xref: /openbmc/linux/kernel/ptrace.c (revision 9d749629)
1 /*
2  * linux/kernel/ptrace.c
3  *
4  * (C) Copyright 1999 Linus Torvalds
5  *
6  * Common interfaces for "ptrace()" which we do not want
7  * to continually duplicate across every architecture.
8  */
9 
10 #include <linux/capability.h>
11 #include <linux/export.h>
12 #include <linux/sched.h>
13 #include <linux/errno.h>
14 #include <linux/mm.h>
15 #include <linux/highmem.h>
16 #include <linux/pagemap.h>
17 #include <linux/ptrace.h>
18 #include <linux/security.h>
19 #include <linux/signal.h>
20 #include <linux/audit.h>
21 #include <linux/pid_namespace.h>
22 #include <linux/syscalls.h>
23 #include <linux/uaccess.h>
24 #include <linux/regset.h>
25 #include <linux/hw_breakpoint.h>
26 #include <linux/cn_proc.h>
27 
28 
29 static int ptrace_trapping_sleep_fn(void *flags)
30 {
31 	schedule();
32 	return 0;
33 }
34 
35 /*
36  * ptrace a task: make the debugger its new parent and
37  * move it to the ptrace list.
38  *
39  * Must be called with the tasklist lock write-held.
40  */
41 void __ptrace_link(struct task_struct *child, struct task_struct *new_parent)
42 {
43 	BUG_ON(!list_empty(&child->ptrace_entry));
44 	list_add(&child->ptrace_entry, &new_parent->ptraced);
45 	child->parent = new_parent;
46 }
47 
48 /**
49  * __ptrace_unlink - unlink ptracee and restore its execution state
50  * @child: ptracee to be unlinked
51  *
52  * Remove @child from the ptrace list, move it back to the original parent,
53  * and restore the execution state so that it conforms to the group stop
54  * state.
55  *
56  * Unlinking can happen via two paths - explicit PTRACE_DETACH or ptracer
57  * exiting.  For PTRACE_DETACH, unless the ptracee has been killed between
58  * ptrace_check_attach() and here, it's guaranteed to be in TASK_TRACED.
59  * If the ptracer is exiting, the ptracee can be in any state.
60  *
61  * After detach, the ptracee should be in a state which conforms to the
62  * group stop.  If the group is stopped or in the process of stopping, the
63  * ptracee should be put into TASK_STOPPED; otherwise, it should be woken
64  * up from TASK_TRACED.
65  *
66  * If the ptracee is in TASK_TRACED and needs to be moved to TASK_STOPPED,
67  * it goes through TRACED -> RUNNING -> STOPPED transition which is similar
68  * to but in the opposite direction of what happens while attaching to a
69  * stopped task.  However, in this direction, the intermediate RUNNING
70  * state is not hidden even from the current ptracer and if it immediately
71  * re-attaches and performs a WNOHANG wait(2), it may fail.
72  *
73  * CONTEXT:
74  * write_lock_irq(tasklist_lock)
75  */
76 void __ptrace_unlink(struct task_struct *child)
77 {
78 	BUG_ON(!child->ptrace);
79 
80 	child->ptrace = 0;
81 	child->parent = child->real_parent;
82 	list_del_init(&child->ptrace_entry);
83 
84 	spin_lock(&child->sighand->siglock);
85 
86 	/*
87 	 * Clear all pending traps and TRAPPING.  TRAPPING should be
88 	 * cleared regardless of JOBCTL_STOP_PENDING.  Do it explicitly.
89 	 */
90 	task_clear_jobctl_pending(child, JOBCTL_TRAP_MASK);
91 	task_clear_jobctl_trapping(child);
92 
93 	/*
94 	 * Reinstate JOBCTL_STOP_PENDING if group stop is in effect and
95 	 * @child isn't dead.
96 	 */
97 	if (!(child->flags & PF_EXITING) &&
98 	    (child->signal->flags & SIGNAL_STOP_STOPPED ||
99 	     child->signal->group_stop_count)) {
100 		child->jobctl |= JOBCTL_STOP_PENDING;
101 
102 		/*
103 		 * This is only possible if this thread was cloned by the
104 		 * traced task running in the stopped group, set the signal
105 		 * for the future reports.
106 		 * FIXME: we should change ptrace_init_task() to handle this
107 		 * case.
108 		 */
109 		if (!(child->jobctl & JOBCTL_STOP_SIGMASK))
110 			child->jobctl |= SIGSTOP;
111 	}
112 
113 	/*
114 	 * If transition to TASK_STOPPED is pending or in TASK_TRACED, kick
115 	 * @child in the butt.  Note that @resume should be used iff @child
116 	 * is in TASK_TRACED; otherwise, we might unduly disrupt
117 	 * TASK_KILLABLE sleeps.
