1 // SPDX-License-Identifier: GPL-2.0-only
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.
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>
19 #include <linux/highmem.h>
20 #include <linux/pagemap.h>
21 #include <linux/ptrace.h>
22 #include <linux/security.h>
23 #include <linux/signal.h>
24 #include <linux/uio.h>
25 #include <linux/audit.h>
26 #include <linux/pid_namespace.h>
27 #include <linux/syscalls.h>
28 #include <linux/uaccess.h>
29 #include <linux/regset.h>
30 #include <linux/hw_breakpoint.h>
31 #include <linux/cn_proc.h>
32 #include <linux/compat.h>
33 #include <linux/sched/signal.h>
34 #include <linux/minmax.h>
36 #include <asm/syscall.h> /* for syscall_get_* */
39 * Access another process' address space via ptrace.
40 * Source/target buffer must be kernel space,
41 * Do not walk the page table directly, use get_user_pages
43 int ptrace_access_vm(struct task_struct *tsk, unsigned long addr,
44 void *buf, int len, unsigned int gup_flags)
49 mm = get_task_mm(tsk);
54 (current != tsk->parent) ||
55 ((get_dumpable(mm) != SUID_DUMP_USER) &&
56 !ptracer_capable(tsk, mm->user_ns))) {
61 ret = __access_remote_vm(mm, addr, buf, len, gup_flags);
68 void __ptrace_link(struct task_struct *child, struct task_struct *new_parent,
69 const struct cred *ptracer_cred)
71 BUG_ON(!list_empty(&child->ptrace_entry));
72 list_add(&child->ptrace_entry, &new_parent->ptraced);
73 child->parent = new_parent;
74 child->ptracer_cred = get_cred(ptracer_cred);
78 * ptrace a task: make the debugger its new parent and
79 * move it to the ptrace list.
81 * Must be called with the tasklist lock write-held.
83 static void ptrace_link(struct task_struct *child, struct task_struct *new_parent)
85 __ptrace_link(child, new_parent, current_cred());
89 * __ptrace_unlink - unlink ptracee and restore its execution state
90 * @child: ptracee to be unlinked
92 * Remove @child from the ptrace list, move it back to the original parent,
93 * and restore the execution state so that it conforms to the group stop
96 * Unlinking can happen via two paths - explicit PTRACE_DETACH or ptracer
97 * exiting. For PTRACE_DETACH, unless the ptracee has been killed between
98 * ptrace_check_attach() and here, it's guaranteed to be in TASK_TRACED.
99 * If the ptracer is exiting, the ptracee can be in any state.
101 * After detach, the ptracee should be in a state which conforms to the
102 * group stop. If the group is stopped or in the process of stopping, the
103 * ptracee should be put into TASK_STOPPED; otherwise, it should be woken
104 * up from TASK_TRACED.
106 * If the ptracee is in TASK_TRACED and needs to be moved to TASK_STOPPED,
107 * it goes through TRACED -> RUNNING -> STOPPED transition which is similar
108 * to but in the opposite direction of what happens while attaching to a
109 * stopped task. However, in this direction, the intermediate RUNNING
110 * state is not hidden even from the current ptracer and if it immediately
111 * re-attaches and performs a WNOHANG wait(2), it may fail.
114 * write_lock_irq(tasklist_lock)
116 void __ptrace_unlink(struct task_struct *child)
118 const struct cred *old_cred;
119 BUG_ON(!child->ptrace);
121 clear_task_syscall_work(child, SYSCALL_TRACE);
122 #if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU)
123 clear_task_syscall_work(child, SYSCALL_EMU);
126 child->parent = child->real_parent;
127 list_del_init(&child->ptrace_entry);
128 old_cred = child->ptracer_cred;
129 child->ptracer_cred = NULL;
132 spin_lock(&child->sighand->siglock);
135 * Clear all pending traps and TRAPPING. TRAPPING should be
136 * cleared regardless of JOBCTL_STOP_PENDING. Do it explicitly.
138 task_clear_jobctl_pending(child, JOBCTL_TRAP_MASK);
139 task_clear_jobctl_trapping(child);
142 * Reinstate JOBCTL_STOP_PENDING if group stop is in effect and
145 if (!(child->flags & PF_EXITING) &&
146 (child->signal->flags & SIGNAL_STOP_STOPPED ||
147 child->signal->group_stop_count)) {
148 child->jobctl |= JOBCTL_STOP_PENDING;
151 * This is only possible if this thread was cloned by the
152 * traced task running in the stopped group, set the signal
153 * for the future reports.
