Linux Audio

Check our new training course

Loading...
v6.2
   1// SPDX-License-Identifier: GPL-2.0-only
   2/*
   3 * linux/kernel/ptrace.c
   4 *
   5 * (C) Copyright 1999 Linus Torvalds
   6 *
   7 * Common interfaces for "ptrace()" which we do not want
   8 * to continually duplicate across every architecture.
   9 */
  10
  11#include <linux/capability.h>
  12#include <linux/export.h>
  13#include <linux/sched.h>
  14#include <linux/sched/mm.h>
  15#include <linux/sched/coredump.h>
  16#include <linux/sched/task.h>
  17#include <linux/errno.h>
  18#include <linux/mm.h>
  19#include <linux/highmem.h>
  20#include <linux/pagemap.h>
  21#include <linux/ptrace.h>
  22#include <linux/security.h>
  23#include <linux/signal.h>
  24#include <linux/uio.h>
  25#include <linux/audit.h>
  26#include <linux/pid_namespace.h>
  27#include <linux/syscalls.h>
  28#include <linux/uaccess.h>
  29#include <linux/regset.h>
  30#include <linux/hw_breakpoint.h>
  31#include <linux/cn_proc.h>
  32#include <linux/compat.h>
  33#include <linux/sched/signal.h>
  34#include <linux/minmax.h>
 
  35
  36#include <asm/syscall.h>	/* for syscall_get_* */
  37
  38/*
  39 * Access another process' address space via ptrace.
  40 * Source/target buffer must be kernel space,
  41 * Do not walk the page table directly, use get_user_pages
  42 */
  43int ptrace_access_vm(struct task_struct *tsk, unsigned long addr,
  44		     void *buf, int len, unsigned int gup_flags)
  45{
  46	struct mm_struct *mm;
  47	int ret;
  48
  49	mm = get_task_mm(tsk);
  50	if (!mm)
  51		return 0;
  52
  53	if (!tsk->ptrace ||
  54	    (current != tsk->parent) ||
  55	    ((get_dumpable(mm) != SUID_DUMP_USER) &&
  56	     !ptracer_capable(tsk, mm->user_ns))) {
  57		mmput(mm);
  58		return 0;
  59	}
  60
  61	ret = __access_remote_vm(mm, addr, buf, len, gup_flags);
  62	mmput(mm);
  63
  64	return ret;
  65}
  66
  67
  68void __ptrace_link(struct task_struct *child, struct task_struct *new_parent,
  69		   const struct cred *ptracer_cred)
  70{
  71	BUG_ON(!list_empty(&child->ptrace_entry));
  72	list_add(&child->ptrace_entry, &new_parent->ptraced);
  73	child->parent = new_parent;
  74	child->ptracer_cred = get_cred(ptracer_cred);
  75}
  76
  77/*
  78 * ptrace a task: make the debugger its new parent and
  79 * move it to the ptrace list.
  80 *
  81 * Must be called with the tasklist lock write-held.
  82 */
  83static void ptrace_link(struct task_struct *child, struct task_struct *new_parent)
  84{
  85	__ptrace_link(child, new_parent, current_cred());
  86}
  87
  88/**
  89 * __ptrace_unlink - unlink ptracee and restore its execution state
  90 * @child: ptracee to be unlinked
  91 *
  92 * Remove @child from the ptrace list, move it back to the original parent,
  93 * and restore the execution state so that it conforms to the group stop
  94 * state.
  95 *
  96 * Unlinking can happen via two paths - explicit PTRACE_DETACH or ptracer
  97 * exiting.  For PTRACE_DETACH, unless the ptracee has been killed between
  98 * ptrace_check_attach() and here, it's guaranteed to be in TASK_TRACED.
  99 * If the ptracer is exiting, the ptracee can be in any state.
 100 *
 101 * After detach, the ptracee should be in a state which conforms to the
 102 * group stop.  If the group is stopped or in the process of stopping, the
 103 * ptracee should be put into TASK_STOPPED; otherwise, it should be woken
 104 * up from TASK_TRACED.
 105 *
 106 * If the ptracee is in TASK_TRACED and needs to be moved to TASK_STOPPED,
 107 * it goes through TRACED -> RUNNING -> STOPPED transition which is similar
 108 * to but in the opposite direction of what happens while attaching to a
 109 * stopped task.  However, in this direction, the intermediate RUNNING
 110 * state is not hidden even from the current ptracer and if it immediately
 111 * re-attaches and performs a WNOHANG wait(2), it may fail.
 112 *
 113 * CONTEXT:
 114 * write_lock_irq(tasklist_lock)
 115 */
 116void __ptrace_unlink(struct task_struct *child)
 117{
 118	const struct cred *old_cred;
 119	BUG_ON(!child->ptrace);
 120
 121	clear_task_syscall_work(child, SYSCALL_TRACE);
 122#if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU)
 123	clear_task_syscall_work(child, SYSCALL_EMU);
 124#endif
 125
 126	child->parent = child->real_parent;
 127	list_del_init(&child->ptrace_entry);
 128	old_cred = child->ptracer_cred;
 129	child->ptracer_cred = NULL;
 130	put_cred(old_cred);
 131
 132	spin_lock(&child->sighand->siglock);
 133	child->ptrace = 0;
 134	/*
 135	 * Clear all pending traps and TRAPPING.  TRAPPING should be
 136	 * cleared regardless of JOBCTL_STOP_PENDING.  Do it explicitly.
 137	 */
 138	task_clear_jobctl_pending(child, JOBCTL_TRAP_MASK);
 139	task_clear_jobctl_trapping(child);
 140
 141	/*
 142	 * Reinstate JOBCTL_STOP_PENDING if group stop is in effect and
 143	 * @child isn't dead.
 144	 */
 145	if (!(child->flags & PF_EXITING) &&
 146	    (child->signal->flags & SIGNAL_STOP_STOPPED ||
 147	     child->signal->group_stop_count)) {
 148		child->jobctl |= JOBCTL_STOP_PENDING;
 149
 150		/*
 151		 * This is only possible if this thread was cloned by the
 152		 * traced task running in the stopped group, set the signal
 153		 * for the future reports.
 154		 * FIXME: we should change ptrace_init_task() to handle this
 155		 * case.
 156		 */
 157		if (!(child->jobctl & JOBCTL_STOP_SIGMASK))
 158			child->jobctl |= SIGSTOP;
 159	}
 160
 161	/*
 162	 * If transition to TASK_STOPPED is pending or in TASK_TRACED, kick
 163	 * @child in the butt.  Note that @resume should be used iff @child
 164	 * is in TASK_TRACED; otherwise, we might unduly disrupt
 165	 * TASK_KILLABLE sleeps.
 166	 */
 167	if (child->jobctl & JOBCTL_STOP_PENDING || task_is_traced(child))
 168		ptrace_signal_wake_up(child, true);
 169
 170	spin_unlock(&child->sighand->siglock);
 171}
 172
 173static bool looks_like_a_spurious_pid(struct task_struct *task)
 174{
 175	if (task->exit_code != ((PTRACE_EVENT_EXEC << 8) | SIGTRAP))
 176		return false;
 177
 178	if (task_pid_vnr(task) == task->ptrace_message)
 179		return false;
 180	/*
 181	 * The tracee changed its pid but the PTRACE_EVENT_EXEC event
 182	 * was not wait()'ed, most probably debugger targets the old
 183	 * leader which was destroyed in de_thread().
 184	 */
 185	return true;
 186}
 187
 188/*
 189 * Ensure that nothing can wake it up, even SIGKILL
 190 *
 191 * A task is switched to this state while a ptrace operation is in progress;
 192 * such that the ptrace operation is uninterruptible.
 193 */
 194static bool ptrace_freeze_traced(struct task_struct *task)
 195{
 196	bool ret = false;
 197
 198	/* Lockless, nobody but us can set this flag */
 199	if (task->jobctl & JOBCTL_LISTENING)
 200		return ret;
 201
 202	spin_lock_irq(&task->sighand->siglock);
 203	if (task_is_traced(task) && !looks_like_a_spurious_pid(task) &&
 204	    !__fatal_signal_pending(task)) {
 205		task->jobctl |= JOBCTL_PTRACE_FROZEN;
 206		ret = true;
 207	}
 208	spin_unlock_irq(&task->sighand->siglock);
 209
 210	return ret;
 211}
 212
 213static void ptrace_unfreeze_traced(struct task_struct *task)
 214{
 215	unsigned long flags;
 216
 217	/*
 218	 * The child may be awake and may have cleared
 219	 * JOBCTL_PTRACE_FROZEN (see ptrace_resume).  The child will
 220	 * not set JOBCTL_PTRACE_FROZEN or enter __TASK_TRACED anew.
 221	 */
 222	if (lock_task_sighand(task, &flags)) {
 223		task->jobctl &= ~JOBCTL_PTRACE_FROZEN;
 224		if (__fatal_signal_pending(task)) {
 225			task->jobctl &= ~JOBCTL_TRACED;
 226			wake_up_state(task, __TASK_TRACED);
 227		}
 228		unlock_task_sighand(task, &flags);
 229	}
 230}
 231
 232/**
 233 * ptrace_check_attach - check whether ptracee is ready for ptrace operation
 234 * @child: ptracee to check for
 235 * @ignore_state: don't check whether @child is currently %TASK_TRACED
 236 *
 237 * Check whether @child is being ptraced by %current and ready for further
 238 * ptrace operations.  If @ignore_state is %false, @child also should be in
 239 * %TASK_TRACED state and on return the child is guaranteed to be traced
 240 * and not executing.  If @ignore_state is %true, @child can be in any
 241 * state.
 242 *
 243 * CONTEXT:
 244 * Grabs and releases tasklist_lock and @child->sighand->siglock.
 245 *
 246 * RETURNS:
 247 * 0 on success, -ESRCH if %child is not ready.
 248 */
 249static int ptrace_check_attach(struct task_struct *child, bool ignore_state)
 250{
 251	int ret = -ESRCH;
 252
 253	/*
 254	 * We take the read lock around doing both checks to close a
 255	 * possible race where someone else was tracing our child and
 256	 * detached between these two checks.  After this locked check,
 257	 * we are sure that this is our traced child and that can only
 258	 * be changed by us so it's not changing right after this.
 259	 */
 260	read_lock(&tasklist_lock);
 261	if (child->ptrace && child->parent == current) {
 262		/*
 263		 * child->sighand can't be NULL, release_task()
 264		 * does ptrace_unlink() before __exit_signal().
 265		 */
 266		if (ignore_state || ptrace_freeze_traced(child))
 267			ret = 0;
 268	}
 269	read_unlock(&tasklist_lock);
 270
 271	if (!ret && !ignore_state &&
 272	    WARN_ON_ONCE(!wait_task_inactive(child, __TASK_TRACED|TASK_FROZEN)))
 273		ret = -ESRCH;
 274
 275	return ret;
 276}
 277
 278static bool ptrace_has_cap(struct user_namespace *ns, unsigned int mode)
 279{
 280	if (mode & PTRACE_MODE_NOAUDIT)
 281		return ns_capable_noaudit(ns, CAP_SYS_PTRACE);
 282	return ns_capable(ns, CAP_SYS_PTRACE);
 283}
 284
 285/* Returns 0 on success, -errno on denial. */
 286static int __ptrace_may_access(struct task_struct *task, unsigned int mode)
 287{
 288	const struct cred *cred = current_cred(), *tcred;
 289	struct mm_struct *mm;
 290	kuid_t caller_uid;
 291	kgid_t caller_gid;
 292
 293	if (!(mode & PTRACE_MODE_FSCREDS) == !(mode & PTRACE_MODE_REALCREDS)) {
 294		WARN(1, "denying ptrace access check without PTRACE_MODE_*CREDS\n");
 295		return -EPERM;
 296	}
 297
 298	/* May we inspect the given task?
 299	 * This check is used both for attaching with ptrace
 300	 * and for allowing access to sensitive information in /proc.
 301	 *
 302	 * ptrace_attach denies several cases that /proc allows
 303	 * because setting up the necessary parent/child relationship
 304	 * or halting the specified task is impossible.
 305	 */
 306
 307	/* Don't let security modules deny introspection */
 308	if (same_thread_group(task, current))
 309		return 0;
 310	rcu_read_lock();
 311	if (mode & PTRACE_MODE_FSCREDS) {
 312		caller_uid = cred->fsuid;
 313		caller_gid = cred->fsgid;
 314	} else {
 315		/*
 316		 * Using the euid would make more sense here, but something
 317		 * in userland might rely on the old behavior, and this
 318		 * shouldn't be a security problem since
 319		 * PTRACE_MODE_REALCREDS implies that the caller explicitly
 320		 * used a syscall that requests access to another process
 321		 * (and not a filesystem syscall to procfs).
 322		 */
 323		caller_uid = cred->uid;
 324		caller_gid = cred->gid;
 325	}
 326	tcred = __task_cred(task);
 327	if (uid_eq(caller_uid, tcred->euid) &&
 328	    uid_eq(caller_uid, tcred->suid) &&
 329	    uid_eq(caller_uid, tcred->uid)  &&
 330	    gid_eq(caller_gid, tcred->egid) &&
 331	    gid_eq(caller_gid, tcred->sgid) &&
 332	    gid_eq(caller_gid, tcred->gid))
 333		goto ok;
 334	if (ptrace_has_cap(tcred->user_ns, mode))
 335		goto ok;
 336	rcu_read_unlock();
 337	return -EPERM;
 338ok:
 339	rcu_read_unlock();
 340	/*
 341	 * If a task drops privileges and becomes nondumpable (through a syscall
 342	 * like setresuid()) while we are trying to access it, we must ensure
 343	 * that the dumpability is read after the credentials; otherwise,
 344	 * we may be able to attach to a task that we shouldn't be able to
 345	 * attach to (as if the task had dropped privileges without becoming
 346	 * nondumpable).
 347	 * Pairs with a write barrier in commit_creds().
 348	 */
 349	smp_rmb();
 350	mm = task->mm;
 351	if (mm &&
 352	    ((get_dumpable(mm) != SUID_DUMP_USER) &&
 353	     !ptrace_has_cap(mm->user_ns, mode)))
 354	    return -EPERM;
 355
 356	return security_ptrace_access_check(task, mode);
 357}
 358
 359bool ptrace_may_access(struct task_struct *task, unsigned int mode)
 360{
 361	int err;
 362	task_lock(task);
 363	err = __ptrace_may_access(task, mode);
 364	task_unlock(task);
 365	return !err;
 366}
 367
 368static int check_ptrace_options(unsigned long data)
 369{
 370	if (data & ~(unsigned long)PTRACE_O_MASK)
 371		return -EINVAL;
 372
 373	if (unlikely(data & PTRACE_O_SUSPEND_SECCOMP)) {
 374		if (!IS_ENABLED(CONFIG_CHECKPOINT_RESTORE) ||
 375		    !IS_ENABLED(CONFIG_SECCOMP))
 376			return -EINVAL;
 377
 378		if (!capable(CAP_SYS_ADMIN))
 379			return -EPERM;
 380
 381		if (seccomp_mode(&current->seccomp) != SECCOMP_MODE_DISABLED ||
 382		    current->ptrace & PT_SUSPEND_SECCOMP)
 383			return -EPERM;
 384	}
 385	return 0;
 386}
 387
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 388static int ptrace_attach(struct task_struct *task, long request,
 389			 unsigned long addr,
 390			 unsigned long flags)
 391{
 392	bool seize = (request == PTRACE_SEIZE);
 393	int retval;
 394
 395	retval = -EIO;
 396	if (seize) {
 397		if (addr != 0)
 398			goto out;
 399		/*
 400		 * This duplicates the check in check_ptrace_options() because
 401		 * ptrace_attach() and ptrace_setoptions() have historically
 402		 * used different error codes for unknown ptrace options.
 403		 */
 404		if (flags & ~(unsigned long)PTRACE_O_MASK)
 405			goto out;
 