118 	 */
119 	if (child->jobctl & JOBCTL_STOP_PENDING || task_is_traced(child))
120 		ptrace_signal_wake_up(child, true);
121 
122 	spin_unlock(&child->sighand->siglock);
123 }
124 
125 /* Ensure that nothing can wake it up, even SIGKILL */
126 static bool ptrace_freeze_traced(struct task_struct *task)
127 {
128 	bool ret = false;
129 
130 	/* Lockless, nobody but us can set this flag */
131 	if (task->jobctl & JOBCTL_LISTENING)
132 		return ret;
133 
134 	spin_lock_irq(&task->sighand->siglock);
135 	if (task_is_traced(task) && !__fatal_signal_pending(task)) {
136 		task->state = __TASK_TRACED;
137 		ret = true;
138 	}
139 	spin_unlock_irq(&task->sighand->siglock);
140 
141 	return ret;
142 }
143 
144 static void ptrace_unfreeze_traced(struct task_struct *task)
145 {
146 	if (task->state != __TASK_TRACED)
147 		return;
148 
149 	WARN_ON(!task->ptrace || task->parent != current);
150 
151 	spin_lock_irq(&task->sighand->siglock);
152 	if (__fatal_signal_pending(task))
153 		wake_up_state(task, __TASK_TRACED);
154 	else
155 		task->state = TASK_TRACED;
156 	spin_unlock_irq(&task->sighand->siglock);
157 }
158 
159 /**
160  * ptrace_check_attach - check whether ptracee is ready for ptrace operation
161  * @child: ptracee to check for
162  * @ignore_state: don't check whether @child is currently %TASK_TRACED
163  *
164  * Check whether @child is being ptraced by %current and ready for further
165  * ptrace operations.  If @ignore_state is %false, @child also should be in
166  * %TASK_TRACED state and on return the child is guaranteed to be traced
167  * and not executing.  If @ignore_state is %true, @child can be in any
168  * state.
169  *
170  * CONTEXT:
171  * Grabs and releases tasklist_lock and @child->sighand->siglock.
172  *
173  * RETURNS:
174  * 0 on success, -ESRCH if %child is not ready.
175  */
176 static int ptrace_check_attach(struct task_struct *child, bool ignore_state)
177 {
178 	int ret = -ESRCH;
179 
180 	/*
181 	 * We take the read lock around doing both checks to close a
182 	 * possible race where someone else was tracing our child and
183 	 * detached between these two checks.  After this locked check,
184 	 * we are sure that this is our traced child and that can only
185 	 * be changed by us so it's not changing right after this.
186 	 */
187 	read_lock(&tasklist_lock);
188 	if (child->ptrace && child->parent == current) {
189 		WARN_ON(child->state == __TASK_TRACED);
190 		/*
191 		 * child->sighand can't be NULL, release_task()
192 		 * does ptrace_unlink() before __exit_signal().
193 		 */
194 		if (ignore_state || ptrace_freeze_traced(child))
195 			ret = 0;
196 	}
197 	read_unlock(&tasklist_lock);
198 
199 	if (!ret && !ignore_state) {
200 		if (!wait_task_inactive(child, __TASK_TRACED)) {
201 			/*
202 			 * This can only happen if may_ptrace_stop() fails and
203 			 * ptrace_stop() changes ->state back to TASK_RUNNING,
204 			 * so we should not worry about leaking __TASK_TRACED.
205 			 */
206 			WARN_ON(child->state == __TASK_TRACED);
207 			ret = -ESRCH;
208 		}
209 	}
210 
211 	return ret;
212 }
213 
214 static int ptrace_has_cap(struct user_namespace *ns, unsigned int mode)
215 {
216 	if (mode & PTRACE_MODE_NOAUDIT)
217 		return has_ns_capability_noaudit(current, ns, CAP_SYS_PTRACE);
218 	else
219 		return has_ns_capability(current, ns, CAP_SYS_PTRACE);
220 }
221 
222 /* Returns 0 on success, -errno on denial. */
223 static int __ptrace_may_access(struct task_struct *task, unsigned int mode)
224 {
225 	const struct cred *cred = current_cred(), *tcred;
226 
227 	/* May we inspect the given task?
228 	 * This check is used both for attaching with ptrace
229 	 * and for allowing access to sensitive information in /proc.
230 	 *
231 	 * ptrace_attach denies several cases that /proc allows
232 	 * because setting up the necessary parent/child relationship
233 	 * or halting the specified task is impossible.
