rtlwifi: rtl8192se: Replace local bit manipulation macros
[linux/fpc-iii.git] / kernel / ptrace.c
blobcb9ddcc08119added2355311507c65db3849b0c4
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * linux/kernel/ptrace.c
5 * (C) Copyright 1999 Linus Torvalds
7 * Common interfaces for "ptrace()" which we do not want
8 * to continually duplicate across every architecture.
9 */
11 #include <linux/capability.h>
12 #include <linux/export.h>
13 #include <linux/sched.h>
14 #include <linux/sched/mm.h>
15 #include <linux/sched/coredump.h>
16 #include <linux/sched/task.h>
17 #include <linux/errno.h>
18 #include <linux/mm.h>
19 #include <linux/highmem.h>
20 #include <linux/pagemap.h>
21 #include <linux/ptrace.h>
22 #include <linux/security.h>
23 #include <linux/signal.h>
24 #include <linux/uio.h>
25 #include <linux/audit.h>
26 #include <linux/pid_namespace.h>
27 #include <linux/syscalls.h>
28 #include <linux/uaccess.h>
29 #include <linux/regset.h>
30 #include <linux/hw_breakpoint.h>
31 #include <linux/cn_proc.h>
32 #include <linux/compat.h>
33 #include <linux/sched/signal.h>
35 #include <asm/syscall.h> /* for syscall_get_* */
38 * Access another process' address space via ptrace.
39 * Source/target buffer must be kernel space,
40 * Do not walk the page table directly, use get_user_pages
42 int ptrace_access_vm(struct task_struct *tsk, unsigned long addr,
43 void *buf, int len, unsigned int gup_flags)
45 struct mm_struct *mm;
46 int ret;
48 mm = get_task_mm(tsk);
49 if (!mm)
50 return 0;
52 if (!tsk->ptrace ||
53 (current != tsk->parent) ||
54 ((get_dumpable(mm) != SUID_DUMP_USER) &&
55 !ptracer_capable(tsk, mm->user_ns))) {
56 mmput(mm);
57 return 0;
60 ret = __access_remote_vm(tsk, mm, addr, buf, len, gup_flags);
61 mmput(mm);
63 return ret;
67 void __ptrace_link(struct task_struct *child, struct task_struct *new_parent,
68 const struct cred *ptracer_cred)
70 BUG_ON(!list_empty(&child->ptrace_entry));
71 list_add(&child->ptrace_entry, &new_parent->ptraced);
72 child->parent = new_parent;
73 child->ptracer_cred = get_cred(ptracer_cred);
77 * ptrace a task: make the debugger its new parent and
78 * move it to the ptrace list.
80 * Must be called with the tasklist lock write-held.
82 static void ptrace_link(struct task_struct *child, struct task_struct *new_parent)
84 __ptrace_link(child, new_parent, current_cred());
87 /**
88 * __ptrace_unlink - unlink ptracee and restore its execution state
89 * @child: ptracee to be unlinked
91 * Remove @child from the ptrace list, move it back to the original parent,
92 * and restore the execution state so that it conforms to the group stop
93 * state.
95 * Unlinking can happen via two paths - explicit PTRACE_DETACH or ptracer
96 * exiting. For PTRACE_DETACH, unless the ptracee has been killed between
97 * ptrace_check_attach() and here, it's guaranteed to be in TASK_TRACED.
98 * If the ptracer is exiting, the ptracee can be in any state.
100 * After detach, the ptracee should be in a state which conforms to the
101 * group stop. If the group is stopped or in the process of stopping, the
102 * ptracee should be put into TASK_STOPPED; otherwise, it should be woken
103 * up from TASK_TRACED.
105 * If the ptracee is in TASK_TRACED and needs to be moved to TASK_STOPPED,
106 * it goes through TRACED -> RUNNING -> STOPPED transition which is similar
107 * to but in the opposite direction of what happens while attaching to a
108 * stopped task. However, in this direction, the intermediate RUNNING
109 * state is not hidden even from the current ptracer and if it immediately
110 * re-attaches and performs a WNOHANG wait(2), it may fail.
112 * CONTEXT:
113 * write_lock_irq(tasklist_lock)
115 void __ptrace_unlink(struct task_struct *child)
117 const struct cred *old_cred;
118 BUG_ON(!child->ptrace);
120 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
121 #ifdef TIF_SYSCALL_EMU
122 clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
123 #endif
125 child->parent = child->real_parent;
126 list_del_init(&child->ptrace_entry);
127 old_cred = child->ptracer_cred;
128 child->ptracer_cred = NULL;
129 put_cred(old_cred);
131 spin_lock(&child->sighand->siglock);
132 child->ptrace = 0;
134 * Clear all pending traps and TRAPPING. TRAPPING should be
135 * cleared regardless of JOBCTL_STOP_PENDING. Do it explicitly.
137 task_clear_jobctl_pending(child, JOBCTL_TRAP_MASK);
138 task_clear_jobctl_trapping(child);
141 * Reinstate JOBCTL_STOP_PENDING if group stop is in effect and
142 * @child isn't dead.
144 if (!(child->flags & PF_EXITING) &&
145 (child->signal->flags & SIGNAL_STOP_STOPPED ||
146 child->signal->group_stop_count)) {
147 child->jobctl |= JOBCTL_STOP_PENDING;
150 * This is only possible if this thread was cloned by the
151 * traced task running in the stopped group, set the signal
152 * for the future reports.
153 * FIXME: we should change ptrace_init_task() to handle this
154 * case.
156 if (!(child->jobctl & JOBCTL_STOP_SIGMASK))
157 child->jobctl |= SIGSTOP;
161 * If transition to TASK_STOPPED is pending or in TASK_TRACED, kick
162 * @child in the butt. Note that @resume should be used iff @child
163 * is in TASK_TRACED; otherwise, we might unduly disrupt
164 * TASK_KILLABLE sleeps.