 406		retval = check_ptrace_options(flags);
 407		if (retval)
 408			return retval;
 409		flags = PT_PTRACED | PT_SEIZED | (flags << PT_OPT_FLAG_SHIFT);
 410	} else {
 411		flags = PT_PTRACED;
 412	}
 413
 414	audit_ptrace(task);
 415
 416	retval = -EPERM;
 417	if (unlikely(task->flags & PF_KTHREAD))
 418		goto out;
 419	if (same_thread_group(task, current))
 420		goto out;
 421
 422	/*
 423	 * Protect exec's credential calculations against our interference;
 424	 * SUID, SGID and LSM creds get determined differently
 425	 * under ptrace.
 426	 */
 427	retval = -ERESTARTNOINTR;
 428	if (mutex_lock_interruptible(&task->signal->cred_guard_mutex))
 429		goto out;
 430
 431	task_lock(task);
 432	retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH_REALCREDS);
 433	task_unlock(task);
 434	if (retval)
 435		goto unlock_creds;
 436
 437	write_lock_irq(&tasklist_lock);
 438	retval = -EPERM;
 439	if (unlikely(task->exit_state))
 440		goto unlock_tasklist;
 441	if (task->ptrace)
 442		goto unlock_tasklist;
 443
 444	task->ptrace = flags;
 445
 446	ptrace_link(task, current);
 447
 448	/* SEIZE doesn't trap tracee on attach */
 449	if (!seize)
 450		send_sig_info(SIGSTOP, SEND_SIG_PRIV, task);
 451
 452	spin_lock(&task->sighand->siglock);
 
