xref: /linux/kernel/ptrace.c (revision 6b1c66c9cca99bf00386481c7b2aa7394c26d8b8)
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/exec_state.h>
17 #include <linux/sched/task.h>
18 #include <linux/errno.h>
19 #include <linux/mm.h>
20 #include <linux/highmem.h>
21 #include <linux/pagemap.h>
22 #include <linux/ptrace.h>
23 #include <linux/security.h>
24 #include <linux/signal.h>
25 #include <linux/uio.h>
26 #include <linux/audit.h>
27 #include <linux/pid_namespace.h>
28 #include <linux/syscalls.h>
29 #include <linux/uaccess.h>
30 #include <linux/regset.h>
31 #include <linux/hw_breakpoint.h>
32 #include <linux/cn_proc.h>
33 #include <linux/compat.h>
34 #include <linux/sched/signal.h>
35 #include <linux/minmax.h>
36 #include <linux/syscall_user_dispatch.h>
37 
38 #include <asm/syscall.h>	/* for syscall_get_* */
39 
40 /**
41  * ptracer_access_allowed - may current peek/poke @tsk's address space?
42  * @tsk: tracee
43  *
44  * Per-access check used by ptrace_access_vm() and architecture-specific
45  * tag/register accessors.  Returns true iff current is the registered
46  * ptracer of @tsk and either @tsk is owner-dumpable or current holds
47  * CAP_SYS_PTRACE in @tsk's exec namespace.  Lighter than
48  * __ptrace_may_access(): it re-validates only dumpability and
49  * capability on every access, without re-running LSM hooks or
50  * cred_cap_issubset() checks performed at attach time.
51  */
52 bool ptracer_access_allowed(struct task_struct *tsk)
53 {
54 	const struct task_exec_state *es;
55 
56 	guard(rcu)();
57 	if (ptrace_parent(tsk) != current)
58 		return false;
59 	es = task_exec_state_rcu(tsk);
60 	return READ_ONCE(es->dumpable) == TASK_DUMPABLE_OWNER ||
61 	       ptracer_capable(tsk, es->user_ns);
62 }
63 
64 /*
65  * Access another process' address space via ptrace.
66  * Source/target buffer must be kernel space,
67  * Do not walk the page table directly, use get_user_pages
68  */
69 int ptrace_access_vm(struct task_struct *tsk, unsigned long addr,
70 		     void *buf, int len, unsigned int gup_flags)
71 {
72 	struct mm_struct *mm;
73 	int ret = 0;
74 
75 	mm = get_task_mm(tsk);
76 	if (!mm)
77 		return 0;
78 
79 	if (ptracer_access_allowed(tsk))
80 		ret = access_remote_vm(mm, addr, buf, len, gup_flags);
81 	mmput(mm);
82 
83 	return ret;
84 }
85 
86 
87 void __ptrace_link(struct task_struct *child, struct task_struct *new_parent,
88 		   const struct cred *ptracer_cred)
89 {
90 	BUG_ON(!list_empty(&child->ptrace_entry));
91 	list_add(&child->ptrace_entry, &new_parent->ptraced);
92 	child->parent = new_parent;
93 	child->ptracer_cred = get_cred(ptracer_cred);
94 }
95 
96 /*
97  * ptrace a task: make the debugger its new parent and
98  * move it to the ptrace list.
99  *
100  * Must be called with the tasklist lock write-held.
101  */
102 static void ptrace_link(struct task_struct *child, struct task_struct *new_parent)
103 {
104 	__ptrace_link(child, new_parent, current_cred());
105 }
106 
107 /**
108  * __ptrace_unlink - unlink ptracee and restore its execution state
109  * @child: ptracee to be unlinked
110  *
111  * Remove @child from the ptrace list, move it back to the original parent,
112  * and restore the execution state so that it conforms to the group stop
113  * state.
114  *
115  * Unlinking can happen via two paths - explicit PTRACE_DETACH or ptracer
116  * exiting.  For PTRACE_DETACH, unless the ptracee has been killed between
117  * ptrace_check_attach() and here, it's guaranteed to be in TASK_TRACED.
118  * If the ptracer is exiting, the ptracee can be in any state.
119  *
120  * After detach, the ptracee should be in a state which conforms to the
121  * group stop.  If the group is stopped or in the process of stopping, the
122  * ptracee should be put into TASK_STOPPED; otherwise, it should be woken
123  * up from TASK_TRACED.
124  *
125  * If the ptracee is in TASK_TRACED and needs to be moved to TASK_STOPPED,
126  * it goes through TRACED -> RUNNING -> STOPPED transition which is similar
127  * to but in the opposite direction of what happens while attaching to a
128  * stopped task.  However, in this direction, the intermediate RUNNING
129  * state is not hidden even from the current ptracer and if it immediately
130  * re-attaches and performs a WNOHANG wait(2), it may fail.
131  *
132  * CONTEXT:
133  * write_lock_irq(tasklist_lock)
134  */
135 void __ptrace_unlink(struct task_struct *child)
136 {
137 	const struct cred *old_cred;
138 	BUG_ON(!child->ptrace);
139 
140 	clear_task_syscall_work(child, SYSCALL_TRACE);
141 #if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU)
142 	clear_task_syscall_work(child, SYSCALL_EMU);
143 #endif
144 
145 	child->parent = child->real_parent;
146 	list_del_init(&child->ptrace_entry);
147 	old_cred = child->ptracer_cred;
148 	child->ptracer_cred = NULL;
149 	put_cred(old_cred);
150 
151 	spin_lock(&child->sighand->siglock);
152 	child->ptrace = 0;
153 	/*
154 	 * Clear all pending traps and TRAPPING.  TRAPPING should be
155 	 * cleared regardless of JOBCTL_STOP_PENDING.  Do it explicitly.
156 	 */
157 	task_clear_jobctl_pending(child, JOBCTL_TRAP_MASK);
158 	task_clear_jobctl_trapping(child);
159 
160 	/*
161 	 * Reinstate JOBCTL_STOP_PENDING if group stop is in effect and
162 	 * @child isn't dead.
163 	 */
164 	if (!(child->flags & PF_EXITING) &&
165 	    (child->signal->flags & SIGNAL_STOP_STOPPED ||
166 	     child->signal->group_stop_count))
167 		child->jobctl |= JOBCTL_STOP_PENDING;
168 
169 	/*
170 	 * If transition to TASK_STOPPED is pending or in TASK_TRACED, kick
171 	 * @child in the butt.  Note that @resume should be used iff @child
172 	 * is in TASK_TRACED; otherwise, we might unduly disrupt
173 	 * TASK_KILLABLE sleeps.
174 	 */
175 	if (child->jobctl & JOBCTL_STOP_PENDING || task_is_traced(child))
176 		ptrace_signal_wake_up(child, true);
177 
178 	spin_unlock(&child->sighand->siglock);
179 }
180 
181 static bool looks_like_a_spurious_pid(struct task_struct *task)
182 {
183 	if (task->exit_code != ((PTRACE_EVENT_EXEC << 8) | SIGTRAP))
184 		return false;
185 
186 	if (task_pid_vnr(task) == task->ptrace_message)
187 		return false;
188 	/*
189 	 * The tracee changed its pid but the PTRACE_EVENT_EXEC event
190 	 * was not wait()'ed, most probably debugger targets the old
191 	 * leader which was destroyed in de_thread().