234 	 */
235 	int dumpable = 0;
236 	/* Don't let security modules deny introspection */
237 	if (task == current)
238 		return 0;
239 	rcu_read_lock();
240 	tcred = __task_cred(task);
241 	if (uid_eq(cred->uid, tcred->euid) &&
242 	    uid_eq(cred->uid, tcred->suid) &&
243 	    uid_eq(cred->uid, tcred->uid)  &&
244 	    gid_eq(cred->gid, tcred->egid) &&
245 	    gid_eq(cred->gid, tcred->sgid) &&
246 	    gid_eq(cred->gid, tcred->gid))
247 		goto ok;
248 	if (ptrace_has_cap(tcred->user_ns, mode))
249 		goto ok;
250 	rcu_read_unlock();
251 	return -EPERM;
252 ok:
253 	rcu_read_unlock();
254 	smp_rmb();
255 	if (task->mm)
256 		dumpable = get_dumpable(task->mm);
257 	rcu_read_lock();
258 	if (!dumpable && !ptrace_has_cap(__task_cred(task)->user_ns, mode)) {
259 		rcu_read_unlock();
260 		return -EPERM;
261 	}
262 	rcu_read_unlock();
263 
264 	return security_ptrace_access_check(task, mode);
265 }
266 
267 bool ptrace_may_access(struct task_struct *task, unsigned int mode)
268 {
269 	int err;
270 	task_lock(task);
271 	err = __ptrace_may_access(task, mode);
272 	task_unlock(task);
273 	return !err;
274 }
275 
276 static int ptrace_attach(struct task_struct *task, long request,
277 			 unsigned long addr,
278 			 unsigned long flags)
279 {
280 	bool seize = (request == PTRACE_SEIZE);
281 	int retval;
282 
283 	retval = -EIO;
284 	if (seize) {
285 		if (addr != 0)
286 			goto out;
287 		if (flags & ~(unsigned long)PTRACE_O_MASK)
288 			goto out;
289 		flags = PT_PTRACED | PT_SEIZED | (flags << PT_OPT_FLAG_SHIFT);
290 	} else {
291 		flags = PT_PTRACED;
292 	}
293 
294 	audit_ptrace(task);
295 
296 	retval = -EPERM;
297 	if (unlikely(task->flags & PF_KTHREAD))
298 		goto out;
299 	if (same_thread_group(task, current))
300 		goto out;
301 
302 	/*
303 	 * Protect exec's credential calculations against our interference;
304 	 * SUID, SGID and LSM creds get determined differently
305 	 * under ptrace.
306 	 */
307 	retval = -ERESTARTNOINTR;
308 	if (mutex_lock_interruptible(&task->signal->cred_guard_mutex))
309 		goto out;
310 
311 	task_lock(task);
312 	retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH);
313 	task_unlock(task);
314 	if (retval)
315 		goto unlock_creds;
316 
317 	write_lock_irq(&tasklist_lock);
318 	retval = -EPERM;
319 	if (unlikely(task->exit_state))
320 		goto unlock_tasklist;
321 	if (task->ptrace)
322 		goto unlock_tasklist;
323 
324 	if (seize)
325 		flags |= PT_SEIZED;
326 	rcu_read_lock();
327 	if (ns_capable(__task_cred(task)->user_ns, CAP_SYS_PTRACE))
328 		flags |= PT_PTRACE_CAP;
329 	rcu_read_unlock();
330 	task->ptrace = flags;
331 
332 	__ptrace_link(task, current);
333 
334 	/* SEIZE doesn't trap tracee on attach */
335 	if (!seize)
336 		send_sig_info(SIGSTOP, SEND_SIG_FORCED, task);
337 
338 	spin_lock(&task->sighand->siglock);
339 
340 	/*
341 	 * If the task is already STOPPED, set JOBCTL_TRAP_STOP and
342 	 * TRAPPING, and kick it so that it transits to TRACED.  TRAPPING
343 	 * will be cleared if the child completes the transition or any
344 	 * event which clears the group stop states happens.  We'll wait
345 	 * for the transition to complete before returning from this
346 	 * function.
347 	 *
348 	 * This hides STOPPED -> RUNNING -> TRACED transition from the
349 	 * attaching thread but a different thread in the same group can
350 	 * still observe the transient RUNNING state.  IOW, if another
351 	 * thread's WNOHANG wait(2) on the stopped tracee races against
352 	 * ATTACH, the wait(2) may fail due to the transient RUNNING.
353 	 *
354 	 * The following task_is_stopped() test is safe as both transitions
355 	 * in and out of STOPPED are protected by siglock.
356 	 */
357 	if (task_is_stopped(task) &&
358 	    task_set_jobctl_pending(task, JOBCTL_TRAP_STOP | JOBCTL_TRAPPING))
359 		signal_wake_up_state(task, __TASK_STOPPED);
360 
361 	spin_unlock(&task->sighand->siglock);
362 
363 	retval = 0;
364 unlock_tasklist:
365 	write_unlock_irq(&tasklist_lock);
366 unlock_creds:
367 	mutex_unlock(&task->signal->cred_guard_mutex);
368 out:
369 	if (!retval) {
370 		wait_on_bit(&task->jobctl, JOBCTL_TRAPPING_BIT,
371 			    ptrace_trapping_sleep_fn, TASK_UNINTERRUPTIBLE);
372 		proc_ptrace_connector(task, PTRACE_ATTACH);
373 	}
374 
375 	return retval;
376 }
377 
378 /**
379  * ptrace_traceme  --  helper for PTRACE_TRACEME
380  *
381  * Performs checks and sets PT_PTRACED.
382  * Should be used by all ptrace implementations for PTRACE_TRACEME.