 
 453
 454	/*
 455	 * If the task is already STOPPED, set JOBCTL_TRAP_STOP and
 456	 * TRAPPING, and kick it so that it transits to TRACED.  TRAPPING
 457	 * will be cleared if the child completes the transition or any
 458	 * event which clears the group stop states happens.  We'll wait
 459	 * for the transition to complete before returning from this
 460	 * function.
 461	 *
 462	 * This hides STOPPED -> RUNNING -> TRACED transition from the
 463	 * attaching thread but a different thread in the same group can
 464	 * still observe the transient RUNNING state.  IOW, if another
 465	 * thread's WNOHANG wait(2) on the stopped tracee races against
 466	 * ATTACH, the wait(2) may fail due to the transient RUNNING.
 467	 *
 468	 * The following task_is_stopped() test is safe as both transitions
 469	 * in and out of STOPPED are protected by siglock.
 470	 */
 471	if (task_is_stopped(task) &&
 472	    task_set_jobctl_pending(task, JOBCTL_TRAP_STOP | JOBCTL_TRAPPING)) {
 473		task->jobctl &= ~JOBCTL_STOPPED;
 474		signal_wake_up_state(task, __TASK_STOPPED);
 475	}
 476
 477	spin_unlock(&task->sighand->siglock);
 478
 479	retval = 0;
 480unlock_tasklist:
 481	write_unlock_irq(&tasklist_lock);
 482unlock_creds:
 483	mutex_unlock(&task->signal->cred_guard_mutex);
 484out:
 485	if (!retval) {
 486		/*
 487		 * We do not bother to change retval or clear JOBCTL_TRAPPING
 488		 * if wait_on_bit() was interrupted by SIGKILL. The tracer will
 489		 * not return to user-mode, it will exit and clear this bit in
 490		 * __ptrace_unlink() if it wasn't already cleared by the tracee;
 491		 * and until then nobody can ptrace this task.
 492		 */
 493		wait_on_bit(&task->jobctl, JOBCTL_TRAPPING_BIT, TASK_KILLABLE);
 494		proc_ptrace_connector(task, PTRACE_ATTACH);
 495	}
 496
 497	return retval;
 498}
 499
 500/**
 501 * ptrace_traceme  --  helper for PTRACE_TRACEME
 502 *
 503 * Performs checks and sets PT_PTRACED.
 504 * Should be used by all ptrace implementations for PTRACE_TRACEME.
 505 */
 506static int ptrace_traceme(void)
 507{
 508	int ret = -EPERM;
 509
 510	write_lock_irq(&tasklist_lock);
 511	/* Are we already being traced? */
 512	if (!current->ptrace) {
 513		ret = security_ptrace_traceme(current->parent);
 514		/*
 515		 * Check PF_EXITING to ensure ->real_parent has not passed
 516		 * exit_ptrace(). Otherwise we don't report the error but
 517		 * pretend ->real_parent untraces us right after return.
 518		 */
 519		if (!ret && !(current->real_parent->flags & PF_EXITING)) {
 520			current->ptrace = PT_PTRACED;
 521			ptrace_link(current, current->real_parent);
 522		}
 523	}
 524	write_unlock_irq(&tasklist_lock);
 525
 526	return ret;
 527}
 528
 529/*
 530 * Called with irqs disabled, returns true if childs should reap themselves.
 531 */
 532static int ignoring_children(struct sighand_struct *sigh)
 533{
 534	int ret;
 535	spin_lock(&sigh->siglock);
 536	ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
 537	      (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
 538	spin_unlock(&sigh->siglock);
 539	return ret;
 540}
 541
 542/*
 543 * Called with tasklist_lock held for writing.
 544 * Unlink a traced task, and clean it up if it was a traced zombie.
 545 * Return true if it needs to be reaped with release_task().
 546 * (We can't call release_task() here because we already hold tasklist_lock.)
 547 *
 548 * If it's a zombie, our attachedness prevented normal parent notification
 549 * or self-reaping.  Do notification now if it would have happened earlier.
 550 * If it should reap itself, return true.
 551 *
 552 * If it's our own child, there is no notification to do. But if our normal
 553 * children self-reap, then this child was prevented by ptrace and we must
 554 * reap it now, in that case we must also wake up sub-threads sleeping in
 555 * do_wait().
 556 */
 557static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
 558{
 559	bool dead;
 560
 561	__ptrace_unlink(p);
 562
 563	if (p->exit_state != EXIT_ZOMBIE)
 564		return false;
 565
 566	dead = !thread_group_leader(p);
 567
 568	if (!dead && thread_group_empty(p)) {
 569		if (!same_thread_group(p->real_parent, tracer))
 570			dead = do_notify_parent(p, p->exit_signal);
 571		else if (ignoring_children(tracer->sighand)) {
 572			__wake_up_parent(p, tracer);
 573			dead = true;
 574		}
 575	}
 576	/* Mark it as in the process of being reaped. */
 577	if (dead)
 578		p->exit_state = EXIT_DEAD;
 579	return dead;
 580}
 581
 582static int ptrace_detach(struct task_struct *child, unsigned int data)
 583{
 584	if (!valid_signal(data))
 585		return -EIO;
 586
 587	/* Architecture-specific hardware disable .. */
 588	ptrace_disable(child);
 589
 590	write_lock_irq(&tasklist_lock);
 591	/*
 592	 * We rely on ptrace_freeze_traced(). It can't be killed and
 593	 * untraced by another thread, it can't be a zombie.
 594	 */
 595	WARN_ON(!child->ptrace || child->exit_state);
 596	/*
 597	 * tasklist_lock avoids the race with wait_task_stopped(), see
 598	 * the comment in ptrace_resume().
 599	 */
 600	child->exit_code = data;
 601	__ptrace_detach(current, child);
 602	write_unlock_irq(&tasklist_lock);
 603
 604	proc_ptrace_connector(child, PTRACE_DETACH);
 605
 606	return 0;
 607}
 608
 609/*
 610 * Detach all tasks we were using ptrace on. Called with tasklist held
 611 * for writing.
 612 */
 613void exit_ptrace(struct task_struct *tracer, struct list_head *dead)
 614{
 615	struct task_struct *p, *n;
 616
 617	list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
 618		if (unlikely(p->ptrace & PT_EXITKILL))
 619			send_sig_info(SIGKILL, SEND_SIG_PRIV, p);
 620
 621		if (__ptrace_detach(tracer, p))
 622			list_add(&p->ptrace_entry, dead);
 623	}
 624}
 625
 626int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
 627{
 628	int copied = 0;
 629
 630	while (len > 0) {
 631		char buf[128];
 632		int this_len, retval;
 633
 634		this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 635		retval = ptrace_access_vm(tsk, src, buf, this_len, FOLL_FORCE);
 636
 637		if (!retval) {
 638			if (copied)
 639				break;
 640			return -EIO;
 641		}
 642		if (copy_to_user(dst, buf, retval))
 643			return -EFAULT;
 644		copied += retval;
 645		src += retval;
 646		dst += retval;
 647		len -= retval;
 648	}
 649	return copied;
 650}
 651
 652int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
 653{
 654	int copied = 0;
 655
 656	while (len > 0) {
 657		char buf[128];
 658		int this_len, retval;
 659
 660		this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 661		if (copy_from_user(buf, src, this_len))
 662			return -EFAULT;
 663		retval = ptrace_access_vm(tsk, dst, buf, this_len,
 664				FOLL_FORCE | FOLL_WRITE);
 665		if (!retval) {
 666			if (copied)
 667				break;
 668			return -EIO;
 669		}
 670		copied += retval;
 671		src += retval;
 672		dst += retval;
 673		len -= retval;
 674	}
 675	return copied;
 676}
 677
 678static int ptrace_setoptions(struct task_struct *child, unsigned long data)
 679{
 680	unsigned flags;
 681	int ret;
 682
 683	ret = check_ptrace_options(data);
 684	if (ret)
 685		return ret;
 686
 687	/* Avoid intermediate state when all opts are cleared */
 688	flags = child->ptrace;
 689	flags &= ~(PTRACE_O_MASK << PT_OPT_FLAG_SHIFT);
 690	flags |= (data << PT_OPT_FLAG_SHIFT);
 691	child->ptrace = flags;
 692
 693	return 0;
 694}
 695
 696static int ptrace_getsiginfo(struct task_struct *child, kernel_siginfo_t *info)
 697{
 698	unsigned long flags;
 699	int error = -ESRCH;
 700
 701	if (lock_task_sighand(child, &flags)) {
 702		error = -EINVAL;
 703		if (likely(child->last_siginfo != NULL)) {
 704			copy_siginfo(info, child->last_siginfo);
 705			error = 0;
 706		}
 707		unlock_task_sighand(child, &flags);
 708	}
 709	return error;
 710}
 711
 712static int ptrace_setsiginfo(struct task_struct *child, const kernel_siginfo_t *info)
 713{
 714	unsigned long flags;
 715	int error = -ESRCH;
 716
 717	if (lock_task_sighand(child, &flags)) {
 718		error = -EINVAL;
 719		if (likely(child->last_siginfo != NULL)) {
 720			copy_siginfo(child->last_siginfo, info);
 721			error = 0;
 722		}
 723		unlock_task_sighand(child, &flags);
 724	}
 725	return error;
 726}
 727
 728static int ptrace_peek_siginfo(struct task_struct *child,
 729				unsigned long addr,
 730				unsigned long data)
 731{
 732	struct ptrace_peeksiginfo_args arg;
 733	struct sigpending *pending;
 734	struct sigqueue *q;
 735	int ret, i;
 736
 737	ret = copy_from_user(&arg, (void __user *) addr,
 738				sizeof(struct ptrace_peeksiginfo_args));
 739	if (ret)
 740		return -EFAULT;
 741
 742	if (arg.flags & ~PTRACE_PEEKSIGINFO_SHARED)
 743		return -EINVAL; /* unknown flags */
 744
 745	if (arg.nr < 0)
 746		return -EINVAL;
 747
 748	/* Ensure arg.off fits in an unsigned long */
 749	if (arg.off > ULONG_MAX)
 750		return 0;
 751
 752	if (arg.flags & PTRACE_PEEKSIGINFO_SHARED)
 753		pending = &child->signal->shared_pending;
 754	else
 755		pending = &child->pending;
 756
 757	for (i = 0; i < arg.nr; ) {
 758		kernel_siginfo_t info;
 759		unsigned long off = arg.off + i;
 760		bool found = false;
 761
 762		spin_lock_irq(&child->sighand->siglock);
 763		list_for_each_entry(q, &pending->list, list) {
 764			if (!off--) {
 765				found = true;
 766				copy_siginfo(&info, &q->info);
 767				break;
 768			}
 769		}
 770		spin_unlock_irq(&child->sighand->siglock);
 771
 772		if (!found) /* beyond the end of the list */
 773			break;
 774
 775#ifdef CONFIG_COMPAT
 776		if (unlikely(in_compat_syscall())) {
 777			compat_siginfo_t __user *uinfo = compat_ptr(data);
 778
 779			if (copy_siginfo_to_user32(uinfo, &info)) {
 780				ret = -EFAULT;
 781				break;
 782			}
 783
 784		} else
 785#endif
 786		{
 787			siginfo_t __user *uinfo = (siginfo_t __user *) data;
 788
 789			if (copy_siginfo_to_user(uinfo, &info)) {
 790				ret = -EFAULT;
 791				break;
 792			}
 793		}
 794
 795		data += sizeof(siginfo_t);
 796		i++;
 797
 798		if (signal_pending(current))
 799			break;
 800
 801		cond_resched();
 802	}
 803
 804	if (i > 0)
 805		return i;
 806
 807	return ret;
 808}
 809
 810#ifdef CONFIG_RSEQ
 811static long ptrace_get_rseq_configuration(struct task_struct *task,
 812					  unsigned long size, void __user *data)
 813{
 814	struct ptrace_rseq_configuration conf = {
 815		.rseq_abi_pointer = (u64)(uintptr_t)task->rseq,
 816		.rseq_abi_size = sizeof(*task->rseq),
 817		.signature = task->rseq_sig,
 818		.flags = 0,
 819	};
 820
 821	size = min_t(unsigned long, size, sizeof(conf));
 822	if (copy_to_user(data, &conf, size))
 823		return -EFAULT;
 824	return sizeof(conf);
 825}
 826#endif
 827
 828#define is_singlestep(request)		((request) == PTRACE_SINGLESTEP)
 829
 830#ifdef PTRACE_SINGLEBLOCK
 831#define is_singleblock(request)		((request) == PTRACE_SINGLEBLOCK)
 832#else
 833#define is_singleblock(request)		0
 834#endif
 835
 836#ifdef PTRACE_SYSEMU
 837#define is_sysemu_singlestep(request)	((request) == PTRACE_SYSEMU_SINGLESTEP)
 838#else
 839#define is_sysemu_singlestep(request)	0
 840#endif
 841
 842static int ptrace_resume(struct task_struct *child, long request,
 843			 unsigned long data)
 844{
 845	if (!valid_signal(data))
 846		return -EIO;
 847
 848	if (request == PTRACE_SYSCALL)
 849		set_task_syscall_work(child, SYSCALL_TRACE);
 850	else
 851		clear_task_syscall_work(child, SYSCALL_TRACE);
 852
 853#if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU)
 854	if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
 855		set_task_syscall_work(child, SYSCALL_EMU);
 856	else
 857		clear_task_syscall_work(child, SYSCALL_EMU);
 858#endif
 859
 860	if (is_singleblock(request)) {
 861		if (unlikely(!arch_has_block_step()))
 862			return -EIO;
 863		user_enable_block_step(child);
 864	} else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
 865		if (unlikely(!arch_has_single_step()))
 866			return -EIO;
 867		user_enable_single_step(child);
 868	} else {
 869		user_disable_single_step(child);
 870	}
 871
 872	/*
 873	 * Change ->exit_code and ->state under siglock to avoid the race
