]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - kernel/ptrace.c
Merge tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
[mirror_ubuntu-bionic-kernel.git] / kernel / ptrace.c
1 /*
2 * linux/kernel/ptrace.c
3 *
4 * (C) Copyright 1999 Linus Torvalds
5 *
6 * Common interfaces for "ptrace()" which we do not want
7 * to continually duplicate across every architecture.
8 */
9
10 #include <linux/capability.h>
11 #include <linux/export.h>
12 #include <linux/sched.h>
13 #include <linux/sched/mm.h>
14 #include <linux/sched/coredump.h>
15 #include <linux/sched/task.h>
16 #include <linux/errno.h>
17 #include <linux/mm.h>
18 #include <linux/highmem.h>
19 #include <linux/pagemap.h>
20 #include <linux/ptrace.h>
21 #include <linux/security.h>
22 #include <linux/signal.h>
23 #include <linux/uio.h>
24 #include <linux/audit.h>
25 #include <linux/pid_namespace.h>
26 #include <linux/syscalls.h>
27 #include <linux/uaccess.h>
28 #include <linux/regset.h>
29 #include <linux/hw_breakpoint.h>
30 #include <linux/cn_proc.h>
31 #include <linux/compat.h>
32
33 /*
34 * Access another process' address space via ptrace.
35 * Source/target buffer must be kernel space,
36 * Do not walk the page table directly, use get_user_pages
37 */
38 int ptrace_access_vm(struct task_struct *tsk, unsigned long addr,
39 void *buf, int len, unsigned int gup_flags)
40 {
41 struct mm_struct *mm;
42 int ret;
43
44 mm = get_task_mm(tsk);
45 if (!mm)
46 return 0;
47
48 if (!tsk->ptrace ||
49 (current != tsk->parent) ||
50 ((get_dumpable(mm) != SUID_DUMP_USER) &&
51 !ptracer_capable(tsk, mm->user_ns))) {
52 mmput(mm);
53 return 0;
54 }
55
56 ret = __access_remote_vm(tsk, mm, addr, buf, len, gup_flags);
57 mmput(mm);
58
59 return ret;
60 }
61
62
63 /*
64 * ptrace a task: make the debugger its new parent and
65 * move it to the ptrace list.
66 *
67 * Must be called with the tasklist lock write-held.
68 */
69 void __ptrace_link(struct task_struct *child, struct task_struct *new_parent)
70 {
71 BUG_ON(!list_empty(&child->ptrace_entry));
72 list_add(&child->ptrace_entry, &new_parent->ptraced);
73 child->parent = new_parent;
74 rcu_read_lock();
75 child->ptracer_cred = get_cred(__task_cred(new_parent));
76 rcu_read_unlock();
77 }
78
79 /**
80 * __ptrace_unlink - unlink ptracee and restore its execution state
81 * @child: ptracee to be unlinked
82 *
83 * Remove @child from the ptrace list, move it back to the original parent,
84 * and restore the execution state so that it conforms to the group stop
85 * state.
86 *
87 * Unlinking can happen via two paths - explicit PTRACE_DETACH or ptracer
88 * exiting. For PTRACE_DETACH, unless the ptracee has been killed between
89 * ptrace_check_attach() and here, it's guaranteed to be in TASK_TRACED.
90 * If the ptracer is exiting, the ptracee can be in any state.
91 *
92 * After detach, the ptracee should be in a state which conforms to the
93 * group stop. If the group is stopped or in the process of stopping, the
94 * ptracee should be put into TASK_STOPPED; otherwise, it should be woken
95 * up from TASK_TRACED.
96 *
97 * If the ptracee is in TASK_TRACED and needs to be moved to TASK_STOPPED,
98 * it goes through TRACED -> RUNNING -> STOPPED transition which is similar
99 * to but in the opposite direction of what happens while attaching to a
100 * stopped task. However, in this direction, the intermediate RUNNING
101 * state is not hidden even from the current ptracer and if it immediately
102 * re-attaches and performs a WNOHANG wait(2), it may fail.
103 *
104 * CONTEXT:
105 * write_lock_irq(tasklist_lock)
106 */
107 void __ptrace_unlink(struct task_struct *child)
108 {
109 const struct cred *old_cred;
110 BUG_ON(!child->ptrace);
111
112 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
113
114 child->parent = child->real_parent;
115 list_del_init(&child->ptrace_entry);
116 old_cred = child->ptracer_cred;
117 child->ptracer_cred = NULL;
118 put_cred(old_cred);
119
120 spin_lock(&child->sighand->siglock);
121 child->ptrace = 0;
122 /*
123 * Clear all pending traps and TRAPPING. TRAPPING should be
124 * cleared regardless of JOBCTL_STOP_PENDING. Do it explicitly.
125 */
126 task_clear_jobctl_pending(child, JOBCTL_TRAP_MASK);
127 task_clear_jobctl_trapping(child);
128
129 /*
130 * Reinstate JOBCTL_STOP_PENDING if group stop is in effect and
131 * @child isn't dead.
132 */
133 if (!(child->flags & PF_EXITING) &&
134 (child->signal->flags & SIGNAL_STOP_STOPPED ||
135 child->signal->group_stop_count)) {
136 child->jobctl |= JOBCTL_STOP_PENDING;
137
138 /*
139 * This is only possible if this thread was cloned by the
140 * traced task running in the stopped group, set the signal
141 * for the future reports.
142 * FIXME: we should change ptrace_init_task() to handle this
143 * case.
