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