 874	 * with wait_task_stopped() in between; a non-zero ->exit_code will
 875	 * wrongly look like another report from tracee.
 876	 *
 877	 * Note that we need siglock even if ->exit_code == data and/or this
 878	 * status was not reported yet, the new status must not be cleared by
 879	 * wait_task_stopped() after resume.
 880	 */
 881	spin_lock_irq(&child->sighand->siglock);
 882	child->exit_code = data;
 883	child->jobctl &= ~JOBCTL_TRACED;
 884	wake_up_state(child, __TASK_TRACED);
 885	spin_unlock_irq(&child->sighand->siglock);
 886
 887	return 0;
 888}
 889
 890#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
 891
 892static const struct user_regset *
 893find_regset(const struct user_regset_view *view, unsigned int type)
 894{
 895	const struct user_regset *regset;
 896	int n;
 897
 898	for (n = 0; n < view->n; ++n) {
 899		regset = view->regsets + n;
 900		if (regset->core_note_type == type)
 901			return regset;
 902	}
 903
 904	return NULL;
 905}
 906
 907static int ptrace_regset(struct task_struct *task, int req, unsigned int type,
 908			 struct iovec *kiov)
 909{
 910	const struct user_regset_view *view = task_user_regset_view(task);
 911	const struct user_regset *regset = find_regset(view, type);
 912	int regset_no;
 913
 914	if (!regset || (kiov->iov_len % regset->size) != 0)
 915		return -EINVAL;
 916
 917	regset_no = regset - view->regsets;
 918	kiov->iov_len = min(kiov->iov_len,
 919			    (__kernel_size_t) (regset->n * regset->size));
 920
 921	if (req == PTRACE_GETREGSET)
 922		return copy_regset_to_user(task, view, regset_no, 0,
 923					   kiov->iov_len, kiov->iov_base);
 924	else
 925		return copy_regset_from_user(task, view, regset_no, 0,
 926					     kiov->iov_len, kiov->iov_base);
 927}
 928
 929/*
 930 * This is declared in linux/regset.h and defined in machine-dependent
 931 * code.  We put the export here, near the primary machine-neutral use,
 932 * to ensure no machine forgets it.
 933 */
 934EXPORT_SYMBOL_GPL(task_user_regset_view);
 935
 936static unsigned long
 937ptrace_get_syscall_info_entry(struct task_struct *child, struct pt_regs *regs,
 938			      struct ptrace_syscall_info *info)
 939{
 940	unsigned long args[ARRAY_SIZE(info->entry.args)];
 941	int i;
 942
 943	info->op = PTRACE_SYSCALL_INFO_ENTRY;
 944	info->entry.nr = syscall_get_nr(child, regs);
 945	syscall_get_arguments(child, regs, args);
 946	for (i = 0; i < ARRAY_SIZE(args); i++)
 947		info->entry.args[i] = args[i];
 948
 949	/* args is the last field in struct ptrace_syscall_info.entry */
 950	return offsetofend(struct ptrace_syscall_info, entry.args);
 951}
 952
 953static unsigned long
 954ptrace_get_syscall_info_seccomp(struct task_struct *child, struct pt_regs *regs,
 955				struct ptrace_syscall_info *info)
 956{
 957	/*
 958	 * As struct ptrace_syscall_info.entry is currently a subset
 959	 * of struct ptrace_syscall_info.seccomp, it makes sense to
 960	 * initialize that subset using ptrace_get_syscall_info_entry().
 961	 * This can be reconsidered in the future if these structures
 962	 * diverge significantly enough.
 963	 */
 964	ptrace_get_syscall_info_entry(child, regs, info);
 965	info->op = PTRACE_SYSCALL_INFO_SECCOMP;
 966	info->seccomp.ret_data = child->ptrace_message;
 967
 968	/* ret_data is the last field in struct ptrace_syscall_info.seccomp */
 969	return offsetofend(struct ptrace_syscall_info, seccomp.ret_data);
 970}
 971
 972static unsigned long
 973ptrace_get_syscall_info_exit(struct task_struct *child, struct pt_regs *regs,
 974			     struct ptrace_syscall_info *info)
 975{
 976	info->op = PTRACE_SYSCALL_INFO_EXIT;
 977	info->exit.rval = syscall_get_error(child, regs);
 978	info->exit.is_error = !!info->exit.rval;
 979	if (!info->exit.is_error)
 980		info->exit.rval = syscall_get_return_value(child, regs);
 981
 982	/* is_error is the last field in struct ptrace_syscall_info.exit */
 983	return offsetofend(struct ptrace_syscall_info, exit.is_error);
 984}
 985
 986static int
 987ptrace_get_syscall_info(struct task_struct *child, unsigned long user_size,
 988			void __user *datavp)
 989{
 990	struct pt_regs *regs = task_pt_regs(child);
 991	struct ptrace_syscall_info info = {
 992		.op = PTRACE_SYSCALL_INFO_NONE,
 993		.arch = syscall_get_arch(child),
 994		.instruction_pointer = instruction_pointer(regs),
 995		.stack_pointer = user_stack_pointer(regs),
 996	};
 997	unsigned long actual_size = offsetof(struct ptrace_syscall_info, entry);
 998	unsigned long write_size;
 999
1000	/*
1001	 * This does not need lock_task_sighand() to access
1002	 * child->last_siginfo because ptrace_freeze_traced()
1003	 * called earlier by ptrace_check_attach() ensures that
1004	 * the tracee cannot go away and clear its last_siginfo.
1005	 */
1006	switch (child->last_siginfo ? child->last_siginfo->si_code : 0) {
1007	case SIGTRAP | 0x80:
1008		switch (child->ptrace_message) {
1009		case PTRACE_EVENTMSG_SYSCALL_ENTRY:
1010			actual_size = ptrace_get_syscall_info_entry(child, regs,
1011								    &info);
1012			break;
1013		case PTRACE_EVENTMSG_SYSCALL_EXIT:
1014			actual_size = ptrace_get_syscall_info_exit(child, regs,
1015								   &info);
1016			break;
1017		}
1018		break;
1019	case SIGTRAP | (PTRACE_EVENT_SECCOMP << 8):
1020		actual_size = ptrace_get_syscall_info_seccomp(child, regs,
1021							      &info);
1022		break;
1023	}
1024
1025	write_size = min(actual_size, user_size);
1026	return copy_to_user(datavp, &info, write_size) ? -EFAULT : actual_size;
1027}
1028#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
1029
1030int ptrace_request(struct task_struct *child, long request,
1031		   unsigned long addr, unsigned long data)
1032{
1033	bool seized = child->ptrace & PT_SEIZED;
1034	int ret = -EIO;
1035	kernel_siginfo_t siginfo, *si;
1036	void __user *datavp = (void __user *) data;
1037	unsigned long __user *datalp = datavp;
1038	unsigned long flags;
1039
1040	switch (request) {
1041	case PTRACE_PEEKTEXT:
1042	case PTRACE_PEEKDATA:
1043		return generic_ptrace_peekdata(child, addr, data);
1044	case PTRACE_POKETEXT:
1045	case PTRACE_POKEDATA:
1046		return generic_ptrace_pokedata(child, addr, data);
1047
1048#ifdef PTRACE_OLDSETOPTIONS
1049	case PTRACE_OLDSETOPTIONS:
1050#endif
1051	case PTRACE_SETOPTIONS:
1052		ret = ptrace_setoptions(child, data);
1053		break;
1054	case PTRACE_GETEVENTMSG:
1055		ret = put_user(child->ptrace_message, datalp);
1056		break;
1057
1058	case PTRACE_PEEKSIGINFO:
1059		ret = ptrace_peek_siginfo(child, addr, data);
1060		break;
1061
1062	case PTRACE_GETSIGINFO:
1063		ret = ptrace_getsiginfo(child, &siginfo);
1064		if (!ret)
1065			ret = copy_siginfo_to_user(datavp, &siginfo);
1066		break;
1067
1068	case PTRACE_SETSIGINFO:
1069		ret = copy_siginfo_from_user(&siginfo, datavp);
1070		if (!ret)
1071			ret = ptrace_setsiginfo(child, &siginfo);
1072		break;
1073
1074	case PTRACE_GETSIGMASK: {
1075		sigset_t *mask;
1076
1077		if (addr != sizeof(sigset_t)) {
1078			ret = -EINVAL;
1079			break;
1080		}
1081
1082		if (test_tsk_restore_sigmask(child))
1083			mask = &child->saved_sigmask;
1084		else
1085			mask = &child->blocked;
1086
1087		if (copy_to_user(datavp, mask, sizeof(sigset_t)))
1088			ret = -EFAULT;
1089		else
1090			ret = 0;
1091
1092		break;
1093	}
1094
1095	case PTRACE_SETSIGMASK: {
1096		sigset_t new_set;
1097
1098		if (addr != sizeof(sigset_t)) {
1099			ret = -EINVAL;
1100			break;
1101		}
1102
1103		if (copy_from_user(&new_set, datavp, sizeof(sigset_t))) {
1104			ret = -EFAULT;
1105			break;
1106		}
1107
1108		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
1109
1110		/*
1111		 * Every thread does recalc_sigpending() after resume, so
1112		 * retarget_shared_pending() and recalc_sigpending() are not
1113		 * called here.
1114		 */
1115		spin_lock_irq(&child->sighand->siglock);
1116		child->blocked = new_set;
1117		spin_unlock_irq(&child->sighand->siglock);
1118
1119		clear_tsk_restore_sigmask(child);
1120
1121		ret = 0;
1122		break;
1123	}
1124
1125	case PTRACE_INTERRUPT:
1126		/*
1127		 * Stop tracee without any side-effect on signal or job
1128		 * control.  At least one trap is guaranteed to happen
1129		 * after this request.  If @child is already trapped, the
1130		 * current trap is not disturbed and another trap will
1131		 * happen after the current trap is ended with PTRACE_CONT.
1132		 *
1133		 * The actual trap might not be PTRACE_EVENT_STOP trap but
1134		 * the pending condition is cleared regardless.
1135		 */
1136		if (unlikely(!seized || !lock_task_sighand(child, &flags)))
1137			break;
1138
1139		/*
1140		 * INTERRUPT doesn't disturb existing trap sans one
1141		 * exception.  If ptracer issued LISTEN for the current
1142		 * STOP, this INTERRUPT should clear LISTEN and re-trap
1143		 * tracee into STOP.
1144		 */
1145		if (likely(task_set_jobctl_pending(child, JOBCTL_TRAP_STOP)))
1146			ptrace_signal_wake_up(child, child->jobctl & JOBCTL_LISTENING);
1147
1148		unlock_task_sighand(child, &flags);
1149		ret = 0;
1150		break;
1151
1152	case PTRACE_LISTEN:
1153		/*
1154		 * Listen for events.  Tracee must be in STOP.  It's not
1155		 * resumed per-se but is not considered to be in TRACED by
1156		 * wait(2) or ptrace(2).  If an async event (e.g. group
1157		 * stop state change) happens, tracee will enter STOP trap
1158		 * again.  Alternatively, ptracer can issue INTERRUPT to
1159		 * finish listening and re-trap tracee into STOP.
1160		 */
1161		if (unlikely(!seized || !lock_task_sighand(child, &flags)))
1162			break;
1163
1164		si = child->last_siginfo;
1165		if (likely(si && (si->si_code >> 8) == PTRACE_EVENT_STOP)) {
1166			child->jobctl |= JOBCTL_LISTENING;
1167			/*
1168			 * If NOTIFY is set, it means event happened between
1169			 * start of this trap and now.  Trigger re-trap.
1170			 */
1171			if (child->jobctl & JOBCTL_TRAP_NOTIFY)
1172				ptrace_signal_wake_up(child, true);
1173			ret = 0;
1174		}
1175		unlock_task_sighand(child, &flags);
1176		break;
1177
1178	case PTRACE_DETACH:	 /* detach a process that was attached. */
1179		ret = ptrace_detach(child, data);
1180		break;
1181
1182#ifdef CONFIG_BINFMT_ELF_FDPIC
1183	case PTRACE_GETFDPIC: {
1184		struct mm_struct *mm = get_task_mm(child);
1185		unsigned long tmp = 0;
1186
1187		ret = -ESRCH;
1188		if (!mm)
1189			break;
1190
1191		switch (addr) {
1192		case PTRACE_GETFDPIC_EXEC:
1193			tmp = mm->context.exec_fdpic_loadmap;
1194			break;
1195		case PTRACE_GETFDPIC_INTERP:
1196			tmp = mm->context.interp_fdpic_loadmap;
1197			break;
1198		default:
1199			break;
1200		}
1201		mmput(mm);
1202
1203		ret = put_user(tmp, datalp);
1204		break;
1205	}
1206#endif
1207
1208	case PTRACE_SINGLESTEP:
1209#ifdef PTRACE_SINGLEBLOCK
1210	case PTRACE_SINGLEBLOCK:
1211#endif
1212#ifdef PTRACE_SYSEMU
1213	case PTRACE_SYSEMU:
1214	case PTRACE_SYSEMU_SINGLESTEP:
1215#endif
1216	case PTRACE_SYSCALL:
1217	case PTRACE_CONT:
1218		return ptrace_resume(child, request, data);
1219
1220	case PTRACE_KILL:
1221		send_sig_info(SIGKILL, SEND_SIG_NOINFO, child);
1222		return 0;
1223
1224#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
1225	case PTRACE_GETREGSET:
1226	case PTRACE_SETREGSET: {
1227		struct iovec kiov;
1228		struct iovec __user *uiov = datavp;
1229
1230		if (!access_ok(uiov, sizeof(*uiov)))
1231			return -EFAULT;
1232
1233		if (__get_user(kiov.iov_base, &uiov->iov_base) ||
1234		    __get_user(kiov.iov_len, &uiov->iov_len))
1235			return -EFAULT;
1236
1237		ret = ptrace_regset(child, request, addr, &kiov);
1238		if (!ret)
1239			ret = __put_user(kiov.iov_len, &uiov->iov_len);
1240		break;
1241	}
1242
1243	case PTRACE_GET_SYSCALL_INFO:
1244		ret = ptrace_get_syscall_info(child, addr, datavp);
1245		break;
1246#endif
1247
1248	case PTRACE_SECCOMP_GET_FILTER:
1249		ret = seccomp_get_filter(child, addr, datavp);
1250		break;
1251
1252	case PTRACE_SECCOMP_GET_METADATA:
1253		ret = seccomp_get_metadata(child, addr, datavp);
1254		break;
1255
1256#ifdef CONFIG_RSEQ
1257	case PTRACE_GET_RSEQ_CONFIGURATION:
1258		ret = ptrace_get_rseq_configuration(child, addr, datavp);
1259		break;
1260#endif
 