144 */
145 if (!(child->jobctl & JOBCTL_STOP_SIGMASK))
146 child->jobctl |= SIGSTOP;
147 }
148
149 /*
150 * If transition to TASK_STOPPED is pending or in TASK_TRACED, kick
151 * @child in the butt. Note that @resume should be used iff @child
152 * is in TASK_TRACED; otherwise, we might unduly disrupt
153 * TASK_KILLABLE sleeps.
154 */
155 if (child->jobctl & JOBCTL_STOP_PENDING || task_is_traced(child))
156 ptrace_signal_wake_up(child, true);
157
158 spin_unlock(&child->sighand->siglock);
159 }
160
161 /* Ensure that nothing can wake it up, even SIGKILL */
162 static bool ptrace_freeze_traced(struct task_struct *task)
163 {
164 bool ret = false;
165
166 /* Lockless, nobody but us can set this flag */
167 if (task->jobctl & JOBCTL_LISTENING)
168 return ret;
169
170 spin_lock_irq(&task->sighand->siglock);
171 if (task_is_traced(task) && !__fatal_signal_pending(task)) {
172 task->state = __TASK_TRACED;
173 ret = true;
174 }
175 spin_unlock_irq(&task->sighand->siglock);
176
177 return ret;
178 }
179
180 static void ptrace_unfreeze_traced(struct task_struct *task)
181 {
182 if (task->state != __TASK_TRACED)
183 return;
184
185 WARN_ON(!task->ptrace || task->parent != current);
186
187 spin_lock_irq(&task->sighand->siglock);
188 if (__fatal_signal_pending(task))
189 wake_up_state(task, __TASK_TRACED);
190 else
191 task->state = TASK_TRACED;
192 spin_unlock_irq(&task->sighand->siglock);
193 }
194
195 /**
196 * ptrace_check_attach - check whether ptracee is ready for ptrace operation
197 * @child: ptracee to check for
198 * @ignore_state: don't check whether @child is currently %TASK_TRACED
199 *
200 * Check whether @child is being ptraced by %current and ready for further
201 * ptrace operations. If @ignore_state is %false, @child also should be in
202 * %TASK_TRACED state and on return the child is guaranteed to be traced
203 * and not executing. If @ignore_state is %true, @child can be in any
204 * state.
205 *
206 * CONTEXT:
207 * Grabs and releases tasklist_lock and @child->sighand->siglock.
208 *
209 * RETURNS:
210 * 0 on success, -ESRCH if %child is not ready.
211 */
212 static int ptrace_check_attach(struct task_struct *child, bool ignore_state)
213 {
214 int ret = -ESRCH;
215
216 /*
217 * We take the read lock around doing both checks to close a
218 * possible race where someone else was tracing our child and
219 * detached between these two checks. After this locked check,
220 * we are sure that this is our traced child and that can only
221 * be changed by us so it's not changing right after this.
222 */
223 read_lock(&tasklist_lock);
224 if (child->ptrace && child->parent == current) {
225 WARN_ON(child->state == __TASK_TRACED);
226 /*
227 * child->sighand can't be NULL, release_task()
228 * does ptrace_unlink() before __exit_signal().
229 */
230 if (ignore_state || ptrace_freeze_traced(child))
231 ret = 0;
232 }
233 read_unlock(&tasklist_lock);
234
235 if (!ret && !ignore_state) {
236 if (!wait_task_inactive(child, __TASK_TRACED)) {
237 /*
238 * This can only happen if may_ptrace_stop() fails and
239 * ptrace_stop() changes ->state back to TASK_RUNNING,
240 * so we should not worry about leaking __TASK_TRACED.
241 */
242 WARN_ON(child->state == __TASK_TRACED);
243 ret = -ESRCH;
244 }
245 }
246
247 return ret;
248 }
249
250 static int ptrace_has_cap(struct user_namespace *ns, unsigned int mode)
251 {
252 if (mode & PTRACE_MODE_NOAUDIT)
253 return has_ns_capability_noaudit(current, ns, CAP_SYS_PTRACE);
254 else
255 return has_ns_capability(current, ns, CAP_SYS_PTRACE);
256 }
257
258 /* Returns 0 on success, -errno on denial. */
259 static int __ptrace_may_access(struct task_struct *task, unsigned int mode)
260 {
261 const struct cred *cred = current_cred(), *tcred;
262 struct mm_struct *mm;
263 kuid_t caller_uid;
264 kgid_t caller_gid;
265
266 if (!(mode & PTRACE_MODE_FSCREDS) == !(mode & PTRACE_MODE_REALCREDS)) {
267 WARN(1, "denying ptrace access check without PTRACE_MODE_*CREDS\n");
268 return -EPERM;
269 }
270
271 /* May we inspect the given task?
272 * This check is used both for attaching with ptrace
273 * and for allowing access to sensitive information in /proc.
274 *
275 * ptrace_attach denies several cases that /proc allows
276 * because setting up the necessary parent/child relationship
277 * or halting the specified task is impossible.
278 */
279
280 /* Don't let security modules deny introspection */
281 if (same_thread_group(task, current))
282 return 0;
283 rcu_read_lock();
284 if (mode & PTRACE_MODE_FSCREDS) {
285 caller_uid = cred->fsuid;
286 caller_gid = cred->fsgid;
287 } else {
288 /*
289 * Using the euid would make more sense here, but something
290 * in userland might rely on the old behavior, and this
291 * shouldn't be a security problem since
292 * PTRACE_MODE_REALCREDS implies that the caller explicitly
293 * used a syscall that requests access to another process
294 * (and not a filesystem syscall to procfs).
