kernel/ptrace.c

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