 
 
 
 
 
 
 
1261
1262	default:
1263		break;
1264	}
1265
1266	return ret;
1267}
1268
1269SYSCALL_DEFINE4(ptrace, long, request, long, pid, unsigned long, addr,
1270		unsigned long, data)
1271{
1272	struct task_struct *child;
1273	long ret;
1274
1275	if (request == PTRACE_TRACEME) {
1276		ret = ptrace_traceme();
1277		goto out;
1278	}
1279
1280	child = find_get_task_by_vpid(pid);
1281	if (!child) {
1282		ret = -ESRCH;
1283		goto out;
1284	}
1285
1286	if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1287		ret = ptrace_attach(child, request, addr, data);
1288		goto out_put_task_struct;
1289	}
1290
1291	ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1292				  request == PTRACE_INTERRUPT);
1293	if (ret < 0)
1294		goto out_put_task_struct;
1295
1296	ret = arch_ptrace(child, request, addr, data);
1297	if (ret || request != PTRACE_DETACH)
1298		ptrace_unfreeze_traced(child);
1299
1300 out_put_task_struct:
1301	put_task_struct(child);
1302 out:
1303	return ret;
1304}
1305
1306int generic_ptrace_peekdata(struct task_struct *tsk, unsigned long addr,
1307			    unsigned long data)
1308{
1309	unsigned long tmp;
1310	int copied;
1311
1312	copied = ptrace_access_vm(tsk, addr, &tmp, sizeof(tmp), FOLL_FORCE);
1313	if (copied != sizeof(tmp))
1314		return -EIO;
1315	return put_user(tmp, (unsigned long __user *)data);
1316}
1317
1318int generic_ptrace_pokedata(struct task_struct *tsk, unsigned long addr,
1319			    unsigned long data)
1320{
1321	int copied;
1322
1323	copied = ptrace_access_vm(tsk, addr, &data, sizeof(data),
1324			FOLL_FORCE | FOLL_WRITE);
1325	return (copied == sizeof(data)) ? 0 : -EIO;
1326}
1327
1328#if defined CONFIG_COMPAT
1329
1330int compat_ptrace_request(struct task_struct *child, compat_long_t request,
1331			  compat_ulong_t addr, compat_ulong_t data)
1332{
1333	compat_ulong_t __user *datap = compat_ptr(data);
1334	compat_ulong_t word;
1335	kernel_siginfo_t siginfo;
1336	int ret;
1337
1338	switch (request) {
1339	case PTRACE_PEEKTEXT:
1340	case PTRACE_PEEKDATA:
1341		ret = ptrace_access_vm(child, addr, &word, sizeof(word),
1342				FOLL_FORCE);
1343		if (ret != sizeof(word))
1344			ret = -EIO;
1345		else
1346			ret = put_user(word, datap);
1347		break;
1348
1349	case PTRACE_POKETEXT:
1350	case PTRACE_POKEDATA:
1351		ret = ptrace_access_vm(child, addr, &data, sizeof(data),
1352				FOLL_FORCE | FOLL_WRITE);
1353		ret = (ret != sizeof(data) ? -EIO : 0);
1354		break;
1355
1356	case PTRACE_GETEVENTMSG:
1357		ret = put_user((compat_ulong_t) child->ptrace_message, datap);
1358		break;
1359
1360	case PTRACE_GETSIGINFO:
1361		ret = ptrace_getsiginfo(child, &siginfo);
1362		if (!ret)
1363			ret = copy_siginfo_to_user32(
1364				(struct compat_siginfo __user *) datap,
1365				&siginfo);
1366		break;
1367
1368	case PTRACE_SETSIGINFO:
1369		ret = copy_siginfo_from_user32(
1370			&siginfo, (struct compat_siginfo __user *) datap);
1371		if (!ret)
1372			ret = ptrace_setsiginfo(child, &siginfo);
1373		break;
1374#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
1375	case PTRACE_GETREGSET:
1376	case PTRACE_SETREGSET:
1377	{
1378		struct iovec kiov;
1379		struct compat_iovec __user *uiov =
1380			(struct compat_iovec __user *) datap;
1381		compat_uptr_t ptr;
1382		compat_size_t len;
1383
1384		if (!access_ok(uiov, sizeof(*uiov)))
1385			return -EFAULT;
1386
1387		if (__get_user(ptr, &uiov->iov_base) ||
1388		    __get_user(len, &uiov->iov_len))
1389			return -EFAULT;
1390
1391		kiov.iov_base = compat_ptr(ptr);
1392		kiov.iov_len = len;
1393
1394		ret = ptrace_regset(child, request, addr, &kiov);
1395		if (!ret)
1396			ret = __put_user(kiov.iov_len, &uiov->iov_len);
1397		break;
1398	}
1399#endif
1400
1401	default:
1402		ret = ptrace_request(child, request, addr, data);
1403	}
1404
1405	return ret;
1406}
1407
1408COMPAT_SYSCALL_DEFINE4(ptrace, compat_long_t, request, compat_long_t, pid,
1409		       compat_long_t, addr, compat_long_t, data)
1410{
1411	struct task_struct *child;
1412	long ret;
1413
1414	if (request == PTRACE_TRACEME) {
1415		ret = ptrace_traceme();
1416		goto out;
1417	}
1418
1419	child = find_get_task_by_vpid(pid);
1420	if (!child) {
1421		ret = -ESRCH;
1422		goto out;
1423	}
1424
1425	if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1426		ret = ptrace_attach(child, request, addr, data);
1427		goto out_put_task_struct;
1428	}
1429
1430	ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1431				  request == PTRACE_INTERRUPT);
1432	if (!ret) {
1433		ret = compat_arch_ptrace(child, request, addr, data);
1434		if (ret || request != PTRACE_DETACH)
1435			ptrace_unfreeze_traced(child);
1436	}
1437
1438 out_put_task_struct:
1439	put_task_struct(child);
1440 out:
1441	return ret;
1442}
1443#endif	/* CONFIG_COMPAT */
v6.8
   1// SPDX-License-Identifier: GPL-2.0-only
   2/*
   3 * linux/kernel/ptrace.c
   4 *
   5 * (C) Copyright 1999 Linus Torvalds
   6 *
   7 * Common interfaces for "ptrace()" which we do not want
   8 * to continually duplicate across every architecture.
   9 */
  10
  11#include <linux/capability.h>
  12#include <linux/export.h>
  13#include <linux/sched.h>
  14#include <linux/sched/mm.h>
  15#include <linux/sched/coredump.h>
  16#include <linux/sched/task.h>
  17#include <linux/errno.h>
  18#include <linux/mm.h>
  19#include <linux/highmem.h>
  20#include <linux/pagemap.h>
  21#include <linux/ptrace.h>
  22#include <linux/security.h>
  23#include <linux/signal.h>
  24#include <linux/uio.h>
  25#include <linux/audit.h>
  26#include <linux/pid_namespace.h>
  27#include <linux/syscalls.h>
  28#include <linux/uaccess.h>
  29#include <linux/regset.h>
  30#include <linux/hw_breakpoint.h>
  31#include <linux/cn_proc.h>
  32#include <linux/compat.h>
  33#include <linux/sched/signal.h>
  34#include <linux/minmax.h>
  35#include <linux/syscall_user_dispatch.h>
  36
  37#include <asm/syscall.h>	/* for syscall_get_* */
  38
  39/*
  40 * Access another process' address space via ptrace.
  41 * Source/target buffer must be kernel space,
  42 * Do not walk the page table directly, use get_user_pages
  43 */
  44int ptrace_access_vm(struct task_struct *tsk, unsigned long addr,
  45		     void *buf, int len, unsigned int gup_flags)
  46{
  47	struct mm_struct *mm;
  48	int ret;
  49
  50	mm = get_task_mm(tsk);
  51	if (!mm)
  52		return 0;
  53
  54	if (!tsk->ptrace ||
  55	    (current != tsk->parent) ||
  56	    ((get_dumpable(mm) != SUID_DUMP_USER) &&
  57	     !ptracer_capable(tsk, mm->user_ns))) {
  58		mmput(mm);
  59		return 0;
  60	}
  61
  62	ret = access_remote_vm(mm, addr, buf, len, gup_flags);
  63	mmput(mm);
  64
  65	return ret;
  66}
  67
  68
  69void __ptrace_link(struct task_struct *child, struct task_struct *new_parent,
  70		   const struct cred *ptracer_cred)
  71{
  72	BUG_ON(!list_empty(&child->ptrace_entry));
  73	list_add(&child->ptrace_entry, &new_parent->ptraced);
  74	child->parent = new_parent;
  75	child->ptracer_cred = get_cred(ptracer_cred);
  76}
  77
  78/*
  79 * ptrace a task: make the debugger its new parent and
  80 * move it to the ptrace list.
  81 *
  82 * Must be called with the tasklist lock write-held.
  83 */
  84static void ptrace_link(struct task_struct *child, struct task_struct *new_parent)
  85{
  86	__ptrace_link(child, new_parent, current_cred());
  87}
  88
  89/**
  90 * __ptrace_unlink - unlink ptracee and restore its execution state
  91 * @child: ptracee to be unlinked
  92 *
  93 * Remove @child from the ptrace list, move it back to the original parent,
  94 * and restore the execution state so that it conforms to the group stop
  95 * state.
  96 *
  97 * Unlinking can happen via two paths - explicit PTRACE_DETACH or ptracer
  98 * exiting.  For PTRACE_DETACH, unless the ptracee has been killed between
  99 * ptrace_check_attach() and here, it's guaranteed to be in TASK_TRACED.
 100 * If the ptracer is exiting, the ptracee can be in any state.
 101 *
 102 * After detach, the ptracee should be in a state which conforms to the
 103 * group stop.  If the group is stopped or in the process of stopping, the
 104 * ptracee should be put into TASK_STOPPED; otherwise, it should be woken
 105 * up from TASK_TRACED.
 106 *
 107 * If the ptracee is in TASK_TRACED and needs to be moved to TASK_STOPPED,
 108 * it goes through TRACED -> RUNNING -> STOPPED transition which is similar
 109 * to but in the opposite direction of what happens while attaching to a
 110 * stopped task.  However, in this direction, the intermediate RUNNING
 111 * state is not hidden even from the current ptracer and if it immediately
 112 * re-attaches and performs a WNOHANG wait(2), it may fail.
 113 *
 114 * CONTEXT:
 115 * write_lock_irq(tasklist_lock)
 116 */
 117void __ptrace_unlink(struct task_struct *child)
 118{
 119	const struct cred *old_cred;
 120	BUG_ON(!child->ptrace);
 121
 122	clear_task_syscall_work(child, SYSCALL_TRACE);
 123#if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU)
 124	clear_task_syscall_work(child, SYSCALL_EMU);
 125#endif
 126
 127	child->parent = child->real_parent;
 128	list_del_init(&child->ptrace_entry);
 129	old_cred = child->ptracer_cred;
 130	child->ptracer_cred = NULL;
 131	put_cred(old_cred);
 132
 133	spin_lock(&child->sighand->siglock);
 134	child->ptrace = 0;
 135	/*
 136	 * Clear all pending traps and TRAPPING.  TRAPPING should be
 137	 * cleared regardless of JOBCTL_STOP_PENDING.  Do it explicitly.
 138	 */
 139	task_clear_jobctl_pending(child, JOBCTL_TRAP_MASK);
 140	task_clear_jobctl_trapping(child);
 141
 142	/*
 143	 * Reinstate JOBCTL_STOP_PENDING if group stop is in effect and
 144	 * @child isn't dead.
 145	 */
 146	if (!(child->flags & PF_EXITING) &&
 147	    (child->signal->flags & SIGNAL_STOP_STOPPED ||
 148	     child->signal->group_stop_count))
 149		child->jobctl |= JOBCTL_STOP_PENDING;
 150
 
 
 
 
 
 
 
 
 
 
 
 151	/*
 152	 * If transition to TASK_STOPPED is pending or in TASK_TRACED, kick
 153	 * @child in the butt.  Note that @resume should be used iff @child
 154	 * is in TASK_TRACED; otherwise, we might unduly disrupt
 155	 * TASK_KILLABLE sleeps.
 156	 */
 157	if (child->jobctl & JOBCTL_STOP_PENDING || task_is_traced(child))
 158		ptrace_signal_wake_up(child, true);
 159
 160	spin_unlock(&child->sighand->siglock);
 161}
 162
 163static bool looks_like_a_spurious_pid(struct task_struct *task)
 164{
 165	if (task->exit_code != ((PTRACE_EVENT_EXEC << 8) | SIGTRAP))
 166		return false;
 167
 168	if (task_pid_vnr(task) == task->ptrace_message)
 169		return false;
 170	/*
 171	 * The tracee changed its pid but the PTRACE_EVENT_EXEC event
 172	 * was not wait()'ed, most probably debugger targets the old
 173	 * leader which was destroyed in de_thread().
 174	 */
 175	return true;
 176}
 177
 178/*
 179 * Ensure that nothing can wake it up, even SIGKILL
 180 *
 181 * A task is switched to this state while a ptrace operation is in progress;
 182 * such that the ptrace operation is uninterruptible.
 183 */
 184static bool ptrace_freeze_traced(struct task_struct *task)
 185{
 186	bool ret = false;
 187
 188	/* Lockless, nobody but us can set this flag */
 189	if (task->jobctl & JOBCTL_LISTENING)
 190		return ret;
 191
 192	spin_lock_irq(&task->sighand->siglock);
 193	if (task_is_traced(task) && !looks_like_a_spurious_pid(task) &&
 194	    !__fatal_signal_pending(task)) {
 195		task->jobctl |= JOBCTL_PTRACE_FROZEN;
 196		ret = true;
 197	}
 198	spin_unlock_irq(&task->sighand->siglock);
 199
 200	return ret;
 201}
 202
 203static void ptrace_unfreeze_traced(struct task_struct *task)
 204{
 205	unsigned long flags;
 206
 207	/*
 208	 * The child may be awake and may have cleared
 209	 * JOBCTL_PTRACE_FROZEN (see ptrace_resume).  The child will
 210	 * not set JOBCTL_PTRACE_FROZEN or enter __TASK_TRACED anew.
 211	 */
 212	if (lock_task_sighand(task, &flags)) {
 213		task->jobctl &= ~JOBCTL_PTRACE_FROZEN;
 214		if (__fatal_signal_pending(task)) {
 215			task->jobctl &= ~JOBCTL_TRACED;
 216			wake_up_state(task, __TASK_TRACED);
 217		}
 218		unlock_task_sighand(task, &flags);
 219	}
 220}
 221
 222/**
 223 * ptrace_check_attach - check whether ptracee is ready for ptrace operation
 224 * @child: ptracee to check for
 225 * @ignore_state: don't check whether @child is currently %TASK_TRACED
 226 *
 227 * Check whether @child is being ptraced by %current and ready for further
 228 * ptrace operations.  If @ignore_state is %false, @child also should be in
 229 * %TASK_TRACED state and on return the child is guaranteed to be traced
 230 * and not executing.  If @ignore_state is %true, @child can be in any
 231 * state.
 232 *
 233 * CONTEXT:
 234 * Grabs and releases tasklist_lock and @child->sighand->siglock.
 235 *
 236 * RETURNS:
 237 * 0 on success, -ESRCH if %child is not ready.
 238 */
 239static int ptrace_check_attach(struct task_struct *child, bool ignore_state)
 240{
 241	int ret = -ESRCH;
 242
 243	/*
 244	 * We take the read lock around doing both checks to close a
 245	 * possible race where someone else was tracing our child and
 246	 * detached between these two checks.  After this locked check,
 247	 * we are sure that this is our traced child and that can only
 248	 * be changed by us so it's not changing right after this.
 249	 */
 250	read_lock(&tasklist_lock);
 251	if (child->ptrace && child->parent == current) {
 252		/*
 253		 * child->sighand can't be NULL, release_task()
 254		 * does ptrace_unlink() before __exit_signal().
 255		 */
 256		if (ignore_state || ptrace_freeze_traced(child))
 257			ret = 0;
 258	}
 259	read_unlock(&tasklist_lock);
 260
 261	if (!ret && !ignore_state &&
 262	    WARN_ON_ONCE(!wait_task_inactive(child, __TASK_TRACED|TASK_FROZEN)))
 263		ret = -ESRCH;
 264
 265	return ret;
 266}
 267
 268static bool ptrace_has_cap(struct user_namespace *ns, unsigned int mode)
 269{
 270	if (mode & PTRACE_MODE_NOAUDIT)
 271		return ns_capable_noaudit(ns, CAP_SYS_PTRACE);
 272	return ns_capable(ns, CAP_SYS_PTRACE);
 273}
 274
 275/* Returns 0 on success, -errno on denial. */
 276static int __ptrace_may_access(struct task_struct *task, unsigned int mode)
 277{
 278	const struct cred *cred = current_cred(), *tcred;
 279	struct mm_struct *mm;
 280	kuid_t caller_uid;
 281	kgid_t caller_gid;
 282
 283	if (!(mode & PTRACE_MODE_FSCREDS) == !(mode & PTRACE_MODE_REALCREDS)) {
 284		WARN(1, "denying ptrace access check without PTRACE_MODE_*CREDS\n");
 285		return -EPERM;
 286	}
 287
 288	/* May we inspect the given task?
 289	 * This check is used both for attaching with ptrace
 290	 * and for allowing access to sensitive information in /proc.
 291	 *
 292	 * ptrace_attach denies several cases that /proc allows
 293	 * because setting up the necessary parent/child relationship
 294	 * or halting the specified task is impossible.
 295	 */
 296
 297	/* Don't let security modules deny introspection */
 298	if (same_thread_group(task, current))
 299		return 0;
 300	rcu_read_lock();
 301	if (mode & PTRACE_MODE_FSCREDS) {
 302		caller_uid = cred->fsuid;
 303		caller_gid = cred->fsgid;
 304	} else {
 305		/*
 306		 * Using the euid would make more sense here, but something
 307		 * in userland might rely on the old behavior, and this
 308		 * shouldn't be a security problem since
 309		 * PTRACE_MODE_REALCREDS implies that the caller explicitly
 310		 * used a syscall that requests access to another process
 311		 * (and not a filesystem syscall to procfs).
 312		 */
 313		caller_uid = cred->uid;
 314		caller_gid = cred->gid;
 315	}
 316	tcred = __task_cred(task);
 317	if (uid_eq(caller_uid, tcred->euid) &&
 318	    uid_eq(caller_uid, tcred->suid) &&
 319	    uid_eq(caller_uid, tcred->uid)  &&
 320	    gid_eq(caller_gid, tcred->egid) &&
 321	    gid_eq(caller_gid, tcred->sgid) &&
 322	    gid_eq(caller_gid, tcred->gid))
 323		goto ok;
 324	if (ptrace_has_cap(tcred->user_ns, mode))
 325		goto ok;
 326	rcu_read_unlock();
 327	return -EPERM;
 328ok:
 329	rcu_read_unlock();
 330	/*
 331	 * If a task drops privileges and becomes nondumpable (through a syscall
 332	 * like setresuid()) while we are trying to access it, we must ensure
 333	 * that the dumpability is read after the credentials; otherwise,
 334	 * we may be able to attach to a task that we shouldn't be able to
 335	 * attach to (as if the task had dropped privileges without becoming
 336	 * nondumpable).
 337	 * Pairs with a write barrier in commit_creds().
 338	 */
 339	smp_rmb();
 340	mm = task->mm;
 341	if (mm &&
 342	    ((get_dumpable(mm) != SUID_DUMP_USER) &&
 343	     !ptrace_has_cap(mm->user_ns, mode)))
 344	    return -EPERM;
 345
 346	return security_ptrace_access_check(task, mode);
 347}
 348
 349bool ptrace_may_access(struct task_struct *task, unsigned int mode)
 350{
 351	int err;
 352	task_lock(task);
 353	err = __ptrace_may_access(task, mode);
 354	task_unlock(task);
 355	return !err;
 356}
 357
 358static int check_ptrace_options(unsigned long data)
 359{
 360	if (data & ~(unsigned long)PTRACE_O_MASK)
 361		return -EINVAL;
 362
 363	if (unlikely(data & PTRACE_O_SUSPEND_SECCOMP)) {
 364		if (!IS_ENABLED(CONFIG_CHECKPOINT_RESTORE) ||
 365		    !IS_ENABLED(CONFIG_SECCOMP))
 366			return -EINVAL;
 367
 368		if (!capable(CAP_SYS_ADMIN))
 369			return -EPERM;
 370
 371		if (seccomp_mode(&current->seccomp) != SECCOMP_MODE_DISABLED ||
 372		    current->ptrace & PT_SUSPEND_SECCOMP)
 373			return -EPERM;
 374	}
 375	return 0;
 376}
 377
 378static inline void ptrace_set_stopped(struct task_struct *task)
 379{
 380	guard(spinlock)(&task->sighand->siglock);
 381
 382	/*
 383	 * If the task is already STOPPED, set JOBCTL_TRAP_STOP and
 384	 * TRAPPING, and kick it so that it transits to TRACED.  TRAPPING
 385	 * will be cleared if the child completes the transition or any
 386	 * event which clears the group stop states happens.  We'll wait
 387	 * for the transition to complete before returning from this
 388	 * function.
 389	 *
 390	 * This hides STOPPED -> RUNNING -> TRACED transition from the
 391	 * attaching thread but a different thread in the same group can
 392	 * still observe the transient RUNNING state.  IOW, if another
 393	 * thread's WNOHANG wait(2) on the stopped tracee races against
 394	 * ATTACH, the wait(2) may fail due to the transient RUNNING.
 395	 *
 396	 * The following task_is_stopped() test is safe as both transitions
 397	 * in and out of STOPPED are protected by siglock.
 398	 */
 399	if (task_is_stopped(task) &&
 400	    task_set_jobctl_pending(task, JOBCTL_TRAP_STOP | JOBCTL_TRAPPING)) {
 401		task->jobctl &= ~JOBCTL_STOPPED;
 402		signal_wake_up_state(task, __TASK_STOPPED);
 403	}
 404}
 405
 406static int ptrace_attach(struct task_struct *task, long request,
 407			 unsigned long addr,
 408			 unsigned long flags)
 409{
 410	bool seize = (request == PTRACE_SEIZE);
 411	int retval;
 412
 