295 */
296 caller_uid = cred->uid;
297 caller_gid = cred->gid;
298 }
299 tcred = __task_cred(task);
300 if (uid_eq(caller_uid, tcred->euid) &&
301 uid_eq(caller_uid, tcred->suid) &&
302 uid_eq(caller_uid, tcred->uid) &&
303 gid_eq(caller_gid, tcred->egid) &&
304 gid_eq(caller_gid, tcred->sgid) &&
305 gid_eq(caller_gid, tcred->gid))
306 goto ok;
307 if (ptrace_has_cap(tcred->user_ns, mode))
308 goto ok;
309 rcu_read_unlock();
310 return -EPERM;
311 ok:
312 rcu_read_unlock();
313 mm = task->mm;
314 if (mm &&
315 ((get_dumpable(mm) != SUID_DUMP_USER) &&
316 !ptrace_has_cap(mm->user_ns, mode)))
317 return -EPERM;
318
319 return security_ptrace_access_check(task, mode);
320 }
321
322 bool ptrace_may_access(struct task_struct *task, unsigned int mode)
323 {
324 int err;
325 task_lock(task);
326 err = __ptrace_may_access(task, mode);
327 task_unlock(task);
328 return !err;
329 }
330
331 static int ptrace_attach(struct task_struct *task, long request,
332 unsigned long addr,
333 unsigned long flags)
334 {
335 bool seize = (request == PTRACE_SEIZE);
336 int retval;
337
338 retval = -EIO;
339 if (seize) {
340 if (addr != 0)
341 goto out;
342 if (flags & ~(unsigned long)PTRACE_O_MASK)
343 goto out;
344 flags = PT_PTRACED | PT_SEIZED | (flags << PT_OPT_FLAG_SHIFT);
345 } else {
346 flags = PT_PTRACED;
347 }
348
349 audit_ptrace(task);
350
351 retval = -EPERM;
352 if (unlikely(task->flags & PF_KTHREAD))
353 goto out;
354 if (same_thread_group(task, current))
355 goto out;
356
357 /*
358 * Protect exec's credential calculations against our interference;
359 * SUID, SGID and LSM creds get determined differently
360 * under ptrace.
361 */
362 retval = -ERESTARTNOINTR;
363 if (mutex_lock_interruptible(&task->signal->cred_guard_mutex))
364 goto out;
365
366 task_lock(task);
367 retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH_REALCREDS);
368 task_unlock(task);
369 if (retval)
370 goto unlock_creds;
371
372 write_lock_irq(&tasklist_lock);
373 retval = -EPERM;
374 if (unlikely(task->exit_state))
375 goto unlock_tasklist;
376 if (task->ptrace)
377 goto unlock_tasklist;
378
379 if (seize)
380 flags |= PT_SEIZED;
381 task->ptrace = flags;
382
383 __ptrace_link(task, current);
384
385 /* SEIZE doesn't trap tracee on attach */
386 if (!seize)
387 send_sig_info(SIGSTOP, SEND_SIG_FORCED, task);
388
389 spin_lock(&task->sighand->siglock);
390
391 /*
392 * If the task is already STOPPED, set JOBCTL_TRAP_STOP and
393 * TRAPPING, and kick it so that it transits to TRACED. TRAPPING
394 * will be cleared if the child completes the transition or any
395 * event which clears the group stop states happens. We'll wait
396 * for the transition to complete before returning from this
397 * function.
398 *
399 * This hides STOPPED -> RUNNING -> TRACED transition from the
400 * attaching thread but a different thread in the same group can
401 * still observe the transient RUNNING state. IOW, if another
402 * thread's WNOHANG wait(2) on the stopped tracee races against
403 * ATTACH, the wait(2) may fail due to the transient RUNNING.
404 *
405 * The following task_is_stopped() test is safe as both transitions
406 * in and out of STOPPED are protected by siglock.
407 */
408 if (task_is_stopped(task) &&
409 task_set_jobctl_pending(task, JOBCTL_TRAP_STOP | JOBCTL_TRAPPING))
410 signal_wake_up_state(task, __TASK_STOPPED);
411
412 spin_unlock(&task->sighand->siglock);
413
414 retval = 0;
415 unlock_tasklist:
416 write_unlock_irq(&tasklist_lock);
417 unlock_creds:
418 mutex_unlock(&task->signal->cred_guard_mutex);
419 out:
420 if (!retval) {
421 /*
422 * We do not bother to change retval or clear JOBCTL_TRAPPING
423 * if wait_on_bit() was interrupted by SIGKILL. The tracer will
424 * not return to user-mode, it will exit and clear this bit in
425 * __ptrace_unlink() if it wasn't already cleared by the tracee;
426 * and until then nobody can ptrace this task.
427 */
428 wait_on_bit(&task->jobctl, JOBCTL_TRAPPING_BIT, TASK_KILLABLE);
429 proc_ptrace_connector(task, PTRACE_ATTACH);
430 }
431
432 return retval;
433 }
434
435 /**
436 * ptrace_traceme -- helper for PTRACE_TRACEME
437 *
438 * Performs checks and sets PT_PTRACED.
439 * Should be used by all ptrace implementations for PTRACE_TRACEME.