166 if (child->jobctl & JOBCTL_STOP_PENDING || task_is_traced(child))
167 ptrace_signal_wake_up(child, true);
169 spin_unlock(&child->sighand->siglock);
172 /* Ensure that nothing can wake it up, even SIGKILL */
173 static bool ptrace_freeze_traced(struct task_struct *task)
175 bool ret = false;
177 /* Lockless, nobody but us can set this flag */
178 if (task->jobctl & JOBCTL_LISTENING)
179 return ret;
181 spin_lock_irq(&task->sighand->siglock);
182 if (task_is_traced(task) && !__fatal_signal_pending(task)) {
183 task->state = __TASK_TRACED;
184 ret = true;
186 spin_unlock_irq(&task->sighand->siglock);
188 return ret;
191 static void ptrace_unfreeze_traced(struct task_struct *task)
193 if (task->state != __TASK_TRACED)
194 return;
196 WARN_ON(!task->ptrace || task->parent != current);
199 * PTRACE_LISTEN can allow ptrace_trap_notify to wake us up remotely.
200 * Recheck state under the lock to close this race.
202 spin_lock_irq(&task->sighand->siglock);
203 if (task->state == __TASK_TRACED) {
204 if (__fatal_signal_pending(task))
205 wake_up_state(task, __TASK_TRACED);
206 else
207 task->state = TASK_TRACED;
209 spin_unlock_irq(&task->sighand->siglock);
213 * ptrace_check_attach - check whether ptracee is ready for ptrace operation
214 * @child: ptracee to check for
215 * @ignore_state: don't check whether @child is currently %TASK_TRACED
217 * Check whether @child is being ptraced by %current and ready for further
218 * ptrace operations. If @ignore_state is %false, @child also should be in
219 * %TASK_TRACED state and on return the child is guaranteed to be traced
220 * and not executing. If @ignore_state is %true, @child can be in any
221 * state.
223 * CONTEXT:
224 * Grabs and releases tasklist_lock and @child->sighand->siglock.
226 * RETURNS:
227 * 0 on success, -ESRCH if %child is not ready.
229 static int ptrace_check_attach(struct task_struct *child, bool ignore_state)
231 int ret = -ESRCH;
234 * We take the read lock around doing both checks to close a
235 * possible race where someone else was tracing our child and
236 * detached between these two checks. After this locked check,
237 * we are sure that this is our traced child and that can only
238 * be changed by us so it's not changing right after this.
240 read_lock(&tasklist_lock);
241 if (child->ptrace && child->parent == current) {
242 WARN_ON(child->state == __TASK_TRACED);
244 * child->sighand can't be NULL, release_task()
245 * does ptrace_unlink() before __exit_signal().
247 if (ignore_state || ptrace_freeze_traced(child))
248 ret = 0;
250 read_unlock(&tasklist_lock);
252 if (!ret && !ignore_state) {
253 if (!wait_task_inactive(child, __TASK_TRACED)) {
255 * This can only happen if may_ptrace_stop() fails and
256 * ptrace_stop() changes ->state back to TASK_RUNNING,
257 * so we should not worry about leaking __TASK_TRACED.
259 WARN_ON(child->state == __TASK_TRACED);
260 ret = -ESRCH;
264 return ret;
267 static int ptrace_has_cap(struct user_namespace *ns, unsigned int mode)
269 if (mode & PTRACE_MODE_NOAUDIT)
270 return has_ns_capability_noaudit(current, ns, CAP_SYS_PTRACE);
271 else
272 return has_ns_capability(current, ns, CAP_SYS_PTRACE);
275 /* Returns 0 on success, -errno on denial. */
276 static int __ptrace_may_access(struct task_struct *task, unsigned int mode)
278 const struct cred *cred = current_cred(), *tcred;
279 struct mm_struct *mm;
280 kuid_t caller_uid;
281 kgid_t caller_gid;
283 if (!(mode & PTRACE_MODE_FSCREDS) == !(mode & PTRACE_MODE_REALCREDS)) {
284 WARN(1, "denying ptrace access check without PTRACE_MODE_*CREDS\n");
285 return -EPERM;
288 /* May we inspect the given task?
289 * This check is used both for attaching with ptrace
290 * and for allowing access to sensitive information in /proc.
292 * ptrace_attach denies several cases that /proc allows
293 * because setting up the necessary parent/child relationship
294 * or halting the specified task is impossible.
297 /* Don't let security modules deny introspection */
298 if (same_thread_group(task, current))
299 return 0;
300 rcu_read_lock();
301 if (mode & PTRACE_MODE_FSCREDS) {
302 caller_uid = cred->fsuid;
303 caller_gid = cred->fsgid;
304 } else {
306 * Using the euid would make more sense here, but something
307 * in userland might rely on the old behavior, and this
308 * shouldn't be a security problem since
309 * PTRACE_MODE_REALCREDS implies that the caller explicitly
310 * used a syscall that requests access to another process
311 * (and not a filesystem syscall to procfs).
313 caller_uid = cred->uid;
314 caller_gid = cred->gid;
316 tcred = __task_cred(task);
317 if (uid_eq(caller_uid, tcred->euid) &&
318 uid_eq(caller_uid, tcred->suid) &&
319 uid_eq(caller_uid, tcred->uid) &&
320 gid_eq(caller_gid, tcred->egid) &&
321 gid_eq(caller_gid, tcred->sgid) &&
322 gid_eq(caller_gid, tcred->gid))
323 goto ok;
324 if (ptrace_has_cap(tcred->user_ns, mode))
325 goto ok;
326 rcu_read_unlock();
327 return -EPERM;
329 rcu_read_unlock();
331 * If a task drops privileges and becomes nondumpable (through a syscall
332 * like setresuid()) while we are trying to access it, we must ensure
333 * that the dumpability is read after the credentials; otherwise,
334 * we may be able to attach to a task that we shouldn't be able to
335 * attach to (as if the task had dropped privileges without becoming
336 * nondumpable).