192 	 */
193 	return true;
194 }
195 
196 /*
197  * Ensure that nothing can wake it up, even SIGKILL
198  *
199  * A task is switched to this state while a ptrace operation is in progress;
200  * such that the ptrace operation is uninterruptible.
201  */
202 static bool ptrace_freeze_traced(struct task_struct *task)
203 {
204 	bool ret = false;
205 
206 	/* Lockless, nobody but us can set this flag */
207 	if (task->jobctl & JOBCTL_LISTENING)
208 		return ret;
209 
210 	spin_lock_irq(&task->sighand->siglock);
211 	if (task_is_traced(task) && !looks_like_a_spurious_pid(task) &&
212 	    !__fatal_signal_pending(task)) {
213 		task->jobctl |= JOBCTL_PTRACE_FROZEN;
214 		ret = true;
215 	}
216 	spin_unlock_irq(&task->sighand->siglock);
217 
218 	return ret;
219 }
220 
221 static void ptrace_unfreeze_traced(struct task_struct *task)
222 {
223 	unsigned long flags;
224 
225 	/*
226 	 * The child may be awake and may have cleared
227 	 * JOBCTL_PTRACE_FROZEN (see ptrace_resume).  The child will
228 	 * not set JOBCTL_PTRACE_FROZEN or enter __TASK_TRACED anew.
229 	 */
230 	if (lock_task_sighand(task, &flags)) {
231 		task->jobctl &= ~JOBCTL_PTRACE_FROZEN;
232 		if (__fatal_signal_pending(task)) {
233 			task->jobctl &= ~JOBCTL_TRACED;
234 			wake_up_state(task, __TASK_TRACED);
235 		}
236 		unlock_task_sighand(task, &flags);
237 	}
238 }
239 
240 /**
241  * ptrace_check_attach - check whether ptracee is ready for ptrace operation
242  * @child: ptracee to check for
243  * @ignore_state: don't check whether @child is currently %TASK_TRACED
244  *
245  * Check whether @child is being ptraced by %current and ready for further
246  * ptrace operations.  If @ignore_state is %false, @child also should be in
247  * %TASK_TRACED state and on return the child is guaranteed to be traced
248  * and not executing.  If @ignore_state is %true, @child can be in any
249  * state.
250  *
251  * CONTEXT:
252  * Grabs and releases tasklist_lock and @child->sighand->siglock.
253  *
254  * RETURNS:
255  * 0 on success, -ESRCH if %child is not ready.
256  */
257 static int ptrace_check_attach(struct task_struct *child, bool ignore_state)
258 {
259 	int ret = -ESRCH;
260 
261 	/*
262 	 * We take the read lock around doing both checks to close a
263 	 * possible race where someone else was tracing our child and
264 	 * detached between these two checks.  After this locked check,
265 	 * we are sure that this is our traced child and that can only
266 	 * be changed by us so it's not changing right after this.
267 	 */
268 	read_lock(&tasklist_lock);
269 	if (child->ptrace && child->parent == current) {
270 		/*
271 		 * child->sighand can't be NULL, release_task()
272 		 * does ptrace_unlink() before __exit_signal().
273 		 */
274 		if (ignore_state || ptrace_freeze_traced(child))
275 			ret = 0;
276 	}
277 	read_unlock(&tasklist_lock);
278 
279 	if (!ret && !ignore_state &&
280 	    WARN_ON_ONCE(!wait_task_inactive(child, __TASK_TRACED|TASK_FROZEN)))
281 		ret = -ESRCH;
282 
283 	return ret;
284 }
285 
286 static bool ptrace_has_cap(struct user_namespace *ns, unsigned int mode)
287 {
288 	if (mode & PTRACE_MODE_NOAUDIT)
289 		return ns_capable_noaudit(ns, CAP_SYS_PTRACE);
290 	return ns_capable(ns, CAP_SYS_PTRACE);
291 }
292 
293 static bool task_still_dumpable(struct task_struct *task, unsigned int mode)
294 {
295 	const struct task_exec_state *exec_state;
296 
297 	guard(rcu)();
298 	exec_state = task_exec_state_rcu(task);
299 	if (READ_ONCE(exec_state->dumpable) == TASK_DUMPABLE_OWNER)
300 		return true;
301 	return ptrace_has_cap(exec_state->user_ns, mode);
302 }
303 
304 /* Returns 0 on success, -errno on denial. */
305 static int __ptrace_may_access(struct task_struct *task, unsigned int mode)
306 {
307 	const struct cred *cred = current_cred(), *tcred;
308 	kuid_t caller_uid;
309 	kgid_t caller_gid;
310 
311 	if (!(mode & PTRACE_MODE_FSCREDS) == !(mode & PTRACE_MODE_REALCREDS)) {
312 		WARN(1, "denying ptrace access check without PTRACE_MODE_*CREDS\n");
313 		return -EPERM;
314 	}
315 
316 	/* May we inspect the given task?
317 	 * This check is used both for attaching with ptrace
318 	 * and for allowing access to sensitive information in /proc.
319 	 *
320 	 * ptrace_attach denies several cases that /proc allows
321 	 * because setting up the necessary parent/child relationship
322 	 * or halting the specified task is impossible.
323 	 */
324 
325 	/* Don't let security modules deny introspection */
326 	if (same_thread_group(task, current))
327 		return 0;
328 	rcu_read_lock();
329 	if (mode & PTRACE_MODE_FSCREDS) {
330 		caller_uid = cred->fsuid;
331 		caller_gid = cred->fsgid;
332 	} else {
333 		/*
334 		 * Using the euid would make more sense here, but something
335 		 * in userland might rely on the old behavior, and this
336 		 * shouldn't be a security problem since
337 		 * PTRACE_MODE_REALCREDS implies that the caller explicitly
338 		 * used a syscall that requests access to another process
339 		 * (and not a filesystem syscall to procfs).
340 		 */
341 		caller_uid = cred->uid;
342 		caller_gid = cred->gid;
343 	}
344 	tcred = __task_cred(task);
345 	if (uid_eq(caller_uid, tcred->euid) &&
346 	    uid_eq(caller_uid, tcred->suid) &&
347 	    uid_eq(caller_uid, tcred->uid)  &&
348 	    gid_eq(caller_gid, tcred->egid) &&
349 	    gid_eq(caller_gid, tcred->sgid) &&
350 	    gid_eq(caller_gid, tcred->gid))
351 		goto ok;
352 	if (ptrace_has_cap(tcred->user_ns, mode))
353 		goto ok;
354 	rcu_read_unlock();
355 	return -EPERM;
356 ok:
357 	rcu_read_unlock();
358 	/*
359 	 * If a task drops privileges and becomes nondumpable (through a syscall
360 	 * like setresuid()) while we are trying to access it, we must ensure
361 	 * that the dumpability is read after the credentials; otherwise,
362 	 * we may be able to attach to a task that we shouldn't be able to
363 	 * attach to (as if the task had dropped privileges without becoming
364 	 * nondumpable).
365 	 * Pairs with a write barrier in commit_creds().