440 */
441 static int ptrace_traceme(void)
442 {
443 int ret = -EPERM;
444
445 write_lock_irq(&tasklist_lock);
446 /* Are we already being traced? */
447 if (!current->ptrace) {
448 ret = security_ptrace_traceme(current->parent);
449 /*
450 * Check PF_EXITING to ensure ->real_parent has not passed
451 * exit_ptrace(). Otherwise we don't report the error but
452 * pretend ->real_parent untraces us right after return.
453 */
454 if (!ret && !(current->real_parent->flags & PF_EXITING)) {
455 current->ptrace = PT_PTRACED;
456 __ptrace_link(current, current->real_parent);
457 }
458 }
459 write_unlock_irq(&tasklist_lock);
460
461 return ret;
462 }
463
464 /*
465 * Called with irqs disabled, returns true if childs should reap themselves.
466 */
467 static int ignoring_children(struct sighand_struct *sigh)
468 {
469 int ret;
470 spin_lock(&sigh->siglock);
471 ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
472 (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
473 spin_unlock(&sigh->siglock);
474 return ret;
475 }
476
477 /*
478 * Called with tasklist_lock held for writing.
479 * Unlink a traced task, and clean it up if it was a traced zombie.
480 * Return true if it needs to be reaped with release_task().
481 * (We can't call release_task() here because we already hold tasklist_lock.)
482 *
483 * If it's a zombie, our attachedness prevented normal parent notification
484 * or self-reaping. Do notification now if it would have happened earlier.
485 * If it should reap itself, return true.
486 *
487 * If it's our own child, there is no notification to do. But if our normal
488 * children self-reap, then this child was prevented by ptrace and we must
489 * reap it now, in that case we must also wake up sub-threads sleeping in
490 * do_wait().
491 */
492 static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
493 {
494 bool dead;
495
496 __ptrace_unlink(p);
497
498 if (p->exit_state != EXIT_ZOMBIE)
499 return false;
500
501 dead = !thread_group_leader(p);
502
503 if (!dead && thread_group_empty(p)) {
504 if (!same_thread_group(p->real_parent, tracer))
505 dead = do_notify_parent(p, p->exit_signal);
506 else if (ignoring_children(tracer->sighand)) {
507 __wake_up_parent(p, tracer);
508 dead = true;
509 }
510 }
511 /* Mark it as in the process of being reaped. */
512 if (dead)
513 p->exit_state = EXIT_DEAD;
514 return dead;
515 }
516
517 static int ptrace_detach(struct task_struct *child, unsigned int data)
518 {
519 if (!valid_signal(data))
520 return -EIO;
521
522 /* Architecture-specific hardware disable .. */
523 ptrace_disable(child);
524
525 write_lock_irq(&tasklist_lock);
526 /*
527 * We rely on ptrace_freeze_traced(). It can't be killed and
528 * untraced by another thread, it can't be a zombie.
529 */
530 WARN_ON(!child->ptrace || child->exit_state);
531 /*
532 * tasklist_lock avoids the race with wait_task_stopped(), see
533 * the comment in ptrace_resume().
534 */
535 child->exit_code = data;
536 __ptrace_detach(current, child);
537 write_unlock_irq(&tasklist_lock);
538
539 proc_ptrace_connector(child, PTRACE_DETACH);
540
541 return 0;
542 }
543
544 /*
545 * Detach all tasks we were using ptrace on. Called with tasklist held
546 * for writing.
547 */
548 void exit_ptrace(struct task_struct *tracer, struct list_head *dead)
549 {
550 struct task_struct *p, *n;
551
552 list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
553 if (unlikely(p->ptrace & PT_EXITKILL))
554 send_sig_info(SIGKILL, SEND_SIG_FORCED, p);
555
556 if (__ptrace_detach(tracer, p))
557 list_add(&p->ptrace_entry, dead);
558 }
559 }
560
561 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
562 {
563 int copied = 0;
564
565 while (len > 0) {
566 char buf[128];
567 int this_len, retval;
568
569 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
570 retval = ptrace_access_vm(tsk, src, buf, this_len, FOLL_FORCE);
571
572 if (!retval) {
573 if (copied)
574 break;
575 return -EIO;
576 }
577 if (copy_to_user(dst, buf, retval))
578 return -EFAULT;
579 copied += retval;
580 src += retval;
581 dst += retval;
582 len -= retval;
583 }
584 return copied;
585 }
586
587 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
588 {
589 int copied = 0;
590
591 while (len > 0) {
592 char buf[128];
593 int this_len, retval;
594
595 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
596 if (copy_from_user(buf, src, this_len))
597 return -EFAULT;
598 retval = ptrace_access_vm(tsk, dst, buf, this_len,
599 FOLL_FORCE | FOLL_WRITE);
600 if (!retval) {
601 if (copied)
602 break;
603 return -EIO;
604 }
605 copied += retval;
606 src += retval;
607 dst += retval;
608 len -= retval;
609 }
610 return copied;
611 }
612
613 static int ptrace_setoptions(struct task_struct *child, unsigned long data)
614 {
615 unsigned flags;
616
617 if (data & ~(unsigned long)PTRACE_O_MASK)
618 return -EINVAL;
619
620 if (unlikely(data & PTRACE_O_SUSPEND_SECCOMP)) {
621 if (!IS_ENABLED(CONFIG_CHECKPOINT_RESTORE) ||