337 * Pairs with a write barrier in commit_creds().
339 smp_rmb();
340 mm = task->mm;
341 if (mm &&
342 ((get_dumpable(mm) != SUID_DUMP_USER) &&
343 !ptrace_has_cap(mm->user_ns, mode)))
344 return -EPERM;
346 return security_ptrace_access_check(task, mode);
349 bool ptrace_may_access(struct task_struct *task, unsigned int mode)
351 int err;
352 task_lock(task);
353 err = __ptrace_may_access(task, mode);
354 task_unlock(task);
355 return !err;
358 static int ptrace_attach(struct task_struct *task, long request,
359 unsigned long addr,
360 unsigned long flags)
362 bool seize = (request == PTRACE_SEIZE);
363 int retval;
365 retval = -EIO;
366 if (seize) {
367 if (addr != 0)
368 goto out;
369 if (flags & ~(unsigned long)PTRACE_O_MASK)
370 goto out;
371 flags = PT_PTRACED | PT_SEIZED | (flags << PT_OPT_FLAG_SHIFT);
372 } else {
373 flags = PT_PTRACED;
376 audit_ptrace(task);
378 retval = -EPERM;
379 if (unlikely(task->flags & PF_KTHREAD))
380 goto out;
381 if (same_thread_group(task, current))
382 goto out;
385 * Protect exec's credential calculations against our interference;
386 * SUID, SGID and LSM creds get determined differently
387 * under ptrace.
389 retval = -ERESTARTNOINTR;
390 if (mutex_lock_interruptible(&task->signal->cred_guard_mutex))
391 goto out;
393 task_lock(task);
394 retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH_REALCREDS);
395 task_unlock(task);
396 if (retval)
397 goto unlock_creds;
399 write_lock_irq(&tasklist_lock);
400 retval = -EPERM;
401 if (unlikely(task->exit_state))
402 goto unlock_tasklist;
403 if (task->ptrace)
404 goto unlock_tasklist;
406 if (seize)
407 flags |= PT_SEIZED;
408 task->ptrace = flags;
410 ptrace_link(task, current);
412 /* SEIZE doesn't trap tracee on attach */
413 if (!seize)
414 send_sig_info(SIGSTOP, SEND_SIG_PRIV, task);
416 spin_lock(&task->sighand->siglock);
419 * If the task is already STOPPED, set JOBCTL_TRAP_STOP and
420 * TRAPPING, and kick it so that it transits to TRACED. TRAPPING
421 * will be cleared if the child completes the transition or any
422 * event which clears the group stop states happens. We'll wait
423 * for the transition to complete before returning from this
424 * function.
426 * This hides STOPPED -> RUNNING -> TRACED transition from the
427 * attaching thread but a different thread in the same group can
428 * still observe the transient RUNNING state. IOW, if another
429 * thread's WNOHANG wait(2) on the stopped tracee races against
430 * ATTACH, the wait(2) may fail due to the transient RUNNING.
432 * The following task_is_stopped() test is safe as both transitions
433 * in and out of STOPPED are protected by siglock.
435 if (task_is_stopped(task) &&
436 task_set_jobctl_pending(task, JOBCTL_TRAP_STOP | JOBCTL_TRAPPING))
437 signal_wake_up_state(task, __TASK_STOPPED);
439 spin_unlock(&task->sighand->siglock);
441 retval = 0;
442 unlock_tasklist:
443 write_unlock_irq(&tasklist_lock);
444 unlock_creds:
445 mutex_unlock(&task->signal->cred_guard_mutex);
446 out:
447 if (!retval) {
449 * We do not bother to change retval or clear JOBCTL_TRAPPING
450 * if wait_on_bit() was interrupted by SIGKILL. The tracer will
451 * not return to user-mode, it will exit and clear this bit in
452 * __ptrace_unlink() if it wasn't already cleared by the tracee;
453 * and until then nobody can ptrace this task.
455 wait_on_bit(&task->jobctl, JOBCTL_TRAPPING_BIT, TASK_KILLABLE);
456 proc_ptrace_connector(task, PTRACE_ATTACH);
459 return retval;
463 * ptrace_traceme -- helper for PTRACE_TRACEME
465 * Performs checks and sets PT_PTRACED.
466 * Should be used by all ptrace implementations for PTRACE_TRACEME.
468 static int ptrace_traceme(void)
470 int ret = -EPERM;
472 write_lock_irq(&tasklist_lock);
473 /* Are we already being traced? */
474 if (!current->ptrace) {
475 ret = security_ptrace_traceme(current->parent);
477 * Check PF_EXITING to ensure ->real_parent has not passed
478 * exit_ptrace(). Otherwise we don't report the error but
479 * pretend ->real_parent untraces us right after return.
481 if (!ret && !(current->real_parent->flags & PF_EXITING)) {
482 current->ptrace = PT_PTRACED;
483 ptrace_link(current, current->real_parent);
486 write_unlock_irq(&tasklist_lock);
488 return ret;
492 * Called with irqs disabled, returns true if childs should reap themselves.
494 static int ignoring_children(struct sighand_struct *sigh)
496 int ret;
497 spin_lock(&sigh->siglock);
498 ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
499 (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
500 spin_unlock(&sigh->siglock);
501 return ret;
505 * Called with tasklist_lock held for writing.