383  */
384 static int ptrace_traceme(void)
385 {
386 	int ret = -EPERM;
387 
388 	write_lock_irq(&tasklist_lock);
389 	/* Are we already being traced? */
390 	if (!current->ptrace) {
391 		ret = security_ptrace_traceme(current->parent);
392 		/*
393 		 * Check PF_EXITING to ensure ->real_parent has not passed
394 		 * exit_ptrace(). Otherwise we don't report the error but
395 		 * pretend ->real_parent untraces us right after return.
396 		 */
397 		if (!ret && !(current->real_parent->flags & PF_EXITING)) {
398 			current->ptrace = PT_PTRACED;
399 			__ptrace_link(current, current->real_parent);
400 		}
401 	}
402 	write_unlock_irq(&tasklist_lock);
403 
404 	return ret;
405 }
406 
407 /*
408  * Called with irqs disabled, returns true if childs should reap themselves.
409  */
410 static int ignoring_children(struct sighand_struct *sigh)
411 {
412 	int ret;
413 	spin_lock(&sigh->siglock);
414 	ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
415 	      (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
416 	spin_unlock(&sigh->siglock);
417 	return ret;
418 }
419 
420 /*
421  * Called with tasklist_lock held for writing.
422  * Unlink a traced task, and clean it up if it was a traced zombie.
423  * Return true if it needs to be reaped with release_task().
424  * (We can't call release_task() here because we already hold tasklist_lock.)
425  *
426  * If it's a zombie, our attachedness prevented normal parent notification
427  * or self-reaping.  Do notification now if it would have happened earlier.
428  * If it should reap itself, return true.
429  *
430  * If it's our own child, there is no notification to do. But if our normal
431  * children self-reap, then this child was prevented by ptrace and we must
432  * reap it now, in that case we must also wake up sub-threads sleeping in
433  * do_wait().
434  */
435 static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
436 {
437 	bool dead;
438 
439 	__ptrace_unlink(p);
440 
441 	if (p->exit_state != EXIT_ZOMBIE)
442 		return false;
443 
444 	dead = !thread_group_leader(p);
445 
446 	if (!dead && thread_group_empty(p)) {
447 		if (!same_thread_group(p->real_parent, tracer))
448 			dead = do_notify_parent(p, p->exit_signal);
449 		else if (ignoring_children(tracer->sighand)) {
450 			__wake_up_parent(p, tracer);
451 			dead = true;
452 		}
453 	}
454 	/* Mark it as in the process of being reaped. */
455 	if (dead)
456 		p->exit_state = EXIT_DEAD;
457 	return dead;
458 }
459 
460 static int ptrace_detach(struct task_struct *child, unsigned int data)
461 {
462 	bool dead = false;
463 
464 	if (!valid_signal(data))
465 		return -EIO;
466 
467 	/* Architecture-specific hardware disable .. */
468 	ptrace_disable(child);
469 	clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
470 
471 	write_lock_irq(&tasklist_lock);
472 	/*
473 	 * This child can be already killed. Make sure de_thread() or
474 	 * our sub-thread doing do_wait() didn't do release_task() yet.
475 	 */
476 	if (child->ptrace) {
477 		child->exit_code = data;
478 		dead = __ptrace_detach(current, child);
479 	}
480 	write_unlock_irq(&tasklist_lock);
481 
482 	proc_ptrace_connector(child, PTRACE_DETACH);
483 	if (unlikely(dead))
484 		release_task(child);
485 
486 	return 0;
487 }
488 
489 /*
490  * Detach all tasks we were using ptrace on. Called with tasklist held
491  * for writing, and returns with it held too. But note it can release
492  * and reacquire the lock.