622 !IS_ENABLED(CONFIG_SECCOMP))
623 return -EINVAL;
624
625 if (!capable(CAP_SYS_ADMIN))
626 return -EPERM;
627
628 if (seccomp_mode(&current->seccomp) != SECCOMP_MODE_DISABLED ||
629 current->ptrace & PT_SUSPEND_SECCOMP)
630 return -EPERM;
631 }
632
633 /* Avoid intermediate state when all opts are cleared */
634 flags = child->ptrace;
635 flags &= ~(PTRACE_O_MASK << PT_OPT_FLAG_SHIFT);
636 flags |= (data << PT_OPT_FLAG_SHIFT);
637 child->ptrace = flags;
638
639 return 0;
640 }
641
642 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t *info)
643 {
644 unsigned long flags;
645 int error = -ESRCH;
646
647 if (lock_task_sighand(child, &flags)) {
648 error = -EINVAL;
649 if (likely(child->last_siginfo != NULL)) {
650 *info = *child->last_siginfo;
651 error = 0;
652 }
653 unlock_task_sighand(child, &flags);
654 }
655 return error;
656 }
657
658 static int ptrace_setsiginfo(struct task_struct *child, const siginfo_t *info)
659 {
660 unsigned long flags;
661 int error = -ESRCH;
662
663 if (lock_task_sighand(child, &flags)) {
664 error = -EINVAL;
665 if (likely(child->last_siginfo != NULL)) {
666 *child->last_siginfo = *info;
667 error = 0;
668 }
669 unlock_task_sighand(child, &flags);
670 }
671 return error;
672 }
673
674 static int ptrace_peek_siginfo(struct task_struct *child,
675 unsigned long addr,
676 unsigned long data)
677 {
678 struct ptrace_peeksiginfo_args arg;
679 struct sigpending *pending;
680 struct sigqueue *q;
681 int ret, i;
682
683 ret = copy_from_user(&arg, (void __user *) addr,
684 sizeof(struct ptrace_peeksiginfo_args));
685 if (ret)
686 return -EFAULT;
687
688 if (arg.flags & ~PTRACE_PEEKSIGINFO_SHARED)
689 return -EINVAL; /* unknown flags */
690
691 if (arg.nr < 0)
692 return -EINVAL;
693
694 if (arg.flags & PTRACE_PEEKSIGINFO_SHARED)
695 pending = &child->signal->shared_pending;
696 else
697 pending = &child->pending;
698
699 for (i = 0; i < arg.nr; ) {
700 siginfo_t info;
701 s32 off = arg.off + i;
702
703 spin_lock_irq(&child->sighand->siglock);
704 list_for_each_entry(q, &pending->list, list) {
705 if (!off--) {
706 copy_siginfo(&info, &q->info);
707 break;
708 }
709 }
710 spin_unlock_irq(&child->sighand->siglock);
711
712 if (off >= 0) /* beyond the end of the list */
713 break;
714
715 #ifdef CONFIG_COMPAT
716 if (unlikely(in_compat_syscall())) {
717 compat_siginfo_t __user *uinfo = compat_ptr(data);
718
719 if (copy_siginfo_to_user32(uinfo, &info) ||
720 __put_user(info.si_code, &uinfo->si_code)) {
721 ret = -EFAULT;
722 break;
723 }
724
725 } else
726 #endif
727 {
728 siginfo_t __user *uinfo = (siginfo_t __user *) data;
729
730 if (copy_siginfo_to_user(uinfo, &info) ||
731 __put_user(info.si_code, &uinfo->si_code)) {
732 ret = -EFAULT;
733 break;
734 }
735 }
736
737 data += sizeof(siginfo_t);
738 i++;
739
740 if (signal_pending(current))
741 break;
742
743 cond_resched();
744 }
745
746 if (i > 0)
747 return i;
748
749 return ret;
750 }
751
752 #ifdef PTRACE_SINGLESTEP
753 #define is_singlestep(request) ((request) == PTRACE_SINGLESTEP)
754 #else
755 #define is_singlestep(request) 0
756 #endif
757
758 #ifdef PTRACE_SINGLEBLOCK
759 #define is_singleblock(request) ((request) == PTRACE_SINGLEBLOCK)
760 #else
761 #define is_singleblock(request) 0
762 #endif
763
764 #ifdef PTRACE_SYSEMU
765 #define is_sysemu_singlestep(request) ((request) == PTRACE_SYSEMU_SINGLESTEP)
766 #else
767 #define is_sysemu_singlestep(request) 0
768 #endif
769
770 static int ptrace_resume(struct task_struct *child, long request,
771 unsigned long data)
772 {
773 bool need_siglock;
774
775 if (!valid_signal(data))
776 return -EIO;
777
778 if (request == PTRACE_SYSCALL)
779 set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
780 else
781 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
782
783 #ifdef TIF_SYSCALL_EMU
784 if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
785 set_tsk_thread_flag(child, TIF_SYSCALL_EMU);
786 else
787 clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
788 #endif
789
790 if (is_singleblock(request)) {
791 if (unlikely(!arch_has_block_step()))
792 return -EIO;
793 user_enable_block_step(child);
794 } else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
795 if (unlikely(!arch_has_single_step()))
796 return -EIO;
797 user_enable_single_step(child);
798 } else {
799 user_disable_single_step(child);
800 }
801
802 /*
803 * Change ->exit_code and ->state under siglock to avoid the race
804 * with wait_task_stopped() in between; a non-zero ->exit_code will
805 * wrongly look like another report from tracee.
806 *
807 * Note that we need siglock even if ->exit_code == data and/or this
808 * status was not reported yet, the new status must not be cleared by
809 * wait_task_stopped() after resume.
810 *
811 * If data == 0 we do not care if wait_task_stopped() reports the old
812 * status and clears the code too; this can't race with the tracee, it
813 * takes siglock after resume.