506 * Unlink a traced task, and clean it up if it was a traced zombie.
507 * Return true if it needs to be reaped with release_task().
508 * (We can't call release_task() here because we already hold tasklist_lock.)
510 * If it's a zombie, our attachedness prevented normal parent notification
511 * or self-reaping. Do notification now if it would have happened earlier.
512 * If it should reap itself, return true.
514 * If it's our own child, there is no notification to do. But if our normal
515 * children self-reap, then this child was prevented by ptrace and we must
516 * reap it now, in that case we must also wake up sub-threads sleeping in
517 * do_wait().
519 static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
521 bool dead;
523 __ptrace_unlink(p);
525 if (p->exit_state != EXIT_ZOMBIE)
526 return false;
528 dead = !thread_group_leader(p);
530 if (!dead && thread_group_empty(p)) {
531 if (!same_thread_group(p->real_parent, tracer))
532 dead = do_notify_parent(p, p->exit_signal);
533 else if (ignoring_children(tracer->sighand)) {
534 __wake_up_parent(p, tracer);
535 dead = true;
538 /* Mark it as in the process of being reaped. */
539 if (dead)
540 p->exit_state = EXIT_DEAD;
541 return dead;
544 static int ptrace_detach(struct task_struct *child, unsigned int data)
546 if (!valid_signal(data))
547 return -EIO;
549 /* Architecture-specific hardware disable .. */
550 ptrace_disable(child);
552 write_lock_irq(&tasklist_lock);
554 * We rely on ptrace_freeze_traced(). It can't be killed and
555 * untraced by another thread, it can't be a zombie.
557 WARN_ON(!child->ptrace || child->exit_state);
559 * tasklist_lock avoids the race with wait_task_stopped(), see
560 * the comment in ptrace_resume().
562 child->exit_code = data;
563 __ptrace_detach(current, child);
564 write_unlock_irq(&tasklist_lock);
566 proc_ptrace_connector(child, PTRACE_DETACH);
568 return 0;
572 * Detach all tasks we were using ptrace on. Called with tasklist held
573 * for writing.
575 void exit_ptrace(struct task_struct *tracer, struct list_head *dead)
577 struct task_struct *p, *n;
579 list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
580 if (unlikely(p->ptrace & PT_EXITKILL))
581 send_sig_info(SIGKILL, SEND_SIG_PRIV, p);
583 if (__ptrace_detach(tracer, p))
584 list_add(&p->ptrace_entry, dead);
588 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
590 int copied = 0;
592 while (len > 0) {
593 char buf[128];
594 int this_len, retval;
596 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
597 retval = ptrace_access_vm(tsk, src, buf, this_len, FOLL_FORCE);
599 if (!retval) {
600 if (copied)
601 break;
602 return -EIO;
604 if (copy_to_user(dst, buf, retval))
605 return -EFAULT;
606 copied += retval;
607 src += retval;
608 dst += retval;
609 len -= retval;
611 return copied;
614 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
616 int copied = 0;
618 while (len > 0) {
619 char buf[128];
620 int this_len, retval;
622 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
623 if (copy_from_user(buf, src, this_len))
624 return -EFAULT;
625 retval = ptrace_access_vm(tsk, dst, buf, this_len,
626 FOLL_FORCE | FOLL_WRITE);
627 if (!retval) {
628 if (copied)
629 break;
630 return -EIO;
632 copied += retval;
633 src += retval;
634 dst += retval;
635 len -= retval;
637 return copied;
640 static int ptrace_setoptions(struct task_struct *child, unsigned long data)
642 unsigned flags;
644 if (data & ~(unsigned long)PTRACE_O_MASK)
645 return -EINVAL;
647 if (unlikely(data & PTRACE_O_SUSPEND_SECCOMP)) {
648 if (!IS_ENABLED(CONFIG_CHECKPOINT_RESTORE) ||
649 !IS_ENABLED(CONFIG_SECCOMP))
650 return -EINVAL;
652 if (!capable(CAP_SYS_ADMIN))
653 return -EPERM;
655 if (seccomp_mode(&current->seccomp) != SECCOMP_MODE_DISABLED ||
656 current->ptrace & PT_SUSPEND_SECCOMP)
657 return -EPERM;
660 /* Avoid intermediate state when all opts are cleared */
661 flags = child->ptrace;
662 flags &= ~(PTRACE_O_MASK << PT_OPT_FLAG_SHIFT);
663 flags |= (data << PT_OPT_FLAG_SHIFT);
664 child->ptrace = flags;
666 return 0;
669 static int ptrace_getsiginfo(struct task_struct *child, kernel_siginfo_t *info)
671 unsigned long flags;
672 int error = -ESRCH;
674 if (lock_task_sighand(child, &flags)) {
675 error = -EINVAL;
676 if (likely(child->last_siginfo != NULL)) {
677 copy_siginfo(info, child->last_siginfo);
678 error = 0;
680 unlock_task_sighand(child, &flags);
682 return error;
685 static int ptrace_setsiginfo(struct task_struct *child, const kernel_siginfo_t *info)
687 unsigned long flags;
688 int error = -ESRCH;
690 if (lock_task_sighand(child, &flags)) {
691 error = -EINVAL;
692 if (likely(child->last_siginfo != NULL)) {
693 copy_siginfo(child->last_siginfo, info);
694 error = 0;
696 unlock_task_sighand(child, &flags);
698 return error;
701 static int ptrace_peek_siginfo(struct task_struct *child,
702 unsigned long addr,
703 unsigned long data)
705 struct ptrace_peeksiginfo_args arg;