366 	 */
367 	smp_rmb();
368 	if (!task_still_dumpable(task, mode))
369 		return -EPERM;
370 
371 	return security_ptrace_access_check(task, mode);
372 }
373 
374 bool ptrace_may_access(struct task_struct *task, unsigned int mode)
375 {
376 	int err;
377 	task_lock(task);
378 	err = __ptrace_may_access(task, mode);
379 	task_unlock(task);
380 	return !err;
381 }
382 
383 static int check_ptrace_options(unsigned long data)
384 {
385 	if (data & ~(unsigned long)PTRACE_O_MASK)
386 		return -EINVAL;
387 
388 	if (unlikely(data & PTRACE_O_SUSPEND_SECCOMP)) {
389 		if (!IS_ENABLED(CONFIG_CHECKPOINT_RESTORE) ||
390 		    !IS_ENABLED(CONFIG_SECCOMP))
391 			return -EINVAL;
392 
393 		if (!capable(CAP_SYS_ADMIN))
394 			return -EPERM;
395 
396 		if (seccomp_mode(&current->seccomp) != SECCOMP_MODE_DISABLED ||
397 		    current->ptrace & PT_SUSPEND_SECCOMP)
398 			return -EPERM;
399 	}
400 	return 0;
401 }
402 
403 static inline void ptrace_set_stopped(struct task_struct *task, bool seize)
404 {
405 	guard(spinlock)(&task->sighand->siglock);
406 
407 	/* SEIZE doesn't trap tracee on attach */
408 	if (!seize)
409 		send_signal_locked(SIGSTOP, SEND_SIG_PRIV, task, PIDTYPE_PID);
410 	/*
411 	 * If the task is already STOPPED, set JOBCTL_TRAP_STOP and
412 	 * TRAPPING, and kick it so that it transits to TRACED.  TRAPPING
413 	 * will be cleared if the child completes the transition or any
414 	 * event which clears the group stop states happens.  We'll wait
415 	 * for the transition to complete before returning from this
416 	 * function.
417 	 *
418 	 * This hides STOPPED -> RUNNING -> TRACED transition from the
419 	 * attaching thread but a different thread in the same group can
420 	 * still observe the transient RUNNING state.  IOW, if another
421 	 * thread's WNOHANG wait(2) on the stopped tracee races against
422 	 * ATTACH, the wait(2) may fail due to the transient RUNNING.
423 	 *
424 	 * The following task_is_stopped() test is safe as both transitions
425 	 * in and out of STOPPED are protected by siglock.
426 	 */
427 	if (task_is_stopped(task) &&
428 	    task_set_jobctl_pending(task, JOBCTL_TRAP_STOP | JOBCTL_TRAPPING)) {
429 		task->jobctl &= ~JOBCTL_STOPPED;
430 		signal_wake_up_state(task, __TASK_STOPPED);
431 	}
432 }
433 
434 static int ptrace_attach(struct task_struct *task, long request,
435 			 unsigned long addr,
436 			 unsigned long flags)
437 {
438 	bool seize = (request == PTRACE_SEIZE);
439 	int retval;
440 
441 	if (seize) {
442 		if (addr != 0)
443 			return -EIO;
444 		/*
445 		 * This duplicates the check in check_ptrace_options() because
446 		 * ptrace_attach() and ptrace_setoptions() have historically
447 		 * used different error codes for unknown ptrace options.
448 		 */
449 		if (flags & ~(unsigned long)PTRACE_O_MASK)
450 			return -EIO;
451 
452 		retval = check_ptrace_options(flags);
453 		if (retval)
454 			return retval;
455 		flags = PT_PTRACED | PT_SEIZED | (flags << PT_OPT_FLAG_SHIFT);
456 	} else {
457 		flags = PT_PTRACED;
458 	}
459 
460 	audit_ptrace(task);
461 
462 	if (unlikely(task->flags & PF_KTHREAD))
463 		return -EPERM;
464 	if (same_thread_group(task, current))
465 		return -EPERM;
466 
467 	/*
468 	 * Protect exec's credential calculations against our interference;
469 	 * SUID, SGID and LSM creds get determined differently
470 	 * under ptrace.
471 	 */
472 	scoped_cond_guard (mutex_intr, return -ERESTARTNOINTR,
473 			   &task->signal->cred_guard_mutex) {
474 
475 		scoped_guard (task_lock, task) {
476 			retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH_REALCREDS);
477 			if (retval)
478 				return retval;
479 		}
480 
481 		scoped_guard (write_lock_irq, &tasklist_lock) {
482 			if (unlikely(task->exit_state))
483 				return -EPERM;
484 			if (task->ptrace)
485 				return -EPERM;
486 
487 			task->ptrace = flags;
488 			ptrace_link(task, current);
489 			ptrace_set_stopped(task, seize);
490 		}
491 	}
492 
493 	/*
494 	 * We do not bother to change retval or clear JOBCTL_TRAPPING
495 	 * if wait_on_bit() was interrupted by SIGKILL. The tracer will
496 	 * not return to user-mode, it will exit and clear this bit in
497 	 * __ptrace_unlink() if it wasn't already cleared by the tracee;
498 	 * and until then nobody can ptrace this task.
499 	 */
500 	wait_on_bit(&task->jobctl, JOBCTL_TRAPPING_BIT, TASK_KILLABLE);
501 	proc_ptrace_connector(task, PTRACE_ATTACH);
502 
503 	return 0;
504 }
505 
506 /**
507  * ptrace_traceme  --  helper for PTRACE_TRACEME
508  *
509  * Performs checks and sets PT_PTRACED.
510  * Should be used by all ptrace implementations for PTRACE_TRACEME.
511  */
512 static int ptrace_traceme(void)
513 {
514 	int ret = -EPERM;
515 
516 	write_lock_irq(&tasklist_lock);
517 	/* Are we already being traced? */
518 	if (!current->ptrace) {
519 		ret = security_ptrace_traceme(current->parent);
520 		/*
521 		 * Check PF_EXITING to ensure ->real_parent has not passed
522 		 * exit_ptrace(). Otherwise we don't report the error but
523 		 * pretend ->real_parent untraces us right after return.
524 		 */
525 		if (!ret && !(current->real_parent->flags & PF_EXITING)) {
526 			current->ptrace = PT_PTRACED;
527 			ptrace_link(current, current->real_parent);
528 		}
529 	}
530 	write_unlock_irq(&tasklist_lock);
531 
532 	return ret;
533 }
534 
535 /*
536  * Called with irqs disabled, returns true if childs should reap themselves.
537  */
538 static int ignoring_children(struct sighand_struct *sigh)
539 {
540 	int ret;
541 	spin_lock(&sigh->siglock);
542 	ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
543 	      (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
544 	spin_unlock(&sigh->siglock);
545 	return ret;
546 }
547 
548 /*
549  * Called with tasklist_lock held for writing.
550  * Unlink a traced task, and clean it up if it was a traced zombie.
551  * Return true if it needs to be reaped with release_task().
552  * (We can't call release_task() here because we already hold tasklist_lock.)
553  *
554  * If it's a zombie, our attachedness prevented normal parent notification
555  * or self-reaping.  Do notification now if it would have happened earlier.
556  * If it should reap itself, return true.
557  *
558  * If it's our own child, there is no notification to do. But if our normal
559  * children self-reap, then this child was prevented by ptrace and we must
560  * reap it now, in that case we must also wake up sub-threads sleeping in
561  * do_wait().