814 */
815 need_siglock = data && !thread_group_empty(current);
816 if (need_siglock)
817 spin_lock_irq(&child->sighand->siglock);
818 child->exit_code = data;
819 wake_up_state(child, __TASK_TRACED);
820 if (need_siglock)
821 spin_unlock_irq(&child->sighand->siglock);
822
823 return 0;
824 }
825
826 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
827
828 static const struct user_regset *
829 find_regset(const struct user_regset_view *view, unsigned int type)
830 {
831 const struct user_regset *regset;
832 int n;
833
834 for (n = 0; n < view->n; ++n) {
835 regset = view->regsets + n;
836 if (regset->core_note_type == type)
837 return regset;
838 }
839
840 return NULL;
841 }
842
843 static int ptrace_regset(struct task_struct *task, int req, unsigned int type,
844 struct iovec *kiov)
845 {
846 const struct user_regset_view *view = task_user_regset_view(task);
847 const struct user_regset *regset = find_regset(view, type);
848 int regset_no;
849
850 if (!regset || (kiov->iov_len % regset->size) != 0)
851 return -EINVAL;
852
853 regset_no = regset - view->regsets;
854 kiov->iov_len = min(kiov->iov_len,
855 (__kernel_size_t) (regset->n * regset->size));
856
857 if (req == PTRACE_GETREGSET)
858 return copy_regset_to_user(task, view, regset_no, 0,
859 kiov->iov_len, kiov->iov_base);
860 else
861 return copy_regset_from_user(task, view, regset_no, 0,
862 kiov->iov_len, kiov->iov_base);
863 }
864
865 /*
866 * This is declared in linux/regset.h and defined in machine-dependent
867 * code. We put the export here, near the primary machine-neutral use,
868 * to ensure no machine forgets it.
869 */
870 EXPORT_SYMBOL_GPL(task_user_regset_view);
871 #endif
872
873 int ptrace_request(struct task_struct *child, long request,
874 unsigned long addr, unsigned long data)
875 {
876 bool seized = child->ptrace & PT_SEIZED;
877 int ret = -EIO;
878 siginfo_t siginfo, *si;
879 void __user *datavp = (void __user *) data;
880 unsigned long __user *datalp = datavp;
881 unsigned long flags;
882
883 switch (request) {
884 case PTRACE_PEEKTEXT:
885 case PTRACE_PEEKDATA:
886 return generic_ptrace_peekdata(child, addr, data);
887 case PTRACE_POKETEXT:
888 case PTRACE_POKEDATA:
889 return generic_ptrace_pokedata(child, addr, data);
890
891 #ifdef PTRACE_OLDSETOPTIONS
892 case PTRACE_OLDSETOPTIONS:
893 #endif
894 case PTRACE_SETOPTIONS:
895 ret = ptrace_setoptions(child, data);
896 break;
897 case PTRACE_GETEVENTMSG:
898 ret = put_user(child->ptrace_message, datalp);
899 break;
900
901 case PTRACE_PEEKSIGINFO:
902 ret = ptrace_peek_siginfo(child, addr, data);
903 break;
904
905 case PTRACE_GETSIGINFO:
906 ret = ptrace_getsiginfo(child, &siginfo);
907 if (!ret)
908 ret = copy_siginfo_to_user(datavp, &siginfo);
909 break;
910
911 case PTRACE_SETSIGINFO:
912 if (copy_from_user(&siginfo, datavp, sizeof siginfo))
913 ret = -EFAULT;
914 else
915 ret = ptrace_setsiginfo(child, &siginfo);
916 break;
917
918 case PTRACE_GETSIGMASK:
919 if (addr != sizeof(sigset_t)) {
920 ret = -EINVAL;
921 break;
922 }
923
924 if (copy_to_user(datavp, &child->blocked, sizeof(sigset_t)))
925 ret = -EFAULT;
926 else
927 ret = 0;
928
929 break;
930
931 case PTRACE_SETSIGMASK: {
932 sigset_t new_set;
933
934 if (addr != sizeof(sigset_t)) {
935 ret = -EINVAL;
936 break;
937 }
938
939 if (copy_from_user(&new_set, datavp, sizeof(sigset_t))) {
940 ret = -EFAULT;
941 break;
942 }
943
944 sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
945
946 /*
947 * Every thread does recalc_sigpending() after resume, so
948 * retarget_shared_pending() and recalc_sigpending() are not
949 * called here.
950 */
951 spin_lock_irq(&child->sighand->siglock);
952 child->blocked = new_set;
953 spin_unlock_irq(&child->sighand->siglock);
954
955 ret = 0;
956 break;
957 }
958
959 case PTRACE_INTERRUPT:
960 /*
961 * Stop tracee without any side-effect on signal or job
962 * control. At least one trap is guaranteed to happen
963 * after this request. If @child is already trapped, the
964 * current trap is not disturbed and another trap will
965 * happen after the current trap is ended with PTRACE_CONT.
966 *
967 * The actual trap might not be PTRACE_EVENT_STOP trap but
968 * the pending condition is cleared regardless.
969 */
970 if (unlikely(!seized || !lock_task_sighand(child, &flags)))
971 break;
972
973 /*
974 * INTERRUPT doesn't disturb existing trap sans one
975 * exception. If ptracer issued LISTEN for the current
976 * STOP, this INTERRUPT should clear LISTEN and re-trap
977 * tracee into STOP.