706 struct sigpending *pending;
707 struct sigqueue *q;
708 int ret, i;
710 ret = copy_from_user(&arg, (void __user *) addr,
711 sizeof(struct ptrace_peeksiginfo_args));
712 if (ret)
713 return -EFAULT;
715 if (arg.flags & ~PTRACE_PEEKSIGINFO_SHARED)
716 return -EINVAL; /* unknown flags */
718 if (arg.nr < 0)
719 return -EINVAL;
721 /* Ensure arg.off fits in an unsigned long */
722 if (arg.off > ULONG_MAX)
723 return 0;
725 if (arg.flags & PTRACE_PEEKSIGINFO_SHARED)
726 pending = &child->signal->shared_pending;
727 else
728 pending = &child->pending;
730 for (i = 0; i < arg.nr; ) {
731 kernel_siginfo_t info;
732 unsigned long off = arg.off + i;
733 bool found = false;
735 spin_lock_irq(&child->sighand->siglock);
736 list_for_each_entry(q, &pending->list, list) {
737 if (!off--) {
738 found = true;
739 copy_siginfo(&info, &q->info);
740 break;
743 spin_unlock_irq(&child->sighand->siglock);
745 if (!found) /* beyond the end of the list */
746 break;
748 #ifdef CONFIG_COMPAT
749 if (unlikely(in_compat_syscall())) {
750 compat_siginfo_t __user *uinfo = compat_ptr(data);
752 if (copy_siginfo_to_user32(uinfo, &info)) {
753 ret = -EFAULT;
754 break;
757 } else
758 #endif
760 siginfo_t __user *uinfo = (siginfo_t __user *) data;
762 if (copy_siginfo_to_user(uinfo, &info)) {
763 ret = -EFAULT;
764 break;
768 data += sizeof(siginfo_t);
769 i++;
771 if (signal_pending(current))
772 break;
774 cond_resched();
777 if (i > 0)
778 return i;
780 return ret;
783 #ifdef PTRACE_SINGLESTEP
784 #define is_singlestep(request) ((request) == PTRACE_SINGLESTEP)
785 #else
786 #define is_singlestep(request) 0
787 #endif
789 #ifdef PTRACE_SINGLEBLOCK
790 #define is_singleblock(request) ((request) == PTRACE_SINGLEBLOCK)
791 #else
792 #define is_singleblock(request) 0
793 #endif
795 #ifdef PTRACE_SYSEMU
796 #define is_sysemu_singlestep(request) ((request) == PTRACE_SYSEMU_SINGLESTEP)
797 #else
798 #define is_sysemu_singlestep(request) 0
799 #endif
801 static int ptrace_resume(struct task_struct *child, long request,
802 unsigned long data)
804 bool need_siglock;
806 if (!valid_signal(data))
807 return -EIO;
809 if (request == PTRACE_SYSCALL)
810 set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
811 else
812 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
814 #ifdef TIF_SYSCALL_EMU
815 if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
816 set_tsk_thread_flag(child, TIF_SYSCALL_EMU);
817 else
818 clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
819 #endif
821 if (is_singleblock(request)) {
822 if (unlikely(!arch_has_block_step()))
823 return -EIO;
824 user_enable_block_step(child);
825 } else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
826 if (unlikely(!arch_has_single_step()))
827 return -EIO;
828 user_enable_single_step(child);
829 } else {
830 user_disable_single_step(child);
834 * Change ->exit_code and ->state under siglock to avoid the race
835 * with wait_task_stopped() in between; a non-zero ->exit_code will
836 * wrongly look like another report from tracee.
838 * Note that we need siglock even if ->exit_code == data and/or this
839 * status was not reported yet, the new status must not be cleared by
840 * wait_task_stopped() after resume.
842 * If data == 0 we do not care if wait_task_stopped() reports the old
843 * status and clears the code too; this can't race with the tracee, it
844 * takes siglock after resume.
846 need_siglock = data && !thread_group_empty(current);
847 if (need_siglock)
848 spin_lock_irq(&child->sighand->siglock);
849 child->exit_code = data;
850 wake_up_state(child, __TASK_TRACED);
851 if (need_siglock)
852 spin_unlock_irq(&child->sighand->siglock);
854 return 0;
857 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
859 static const struct user_regset *
860 find_regset(const struct user_regset_view *view, unsigned int type)
862 const struct user_regset *regset;
863 int n;
865 for (n = 0; n < view->n; ++n) {
866 regset = view->regsets + n;
867 if (regset->core_note_type == type)
868 return regset;
871 return NULL;
874 static int ptrace_regset(struct task_struct *task, int req, unsigned int type,
875 struct iovec *kiov)
877 const struct user_regset_view *view = task_user_regset_view(task);
878 const struct user_regset *regset = find_regset(view, type);
879 int regset_no;
881 if (!regset || (kiov->iov_len % regset->size) != 0)
882 return -EINVAL;
884 regset_no = regset - view->regsets;
885 kiov->iov_len = min(kiov->iov_len,
886 (__kernel_size_t) (regset->n * regset->size));
888 if (req == PTRACE_GETREGSET)
889 return copy_regset_to_user(task, view, regset_no, 0,
890 kiov->iov_len, kiov->iov_base);
891 else
892 return copy_regset_from_user(task, view, regset_no, 0,
893 kiov->iov_len, kiov->iov_base);
897 * This is declared in linux/regset.h and defined in machine-dependent
898 * code. We put the export here, near the primary machine-neutral use,
899 * to ensure no machine forgets it.