 413	if (seize) {
 414		if (addr != 0)
 415			return -EIO;
 416		/*
 417		 * This duplicates the check in check_ptrace_options() because
 418		 * ptrace_attach() and ptrace_setoptions() have historically
 419		 * used different error codes for unknown ptrace options.
 420		 */
 421		if (flags & ~(unsigned long)PTRACE_O_MASK)
 422			return -EIO;
 423
 424		retval = check_ptrace_options(flags);
 425		if (retval)
 426			return retval;
 427		flags = PT_PTRACED | PT_SEIZED | (flags << PT_OPT_FLAG_SHIFT);
 428	} else {
 429		flags = PT_PTRACED;
 430	}
 431
 432	audit_ptrace(task);
 433
 
 434	if (unlikely(task->flags & PF_KTHREAD))
 435		return -EPERM;
 436	if (same_thread_group(task, current))
 437		return -EPERM;
 438
 439	/*
 440	 * Protect exec's credential calculations against our interference;
 441	 * SUID, SGID and LSM creds get determined differently
 442	 * under ptrace.
 443	 */
 444	scoped_cond_guard (mutex_intr, return -ERESTARTNOINTR,
 445			   &task->signal->cred_guard_mutex) {
 
 446
 447		scoped_guard (task_lock, task) {
 448			retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH_REALCREDS);
 449			if (retval)
 450				return retval;
 451		}
 452
 453		scoped_guard (write_lock_irq, &tasklist_lock) {
 454			if (unlikely(task->exit_state))
 455				return -EPERM;
 456			if (task->ptrace)
 457				return -EPERM;
 
 458
 459			task->ptrace = flags;
 460
 461			ptrace_link(task, current);
 462
 463			/* SEIZE doesn't trap tracee on attach */
 464			if (!seize)
 465				send_sig_info(SIGSTOP, SEND_SIG_PRIV, task);
 466
 467			ptrace_set_stopped(task);
 468		}
 469	}
 470
 471	/*
 472	 * We do not bother to change retval or clear JOBCTL_TRAPPING
 473	 * if wait_on_bit() was interrupted by SIGKILL. The tracer will
 474	 * not return to user-mode, it will exit and clear this bit in
 475	 * __ptrace_unlink() if it wasn't already cleared by the tracee;
 476	 * and until then nobody can ptrace this task.
 
 
 
 
 
 
 
 
 