978 */
979 if (likely(task_set_jobctl_pending(child, JOBCTL_TRAP_STOP)))
980 ptrace_signal_wake_up(child, child->jobctl & JOBCTL_LISTENING);
981
982 unlock_task_sighand(child, &flags);
983 ret = 0;
984 break;
985
986 case PTRACE_LISTEN:
987 /*
988 * Listen for events. Tracee must be in STOP. It's not
989 * resumed per-se but is not considered to be in TRACED by
990 * wait(2) or ptrace(2). If an async event (e.g. group
991 * stop state change) happens, tracee will enter STOP trap
992 * again. Alternatively, ptracer can issue INTERRUPT to
993 * finish listening and re-trap tracee into STOP.
994 */
995 if (unlikely(!seized || !lock_task_sighand(child, &flags)))
996 break;
997
998 si = child->last_siginfo;
999 if (likely(si && (si->si_code >> 8) == PTRACE_EVENT_STOP)) {
1000 child->jobctl |= JOBCTL_LISTENING;
1001 /*
1002 * If NOTIFY is set, it means event happened between
1003 * start of this trap and now. Trigger re-trap.
1004 */
1005 if (child->jobctl & JOBCTL_TRAP_NOTIFY)
1006 ptrace_signal_wake_up(child, true);
1007 ret = 0;
1008 }
1009 unlock_task_sighand(child, &flags);
1010 break;
1011
1012 case PTRACE_DETACH: /* detach a process that was attached. */
1013 ret = ptrace_detach(child, data);
1014 break;
1015
1016 #ifdef CONFIG_BINFMT_ELF_FDPIC
1017 case PTRACE_GETFDPIC: {
1018 struct mm_struct *mm = get_task_mm(child);
1019 unsigned long tmp = 0;
1020
1021 ret = -ESRCH;
1022 if (!mm)
1023 break;
1024
1025 switch (addr) {
1026 case PTRACE_GETFDPIC_EXEC:
1027 tmp = mm->context.exec_fdpic_loadmap;
1028 break;
1029 case PTRACE_GETFDPIC_INTERP:
1030 tmp = mm->context.interp_fdpic_loadmap;
1031 break;
1032 default:
1033 break;
1034 }
1035 mmput(mm);
1036
1037 ret = put_user(tmp, datalp);
1038 break;
1039 }
1040 #endif
1041
1042 #ifdef PTRACE_SINGLESTEP
1043 case PTRACE_SINGLESTEP:
1044 #endif
1045 #ifdef PTRACE_SINGLEBLOCK
1046 case PTRACE_SINGLEBLOCK:
1047 #endif
1048 #ifdef PTRACE_SYSEMU
1049 case PTRACE_SYSEMU:
1050 case PTRACE_SYSEMU_SINGLESTEP:
1051 #endif
1052 case PTRACE_SYSCALL:
1053 case PTRACE_CONT:
1054 return ptrace_resume(child, request, data);
1055
1056 case PTRACE_KILL:
1057 if (child->exit_state) /* already dead */
1058 return 0;
1059 return ptrace_resume(child, request, SIGKILL);
1060
1061 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
1062 case PTRACE_GETREGSET:
1063 case PTRACE_SETREGSET: {
1064 struct iovec kiov;
1065 struct iovec __user *uiov = datavp;
1066
1067 if (!access_ok(VERIFY_WRITE, uiov, sizeof(*uiov)))
1068 return -EFAULT;
1069
1070 if (__get_user(kiov.iov_base, &uiov->iov_base) ||
1071 __get_user(kiov.iov_len, &uiov->iov_len))
1072 return -EFAULT;
1073
1074 ret = ptrace_regset(child, request, addr, &kiov);
1075 if (!ret)
1076 ret = __put_user(kiov.iov_len, &uiov->iov_len);
1077 break;
1078 }
1079 #endif
1080
1081 case PTRACE_SECCOMP_GET_FILTER:
1082 ret = seccomp_get_filter(child, addr, datavp);
1083 break;
1084
1085 default:
1086 break;
1087 }
1088
1089 return ret;
1090 }
1091
1092 static struct task_struct *ptrace_get_task_struct(pid_t pid)
1093 {
1094 struct task_struct *child;
1095
1096 rcu_read_lock();
1097 child = find_task_by_vpid(pid);
1098 if (child)
1099 get_task_struct(child);
1100 rcu_read_unlock();
1101
1102 if (!child)
1103 return ERR_PTR(-ESRCH);
1104 return child;
1105 }
1106
1107 #ifndef arch_ptrace_attach
1108 #define arch_ptrace_attach(child) do { } while (0)
1109 #endif
1110
1111 SYSCALL_DEFINE4(ptrace, long, request, long, pid, unsigned long, addr,
1112 unsigned long, data)
1113 {
1114 struct task_struct *child;
1115 long ret;
1116
1117 if (request == PTRACE_TRACEME) {
1118 ret = ptrace_traceme();
1119 if (!ret)
1120 arch_ptrace_attach(current);
1121 goto out;
1122 }
1123
1124 child = ptrace_get_task_struct(pid);
1125 if (IS_ERR(child)) {
1126 ret = PTR_ERR(child);
1127 goto out;
1128 }
1129
1130 if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1131 ret = ptrace_attach(child, request, addr, data);
1132 /*
1133 * Some architectures need to do book-keeping after
1134 * a ptrace attach.