901 EXPORT_SYMBOL_GPL(task_user_regset_view);
903 static unsigned long
904 ptrace_get_syscall_info_entry(struct task_struct *child, struct pt_regs *regs,
905 struct ptrace_syscall_info *info)
907 unsigned long args[ARRAY_SIZE(info->entry.args)];
908 int i;
910 info->op = PTRACE_SYSCALL_INFO_ENTRY;
911 info->entry.nr = syscall_get_nr(child, regs);
912 syscall_get_arguments(child, regs, args);
913 for (i = 0; i < ARRAY_SIZE(args); i++)
914 info->entry.args[i] = args[i];
916 /* args is the last field in struct ptrace_syscall_info.entry */
917 return offsetofend(struct ptrace_syscall_info, entry.args);
920 static unsigned long
921 ptrace_get_syscall_info_seccomp(struct task_struct *child, struct pt_regs *regs,
922 struct ptrace_syscall_info *info)
925 * As struct ptrace_syscall_info.entry is currently a subset
926 * of struct ptrace_syscall_info.seccomp, it makes sense to
927 * initialize that subset using ptrace_get_syscall_info_entry().
928 * This can be reconsidered in the future if these structures
929 * diverge significantly enough.
931 ptrace_get_syscall_info_entry(child, regs, info);
932 info->op = PTRACE_SYSCALL_INFO_SECCOMP;
933 info->seccomp.ret_data = child->ptrace_message;
935 /* ret_data is the last field in struct ptrace_syscall_info.seccomp */
936 return offsetofend(struct ptrace_syscall_info, seccomp.ret_data);
939 static unsigned long
940 ptrace_get_syscall_info_exit(struct task_struct *child, struct pt_regs *regs,
941 struct ptrace_syscall_info *info)
943 info->op = PTRACE_SYSCALL_INFO_EXIT;
944 info->exit.rval = syscall_get_error(child, regs);
945 info->exit.is_error = !!info->exit.rval;
946 if (!info->exit.is_error)
947 info->exit.rval = syscall_get_return_value(child, regs);
949 /* is_error is the last field in struct ptrace_syscall_info.exit */
950 return offsetofend(struct ptrace_syscall_info, exit.is_error);
953 static int
954 ptrace_get_syscall_info(struct task_struct *child, unsigned long user_size,
955 void __user *datavp)
957 struct pt_regs *regs = task_pt_regs(child);
958 struct ptrace_syscall_info info = {
959 .op = PTRACE_SYSCALL_INFO_NONE,
960 .arch = syscall_get_arch(child),
961 .instruction_pointer = instruction_pointer(regs),
962 .stack_pointer = user_stack_pointer(regs),
964 unsigned long actual_size = offsetof(struct ptrace_syscall_info, entry);
965 unsigned long write_size;
968 * This does not need lock_task_sighand() to access
969 * child->last_siginfo because ptrace_freeze_traced()
970 * called earlier by ptrace_check_attach() ensures that
971 * the tracee cannot go away and clear its last_siginfo.
973 switch (child->last_siginfo ? child->last_siginfo->si_code : 0) {
974 case SIGTRAP | 0x80:
975 switch (child->ptrace_message) {
976 case PTRACE_EVENTMSG_SYSCALL_ENTRY:
977 actual_size = ptrace_get_syscall_info_entry(child, regs,
978 &info);
979 break;
980 case PTRACE_EVENTMSG_SYSCALL_EXIT:
981 actual_size = ptrace_get_syscall_info_exit(child, regs,
982 &info);
983 break;
985 break;
986 case SIGTRAP | (PTRACE_EVENT_SECCOMP << 8):
987 actual_size = ptrace_get_syscall_info_seccomp(child, regs,
988 &info);
989 break;
992 write_size = min(actual_size, user_size);
993 return copy_to_user(datavp, &info, write_size) ? -EFAULT : actual_size;
995 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
997 int ptrace_request(struct task_struct *child, long request,
998 unsigned long addr, unsigned long data)
1000 bool seized = child->ptrace & PT_SEIZED;
1001 int ret = -EIO;
1002 kernel_siginfo_t siginfo, *si;
1003 void __user *datavp = (void __user *) data;
1004 unsigned long __user *datalp = datavp;
1005 unsigned long flags;
1007 switch (request) {
1008 case PTRACE_PEEKTEXT:
1009 case PTRACE_PEEKDATA:
1010 return generic_ptrace_peekdata(child, addr, data);
1011 case PTRACE_POKETEXT:
1012 case PTRACE_POKEDATA:
1013 return generic_ptrace_pokedata(child, addr, data);
1015 #ifdef PTRACE_OLDSETOPTIONS
1016 case PTRACE_OLDSETOPTIONS:
1017 #endif
1018 case PTRACE_SETOPTIONS:
1019 ret = ptrace_setoptions(child, data);
1020 break;
1021 case PTRACE_GETEVENTMSG:
1022 ret = put_user(child->ptrace_message, datalp);
1023 break;
1025 case PTRACE_PEEKSIGINFO:
1026 ret = ptrace_peek_siginfo(child, addr, data);
1027 break;
1029 case PTRACE_GETSIGINFO:
1030 ret = ptrace_getsiginfo(child, &siginfo);
1031 if (!ret)
1032 ret = copy_siginfo_to_user(datavp, &siginfo);
1033 break;
1035 case PTRACE_SETSIGINFO:
1036 ret = copy_siginfo_from_user(&siginfo, datavp);
1037 if (!ret)
1038 ret = ptrace_setsiginfo(child, &siginfo);
1039 break;
1041 case PTRACE_GETSIGMASK: {
1042 sigset_t *mask;
1044 if (addr != sizeof(sigset_t)) {
1045 ret = -EINVAL;
1046 break;
1049 if (test_tsk_restore_sigmask(child))
1050 mask = &child->saved_sigmask;
1051 else
1052 mask = &child->blocked;
1054 if (copy_to_user(datavp, mask, sizeof(sigset_t)))
1055 ret = -EFAULT;
1056 else
1057 ret = 0;
1059 break;
1062 case PTRACE_SETSIGMASK: {
1063 sigset_t new_set;
1065 if (addr != sizeof(sigset_t)) {
1066 ret = -EINVAL;
1067 break;
1070 if (copy_from_user(&new_set, datavp, sizeof(sigset_t))) {
1071 ret = -EFAULT;
1072 break;
1075 sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
1078 * Every thread does recalc_sigpending() after resume, so
1079 * retarget_shared_pending() and recalc_sigpending() are not
1080 * called here.