562  */
563 static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
564 {
565 	bool dead;
566 
567 	__ptrace_unlink(p);
568 
569 	if (p->exit_state != EXIT_ZOMBIE)
570 		return false;
571 
572 	dead = !thread_group_leader(p);
573 
574 	if (!dead && thread_group_empty(p)) {
575 		if (!same_thread_group(p->real_parent, tracer))
576 			dead = do_notify_parent(p, p->exit_signal);
577 		else if (ignoring_children(tracer->sighand) ||
578 			 p->signal->autoreap) {
579 			__wake_up_parent(p, tracer);
580 			dead = true;
581 		}
582 	}
583 	/* Mark it as in the process of being reaped. */
584 	if (dead)
585 		p->exit_state = EXIT_DEAD;
586 	return dead;
587 }
588 
589 static int ptrace_detach(struct task_struct *child, unsigned int data)
590 {
591 	if (!valid_signal(data))
592 		return -EIO;
593 
594 	/* Architecture-specific hardware disable .. */
595 	ptrace_disable(child);
596 
597 	write_lock_irq(&tasklist_lock);
598 	/*
599 	 * We rely on ptrace_freeze_traced(). It can't be killed and
600 	 * untraced by another thread, it can't be a zombie.
601 	 */
602 	WARN_ON(!child->ptrace || child->exit_state);
603 	/*
604 	 * tasklist_lock avoids the race with wait_task_stopped(), see
605 	 * the comment in ptrace_resume().
606 	 */
607 	child->exit_code = data;
608 	__ptrace_detach(current, child);
609 	write_unlock_irq(&tasklist_lock);
610 
611 	proc_ptrace_connector(child, PTRACE_DETACH);
612 
613 	return 0;
614 }
615 
616 /*
617  * Detach all tasks we were using ptrace on. Called with tasklist held
618  * for writing.
619  */
620 void exit_ptrace(struct task_struct *tracer, struct list_head *dead)
621 {
622 	struct task_struct *p, *n;
623 
624 	list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
625 		if (unlikely(p->ptrace & PT_EXITKILL))
626 			send_sig_info(SIGKILL, SEND_SIG_PRIV, p);
627 
628 		if (__ptrace_detach(tracer, p))
629 			list_add(&p->ptrace_entry, dead);
630 	}
631 }
632 
633 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
634 {
635 	int copied = 0;
636 
637 	while (len > 0) {
638 		char buf[128];
639 		int this_len, retval;
640 
641 		this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
642 		retval = ptrace_access_vm(tsk, src, buf, this_len, FOLL_FORCE);
643 
644 		if (!retval) {
645 			if (copied)
646 				break;
647 			return -EIO;
648 		}
649 		if (copy_to_user(dst, buf, retval))
650 			return -EFAULT;
651 		copied += retval;
652 		src += retval;
653 		dst += retval;
654 		len -= retval;
655 	}
656 	return copied;
657 }
658 
659 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
660 {
661 	int copied = 0;
662 
663 	while (len > 0) {
664 		char buf[128];
665 		int this_len, retval;
666 
667 		this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
668 		if (copy_from_user(buf, src, this_len))
669 			return -EFAULT;
670 		retval = ptrace_access_vm(tsk, dst, buf, this_len,
671 				FOLL_FORCE | FOLL_WRITE);
672 		if (!retval) {
673 			if (copied)
674 				break;
675 			return -EIO;
676 		}
677 		copied += retval;
678 		src += retval;
679 		dst += retval;
680 		len -= retval;
681 	}
682 	return copied;
683 }
684 
685 static int ptrace_setoptions(struct task_struct *child, unsigned long data)
686 {
687 	unsigned flags;
688 	int ret;
689 
690 	ret = check_ptrace_options(data);
691 	if (ret)
692 		return ret;
693 
694 	/* Avoid intermediate state when all opts are cleared */
695 	flags = child->ptrace;
696 	flags &= ~(PTRACE_O_MASK << PT_OPT_FLAG_SHIFT);
697 	flags |= (data << PT_OPT_FLAG_SHIFT);
698 	child->ptrace = flags;
699 
700 	return 0;
701 }
702 
703 static int ptrace_getsiginfo(struct task_struct *child, kernel_siginfo_t *info)
704 {
705 	unsigned long flags;
706 	int error = -ESRCH;
707 
708 	if (lock_task_sighand(child, &flags)) {
709 		error = -EINVAL;
710 		if (likely(child->last_siginfo != NULL)) {
711 			copy_siginfo(info, child->last_siginfo);
712 			error = 0;
713 		}
714 		unlock_task_sighand(child, &flags);
715 	}
716 	return error;
717 }
718 
719 static int ptrace_setsiginfo(struct task_struct *child, const kernel_siginfo_t *info)
720 {
721 	unsigned long flags;
722 	int error = -ESRCH;
723 
724 	if (lock_task_sighand(child, &flags)) {
725 		error = -EINVAL;
726 		if (likely(child->last_siginfo != NULL)) {
727 			copy_siginfo(child->last_siginfo, info);
728 			error = 0;
729 		}
730 		unlock_task_sighand(child, &flags);
731 	}
732 	return error;
733 }
734 
735 static int ptrace_peek_siginfo(struct task_struct *child,
736 				unsigned long addr,
737 				unsigned long data)
738 {
739 	struct ptrace_peeksiginfo_args arg;
740 	struct sigpending *pending;
741 	struct sigqueue *q;
742 	int ret, i;
743 
744 	ret = copy_from_user(&arg, (void __user *) addr,
745 				sizeof(struct ptrace_peeksiginfo_args));
746 	if (ret)
747 		return -EFAULT;
748 
749 	if (arg.flags & ~PTRACE_PEEKSIGINFO_SHARED)
750 		return -EINVAL; /* unknown flags */
751 
752 	if (arg.nr < 0)
753 		return -EINVAL;
754 
755 	/* Ensure arg.off fits in an unsigned long */
756 	if (arg.off > ULONG_MAX)
757 		return 0;
758 
759 	if (arg.flags & PTRACE_PEEKSIGINFO_SHARED)
760 		pending = &child->signal->shared_pending;
761 	else
762 		pending = &child->pending;
763 
764 	for (i = 0; i < arg.nr; ) {
765 		kernel_siginfo_t info;
766 		unsigned long off = arg.off + i;
767 		bool found = false;
768 
769 		spin_lock_irq(&child->sighand->siglock);
770 		list_for_each_entry(q, &pending->list, list) {
771 			if (!off--) {
772 				found = true;
773 				copy_siginfo(&info, &q->info);
774 				break;
775 			}
776 		}
777 		spin_unlock_irq(&child->sighand->siglock);
778 
779 		if (!found) /* beyond the end of the list */
780 			break;
781 
782 #ifdef CONFIG_COMPAT
783 		if (unlikely(in_compat_syscall())) {
784 			compat_siginfo_t __user *uinfo = compat_ptr(data);
785 
786 			if (copy_siginfo_to_user32(uinfo, &info)) {
787 				ret = -EFAULT;
788 				break;
789 			}
790 
791 		} else
792 #endif
793 		{
794 			siginfo_t __user *uinfo = (siginfo_t __user *) data;
795 
796 			if (copy_siginfo_to_user(uinfo, &info)) {
797 				ret = -EFAULT;
798 				break;
799 			}
800 		}
801 
802 		data += sizeof(siginfo_t);
803 		i++;
804 
805 		if (signal_pending(current))
806 			break;
807 
808 		cond_resched();
809 	}
810 