493  */
494 void exit_ptrace(struct task_struct *tracer)
495 	__releases(&tasklist_lock)
496 	__acquires(&tasklist_lock)
497 {
498 	struct task_struct *p, *n;
499 	LIST_HEAD(ptrace_dead);
500 
501 	if (likely(list_empty(&tracer->ptraced)))
502 		return;
503 
504 	list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
505 		if (unlikely(p->ptrace & PT_EXITKILL))
506 			send_sig_info(SIGKILL, SEND_SIG_FORCED, p);
507 
508 		if (__ptrace_detach(tracer, p))
509 			list_add(&p->ptrace_entry, &ptrace_dead);
510 	}
511 
512 	write_unlock_irq(&tasklist_lock);
513 	BUG_ON(!list_empty(&tracer->ptraced));
514 
515 	list_for_each_entry_safe(p, n, &ptrace_dead, ptrace_entry) {
516 		list_del_init(&p->ptrace_entry);
517 		release_task(p);
518 	}
519 
520 	write_lock_irq(&tasklist_lock);
521 }
522 
523 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
524 {
525 	int copied = 0;
526 
527 	while (len > 0) {
528 		char buf[128];
529 		int this_len, retval;
530 
531 		this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
532 		retval = access_process_vm(tsk, src, buf, this_len, 0);
533 		if (!retval) {
534 			if (copied)
535 				break;
536 			return -EIO;
537 		}
538 		if (copy_to_user(dst, buf, retval))
539 			return -EFAULT;
540 		copied += retval;
541 		src += retval;
542 		dst += retval;
543 		len -= retval;
544 	}
545 	return copied;
546 }
547 
548 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
549 {
550 	int copied = 0;
551 
552 	while (len > 0) {
553 		char buf[128];
554 		int this_len, retval;
555 
556 		this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
557 		if (copy_from_user(buf, src, this_len))
558 			return -EFAULT;
559 		retval = access_process_vm(tsk, dst, buf, this_len, 1);
560 		if (!retval) {
561 			if (copied)
562 				break;
563 			return -EIO;
564 		}
565 		copied += retval;
566 		src += retval;
567 		dst += retval;
568 		len -= retval;
569 	}
570 	return copied;
571 }
572 
573 static int ptrace_setoptions(struct task_struct *child, unsigned long data)
574 {
575 	unsigned flags;
576 
577 	if (data & ~(unsigned long)PTRACE_O_MASK)
578 		return -EINVAL;
579 
580 	/* Avoid intermediate state when all opts are cleared */
581 	flags = child->ptrace;
582 	flags &= ~(PTRACE_O_MASK << PT_OPT_FLAG_SHIFT);
583 	flags |= (data << PT_OPT_FLAG_SHIFT);
584 	child->ptrace = flags;
585 
586 	return 0;
587 }
588 
589 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t *info)
590 {
591 	unsigned long flags;
592 	int error = -ESRCH;
593 
594 	if (lock_task_sighand(child, &flags)) {
595 		error = -EINVAL;
596 		if (likely(child->last_siginfo != NULL)) {
597 			*info = *child->last_siginfo;
598 			error = 0;
599 		}
600 		unlock_task_sighand(child, &flags);
601 	}
602 	return error;
603 }
604 
605 static int ptrace_setsiginfo(struct task_struct *child, const siginfo_t *info)
606 {
607 	unsigned long flags;
608 	int error = -ESRCH;
609 
610 	if (lock_task_sighand(child, &flags)) {
611 		error = -EINVAL;
612 		if (likely(child->last_siginfo != NULL)) {
613 			*child->last_siginfo = *info;
614 			error = 0;
615 		}
616 		unlock_task_sighand(child, &flags);
617 	}
618 	return error;
619 }
620 
621 
622 #ifdef PTRACE_SINGLESTEP
623 #define is_singlestep(request)		((request) == PTRACE_SINGLESTEP)
624 #else
625 #define is_singlestep(request)		0
626 #endif
627 
628 #ifdef PTRACE_SINGLEBLOCK
629 #define is_singleblock(request)		((request) == PTRACE_SINGLEBLOCK)
630 #else
631 #define is_singleblock(request)		0
632 #endif
633 
634 #ifdef PTRACE_SYSEMU
635 #define is_sysemu_singlestep(request)	((request) == PTRACE_SYSEMU_SINGLESTEP)
636 #else
637 #define is_sysemu_singlestep(request)	0
638 #endif
639 
640 static int ptrace_resume(struct task_struct *child, long request,
641 			 unsigned long data)
642 {
643 	if (!valid_signal(data))
644 		return -EIO;
645 
646 	if (request == PTRACE_SYSCALL)
647 		set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
648 	else
649 		clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
650 
651 #ifdef TIF_SYSCALL_EMU
652 	if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
653 		set_tsk_thread_flag(child, TIF_SYSCALL_EMU);
654 	else
655 		clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
656 #endif
657 
658 	if (is_singleblock(request)) {
659 		if (unlikely(!arch_has_block_step()))
660 			return -EIO;
661 		user_enable_block_step(child);
662 	} else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
663 		if (unlikely(!arch_has_single_step()))
664 			return -EIO;
665 		user_enable_single_step(child);
666 	} else {
667 		user_disable_single_step(child);
668 	}
669 
670 	child->exit_code = data;
671 	wake_up_state(child, __TASK_TRACED);
672 
673 	return 0;
674 }
675 
676 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
677 
678 static const struct user_regset *
679 find_regset(const struct user_regset_view *view, unsigned int type)
680 {
681 	const struct user_regset *regset;
682 	int n;
683 
684 	for (n = 0; n < view->n; ++n) {
685 		regset = view->regsets + n;
686 		if (regset->core_note_type == type)
687 			return regset;
688 	}
689 
690 	return NULL;
691 }
692 
693 static int ptrace_regset(struct task_struct *task, int req, unsigned int type,
694 			 struct iovec *kiov)
695 {
696 	const struct user_regset_view *view = task_user_regset_view(task);
697 	const struct user_regset *regset = find_regset(view, type);
698 	int regset_no;
699 
700 	if (!regset || (kiov->iov_len % regset->size) != 0)
701 		return -EINVAL;
702 
703 	regset_no = regset - view->regsets;
704 	kiov->iov_len = min(kiov->iov_len,
705 			    (__kernel_size_t) (regset->n * regset->size));
706 
707 	if (req == PTRACE_GETREGSET)
708 		return copy_regset_to_user(task, view, regset_no, 0,
709 					   kiov->iov_len, kiov->iov_base);
710 	else
711 		return copy_regset_from_user(task, view, regset_no, 0,
712 					     kiov->iov_len, kiov->iov_base);
713 }
714 
715 /*
716  * This is declared in linux/regset.h and defined in machine-dependent
717  * code.  We put the export here, near the primary machine-neutral use,
718  * to ensure no machine forgets it.