1082 spin_lock_irq(&child->sighand->siglock);
1083 child->blocked = new_set;
1084 spin_unlock_irq(&child->sighand->siglock);
1086 clear_tsk_restore_sigmask(child);
1088 ret = 0;
1089 break;
1092 case PTRACE_INTERRUPT:
1094 * Stop tracee without any side-effect on signal or job
1095 * control. At least one trap is guaranteed to happen
1096 * after this request. If @child is already trapped, the
1097 * current trap is not disturbed and another trap will
1098 * happen after the current trap is ended with PTRACE_CONT.
1100 * The actual trap might not be PTRACE_EVENT_STOP trap but
1101 * the pending condition is cleared regardless.
1103 if (unlikely(!seized || !lock_task_sighand(child, &flags)))
1104 break;
1107 * INTERRUPT doesn't disturb existing trap sans one
1108 * exception. If ptracer issued LISTEN for the current
1109 * STOP, this INTERRUPT should clear LISTEN and re-trap
1110 * tracee into STOP.
1112 if (likely(task_set_jobctl_pending(child, JOBCTL_TRAP_STOP)))
1113 ptrace_signal_wake_up(child, child->jobctl & JOBCTL_LISTENING);
1115 unlock_task_sighand(child, &flags);
1116 ret = 0;
1117 break;
1119 case PTRACE_LISTEN:
1121 * Listen for events. Tracee must be in STOP. It's not
1122 * resumed per-se but is not considered to be in TRACED by
1123 * wait(2) or ptrace(2). If an async event (e.g. group
1124 * stop state change) happens, tracee will enter STOP trap
1125 * again. Alternatively, ptracer can issue INTERRUPT to
1126 * finish listening and re-trap tracee into STOP.
1128 if (unlikely(!seized || !lock_task_sighand(child, &flags)))
1129 break;
1131 si = child->last_siginfo;
1132 if (likely(si && (si->si_code >> 8) == PTRACE_EVENT_STOP)) {
1133 child->jobctl |= JOBCTL_LISTENING;
1135 * If NOTIFY is set, it means event happened between
1136 * start of this trap and now. Trigger re-trap.
1138 if (child->jobctl & JOBCTL_TRAP_NOTIFY)
1139 ptrace_signal_wake_up(child, true);
1140 ret = 0;
1142 unlock_task_sighand(child, &flags);
1143 break;
1145 case PTRACE_DETACH: /* detach a process that was attached. */
1146 ret = ptrace_detach(child, data);
1147 break;
1149 #ifdef CONFIG_BINFMT_ELF_FDPIC
1150 case PTRACE_GETFDPIC: {
1151 struct mm_struct *mm = get_task_mm(child);
1152 unsigned long tmp = 0;
1154 ret = -ESRCH;
1155 if (!mm)
1156 break;
1158 switch (addr) {
1159 case PTRACE_GETFDPIC_EXEC:
1160 tmp = mm->context.exec_fdpic_loadmap;
1161 break;
1162 case PTRACE_GETFDPIC_INTERP:
1163 tmp = mm->context.interp_fdpic_loadmap;
1164 break;
1165 default:
1166 break;
1168 mmput(mm);
1170 ret = put_user(tmp, datalp);
1171 break;
1173 #endif
1175 #ifdef PTRACE_SINGLESTEP
1176 case PTRACE_SINGLESTEP:
1177 #endif
1178 #ifdef PTRACE_SINGLEBLOCK
1179 case PTRACE_SINGLEBLOCK:
1180 #endif
1181 #ifdef PTRACE_SYSEMU
1182 case PTRACE_SYSEMU:
1183 case PTRACE_SYSEMU_SINGLESTEP:
1184 #endif
1185 case PTRACE_SYSCALL:
1186 case PTRACE_CONT:
1187 return ptrace_resume(child, request, data);
1189 case PTRACE_KILL:
1190 if (child->exit_state) /* already dead */
1191 return 0;
1192 return ptrace_resume(child, request, SIGKILL);
1194 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
1195 case PTRACE_GETREGSET:
1196 case PTRACE_SETREGSET: {
1197 struct iovec kiov;
1198 struct iovec __user *uiov = datavp;
1200 if (!access_ok(uiov, sizeof(*uiov)))
1201 return -EFAULT;
1203 if (__get_user(kiov.iov_base, &uiov->iov_base) ||
1204 __get_user(kiov.iov_len, &uiov->iov_len))
1205 return -EFAULT;
1207 ret = ptrace_regset(child, request, addr, &kiov);
1208 if (!ret)
1209 ret = __put_user(kiov.iov_len, &uiov->iov_len);
1210 break;
1213 case PTRACE_GET_SYSCALL_INFO:
1214 ret = ptrace_get_syscall_info(child, addr, datavp);
1215 break;
1216 #endif
1218 case PTRACE_SECCOMP_GET_FILTER:
1219 ret = seccomp_get_filter(child, addr, datavp);
1220 break;
1222 case PTRACE_SECCOMP_GET_METADATA:
1223 ret = seccomp_get_metadata(child, addr, datavp);
1224 break;
1226 default:
1227 break;
1230 return ret;
1233 #ifndef arch_ptrace_attach
1234 #define arch_ptrace_attach(child) do { } while (0)
1235 #endif
1237 SYSCALL_DEFINE4(ptrace, long, request, long, pid, unsigned long, addr,
1238 unsigned long, data)
1240 struct task_struct *child;
1241 long ret;
1243 if (request == PTRACE_TRACEME) {
1244 ret = ptrace_traceme();
1245 if (!ret)
1246 arch_ptrace_attach(current);
1247 goto out;
1250 child = find_get_task_by_vpid(pid);
1251 if (!child) {
1252 ret = -ESRCH;
1253 goto out;
1256 if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1257 ret = ptrace_attach(child, request, addr, data);
1259 * Some architectures need to do book-keeping after
1260 * a ptrace attach.