811 	if (i > 0)
812 		return i;
813 
814 	return ret;
815 }
816 
817 #ifdef CONFIG_RSEQ
818 static long ptrace_get_rseq_configuration(struct task_struct *task,
819 					  unsigned long size, void __user *data)
820 {
821 	struct ptrace_rseq_configuration conf = {
822 		.rseq_abi_pointer = (u64)(uintptr_t)task->rseq.usrptr,
823 		.rseq_abi_size = task->rseq.len,
824 		.signature = task->rseq.sig,
825 		.flags = 0,
826 	};
827 
828 	size = min_t(unsigned long, size, sizeof(conf));
829 	if (copy_to_user(data, &conf, size))
830 		return -EFAULT;
831 	return sizeof(conf);
832 }
833 #endif
834 
835 #define is_singlestep(request)		((request) == PTRACE_SINGLESTEP)
836 
837 #ifdef PTRACE_SINGLEBLOCK
838 #define is_singleblock(request)		((request) == PTRACE_SINGLEBLOCK)
839 #else
840 #define is_singleblock(request)		0
841 #endif
842 
843 #ifdef PTRACE_SYSEMU
844 #define is_sysemu_singlestep(request)	((request) == PTRACE_SYSEMU_SINGLESTEP)
845 #else
846 #define is_sysemu_singlestep(request)	0
847 #endif
848 
849 static int ptrace_resume(struct task_struct *child, long request,
850 			 unsigned long data)
851 {
852 	if (!valid_signal(data))
853 		return -EIO;
854 
855 	if (request == PTRACE_SYSCALL)
856 		set_task_syscall_work(child, SYSCALL_TRACE);
857 	else
858 		clear_task_syscall_work(child, SYSCALL_TRACE);
859 
860 #if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU)
861 	if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
862 		set_task_syscall_work(child, SYSCALL_EMU);
863 	else
864 		clear_task_syscall_work(child, SYSCALL_EMU);
865 #endif
866 
867 	if (is_singleblock(request)) {
868 		if (unlikely(!arch_has_block_step()))
869 			return -EIO;
870 		user_enable_block_step(child);
871 	} else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
872 		if (unlikely(!arch_has_single_step()))
873 			return -EIO;
874 		user_enable_single_step(child);
875 	} else {
876 		user_disable_single_step(child);
877 	}
878 
879 	/*
880 	 * Change ->exit_code and ->state under siglock to avoid the race
881 	 * with wait_task_stopped() in between; a non-zero ->exit_code will
882 	 * wrongly look like another report from tracee.
883 	 *
884 	 * Note that we need siglock even if ->exit_code == data and/or this
885 	 * status was not reported yet, the new status must not be cleared by
886 	 * wait_task_stopped() after resume.
887 	 */
888 	spin_lock_irq(&child->sighand->siglock);
889 	child->exit_code = data;
890 	child->jobctl &= ~JOBCTL_TRACED;
891 	wake_up_state(child, __TASK_TRACED);
892 	spin_unlock_irq(&child->sighand->siglock);
893 
894 	return 0;
895 }
896 
897 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
898 
899 static const struct user_regset *
900 find_regset(const struct user_regset_view *view, unsigned int type)
901 {
902 	const struct user_regset *regset;
903 	int n;
904 
905 	for (n = 0; n < view->n; ++n) {
906 		regset = view->regsets + n;
907 		if (regset->core_note_type == type)
908 			return regset;
909 	}
910 
911 	return NULL;
912 }
913 
914 static int ptrace_regset(struct task_struct *task, int req, unsigned int type,
915 			 struct iovec *kiov)
916 {
917 	const struct user_regset_view *view = task_user_regset_view(task);
918 	const struct user_regset *regset = find_regset(view, type);
919 	int regset_no;
920 
921 	if (!regset || (kiov->iov_len % regset->size) != 0)
922 		return -EINVAL;
923 
924 	regset_no = regset - view->regsets;
925 	kiov->iov_len = min(kiov->iov_len,
926 			    (__kernel_size_t) (regset->n * regset->size));
927 
928 	if (req == PTRACE_GETREGSET)
929 		return copy_regset_to_user(task, view, regset_no, 0,
930 					   kiov->iov_len, kiov->iov_base);
931 	else
932 		return copy_regset_from_user(task, view, regset_no, 0,
933 					     kiov->iov_len, kiov->iov_base);
934 }
935 
936 /*
937  * This is declared in linux/regset.h and defined in machine-dependent
938  * code.  We put the export here, near the primary machine-neutral use,
939  * to ensure no machine forgets it.
940  */
941 EXPORT_SYMBOL_GPL(task_user_regset_view);
942 
943 static unsigned long
944 ptrace_get_syscall_info_entry(struct task_struct *child, struct pt_regs *regs,
945 			      struct ptrace_syscall_info *info)
946 {
947 	unsigned long args[ARRAY_SIZE(info->entry.args)];
948 	int i;
949 
950 	info->entry.nr = syscall_get_nr(child, regs);
951 	syscall_get_arguments(child, regs, args);
952 	for (i = 0; i < ARRAY_SIZE(args); i++)
953 		info->entry.args[i] = args[i];
954 
955 	/* args is the last field in struct ptrace_syscall_info.entry */
956 	return offsetofend(struct ptrace_syscall_info, entry.args);
957 }
958 
959 static unsigned long
960 ptrace_get_syscall_info_seccomp(struct task_struct *child, struct pt_regs *regs,
961 				struct ptrace_syscall_info *info)
962 {
963 	/*
964 	 * As struct ptrace_syscall_info.entry is currently a subset
965 	 * of struct ptrace_syscall_info.seccomp, it makes sense to
966 	 * initialize that subset using ptrace_get_syscall_info_entry().
967 	 * This can be reconsidered in the future if these structures
968 	 * diverge significantly enough.
969 	 */
970 	ptrace_get_syscall_info_entry(child, regs, info);
971 	info->seccomp.ret_data = child->ptrace_message;
972 
973 	/*
974 	 * ret_data is the last non-reserved field
975 	 * in struct ptrace_syscall_info.seccomp
976 	 */
977 	return offsetofend(struct ptrace_syscall_info, seccomp.ret_data);
978 }
979 
980 static unsigned long
981 ptrace_get_syscall_info_exit(struct task_struct *child, struct pt_regs *regs,
982 			     struct ptrace_syscall_info *info)
983 {
984 	info->exit.rval = syscall_get_error(child, regs);
985 	info->exit.is_error = !!info->exit.rval;
986 	if (!info->exit.is_error)
987 		info->exit.rval = syscall_get_return_value(child, regs);
988 
989 	/* is_error is the last field in struct ptrace_syscall_info.exit */
990 	return offsetofend(struct ptrace_syscall_info, exit.is_error);
991 }
992 
993 static int
994 ptrace_get_syscall_info_op(struct task_struct *child)
995 {
996 	/*
997 	 * This does not need lock_task_sighand() to access
998 	 * child->last_siginfo because ptrace_freeze_traced()
999 	 * called earlier by ptrace_check_attach() ensures that
1000 	 * the tracee cannot go away and clear its last_siginfo.