719  */
720 EXPORT_SYMBOL_GPL(task_user_regset_view);
721 #endif
722 
723 int ptrace_request(struct task_struct *child, long request,
724 		   unsigned long addr, unsigned long data)
725 {
726 	bool seized = child->ptrace & PT_SEIZED;
727 	int ret = -EIO;
728 	siginfo_t siginfo, *si;
729 	void __user *datavp = (void __user *) data;
730 	unsigned long __user *datalp = datavp;
731 	unsigned long flags;
732 
733 	switch (request) {
734 	case PTRACE_PEEKTEXT:
735 	case PTRACE_PEEKDATA:
736 		return generic_ptrace_peekdata(child, addr, data);
737 	case PTRACE_POKETEXT:
738 	case PTRACE_POKEDATA:
739 		return generic_ptrace_pokedata(child, addr, data);
740 
741 #ifdef PTRACE_OLDSETOPTIONS
742 	case PTRACE_OLDSETOPTIONS:
743 #endif
744 	case PTRACE_SETOPTIONS:
745 		ret = ptrace_setoptions(child, data);
746 		break;
747 	case PTRACE_GETEVENTMSG:
748 		ret = put_user(child->ptrace_message, datalp);
749 		break;
750 
751 	case PTRACE_GETSIGINFO:
752 		ret = ptrace_getsiginfo(child, &siginfo);
753 		if (!ret)
754 			ret = copy_siginfo_to_user(datavp, &siginfo);
755 		break;
756 
757 	case PTRACE_SETSIGINFO:
758 		if (copy_from_user(&siginfo, datavp, sizeof siginfo))
759 			ret = -EFAULT;
760 		else
761 			ret = ptrace_setsiginfo(child, &siginfo);
762 		break;
763 
764 	case PTRACE_INTERRUPT:
765 		/*
766 		 * Stop tracee without any side-effect on signal or job
767 		 * control.  At least one trap is guaranteed to happen
768 		 * after this request.  If @child is already trapped, the
769 		 * current trap is not disturbed and another trap will
770 		 * happen after the current trap is ended with PTRACE_CONT.
771 		 *
772 		 * The actual trap might not be PTRACE_EVENT_STOP trap but
773 		 * the pending condition is cleared regardless.
774 		 */
775 		if (unlikely(!seized || !lock_task_sighand(child, &flags)))
776 			break;
777 
778 		/*
779 		 * INTERRUPT doesn't disturb existing trap sans one
780 		 * exception.  If ptracer issued LISTEN for the current
781 		 * STOP, this INTERRUPT should clear LISTEN and re-trap
782 		 * tracee into STOP.
783 		 */
784 		if (likely(task_set_jobctl_pending(child, JOBCTL_TRAP_STOP)))
785 			ptrace_signal_wake_up(child, child->jobctl & JOBCTL_LISTENING);
786 
787 		unlock_task_sighand(child, &flags);
788 		ret = 0;
789 		break;
790 
791 	case PTRACE_LISTEN:
792 		/*
793 		 * Listen for events.  Tracee must be in STOP.  It's not
794 		 * resumed per-se but is not considered to be in TRACED by
795 		 * wait(2) or ptrace(2).  If an async event (e.g. group
796 		 * stop state change) happens, tracee will enter STOP trap
797 		 * again.  Alternatively, ptracer can issue INTERRUPT to
798 		 * finish listening and re-trap tracee into STOP.
799 		 */
800 		if (unlikely(!seized || !lock_task_sighand(child, &flags)))
801 			break;
802 
803 		si = child->last_siginfo;
804 		if (likely(si && (si->si_code >> 8) == PTRACE_EVENT_STOP)) {
805 			child->jobctl |= JOBCTL_LISTENING;
806 			/*
807 			 * If NOTIFY is set, it means event happened between
808 			 * start of this trap and now.  Trigger re-trap.