 
 477	 */
 478	wait_on_bit(&task->jobctl, JOBCTL_TRAPPING_BIT, TASK_KILLABLE);
 479	proc_ptrace_connector(task, PTRACE_ATTACH);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 480
 481	return 0;
 482}
 483
 484/**
 485 * ptrace_traceme  --  helper for PTRACE_TRACEME
 486 *
 487 * Performs checks and sets PT_PTRACED.
 488 * Should be used by all ptrace implementations for PTRACE_TRACEME.
 489 */
 490static int ptrace_traceme(void)
 491{
 492	int ret = -EPERM;
 493
 494	write_lock_irq(&tasklist_lock);
 495	/* Are we already being traced? */
 496	if (!current->ptrace) {
 497		ret = security_ptrace_traceme(current->parent);
 498		/*
 499		 * Check PF_EXITING to ensure ->real_parent has not passed
 500		 * exit_ptrace(). Otherwise we don't report the error but
 501		 * pretend ->real_parent untraces us right after return.
 502		 */
 503		if (!ret && !(current->real_parent->flags & PF_EXITING)) {
 504			current->ptrace = PT_PTRACED;
 505			ptrace_link(current, current->real_parent);
 506		}
 507	}
 508	write_unlock_irq(&tasklist_lock);
 509
 510	return ret;
 511}
 512
 513/*
 514 * Called with irqs disabled, returns true if childs should reap themselves.
 515 */
 516static int ignoring_children(struct sighand_struct *sigh)
 517{
 518	int ret;
 519	spin_lock(&sigh->siglock);
 520	ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
 521	      (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
 522	spin_unlock(&sigh->siglock);
 523	return ret;
 524}
 525
 526/*
 527 * Called with tasklist_lock held for writing.
 528 * Unlink a traced task, and clean it up if it was a traced zombie.
 529 * Return true if it needs to be reaped with release_task().
 530 * (We can't call release_task() here because we already hold tasklist_lock.)
 531 *
 532 * If it's a zombie, our attachedness prevented normal parent notification
 533 * or self-reaping.  Do notification now if it would have happened earlier.
 534 * If it should reap itself, return true.
 535 *
 536 * If it's our own child, there is no notification to do. But if our normal
 537 * children self-reap, then this child was prevented by ptrace and we must
 538 * reap it now, in that case we must also wake up sub-threads sleeping in
 539 * do_wait().
 540 */
 541static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
 542{
 543	bool dead;
 544
 545	__ptrace_unlink(p);
 546
 547	if (p->exit_state != EXIT_ZOMBIE)
 548		return false;
 549
 550	dead = !thread_group_leader(p);
 551
 552	if (!dead && thread_group_empty(p)) {
 553		if (!same_thread_group(p->real_parent, tracer))
 554			dead = do_notify_parent(p, p->exit_signal);
 555		else if (ignoring_children(tracer->sighand)) {
 556			__wake_up_parent(p, tracer);
 557			dead = true;
 558		}
 559	}
 560	/* Mark it as in the process of being reaped. */
 561	if (dead)
 562		p->exit_state = EXIT_DEAD;
 563	return dead;
 564}
 565
 566static int ptrace_detach(struct task_struct *child, unsigned int data)
 567{
 568	if (!valid_signal(data))
 569		return -EIO;
 570
 571	/* Architecture-specific hardware disable .. */
 572	ptrace_disable(child);
 573
 574	write_lock_irq(&tasklist_lock);
 575	/*
 576	 * We rely on ptrace_freeze_traced(). It can't be killed and
 577	 * untraced by another thread, it can't be a zombie.
 578	 */
 579	WARN_ON(!child->ptrace || child->exit_state);
 580	/*
 581	 * tasklist_lock avoids the race with wait_task_stopped(), see
 582	 * the comment in ptrace_resume().
 583	 */
 584	child->exit_code = data;
 585	__ptrace_detach(current, child);
 586	write_unlock_irq(&tasklist_lock);
 587
 588	proc_ptrace_connector(child, PTRACE_DETACH);
 589
 590	return 0;
 591}
 592
 593/*
 594 * Detach all tasks we were using ptrace on. Called with tasklist held
 595 * for writing.
 596 */
 597void exit_ptrace(struct task_struct *tracer, struct list_head *dead)
 598{
 599	struct task_struct *p, *n;
 600
 601	list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
 602		if (unlikely(p->ptrace & PT_EXITKILL))
 603			send_sig_info(SIGKILL, SEND_SIG_PRIV, p);
 604
 605		if (__ptrace_detach(tracer, p))
 606			list_add(&p->ptrace_entry, dead);
 607	}
 608}
 609
 610int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
 611{
 612	int copied = 0;
 613
 614	while (len > 0) {
 615		char buf[128];
 616		int this_len, retval;
 617
 618		this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 619		retval = ptrace_access_vm(tsk, src, buf, this_len, FOLL_FORCE);
 620
 621		if (!retval) {
 622			if (copied)
 623				break;
 624			return -EIO;
 625		}
 626		if (copy_to_user(dst, buf, retval))
 627			return -EFAULT;
 628		copied += retval;
 629		src += retval;
 630		dst += retval;
 631		len -= retval;
 632	}
 633	return copied;
 634}
 635
 636int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
 637{
 638	int copied = 0;
 639
 640	while (len > 0) {
 641		char buf[128];
 642		int this_len, retval;
 643
 644		this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 645		if (copy_from_user(buf, src, this_len))
 646			return -EFAULT;
 647		retval = ptrace_access_vm(tsk, dst, buf, this_len,
 648				FOLL_FORCE | FOLL_WRITE);
 649		if (!retval) {
 650			if (copied)
 651				break;
 652			return -EIO;
 653		}
 654		copied += retval;
 655		src += retval;
 656		dst += retval;
 657		len -= retval;
 658	}
 659	return copied;
 660}
 661
 662static int ptrace_setoptions(struct task_struct *child, unsigned long data)
 663{
 664	unsigned flags;
 665	int ret;
 666
 667	ret = check_ptrace_options(data);
 668	if (ret)
 669		return ret;
 670
 671	/* Avoid intermediate state when all opts are cleared */
 672	flags = child->ptrace;
 673	flags &= ~(PTRACE_O_MASK << PT_OPT_FLAG_SHIFT);
 674	flags |= (data << PT_OPT_FLAG_SHIFT);
 675	child->ptrace = flags;
 676
 677	return 0;
 678}
 679
 680static int ptrace_getsiginfo(struct task_struct *child, kernel_siginfo_t *info)
 681{
 682	unsigned long flags;
 683	int error = -ESRCH;
 684
 685	if (lock_task_sighand(child, &flags)) {
 686		error = -EINVAL;
 687		if (likely(child->last_siginfo != NULL)) {
 688			copy_siginfo(info, child->last_siginfo);
 689			error = 0;
 690		}
 691		unlock_task_sighand(child, &flags);
 692	}
 693	return error;
 694}
 695
 696static int ptrace_setsiginfo(struct task_struct *child, const kernel_siginfo_t *info)
 697{
 698	unsigned long flags;
 699	int error = -ESRCH;
 700
 701	if (lock_task_sighand(child, &flags)) {
 702		error = -EINVAL;
 703		if (likely(child->last_siginfo != NULL)) {
 704			copy_siginfo(child->last_siginfo, info);
 705			error = 0;
 706		}
 707		unlock_task_sighand(child, &flags);
 708	}
 709	return error;
 710}
 711
 712static int ptrace_peek_siginfo(struct task_struct *child,
 713				unsigned long addr,
 714				unsigned long data)
 715{
 716	struct ptrace_peeksiginfo_args arg;
 717	struct sigpending *pending;
 718	struct sigqueue *q;
 719	int ret, i;
 720
 721	ret = copy_from_user(&arg, (void __user *) addr,
 722				sizeof(struct ptrace_peeksiginfo_args));
 723	if (ret)
 724		return -EFAULT;
 725
 726	if (arg.flags & ~PTRACE_PEEKSIGINFO_SHARED)
 727		return -EINVAL; /* unknown flags */
 728
 729	if (arg.nr < 0)
 730		return -EINVAL;
 731
 732	/* Ensure arg.off fits in an unsigned long */
 733	if (arg.off > ULONG_MAX)
 734		return 0;
 735
 736	if (arg.flags & PTRACE_PEEKSIGINFO_SHARED)
 737		pending = &child->signal->shared_pending;
 738	else
 739		pending = &child->pending;
 740
 741	for (i = 0; i < arg.nr; ) {
 742		kernel_siginfo_t info;
 743		unsigned long off = arg.off + i;
 744		bool found = false;
 745
 746		spin_lock_irq(&child->sighand->siglock);
 747		list_for_each_entry(q, &pending->list, list) {
 748			if (!off--) {
 749				found = true;
 750				copy_siginfo(&info, &q->info);
 751				break;
 752			}
 753		}
 754		spin_unlock_irq(&child->sighand->siglock);
 755
 756		if (!found) /* beyond the end of the list */
 757			break;
 758
 759#ifdef CONFIG_COMPAT
 760		if (unlikely(in_compat_syscall())) {
 761			compat_siginfo_t __user *uinfo = compat_ptr(data);
 762
 763			if (copy_siginfo_to_user32(uinfo, &info)) {
 764				ret = -EFAULT;
 765				break;
 766			}
 767
 768		} else
 769#endif
 770		{
 771			siginfo_t __user *uinfo = (siginfo_t __user *) data;
 772
 773			if (copy_siginfo_to_user(uinfo, &info)) {
 774				ret = -EFAULT;
 775				break;
 776			}
 777		}
 778
 779		data += sizeof(siginfo_t);
 780		i++;
 781
 782		if (signal_pending(current))
 783			break;
 784
 785		cond_resched();
 786	}
 787
 788	if (i > 0)
 789		return i;
 790
 791	return ret;
 792}
 793
 794#ifdef CONFIG_RSEQ
 795static long ptrace_get_rseq_configuration(struct task_struct *task,
 796					  unsigned long size, void __user *data)
 797{
 798	struct ptrace_rseq_configuration conf = {
 799		.rseq_abi_pointer = (u64)(uintptr_t)task->rseq,
 800		.rseq_abi_size = task->rseq_len,
 801		.signature = task->rseq_sig,
 802		.flags = 0,
 803	};
 804
 805	size = min_t(unsigned long, size, sizeof(conf));
 806	if (copy_to_user(data, &conf, size))
 807		return -EFAULT;
 808	return sizeof(conf);
 809}
 810#endif
 811
 812#define is_singlestep(request)		((request) == PTRACE_SINGLESTEP)
 813
 814#ifdef PTRACE_SINGLEBLOCK
 815#define is_singleblock(request)		((request) == PTRACE_SINGLEBLOCK)
 816#else
 817#define is_singleblock(request)		0
 818#endif
 819
 820#ifdef PTRACE_SYSEMU
 821#define is_sysemu_singlestep(request)	((request) == PTRACE_SYSEMU_SINGLESTEP)
 822#else
 823#define is_sysemu_singlestep(request)	0
 824#endif
 825
 826static int ptrace_resume(struct task_struct *child, long request,
 827			 unsigned long data)
 828{
 829	if (!valid_signal(data))
 830		return -EIO;
 831
 832	if (request == PTRACE_SYSCALL)
 833		set_task_syscall_work(child, SYSCALL_TRACE);
 834	else
 835		clear_task_syscall_work(child, SYSCALL_TRACE);
 836
 837#if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU)
 838	if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
 839		set_task_syscall_work(child, SYSCALL_EMU);
 840	else
 841		clear_task_syscall_work(child, SYSCALL_EMU);
 842#endif
 843
 844	if (is_singleblock(request)) {
 845		if (unlikely(!arch_has_block_step()))
 846			return -EIO;
 847		user_enable_block_step(child);
 848	} else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
 849		if (unlikely(!arch_has_single_step()))
 850			return -EIO;
 851		user_enable_single_step(child);
 852	} else {
 853		user_disable_single_step(child);
 854	}
 855
 856	/*
 857	 * Change ->exit_code and ->state under siglock to avoid the race
 858	 * with wait_task_stopped() in between; a non-zero ->exit_code will
 859	 * wrongly look like another report from tracee.
 860	 *
 861	 * Note that we need siglock even if ->exit_code == data and/or this
 862	 * status was not reported yet, the new status must not be cleared by
 863	 * wait_task_stopped() after resume.
 864	 */
 865	spin_lock_irq(&child->sighand->siglock);
 866	child->exit_code = data;
 867	child->jobctl &= ~JOBCTL_TRACED;
 868	wake_up_state(child, __TASK_TRACED);
 869	spin_unlock_irq(&child->sighand->siglock);
 870
 871	return 0;
 872}
 873
 874#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
 875
 876static const struct user_regset *
 877find_regset(const struct user_regset_view *view, unsigned int type)
 878{
 879	const struct user_regset *regset;
 880	int n;
 881
 882	for (n = 0; n < view->n; ++n) {
 883		regset = view->regsets + n;
 884		if (regset->core_note_type == type)
 885			return regset;
 886	}
 887
 888	return NULL;
 889}
 890
 891static int ptrace_regset(struct task_struct *task, int req, unsigned int type,
 892			 struct iovec *kiov)
 893{
 894	const struct user_regset_view *view = task_user_regset_view(task);
 895	const struct user_regset *regset = find_regset(view, type);
 896	int regset_no;
 897
 898	if (!regset || (kiov->iov_len % regset->size) != 0)
 899		return -EINVAL;
 900
 901	regset_no = regset - view->regsets;
 902	kiov->iov_len = min(kiov->iov_len,
 903			    (__kernel_size_t) (regset->n * regset->size));
 904
 905	if (req == PTRACE_GETREGSET)
 906		return copy_regset_to_user(task, view, regset_no, 0,
 907					   kiov->iov_len, kiov->iov_base);
 908	else
 909		return copy_regset_from_user(task, view, regset_no, 0,
 910					     kiov->iov_len, kiov->iov_base);
 911}
 912
 913/*
 914 * This is declared in linux/regset.h and defined in machine-dependent
 915 * code.  We put the export here, near the primary machine-neutral use,
 916 * to ensure no machine forgets it.
 917 */
 918EXPORT_SYMBOL_GPL(task_user_regset_view);
 919
 920static unsigned long
 921ptrace_get_syscall_info_entry(struct task_struct *child, struct pt_regs *regs,
 922			      struct ptrace_syscall_info *info)
 923{
 924	unsigned long args[ARRAY_SIZE(info->entry.args)];
 925	int i;
 926
 927	info->op = PTRACE_SYSCALL_INFO_ENTRY;
 928	info->entry.nr = syscall_get_nr(child, regs);
 929	syscall_get_arguments(child, regs, args);
 930	for (i = 0; i < ARRAY_SIZE(args); i++)
 931		info->entry.args[i] = args[i];
 932
 933	/* args is the last field in struct ptrace_syscall_info.entry */
 934	return offsetofend(struct ptrace_syscall_info, entry.args);
 935}
 936
 937static unsigned long
 938ptrace_get_syscall_info_seccomp(struct task_struct *child, struct pt_regs *regs,
 939				struct ptrace_syscall_info *info)
 940{
 941	/*
 942	 * As struct ptrace_syscall_info.entry is currently a subset
 943	 * of struct ptrace_syscall_info.seccomp, it makes sense to
 944	 * initialize that subset using ptrace_get_syscall_info_entry().
 945	 * This can be reconsidered in the future if these structures
 946	 * diverge significantly enough.
 947	 */
 948	ptrace_get_syscall_info_entry(child, regs, info);
 949	info->op = PTRACE_SYSCALL_INFO_SECCOMP;
 950	info->seccomp.ret_data = child->ptrace_message;
 951
 952	/* ret_data is the last field in struct ptrace_syscall_info.seccomp */