1001 	 */
1002 	switch (child->last_siginfo ? child->last_siginfo->si_code : 0) {
1003 	case SIGTRAP | 0x80:
1004 		switch (child->ptrace_message) {
1005 		case PTRACE_EVENTMSG_SYSCALL_ENTRY:
1006 			return PTRACE_SYSCALL_INFO_ENTRY;
1007 		case PTRACE_EVENTMSG_SYSCALL_EXIT:
1008 			return PTRACE_SYSCALL_INFO_EXIT;
1009 		default:
1010 			return PTRACE_SYSCALL_INFO_NONE;
1011 		}
1012 	case SIGTRAP | (PTRACE_EVENT_SECCOMP << 8):
1013 		return PTRACE_SYSCALL_INFO_SECCOMP;
1014 	default:
1015 		return PTRACE_SYSCALL_INFO_NONE;
1016 	}
1017 }
1018 
1019 static int
1020 ptrace_get_syscall_info(struct task_struct *child, unsigned long user_size,
1021 			void __user *datavp)
1022 {
1023 	struct pt_regs *regs = task_pt_regs(child);
1024 	struct ptrace_syscall_info info = {
1025 		.op = ptrace_get_syscall_info_op(child),
1026 		.arch = syscall_get_arch(child),
1027 		.instruction_pointer = instruction_pointer(regs),
1028 		.stack_pointer = user_stack_pointer(regs),
1029 	};
1030 	unsigned long actual_size = offsetof(struct ptrace_syscall_info, entry);
1031 	unsigned long write_size;
1032 
1033 	switch (info.op) {
1034 	case PTRACE_SYSCALL_INFO_ENTRY:
1035 		actual_size = ptrace_get_syscall_info_entry(child, regs, &info);
1036 		break;
1037 	case PTRACE_SYSCALL_INFO_EXIT:
1038 		actual_size = ptrace_get_syscall_info_exit(child, regs, &info);
1039 		break;
1040 	case PTRACE_SYSCALL_INFO_SECCOMP:
1041 		actual_size = ptrace_get_syscall_info_seccomp(child, regs, &info);
1042 		break;
1043 	}
1044 
1045 	write_size = min(actual_size, user_size);
1046 	return copy_to_user(datavp, &info, write_size) ? -EFAULT : actual_size;
1047 }
1048 
1049 static int
1050 ptrace_set_syscall_info_entry(struct task_struct *child, struct pt_regs *regs,
1051 			      struct ptrace_syscall_info *info)
1052 {
1053 	unsigned long args[ARRAY_SIZE(info->entry.args)];
1054 	int nr = info->entry.nr;
1055 	int i;
1056 
1057 	/*
1058 	 * Check that the syscall number specified in info->entry.nr
1059 	 * is either a value of type "int" or a sign-extended value
1060 	 * of type "int".
1061 	 */
1062 	if (nr != info->entry.nr)
1063 		return -ERANGE;
1064 
1065 	for (i = 0; i < ARRAY_SIZE(args); i++) {
1066 		args[i] = info->entry.args[i];
1067 		/*
1068 		 * Check that the syscall argument specified in
1069 		 * info->entry.args[i] is either a value of type
1070 		 * "unsigned long" or a sign-extended value of type "long".
1071 		 */
1072 		if (args[i] != info->entry.args[i])
1073 			return -ERANGE;
1074 	}
1075 
1076 	syscall_set_nr(child, regs, nr);
1077 	/*
1078 	 * If the syscall number is set to -1, setting syscall arguments is not
1079 	 * just pointless, it would also clobber the syscall return value on
1080 	 * those architectures that share the same register both for the first
1081 	 * argument of syscall and its return value.
1082 	 */
1083 	if (nr != -1)
1084 		syscall_set_arguments(child, regs, args);
1085 
1086 	return 0;
1087 }
1088 
1089 static int
1090 ptrace_set_syscall_info_seccomp(struct task_struct *child, struct pt_regs *regs,
1091 				struct ptrace_syscall_info *info)
1092 {
1093 	/*
1094 	 * info->entry is currently a subset of info->seccomp,
1095 	 * info->seccomp.ret_data is currently ignored.
1096 	 */
1097 	return ptrace_set_syscall_info_entry(child, regs, info);
1098 }
1099 
1100 static int
1101 ptrace_set_syscall_info_exit(struct task_struct *child, struct pt_regs *regs,
1102 			     struct ptrace_syscall_info *info)
1103 {
1104 	long rval = info->exit.rval;
1105 
1106 	/*
1107 	 * Check that the return value specified in info->exit.rval
1108 	 * is either a value of type "long" or a sign-extended value
1109 	 * of type "long".
1110 	 */
1111 	if (rval != info->exit.rval)
1112 		return -ERANGE;
1113 
1114 	if (info->exit.is_error)
1115 		syscall_set_return_value(child, regs, rval, 0);
1116 	else
1117 		syscall_set_return_value(child, regs, 0, rval);
1118 
1119 	return 0;
1120 }
1121 
1122 static int
1123 ptrace_set_syscall_info(struct task_struct *child, unsigned long user_size,
1124 			const void __user *datavp)
1125 {
1126 	struct pt_regs *regs = task_pt_regs(child);
1127 	struct ptrace_syscall_info info;
1128 
1129 	if (user_size < sizeof(info))
1130 		return -EINVAL;
1131 
1132 	/*
1133 	 * The compatibility is tracked by info.op and info.flags: if user-space
1134 	 * does not instruct us to use unknown extra bits from future versions
1135 	 * of ptrace_syscall_info, we are not going to read them either.
1136 	 */
1137 	if (copy_from_user(&info, datavp, sizeof(info)))
1138 		return -EFAULT;
1139 
1140 	/* Reserved for future use. */
1141 	if (info.flags || info.reserved)
1142 		return -EINVAL;
1143 
1144 	/* Changing the type of the system call stop is not supported yet. */
1145 	if (ptrace_get_syscall_info_op(child) != info.op)
1146 		return -EINVAL;
1147 
1148 	switch (info.op) {
1149 	case PTRACE_SYSCALL_INFO_ENTRY:
1150 		return ptrace_set_syscall_info_entry(child, regs, &info);
1151 	case PTRACE_SYSCALL_INFO_EXIT:
1152 		return ptrace_set_syscall_info_exit(child, regs, &info);
1153 	case PTRACE_SYSCALL_INFO_SECCOMP:
1154 		return ptrace_set_syscall_info_seccomp(child, regs, &info);
1155 	default:
1156 		/* Other types of system call stops are not supported yet. */
1157 		return -EINVAL;
1158 	}
1159 }
1160 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
1161 
1162 int ptrace_request(struct task_struct *child, long request,
1163 		   unsigned long addr, unsigned long data)
1164 {
1165 	bool seized = child->ptrace & PT_SEIZED;
1166 	int ret = -EIO;
1167 	kernel_siginfo_t siginfo, *si;
1168 	void __user *datavp = (void __user *) data;
1169 	unsigned long __user *datalp = datavp;
1170 	unsigned long flags;
1171 
1172 	switch (request) {
1173 	case PTRACE_PEEKTEXT:
1174 	case PTRACE_PEEKDATA:
1175 		return generic_ptrace_peekdata(child, addr, data);
1176 	case PTRACE_POKETEXT:
1177 	case PTRACE_POKEDATA:
1178 		return generic_ptrace_pokedata(child, addr, data);
1179 
1180 #ifdef PTRACE_OLDSETOPTIONS
1181 	case PTRACE_OLDSETOPTIONS:
1182 #endif
1183 	case PTRACE_SETOPTIONS:
1184 		ret = ptrace_setoptions(child, data);
1185 		break;
1186 	case PTRACE_GETEVENTMSG:
1187 		ret = put_user(child->ptrace_message, datalp);
1188 		break;
1189 
1190 	case PTRACE_PEEKSIGINFO:
1191 		ret = ptrace_peek_siginfo(child, addr, data);
1192 		break;
1193 
1194 	case PTRACE_GETSIGINFO:
1195 		ret = ptrace_getsiginfo(child, &siginfo);
1196 		if (!ret)
1197 			ret = copy_siginfo_to_user(datavp, &siginfo);
1198 		break;
1199 
1200 	case PTRACE_SETSIGINFO:
1201 		ret = copy_siginfo_from_user(&siginfo, datavp);
1202 		if (!ret)
1203 			ret = ptrace_setsiginfo(child, &siginfo);
1204 		break;
1205 
1206 	case PTRACE_GETSIGMASK: {
1207 		sigset_t *mask;
1208 
1209 		if (addr != sizeof(sigset_t)) {
1210 			ret = -EINVAL;
1211 			break;
1212 		}
1213 
1214 		if (test_tsk_restore_sigmask(child))
1215 			mask = &child->saved_sigmask;
1216 		else
1217 			mask = &child->blocked;
1218 
1219 		if (copy_to_user(datavp, mask, sizeof(sigset_t)))
1220 			ret = -EFAULT;
1221 		else
1222 			ret = 0;
1223 
1224 		break;
1225 	}
1226 
1227 	case PTRACE_SETSIGMASK: {
1228 		sigset_t new_set;
1229 
1230 		if (addr != sizeof(sigset_t)) {
1231 			ret = -EINVAL;
1232 			break;
1233 		}
1234 
1235 		if (copy_from_user(&new_set, datavp, sizeof(sigset_t))) {
1236 			ret = -EFAULT;
1237 			break;
1238 		}
1239 
1240 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
1241 
1242 		/*
1243 		 * Every thread does recalc_sigpending() after resume, so
1244 		 * retarget_shared_pending() and recalc_sigpending() are not
1245 		 * called here.