809 			 */
810 			if (child->jobctl & JOBCTL_TRAP_NOTIFY)
811 				ptrace_signal_wake_up(child, true);
812 			ret = 0;
813 		}
814 		unlock_task_sighand(child, &flags);
815 		break;
816 
817 	case PTRACE_DETACH:	 /* detach a process that was attached. */
818 		ret = ptrace_detach(child, data);
819 		break;
820 
821 #ifdef CONFIG_BINFMT_ELF_FDPIC
822 	case PTRACE_GETFDPIC: {
823 		struct mm_struct *mm = get_task_mm(child);
824 		unsigned long tmp = 0;
825 
826 		ret = -ESRCH;
827 		if (!mm)
828 			break;
829 
830 		switch (addr) {
831 		case PTRACE_GETFDPIC_EXEC:
832 			tmp = mm->context.exec_fdpic_loadmap;
833 			break;
834 		case PTRACE_GETFDPIC_INTERP:
835 			tmp = mm->context.interp_fdpic_loadmap;
836 			break;
837 		default:
838 			break;
839 		}
840 		mmput(mm);
841 
842 		ret = put_user(tmp, datalp);
843 		break;
844 	}
845 #endif
846 
847 #ifdef PTRACE_SINGLESTEP
848 	case PTRACE_SINGLESTEP:
849 #endif
850 #ifdef PTRACE_SINGLEBLOCK
851 	case PTRACE_SINGLEBLOCK:
852 #endif
853 #ifdef PTRACE_SYSEMU
854 	case PTRACE_SYSEMU:
855 	case PTRACE_SYSEMU_SINGLESTEP:
856 #endif
857 	case PTRACE_SYSCALL:
858 	case PTRACE_CONT:
859 		return ptrace_resume(child, request, data);
860 
861 	case PTRACE_KILL:
862 		if (child->exit_state)	/* already dead */
863 			return 0;
864 		return ptrace_resume(child, request, SIGKILL);
865 
866 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
867 	case PTRACE_GETREGSET:
868 	case PTRACE_SETREGSET:
869 	{
870 		struct iovec kiov;
871 		struct iovec __user *uiov = datavp;
872 
873 		if (!access_ok(VERIFY_WRITE, uiov, sizeof(*uiov)))
874 			return -EFAULT;
875 
876 		if (__get_user(kiov.iov_base, &uiov->iov_base) ||
877 		    __get_user(kiov.iov_len, &uiov->iov_len))
878 			return -EFAULT;
879 
880 		ret = ptrace_regset(child, request, addr, &kiov);
881 		if (!ret)
882 			ret = __put_user(kiov.iov_len, &uiov->iov_len);
883 		break;
884 	}
885 #endif
886 	default:
887 		break;
888 	}
889 
890 	return ret;
891 }
892 
893 static struct task_struct *ptrace_get_task_struct(pid_t pid)
894 {
895 	struct task_struct *child;
896 
897 	rcu_read_lock();
898 	child = find_task_by_vpid(pid);
899 	if (child)
900 		get_task_struct(child);
901 	rcu_read_unlock();
902 
903 	if (!child)
904 		return ERR_PTR(-ESRCH);
905 	return child;
906 }
907 
908 #ifndef arch_ptrace_attach
909 #define arch_ptrace_attach(child)	do { } while (0)
910 #endif
911 
912 SYSCALL_DEFINE4(ptrace, long, request, long, pid, unsigned long, addr,
913 		unsigned long, data)
914 {
915 	struct task_struct *child;
916 	long ret;
917 
918 	if (request == PTRACE_TRACEME) {
919 		ret = ptrace_traceme();
920 		if (!ret)
921 			arch_ptrace_attach(current);
922 		goto out;
923 	}
924 
925 	child = ptrace_get_task_struct(pid);
926 	if (IS_ERR(child)) {
927 		ret = PTR_ERR(child);
928 		goto out;
929 	}
930 
931 	if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
932 		ret = ptrace_attach(child, request, addr, data);
933 		/*
934 		 * Some architectures need to do book-keeping after
935 		 * a ptrace attach.
936 		 */
937 		if (!ret)
938 			arch_ptrace_attach(child);
939 		goto out_put_task_struct;
940 	}
941 
942 	ret = ptrace_check_attach(child, request == PTRACE_KILL ||
943 				  request == PTRACE_INTERRUPT);
944 	if (ret < 0)
945 		goto out_put_task_struct;
946 
947 	ret = arch_ptrace(child, request, addr, data);
948 	if (ret || request != PTRACE_DETACH)
949 		ptrace_unfreeze_traced(child);
950 
951  out_put_task_struct:
952 	put_task_struct(child);
953  out:
954 	return ret;
955 }
956 
957 int generic_ptrace_peekdata(struct task_struct *tsk, unsigned long addr,
958 			    unsigned long data)
959 {
960 	unsigned long tmp;
961 	int copied;
962 
963 	copied = access_process_vm(tsk, addr, &tmp, sizeof(tmp), 0);
964 	if (copied != sizeof(tmp))
965 		return -EIO;
966 	return put_user(tmp, (unsigned long __user *)data);
967 }
968 
969 int generic_ptrace_pokedata(struct task_struct *tsk, unsigned long addr,
970 			    unsigned long data)
971 {