1262 if (!ret)
1263 arch_ptrace_attach(child);
1264 goto out_put_task_struct;
1267 ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1268 request == PTRACE_INTERRUPT);
1269 if (ret < 0)
1270 goto out_put_task_struct;
1272 ret = arch_ptrace(child, request, addr, data);
1273 if (ret || request != PTRACE_DETACH)
1274 ptrace_unfreeze_traced(child);
1276 out_put_task_struct:
1277 put_task_struct(child);
1278 out:
1279 return ret;
1282 int generic_ptrace_peekdata(struct task_struct *tsk, unsigned long addr,
1283 unsigned long data)
1285 unsigned long tmp;
1286 int copied;
1288 copied = ptrace_access_vm(tsk, addr, &tmp, sizeof(tmp), FOLL_FORCE);
1289 if (copied != sizeof(tmp))
1290 return -EIO;
1291 return put_user(tmp, (unsigned long __user *)data);
1294 int generic_ptrace_pokedata(struct task_struct *tsk, unsigned long addr,
1295 unsigned long data)
1297 int copied;
1299 copied = ptrace_access_vm(tsk, addr, &data, sizeof(data),
1300 FOLL_FORCE | FOLL_WRITE);
1301 return (copied == sizeof(data)) ? 0 : -EIO;
1304 #if defined CONFIG_COMPAT
1306 int compat_ptrace_request(struct task_struct *child, compat_long_t request,
1307 compat_ulong_t addr, compat_ulong_t data)
1309 compat_ulong_t __user *datap = compat_ptr(data);
1310 compat_ulong_t word;
1311 kernel_siginfo_t siginfo;
1312 int ret;
1314 switch (request) {
1315 case PTRACE_PEEKTEXT:
1316 case PTRACE_PEEKDATA:
1317 ret = ptrace_access_vm(child, addr, &word, sizeof(word),
1318 FOLL_FORCE);
1319 if (ret != sizeof(word))
1320 ret = -EIO;
1321 else
1322 ret = put_user(word, datap);
1323 break;
1325 case PTRACE_POKETEXT:
1326 case PTRACE_POKEDATA:
1327 ret = ptrace_access_vm(child, addr, &data, sizeof(data),
1328 FOLL_FORCE | FOLL_WRITE);
1329 ret = (ret != sizeof(data) ? -EIO : 0);
1330 break;
1332 case PTRACE_GETEVENTMSG:
1333 ret = put_user((compat_ulong_t) child->ptrace_message, datap);
1334 break;
1336 case PTRACE_GETSIGINFO:
1337 ret = ptrace_getsiginfo(child, &siginfo);
1338 if (!ret)
1339 ret = copy_siginfo_to_user32(
1340 (struct compat_siginfo __user *) datap,
1341 &siginfo);
1342 break;
1344 case PTRACE_SETSIGINFO:
1345 ret = copy_siginfo_from_user32(
1346 &siginfo, (struct compat_siginfo __user *) datap);
1347 if (!ret)
1348 ret = ptrace_setsiginfo(child, &siginfo);
1349 break;
1350 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
1351 case PTRACE_GETREGSET:
1352 case PTRACE_SETREGSET:
1354 struct iovec kiov;
1355 struct compat_iovec __user *uiov =
1356 (struct compat_iovec __user *) datap;
1357 compat_uptr_t ptr;
1358 compat_size_t len;
1360 if (!access_ok(uiov, sizeof(*uiov)))
1361 return -EFAULT;
1363 if (__get_user(ptr, &uiov->iov_base) ||
1364 __get_user(len, &uiov->iov_len))
1365 return -EFAULT;
1367 kiov.iov_base = compat_ptr(ptr);
1368 kiov.iov_len = len;
1370 ret = ptrace_regset(child, request, addr, &kiov);
1371 if (!ret)
1372 ret = __put_user(kiov.iov_len, &uiov->iov_len);
1373 break;
1375 #endif
1377 default:
1378 ret = ptrace_request(child, request, addr, data);
1381 return ret;
1384 COMPAT_SYSCALL_DEFINE4(ptrace, compat_long_t, request, compat_long_t, pid,
1385 compat_long_t, addr, compat_long_t, data)
1387 struct task_struct *child;
1388 long ret;
1390 if (request == PTRACE_TRACEME) {
1391 ret = ptrace_traceme();
1392 goto out;
1395 child = find_get_task_by_vpid(pid);
1396 if (!child) {
1397 ret = -ESRCH;
1398 goto out;
1401 if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1402 ret = ptrace_attach(child, request, addr, data);
1404 * Some architectures need to do book-keeping after
1405 * a ptrace attach.
1407 if (!ret)
1408 arch_ptrace_attach(child);
1409 goto out_put_task_struct;
1412 ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1413 request == PTRACE_INTERRUPT);
1414 if (!ret) {
1415 ret = compat_arch_ptrace(child, request, addr, data);
1416 if (ret || request != PTRACE_DETACH)
1417 ptrace_unfreeze_traced(child);
1420 out_put_task_struct:
1421 put_task_struct(child);
1422 out:
1423 return ret;
1425 #endif /* CONFIG_COMPAT */