1246 		 */
1247 		spin_lock_irq(&child->sighand->siglock);
1248 		child->blocked = new_set;
1249 		spin_unlock_irq(&child->sighand->siglock);
1250 
1251 		clear_tsk_restore_sigmask(child);
1252 
1253 		ret = 0;
1254 		break;
1255 	}
1256 
1257 	case PTRACE_INTERRUPT:
1258 		/*
1259 		 * Stop tracee without any side-effect on signal or job
1260 		 * control.  At least one trap is guaranteed to happen
1261 		 * after this request.  If @child is already trapped, the
1262 		 * current trap is not disturbed and another trap will
1263 		 * happen after the current trap is ended with PTRACE_CONT.
1264 		 *
1265 		 * The actual trap might not be PTRACE_EVENT_STOP trap but
1266 		 * the pending condition is cleared regardless.
1267 		 */
1268 		if (unlikely(!seized || !lock_task_sighand(child, &flags)))
1269 			break;
1270 
1271 		/*
1272 		 * INTERRUPT doesn't disturb existing trap sans one
1273 		 * exception.  If ptracer issued LISTEN for the current
1274 		 * STOP, this INTERRUPT should clear LISTEN and re-trap
1275 		 * tracee into STOP.
1276 		 */
1277 		if (likely(task_set_jobctl_pending(child, JOBCTL_TRAP_STOP)))
1278 			ptrace_signal_wake_up(child, child->jobctl & JOBCTL_LISTENING);
1279 
1280 		unlock_task_sighand(child, &flags);
1281 		ret = 0;
1282 		break;
1283 
1284 	case PTRACE_LISTEN:
1285 		/*
1286 		 * Listen for events.  Tracee must be in STOP.  It's not
1287 		 * resumed per-se but is not considered to be in TRACED by
1288 		 * wait(2) or ptrace(2).  If an async event (e.g. group
1289 		 * stop state change) happens, tracee will enter STOP trap
1290 		 * again.  Alternatively, ptracer can issue INTERRUPT to
1291 		 * finish listening and re-trap tracee into STOP.
1292 		 */
1293 		if (unlikely(!seized || !lock_task_sighand(child, &flags)))
1294 			break;
1295 
1296 		si = child->last_siginfo;
1297 		if (likely(si && (si->si_code >> 8) == PTRACE_EVENT_STOP)) {
1298 			child->jobctl |= JOBCTL_LISTENING;
1299 			/*
1300 			 * If NOTIFY is set, it means event happened between
1301 			 * start of this trap and now.  Trigger re-trap.
1302 			 */
1303 			if (child->jobctl & JOBCTL_TRAP_NOTIFY)
1304 				ptrace_signal_wake_up(child, true);
1305 			ret = 0;
1306 		}
1307 		unlock_task_sighand(child, &flags);
1308 		break;
1309 
1310 	case PTRACE_DETACH:	 /* detach a process that was attached. */
1311 		ret = ptrace_detach(child, data);
1312 		break;
1313 
1314 #ifdef CONFIG_BINFMT_ELF_FDPIC
1315 	case PTRACE_GETFDPIC: {
1316 		struct mm_struct *mm = get_task_mm(child);
1317 		unsigned long tmp = 0;
1318 
1319 		ret = -ESRCH;
1320 		if (!mm)
1321 			break;
1322 
1323 		switch (addr) {
1324 		case PTRACE_GETFDPIC_EXEC:
1325 			tmp = mm->context.exec_fdpic_loadmap;
1326 			break;
1327 		case PTRACE_GETFDPIC_INTERP:
1328 			tmp = mm->context.interp_fdpic_loadmap;
1329 			break;
1330 		default:
1331 			break;
1332 		}
1333 		mmput(mm);
1334 
1335 		ret = put_user(tmp, datalp);
1336 		break;
1337 	}
1338 #endif
1339 
1340 	case PTRACE_SINGLESTEP:
1341 #ifdef PTRACE_SINGLEBLOCK
1342 	case PTRACE_SINGLEBLOCK:
1343 #endif
1344 #ifdef PTRACE_SYSEMU
1345 	case PTRACE_SYSEMU:
1346 	case PTRACE_SYSEMU_SINGLESTEP:
1347 #endif
1348 	case PTRACE_SYSCALL:
1349 	case PTRACE_CONT:
1350 		return ptrace_resume(child, request, data);
1351 
1352 	case PTRACE_KILL:
1353 		send_sig_info(SIGKILL, SEND_SIG_NOINFO, child);
1354 		return 0;
1355 
1356 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
1357 	case PTRACE_GETREGSET:
1358 	case PTRACE_SETREGSET: {
1359 		struct iovec kiov;
1360 		struct iovec __user *uiov = datavp;
1361 
1362 		if (!access_ok(uiov, sizeof(*uiov)))
1363 			return -EFAULT;
1364 
1365 		if (__get_user(kiov.iov_base, &uiov->iov_base) ||
1366 		    __get_user(kiov.iov_len, &uiov->iov_len))
1367 			return -EFAULT;
1368 
1369 		ret = ptrace_regset(child, request, addr, &kiov);
1370 		if (!ret)
1371 			ret = __put_user(kiov.iov_len, &uiov->iov_len);
1372 		break;
1373 	}
1374 
1375 	case PTRACE_GET_SYSCALL_INFO:
1376 		ret = ptrace_get_syscall_info(child, addr, datavp);
1377 		break;
1378 
1379 	case PTRACE_SET_SYSCALL_INFO:
1380 		ret = ptrace_set_syscall_info(child, addr, datavp);
1381 		break;
1382 #endif
1383 
1384 	case PTRACE_SECCOMP_GET_FILTER:
1385 		ret = seccomp_get_filter(child, addr, datavp);
1386 		break;
1387 
1388 	case PTRACE_SECCOMP_GET_METADATA:
1389 		ret = seccomp_get_metadata(child, addr, datavp);
1390 		break;
1391 
1392 #ifdef CONFIG_RSEQ
1393 	case PTRACE_GET_RSEQ_CONFIGURATION:
1394 		ret = ptrace_get_rseq_configuration(child, addr, datavp);
1395 		break;
1396 #endif
1397 
1398 	case PTRACE_SET_SYSCALL_USER_DISPATCH_CONFIG:
1399 		ret = syscall_user_dispatch_set_config(child, addr, datavp);
1400 		break;
1401 
1402 	case PTRACE_GET_SYSCALL_USER_DISPATCH_CONFIG:
1403 		ret = syscall_user_dispatch_get_config(child, addr, datavp);
1404 		break;
1405 
1406 	default:
1407 		break;
1408 	}
1409 
1410 	return ret;
1411 }
1412 
1413 SYSCALL_DEFINE4(ptrace, long, request, long, pid, unsigned long, addr,
1414 		unsigned long, data)
1415 {
1416 	struct task_struct *child;
1417 	long ret;
1418 
1419 	if (request == PTRACE_TRACEME) {
1420 		ret = ptrace_traceme();
1421 		goto out;
1422 	}
1423 
1424 	child = find_get_task_by_vpid(pid);
1425 	if (!child) {
1426 		ret = -ESRCH;
1427 		goto out;
1428 	}
1429 
1430 	if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1431 		ret = ptrace_attach(child, request, addr, data);
1432 		goto out_put_task_struct;
1433 	}
1434 
1435 	ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1436 				  request == PTRACE_INTERRUPT);
1437 	if (ret < 0)
1438 		goto out_put_task_struct;
1439 
1440 	ret = arch_ptrace(child, request, addr, data);
1441 	if (ret || request != PTRACE_DETACH)
1442 		ptrace_unfreeze_traced(child);
1443 
1444  out_put_task_struct:
1445 	put_task_struct(child);
1446  out:
1447 	return ret;
1448 }
1449 
1450 int generic_ptrace_peekdata(struct task_struct *tsk, unsigned long addr,
1451 			    unsigned long data)
1452 {
1453 	unsigned long tmp;
1454 	int copied;
1455 
1456 	copied = ptrace_access_vm(tsk, addr, &tmp, sizeof(tmp), FOLL_FORCE);
1457 	if (copied != sizeof(tmp))
1458 		return -EIO;
1459 	return put_user(tmp, (unsigned long __user *)data);
1460 }
1461 
1462 int generic_ptrace_pokedata(struct task_struct *tsk, unsigned long addr,
1463 			    unsigned long data)
1464 {
1465 	int copied;
1466 
1467 	copied = ptrace_access_vm(tsk, addr, &data, sizeof(data),
1468 			FOLL_FORCE | FOLL_WRITE);
1469 	return (copied == sizeof(data)) ? 0 : -EIO;
1470 }
1471 
1472 #if defined CONFIG_COMPAT
1473 
1474 int compat_ptrace_request(struct task_struct *child, compat_long_t request,
1475 			  compat_ulong_t addr, compat_ulong_t data)
1476 {
1477 	compat_ulong_t __user *datap = compat_ptr(data);
1478 	compat_ulong_t word;
1479 	kernel_siginfo_t siginfo;
1480 	int ret;
1481 
1482 	switch (request) {
1483 	case PTRACE_PEEKTEXT:
1484 	case PTRACE_PEEKDATA:
1485 		ret = ptrace_access_vm(child, addr, &word, sizeof(word),
1486 				FOLL_FORCE);
1487 		if (ret != sizeof(word))
1488 			ret = -EIO;
1489 		else
1490 			ret = put_user(word, datap);
1491 		break;
1492 
1493 	case PTRACE_POKETEXT:
1494 	case PTRACE_POKEDATA:
1495 		ret = ptrace_access_vm(child, addr, &data, sizeof(data),
1496 				FOLL_FORCE | FOLL_WRITE);
1497 		ret = (ret != sizeof(data) ? -EIO : 0);
1498 		break;
1499 
1500 	case PTRACE_GETEVENTMSG:
1501 		ret = put_user((compat_ulong_t) child->ptrace_message, datap);
1502 		break;
1503 
1504 	case PTRACE_GETSIGINFO:
1505 		ret = ptrace_getsiginfo(child, &siginfo);
1506 		if (!ret)
1507 			ret = copy_siginfo_to_user32(
1508 				(struct compat_siginfo __user *) datap,
1509 				&siginfo);
1510 		break;
1511 
1512 	case PTRACE_SETSIGINFO:
1513 		ret = copy_siginfo_from_user32(
1514 			&siginfo, (struct compat_siginfo __user *) datap);
1515 		if (!ret)
1516 			ret = ptrace_setsiginfo(child, &siginfo);
1517 		break;
1518 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
1519 	case PTRACE_GETREGSET:
1520 	case PTRACE_SETREGSET:
1521 	{
1522 		struct iovec kiov;
1523 		struct compat_iovec __user *uiov =
1524 			(struct compat_iovec __user *) datap;
1525 		compat_uptr_t ptr;
1526 		compat_size_t len;
1527 
1528 		if (!access_ok(uiov, sizeof(*uiov)))
1529 			return -EFAULT;
1530 
1531 		if (__get_user(ptr, &uiov->iov_base) ||
1532 		    __get_user(len, &uiov->iov_len))
1533 			return -EFAULT;
1534 
1535 		kiov.iov_base = compat_ptr(ptr);
1536 		kiov.iov_len = len;
1537 
1538 		ret = ptrace_regset(child, request, addr, &kiov);
1539 		if (!ret)
1540 			ret = __put_user(kiov.iov_len, &uiov->iov_len);
1541 		break;
1542 	}
1543 #endif
1544 
1545 	default:
1546 		ret = ptrace_request(child, request, addr, data);
1547 	}
1548 
1549 	return ret;
1550 }
1551 
1552 COMPAT_SYSCALL_DEFINE4(ptrace, compat_long_t, request, compat_long_t, pid,
1553 		       compat_long_t, addr, compat_long_t, data)
1554 {
1555 	struct task_struct *child;
1556 	long ret;
1557 
1558 	if (request == PTRACE_TRACEME) {
1559 		ret = ptrace_traceme();
1560 		goto out;
1561 	}
1562 
1563 	child = find_get_task_by_vpid(pid);
1564 	if (!child) {
1565 		ret = -ESRCH;
1566 		goto out;
1567 	}
1568 
1569 	if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1570 		ret = ptrace_attach(child, request, addr, data);
1571 		goto out_put_task_struct;
1572 	}
1573 
1574 	ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1575 				  request == PTRACE_INTERRUPT);
1576 	if (!ret) {
1577 		ret = compat_arch_ptrace(child, request, addr, data);
1578 		if (ret || request != PTRACE_DETACH)
1579 			ptrace_unfreeze_traced(child);
1580 	}
1581 
1582  out_put_task_struct:
1583 	put_task_struct(child);
1584  out:
1585 	return ret;
1586 }
1587 #endif	/* CONFIG_COMPAT */
1588