972 	int copied;
973 
974 	copied = access_process_vm(tsk, addr, &data, sizeof(data), 1);
975 	return (copied == sizeof(data)) ? 0 : -EIO;
976 }
977 
978 #if defined CONFIG_COMPAT
979 #include <linux/compat.h>
980 
981 int compat_ptrace_request(struct task_struct *child, compat_long_t request,
982 			  compat_ulong_t addr, compat_ulong_t data)
983 {
984 	compat_ulong_t __user *datap = compat_ptr(data);
985 	compat_ulong_t word;
986 	siginfo_t siginfo;
987 	int ret;
988 
989 	switch (request) {
990 	case PTRACE_PEEKTEXT:
991 	case PTRACE_PEEKDATA:
992 		ret = access_process_vm(child, addr, &word, sizeof(word), 0);
993 		if (ret != sizeof(word))
994 			ret = -EIO;
995 		else
996 			ret = put_user(word, datap);
997 		break;
998 
999 	case PTRACE_POKETEXT:
1000 	case PTRACE_POKEDATA:
1001 		ret = access_process_vm(child, addr, &data, sizeof(data), 1);
1002 		ret = (ret != sizeof(data) ? -EIO : 0);
1003 		break;
1004 
1005 	case PTRACE_GETEVENTMSG:
1006 		ret = put_user((compat_ulong_t) child->ptrace_message, datap);
1007 		break;
1008 
1009 	case PTRACE_GETSIGINFO:
1010 		ret = ptrace_getsiginfo(child, &siginfo);
1011 		if (!ret)
1012 			ret = copy_siginfo_to_user32(
1013 				(struct compat_siginfo __user *) datap,
1014 				&siginfo);
1015 		break;
1016 
1017 	case PTRACE_SETSIGINFO:
1018 		memset(&siginfo, 0, sizeof siginfo);
1019 		if (copy_siginfo_from_user32(
1020 			    &siginfo, (struct compat_siginfo __user *) datap))
1021 			ret = -EFAULT;
1022 		else
1023 			ret = ptrace_setsiginfo(child, &siginfo);
1024 		break;
1025 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
1026 	case PTRACE_GETREGSET:
1027 	case PTRACE_SETREGSET:
1028 	{
1029 		struct iovec kiov;
1030 		struct compat_iovec __user *uiov =
1031 			(struct compat_iovec __user *) datap;
1032 		compat_uptr_t ptr;
1033 		compat_size_t len;
1034 
1035 		if (!access_ok(VERIFY_WRITE, uiov, sizeof(*uiov)))
1036 			return -EFAULT;
1037 
1038 		if (__get_user(ptr, &uiov->iov_base) ||
1039 		    __get_user(len, &uiov->iov_len))
1040 			return -EFAULT;
1041 
1042 		kiov.iov_base = compat_ptr(ptr);
1043 		kiov.iov_len = len;
1044 
1045 		ret = ptrace_regset(child, request, addr, &kiov);
1046 		if (!ret)
1047 			ret = __put_user(kiov.iov_len, &uiov->iov_len);
1048 		break;
1049 	}
1050 #endif
1051 
1052 	default:
1053 		ret = ptrace_request(child, request, addr, data);
1054 	}
1055 
1056 	return ret;
1057 }
1058 
1059 asmlinkage long compat_sys_ptrace(compat_long_t request, compat_long_t pid,
1060 				  compat_long_t addr, compat_long_t data)
1061 {
1062 	struct task_struct *child;
1063 	long ret;
1064 
1065 	if (request == PTRACE_TRACEME) {
1066 		ret = ptrace_traceme();
1067 		goto out;
1068 	}
1069 
1070 	child = ptrace_get_task_struct(pid);
1071 	if (IS_ERR(child)) {
1072 		ret = PTR_ERR(child);
1073 		goto out;
1074 	}
1075 
1076 	if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1077 		ret = ptrace_attach(child, request, addr, data);
1078 		/*
1079 		 * Some architectures need to do book-keeping after
1080 		 * a ptrace attach.
1081 		 */
1082 		if (!ret)
1083 			arch_ptrace_attach(child);
1084 		goto out_put_task_struct;
1085 	}
1086 
1087 	ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1088 				  request == PTRACE_INTERRUPT);
1089 	if (!ret) {
1090 		ret = compat_arch_ptrace(child, request, addr, data);
1091 		if (ret || request != PTRACE_DETACH)
1092 			ptrace_unfreeze_traced(child);
1093 	}
1094 
1095  out_put_task_struct:
1096 	put_task_struct(child);
1097  out:
1098 	return ret;
1099 }
1100 #endif	/* CONFIG_COMPAT */
1101 
1102 #ifdef CONFIG_HAVE_HW_BREAKPOINT
1103 int ptrace_get_breakpoints(struct task_struct *tsk)
1104 {
1105 	if (atomic_inc_not_zero(&tsk->ptrace_bp_refcnt))
1106 		return 0;
1107 
1108 	return -1;
1109 }
1110 
1111 void ptrace_put_breakpoints(struct task_struct *tsk)
1112 {
1113 	if (atomic_dec_and_test(&tsk->ptrace_bp_refcnt))
1114 		flush_ptrace_hw_breakpoint(tsk);
1115 }
1116 #endif /* CONFIG_HAVE_HW_BREAKPOINT */
1117