tmp, (unsigned long __user *)data);
1050}
1051
1052int generic_ptrace_pokedata(struct task_struct *tsk, unsigned long addr,
1053                            unsigned long data)
1054{
1055        int copied;
1056
1057        copied = access_process_vm(tsk, addr, &data, sizeof(data), 1);
1058        return (copied == sizeof(data)) ? 0 : -EIO;
1059}
1060
1061#if defined CONFIG_COMPAT
1062#include <linux/compat.h>
1063
1064int compat_ptrace_request(struct task_struct *child, compat_long_t request,
1065                          compat_ulong_t addr, compat_ulong_t data)
1066{
1067        compat_ulong_t __user *datap = compat_ptr(data);
1068        compat_ulong_t word;
1069        siginfo_t siginfo;
1070        int ret;
1071
1072        switch (request) {
1073        case PTRACE_PEEKTEXT:
1074        case PTRACE_PEEKDATA:
1075                ret = access_process_vm(child, addr, &word, sizeof(word), 0);
1076                if (ret != sizeof(word))
1077                        ret = -EIO;
1078                else
1079                        ret = put_user(word, datap);
1080                break;
1081
1082        case PTRACE_POKETEXT:
1083        case PTRACE_POKEDATA:
1084                ret = access_process_vm(child, addr, &data, sizeof(data), 1);
1085                ret = (ret != sizeof(data) ? -EIO : 0);
1086                break;
1087
1088        case PTRACE_GETEVENTMSG:
1089                ret = put_user((compat_ulong_t) child->ptrace_message, datap);
1090                break;
1091
1092        case PTRACE_GETSIGINFO:
1093                ret = ptrace_getsiginfo(child, &siginfo);
1094                if (!ret)
1095                        ret = copy_siginfo_to_user32(
1096                                (struct compat_siginfo __user *) datap,
1097                                &siginfo);
1098                break;
1099
1100        case PTRACE_SETSIGINFO:
1101                memset(&siginfo, 0, sizeof siginfo);
1102                if (copy_siginfo_from_user32(
1103                            &siginfo, (struct compat_siginfo __user *) datap))
1104                        ret = -EFAULT;
1105                else
1106                        ret = ptrace_setsiginfo(child, &siginfo);
1107                break;
1108#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
1109        case PTRACE_GETREGSET:
1110        case PTRACE_SETREGSET:
1111        {
1112                struct iovec kiov;
1113                struct compat_iovec __user *uiov =
1114                        (struct compat_iovec __user *) datap;
1115                compat_uptr_t ptr;
1116                compat_size_t len;
1117
1118                if (!access_ok(VERIFY_WRITE, uiov, sizeof(*uiov)))
1119                        return -EFAULT;
1120
1121                if (__get_user(ptr, &uiov->iov_base) ||
1122                    __get_user(len, &uiov->iov_len))
1123                        return -EFAULT;
1124
1125                kiov.iov_base = compat_ptr(ptr);
1126                kiov.iov_len = len;
1127
1128                ret = ptrace_regset(child, request, addr, &kiov);
1129                if (!ret)
1130                        ret = __put_user(kiov.iov_len, &uiov->iov_len);
1131                break;
1132        }
1133#endif
1134
1135        default:
1136                ret = ptrace_request(child, request, addr, data);
1137        }
1138
1139        return ret;
1140}
1141
1142asmlinkage long compat_sys_ptrace(compat_long_t request, compat_long_t pid,
1143                                  compat_long_t addr, compat_long_t data)
1144{
1145        struct task_struct *child;
1146        long ret;
1147
1148        if (request == PTRACE_TRACEME) {
1149                ret = ptrace_traceme();
1150                goto out;
1151        }
1152
1153        child = ptrace_get_task_struct(pid);
1154        if (IS_ERR(child)) {
1155                ret = PTR_ERR(child);
1156                goto out;
1157        }
1158
1159        if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1160                ret = ptrace_attach(child, request, addr, data);
1161                /*
1162                 * Some architectures need to do book-keeping after
1163                 * a ptrace attach.
1164                 */
1165                if (!ret)
1166                        arch_ptrace_attach(child);
1167                goto out_put_task_struct;
1168        }
1169
1170        ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1171                                  request == PTRACE_INTERRUPT);
1172        if (!ret) {
1173                ret = compat_arch_ptrace(child, request, addr, data);
1174                if (ret || request != PTRACE_DETACH)
1175                        ptrace_unfreeze_traced(child);
1176        }
1177
1178 out_put_task_struct:
1179        put_task_struct(child);
1180 out:
1181        return ret;
1182}
1183#endif  /* CONFIG_COMPAT */
1184
1185#ifdef CONFIG_HAVE_HW_BREAKPOINT
1186int ptrace_get_breakpoints(struct task_struct *tsk)
1187{
1188        if (atomic_inc_not_zero(&tsk->ptrace_bp_refcnt))
1189                return 0;
1190
1191        return -1;
1192}
1193
1194void ptrace_put_breakpoints(struct task_struct *tsk)
1195{
1196        if (atomic_dec_and_test(&tsk->ptrace_bp_refcnt))
1197                flush_ptrace_hw_breakpoint(tsk);
1198}
1199#endif /* CONFIG_HAVE_HW_BREAKPOINT */
1200