1135 */
1136 if (!ret)
1137 arch_ptrace_attach(child);
1138 goto out_put_task_struct;
1139 }
1140
1141 ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1142 request == PTRACE_INTERRUPT);
1143 if (ret < 0)
1144 goto out_put_task_struct;
1145
1146 ret = arch_ptrace(child, request, addr, data);
1147 if (ret || request != PTRACE_DETACH)
1148 ptrace_unfreeze_traced(child);
1149
1150 out_put_task_struct:
1151 put_task_struct(child);
1152 out:
1153 return ret;
1154 }
1155
1156 int generic_ptrace_peekdata(struct task_struct *tsk, unsigned long addr,
1157 unsigned long data)
1158 {
1159 unsigned long tmp;
1160 int copied;
1161
1162 copied = ptrace_access_vm(tsk, addr, &tmp, sizeof(tmp), FOLL_FORCE);
1163 if (copied != sizeof(tmp))
1164 return -EIO;
1165 return put_user(tmp, (unsigned long __user *)data);
1166 }
1167
1168 int generic_ptrace_pokedata(struct task_struct *tsk, unsigned long addr,
1169 unsigned long data)
1170 {
1171 int copied;
1172
1173 copied = ptrace_access_vm(tsk, addr, &data, sizeof(data),
1174 FOLL_FORCE | FOLL_WRITE);
1175 return (copied == sizeof(data)) ? 0 : -EIO;
1176 }
1177
1178 #if defined CONFIG_COMPAT
1179
1180 int compat_ptrace_request(struct task_struct *child, compat_long_t request,
1181 compat_ulong_t addr, compat_ulong_t data)
1182 {
1183 compat_ulong_t __user *datap = compat_ptr(data);
1184 compat_ulong_t word;
1185 siginfo_t siginfo;
1186 int ret;
1187
1188 switch (request) {
1189 case PTRACE_PEEKTEXT:
1190 case PTRACE_PEEKDATA:
1191 ret = ptrace_access_vm(child, addr, &word, sizeof(word),
1192 FOLL_FORCE);
1193 if (ret != sizeof(word))
1194 ret = -EIO;
1195 else
1196 ret = put_user(word, datap);
1197 break;
1198
1199 case PTRACE_POKETEXT:
1200 case PTRACE_POKEDATA:
1201 ret = ptrace_access_vm(child, addr, &data, sizeof(data),
1202 FOLL_FORCE | FOLL_WRITE);
1203 ret = (ret != sizeof(data) ? -EIO : 0);
1204 break;
1205
1206 case PTRACE_GETEVENTMSG:
1207 ret = put_user((compat_ulong_t) child->ptrace_message, datap);
1208 break;
1209
1210 case PTRACE_GETSIGINFO:
1211 ret = ptrace_getsiginfo(child, &siginfo);
1212 if (!ret)
1213 ret = copy_siginfo_to_user32(
1214 (struct compat_siginfo __user *) datap,
1215 &siginfo);
1216 break;
1217
1218 case PTRACE_SETSIGINFO:
1219 memset(&siginfo, 0, sizeof siginfo);
1220 if (copy_siginfo_from_user32(
1221 &siginfo, (struct compat_siginfo __user *) datap))
1222 ret = -EFAULT;
1223 else
1224 ret = ptrace_setsiginfo(child, &siginfo);
1225 break;
1226 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
1227 case PTRACE_GETREGSET:
1228 case PTRACE_SETREGSET:
1229 {
1230 struct iovec kiov;
1231 struct compat_iovec __user *uiov =
1232 (struct compat_iovec __user *) datap;
1233 compat_uptr_t ptr;
1234 compat_size_t len;
1235
1236 if (!access_ok(VERIFY_WRITE, uiov, sizeof(*uiov)))
1237 return -EFAULT;
1238
1239 if (__get_user(ptr, &uiov->iov_base) ||
1240 __get_user(len, &uiov->iov_len))
1241 return -EFAULT;
1242
1243 kiov.iov_base = compat_ptr(ptr);
1244 kiov.iov_len = len;
1245
1246 ret = ptrace_regset(child, request, addr, &kiov);
1247 if (!ret)
1248 ret = __put_user(kiov.iov_len, &uiov->iov_len);
1249 break;
1250 }
1251 #endif
1252
1253 default:
1254 ret = ptrace_request(child, request, addr, data);
1255 }
1256
1257 return ret;
1258 }
1259
1260 COMPAT_SYSCALL_DEFINE4(ptrace, compat_long_t, request, compat_long_t, pid,
1261 compat_long_t, addr, compat_long_t, data)
1262 {
1263 struct task_struct *child;
1264 long ret;
1265
1266 if (request == PTRACE_TRACEME) {
1267 ret = ptrace_traceme();
1268 goto out;
1269 }
1270
1271 child = ptrace_get_task_struct(pid);
1272 if (IS_ERR(child)) {
1273 ret = PTR_ERR(child);
1274 goto out;
1275 }
1276
1277 if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1278 ret = ptrace_attach(child, request, addr, data);
1279 /*
1280 * Some architectures need to do book-keeping after
1281 * a ptrace attach.
1282 */
1283 if (!ret)
1284 arch_ptrace_attach(child);
1285 goto out_put_task_struct;
1286 }
1287
1288 ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1289 request == PTRACE_INTERRUPT);
1290 if (!ret) {
1291 ret = compat_arch_ptrace(child, request, addr, data);
1292 if (ret || request != PTRACE_DETACH)
1293 ptrace_unfreeze_traced(child);
1294 }
1295
1296 out_put_task_struct:
1297 put_task_struct(child);
1298 out:
1299 return ret;
1300 }
1301 #endif /* CONFIG_COMPAT */