 953	return offsetofend(struct ptrace_syscall_info, seccomp.ret_data);
 954}
 955
 956static unsigned long
 957ptrace_get_syscall_info_exit(struct task_struct *child, struct pt_regs *regs,
 958			     struct ptrace_syscall_info *info)
 959{
 960	info->op = PTRACE_SYSCALL_INFO_EXIT;
 961	info->exit.rval = syscall_get_error(child, regs);
 962	info->exit.is_error = !!info->exit.rval;
 963	if (!info->exit.is_error)
 964		info->exit.rval = syscall_get_return_value(child, regs);
 965
 966	/* is_error is the last field in struct ptrace_syscall_info.exit */
 967	return offsetofend(struct ptrace_syscall_info, exit.is_error);
 968}
 969
 970static int
 971ptrace_get_syscall_info(struct task_struct *child, unsigned long user_size,
 972			void __user *datavp)
 973{
 974	struct pt_regs *regs = task_pt_regs(child);
 975	struct ptrace_syscall_info info = {
 976		.op = PTRACE_SYSCALL_INFO_NONE,
 977		.arch = syscall_get_arch(child),
 978		.instruction_pointer = instruction_pointer(regs),
 979		.stack_pointer = user_stack_pointer(regs),
 980	};
 981	unsigned long actual_size = offsetof(struct ptrace_syscall_info, entry);
 982	unsigned long write_size;
 983
 984	/*
 985	 * This does not need lock_task_sighand() to access
 986	 * child->last_siginfo because ptrace_freeze_traced()
 987	 * called earlier by ptrace_check_attach() ensures that
 988	 * the tracee cannot go away and clear its last_siginfo.
 989	 */
 990	switch (child->last_siginfo ? child->last_siginfo->si_code : 0) {
 991	case SIGTRAP | 0x80:
 992		switch (child->ptrace_message) {
 993		case PTRACE_EVENTMSG_SYSCALL_ENTRY:
 994			actual_size = ptrace_get_syscall_info_entry(child, regs,
 995								    &info);
 996			break;
 997		case PTRACE_EVENTMSG_SYSCALL_EXIT:
 998			actual_size = ptrace_get_syscall_info_exit(child, regs,
 999								   &info);
1000			break;
1001		}
1002		break;
1003	case SIGTRAP | (PTRACE_EVENT_SECCOMP << 8):
1004		actual_size = ptrace_get_syscall_info_seccomp(child, regs,
1005							      &info);
1006		break;
1007	}
1008
1009	write_size = min(actual_size, user_size);
1010	return copy_to_user(datavp, &info, write_size) ? -EFAULT : actual_size;
1011}
1012#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
1013
1014int ptrace_request(struct task_struct *child, long request,
1015		   unsigned long addr, unsigned long data)
1016{
1017	bool seized = child->ptrace & PT_SEIZED;
1018	int ret = -EIO;
1019	kernel_siginfo_t siginfo, *si;
1020	void __user *datavp = (void __user *) data;
1021	unsigned long __user *datalp = datavp;
1022	unsigned long flags;
1023
1024	switch (request) {
1025	case PTRACE_PEEKTEXT:
1026	case PTRACE_PEEKDATA:
1027		return generic_ptrace_peekdata(child, addr, data);
1028	case PTRACE_POKETEXT:
1029	case PTRACE_POKEDATA:
1030		return generic_ptrace_pokedata(child, addr, data);
1031
1032#ifdef PTRACE_OLDSETOPTIONS
1033	case PTRACE_OLDSETOPTIONS:
1034#endif
1035	case PTRACE_SETOPTIONS:
1036		ret = ptrace_setoptions(child, data);
1037		break;
1038	case PTRACE_GETEVENTMSG:
1039		ret = put_user(child->ptrace_message, datalp);
1040		break;
1041
1042	case PTRACE_PEEKSIGINFO:
1043		ret = ptrace_peek_siginfo(child, addr, data);
1044		break;
1045
1046	case PTRACE_GETSIGINFO:
1047		ret = ptrace_getsiginfo(child, &siginfo);
1048		if (!ret)
1049			ret = copy_siginfo_to_user(datavp, &siginfo);
1050		break;
1051
1052	case PTRACE_SETSIGINFO:
1053		ret = copy_siginfo_from_user(&siginfo, datavp);
1054		if (!ret)
1055			ret = ptrace_setsiginfo(child, &siginfo);
1056		break;
1057
1058	case PTRACE_GETSIGMASK: {
1059		sigset_t *mask;
1060
1061		if (addr != sizeof(sigset_t)) {
1062			ret = -EINVAL;
1063			break;
1064		}
1065
1066		if (test_tsk_restore_sigmask(child))
1067			mask = &child->saved_sigmask;
1068		else
1069			mask = &child->blocked;
1070
1071		if (copy_to_user(datavp, mask, sizeof(sigset_t)))
1072			ret = -EFAULT;
1073		else
1074			ret = 0;
1075
1076		break;
1077	}
1078
1079	case PTRACE_SETSIGMASK: {
1080		sigset_t new_set;
1081
1082		if (addr != sizeof(sigset_t)) {
1083			ret = -EINVAL;
1084			break;
1085		}
1086
1087		if (copy_from_user(&new_set, datavp, sizeof(sigset_t))) {
1088			ret = -EFAULT;
1089			break;
1090		}
1091
1092		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
1093
1094		/*
1095		 * Every thread does recalc_sigpending() after resume, so
1096		 * retarget_shared_pending() and recalc_sigpending() are not
1097		 * called here.
1098		 */
1099		spin_lock_irq(&child->sighand->siglock);
1100		child->blocked = new_set;
1101		spin_unlock_irq(&child->sighand->siglock);
1102
1103		clear_tsk_restore_sigmask(child);
1104
1105		ret = 0;
1106		break;
1107	}
1108
1109	case PTRACE_INTERRUPT:
1110		/*
1111		 * Stop tracee without any side-effect on signal or job
1112		 * control.  At least one trap is guaranteed to happen
1113		 * after this request.  If @child is already trapped, the
1114		 * current trap is not disturbed and another trap will
1115		 * happen after the current trap is ended with PTRACE_CONT.
1116		 *
1117		 * The actual trap might not be PTRACE_EVENT_STOP trap but
1118		 * the pending condition is cleared regardless.
1119		 */
1120		if (unlikely(!seized || !lock_task_sighand(child, &flags)))
1121			break;
1122
1123		/*
1124		 * INTERRUPT doesn't disturb existing trap sans one
1125		 * exception.  If ptracer issued LISTEN for the current
1126		 * STOP, this INTERRUPT should clear LISTEN and re-trap
1127		 * tracee into STOP.
1128		 */
1129		if (likely(task_set_jobctl_pending(child, JOBCTL_TRAP_STOP)))
1130			ptrace_signal_wake_up(child, child->jobctl & JOBCTL_LISTENING);
1131
1132		unlock_task_sighand(child, &flags);
1133		ret = 0;
1134		break;
1135
1136	case PTRACE_LISTEN:
1137		/*
1138		 * Listen for events.  Tracee must be in STOP.  It's not
1139		 * resumed per-se but is not considered to be in TRACED by
1140		 * wait(2) or ptrace(2).  If an async event (e.g. group
1141		 * stop state change) happens, tracee will enter STOP trap
1142		 * again.  Alternatively, ptracer can issue INTERRUPT to
1143		 * finish listening and re-trap tracee into STOP.
1144		 */
1145		if (unlikely(!seized || !lock_task_sighand(child, &flags)))
1146			break;
1147
1148		si = child->last_siginfo;
1149		if (likely(si && (si->si_code >> 8) == PTRACE_EVENT_STOP)) {
1150			child->jobctl |= JOBCTL_LISTENING;
1151			/*
1152			 * If NOTIFY is set, it means event happened between
1153			 * start of this trap and now.  Trigger re-trap.
1154			 */
1155			if (child->jobctl & JOBCTL_TRAP_NOTIFY)
1156				ptrace_signal_wake_up(child, true);
1157			ret = 0;
1158		}
1159		unlock_task_sighand(child, &flags);
1160		break;
1161
1162	case PTRACE_DETACH:	 /* detach a process that was attached. */
1163		ret = ptrace_detach(child, data);
1164		break;
1165
1166#ifdef CONFIG_BINFMT_ELF_FDPIC
1167	case PTRACE_GETFDPIC: {
1168		struct mm_struct *mm = get_task_mm(child);
1169		unsigned long tmp = 0;
1170
1171		ret = -ESRCH;
1172		if (!mm)
1173			break;
1174
1175		switch (addr) {
1176		case PTRACE_GETFDPIC_EXEC:
1177			tmp = mm->context.exec_fdpic_loadmap;
1178			break;
1179		case PTRACE_GETFDPIC_INTERP:
1180			tmp = mm->context.interp_fdpic_loadmap;
1181			break;
1182		default:
1183			break;
1184		}
1185		mmput(mm);
1186
1187		ret = put_user(tmp, datalp);
1188		break;
1189	}
1190#endif
1191
1192	case PTRACE_SINGLESTEP:
1193#ifdef PTRACE_SINGLEBLOCK
1194	case PTRACE_SINGLEBLOCK:
1195#endif
1196#ifdef PTRACE_SYSEMU
1197	case PTRACE_SYSEMU:
1198	case PTRACE_SYSEMU_SINGLESTEP:
1199#endif
1200	case PTRACE_SYSCALL:
1201	case PTRACE_CONT:
1202		return ptrace_resume(child, request, data);
1203
1204	case PTRACE_KILL:
1205		send_sig_info(SIGKILL, SEND_SIG_NOINFO, child);
1206		return 0;
1207
1208#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
1209	case PTRACE_GETREGSET:
1210	case PTRACE_SETREGSET: {
1211		struct iovec kiov;
1212		struct iovec __user *uiov = datavp;
1213
1214		if (!access_ok(uiov, sizeof(*uiov)))
1215			return -EFAULT;
1216
1217		if (__get_user(kiov.iov_base, &uiov->iov_base) ||
1218		    __get_user(kiov.iov_len, &uiov->iov_len))
1219			return -EFAULT;
1220
1221		ret = ptrace_regset(child, request, addr, &kiov);
1222		if (!ret)
1223			ret = __put_user(kiov.iov_len, &uiov->iov_len);
1224		break;
1225	}
1226
1227	case PTRACE_GET_SYSCALL_INFO:
1228		ret = ptrace_get_syscall_info(child, addr, datavp);
1229		break;
1230#endif
1231
1232	case PTRACE_SECCOMP_GET_FILTER:
1233		ret = seccomp_get_filter(child, addr, datavp);
1234		break;
1235
1236	case PTRACE_SECCOMP_GET_METADATA:
1237		ret = seccomp_get_metadata(child, addr, datavp);
1238		break;
1239
1240#ifdef CONFIG_RSEQ
1241	case PTRACE_GET_RSEQ_CONFIGURATION:
1242		ret = ptrace_get_rseq_configuration(child, addr, datavp);
1243		break;
1244#endif
1245
1246	case PTRACE_SET_SYSCALL_USER_DISPATCH_CONFIG:
1247		ret = syscall_user_dispatch_set_config(child, addr, datavp);
1248		break;
1249
1250	case PTRACE_GET_SYSCALL_USER_DISPATCH_CONFIG:
1251		ret = syscall_user_dispatch_get_config(child, addr, datavp);
1252		break;
1253
1254	default:
1255		break;
1256	}
1257
1258	return ret;
1259}
1260
1261SYSCALL_DEFINE4(ptrace, long, request, long, pid, unsigned long, addr,
1262		unsigned long, data)
1263{
1264	struct task_struct *child;
1265	long ret;
1266
1267	if (request == PTRACE_TRACEME) {
1268		ret = ptrace_traceme();
1269		goto out;
1270	}
1271
1272	child = find_get_task_by_vpid(pid);
1273	if (!child) {
1274		ret = -ESRCH;
1275		goto out;
1276	}
1277
1278	if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1279		ret = ptrace_attach(child, request, addr, data);
1280		goto out_put_task_struct;
1281	}
1282
1283	ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1284				  request == PTRACE_INTERRUPT);
1285	if (ret < 0)
1286		goto out_put_task_struct;
1287
1288	ret = arch_ptrace(child, request, addr, data);
1289	if (ret || request != PTRACE_DETACH)
1290		ptrace_unfreeze_traced(child);
1291
1292 out_put_task_struct:
1293	put_task_struct(child);
1294 out:
1295	return ret;
1296}
1297
1298int generic_ptrace_peekdata(struct task_struct *tsk, unsigned long addr,
1299			    unsigned long data)
1300{
1301	unsigned long tmp;
1302	int copied;
1303
1304	copied = ptrace_access_vm(tsk, addr, &tmp, sizeof(tmp), FOLL_FORCE);
1305	if (copied != sizeof(tmp))
1306		return -EIO;
1307	return put_user(tmp, (unsigned long __user *)data);
1308}
1309
1310int generic_ptrace_pokedata(struct task_struct *tsk, unsigned long addr,
1311			    unsigned long data)
1312{
1313	int copied;
1314
1315	copied = ptrace_access_vm(tsk, addr, &data, sizeof(data),
1316			FOLL_FORCE | FOLL_WRITE);
1317	return (copied == sizeof(data)) ? 0 : -EIO;
1318}
1319
1320#if defined CONFIG_COMPAT
1321
1322int compat_ptrace_request(struct task_struct *child, compat_long_t request,
1323			  compat_ulong_t addr, compat_ulong_t data)
1324{
1325	compat_ulong_t __user *datap = compat_ptr(data);
1326	compat_ulong_t word;
1327	kernel_siginfo_t siginfo;
1328	int ret;
1329
1330	switch (request) {
1331	case PTRACE_PEEKTEXT:
1332	case PTRACE_PEEKDATA:
1333		ret = ptrace_access_vm(child, addr, &word, sizeof(word),
1334				FOLL_FORCE);
1335		if (ret != sizeof(word))
1336			ret = -EIO;
1337		else
1338			ret = put_user(word, datap);
1339		break;
1340
1341	case PTRACE_POKETEXT:
1342	case PTRACE_POKEDATA:
1343		ret = ptrace_access_vm(child, addr, &data, sizeof(data),
1344				FOLL_FORCE | FOLL_WRITE);
1345		ret = (ret != sizeof(data) ? -EIO : 0);
1346		break;
1347
1348	case PTRACE_GETEVENTMSG:
1349		ret = put_user((compat_ulong_t) child->ptrace_message, datap);
1350		break;
1351
1352	case PTRACE_GETSIGINFO:
1353		ret = ptrace_getsiginfo(child, &siginfo);
1354		if (!ret)
1355			ret = copy_siginfo_to_user32(
1356				(struct compat_siginfo __user *) datap,
1357				&siginfo);
1358		break;
1359
1360	case PTRACE_SETSIGINFO:
1361		ret = copy_siginfo_from_user32(
1362			&siginfo, (struct compat_siginfo __user *) datap);
1363		if (!ret)
1364			ret = ptrace_setsiginfo(child, &siginfo);
1365		break;
1366#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
1367	case PTRACE_GETREGSET:
1368	case PTRACE_SETREGSET:
1369	{
1370		struct iovec kiov;
1371		struct compat_iovec __user *uiov =
1372			(struct compat_iovec __user *) datap;
1373		compat_uptr_t ptr;
1374		compat_size_t len;
1375
1376		if (!access_ok(uiov, sizeof(*uiov)))
1377			return -EFAULT;
1378
1379		if (__get_user(ptr, &uiov->iov_base) ||
1380		    __get_user(len, &uiov->iov_len))
1381			return -EFAULT;
1382
1383		kiov.iov_base = compat_ptr(ptr);
1384		kiov.iov_len = len;
1385
1386		ret = ptrace_regset(child, request, addr, &kiov);
1387		if (!ret)
1388			ret = __put_user(kiov.iov_len, &uiov->iov_len);
1389		break;
1390	}
1391#endif
1392
1393	default:
1394		ret = ptrace_request(child, request, addr, data);
1395	}
1396
1397	return ret;
1398}
1399
1400COMPAT_SYSCALL_DEFINE4(ptrace, compat_long_t, request, compat_long_t, pid,
1401		       compat_long_t, addr, compat_long_t, data)
1402{
1403	struct task_struct *child;
1404	long ret;
1405
1406	if (request == PTRACE_TRACEME) {
1407		ret = ptrace_traceme();
1408		goto out;
1409	}
1410
1411	child = find_get_task_by_vpid(pid);
1412	if (!child) {
1413		ret = -ESRCH;
1414		goto out;
1415	}
1416
1417	if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1418		ret = ptrace_attach(child, request, addr, data);
1419		goto out_put_task_struct;
1420	}
1421
1422	ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1423				  request == PTRACE_INTERRUPT);
1424	if (!ret) {
1425		ret = compat_arch_ptrace(child, request, addr, data);
1426		if (ret || request != PTRACE_DETACH)
1427			ptrace_unfreeze_traced(child);
1428	}
1429
1430 out_put_task_struct:
1431	put_task_struct(child);
1432 out:
1433	return ret;
1434}
1435#endif	/* CONFIG_COMPAT */