154 * FIXME: we should change ptrace_init_task() to handle this
157 if (!(child->jobctl & JOBCTL_STOP_SIGMASK))
158 child->jobctl |= SIGSTOP;
162 * If transition to TASK_STOPPED is pending or in TASK_TRACED, kick
163 * @child in the butt. Note that @resume should be used iff @child
164 * is in TASK_TRACED; otherwise, we might unduly disrupt
165 * TASK_KILLABLE sleeps.
167 if (child->jobctl & JOBCTL_STOP_PENDING || task_is_traced(child))
168 ptrace_signal_wake_up(child, true);
170 spin_unlock(&child->sighand->siglock);
173 static bool looks_like_a_spurious_pid(struct task_struct *task)
175 if (task->exit_code != ((PTRACE_EVENT_EXEC << 8) | SIGTRAP))
178 if (task_pid_vnr(task) == task->ptrace_message)
181 * The tracee changed its pid but the PTRACE_EVENT_EXEC event
182 * was not wait()'ed, most probably debugger targets the old
183 * leader which was destroyed in de_thread().
188 /* Ensure that nothing can wake it up, even SIGKILL */
189 static bool ptrace_freeze_traced(struct task_struct *task)
193 /* Lockless, nobody but us can set this flag */
194 if (task->jobctl & JOBCTL_LISTENING)
197 spin_lock_irq(&task->sighand->siglock);
198 if (task_is_traced(task) && !looks_like_a_spurious_pid(task) &&
199 !__fatal_signal_pending(task)) {
200 task->state = __TASK_TRACED;
203 spin_unlock_irq(&task->sighand->siglock);
208 static void ptrace_unfreeze_traced(struct task_struct *task)
210 if (task->state != __TASK_TRACED)
213 WARN_ON(!task->ptrace || task->parent != current);
216 * PTRACE_LISTEN can allow ptrace_trap_notify to wake us up remotely.
217 * Recheck state under the lock to close this race.
219 spin_lock_irq(&task->sighand->siglock);
220 if (task->state == __TASK_TRACED) {
221 if (__fatal_signal_pending(task))
222 wake_up_state(task, __TASK_TRACED);
224 task->state = TASK_TRACED;
226 spin_unlock_irq(&task->sighand->siglock);
230 * ptrace_check_attach - check whether ptracee is ready for ptrace operation
231 * @child: ptracee to check for
232 * @ignore_state: don't check whether @child is currently %TASK_TRACED
234 * Check whether @child is being ptraced by %current and ready for further
235 * ptrace operations. If @ignore_state is %false, @child also should be in
236 * %TASK_TRACED state and on return the child is guaranteed to be traced
237 * and not executing. If @ignore_state is %true, @child can be in any
241 * Grabs and releases tasklist_lock and @child->sighand->siglock.
244 * 0 on success, -ESRCH if %child is not ready.
246 static int ptrace_check_attach(struct task_struct *child, bool ignore_state)
251 * We take the read lock around doing both checks to close a
252 * possible race where someone else was tracing our child and
253 * detached between these two checks. After this locked check,
254 * we are sure that this is our traced child and that can only
255 * be changed by us so it's not changing right after this.
257 read_lock(&tasklist_lock);
258 if (child->ptrace && child->parent == current) {
259 WARN_ON(child->state == __TASK_TRACED);
261 * child->sighand can't be NULL, release_task()
262 * does ptrace_unlink() before __exit_signal().
264 if (ignore_state || ptrace_freeze_traced(child))
267 read_unlock(&tasklist_lock);
269 if (!ret && !ignore_state) {
270 if (!wait_task_inactive(child, __TASK_TRACED)) {
272 * This can only happen if may_ptrace_stop() fails and
273 * ptrace_stop() changes ->state back to TASK_RUNNING,
274 * so we should not worry about leaking __TASK_TRACED.
276 WARN_ON(child->state == __TASK_TRACED);
284 static bool ptrace_has_cap(struct user_namespace *ns, unsigned int mode)
286 if (mode & PTRACE_MODE_NOAUDIT)
287 return ns_capable_noaudit(ns, CAP_SYS_PTRACE);
288 return ns_capable(ns, CAP_SYS_PTRACE);
291 /* Returns 0 on success, -errno on denial. */
292 static int __ptrace_may_access(struct task_struct *task, unsigned int mode)
294 const struct cred *cred = current_cred(), *tcred;
295 struct mm_struct *mm;
299 if (!(mode & PTRACE_MODE_FSCREDS) == !(mode & PTRACE_MODE_REALCREDS)) {
300 WARN(1, "denying ptrace access check without PTRACE_MODE_*CREDS\n");
304 /* May we inspect the given task?
305 * This check is used both for attaching with ptrace
306 * and for allowing access to sensitive information in /proc.
308 * ptrace_attach denies several cases that /proc allows
309 * because setting up the necessary parent/child relationship
310 * or halting the specified task is impossible.
313 /* Don't let security modules deny introspection */
314 if (same_thread_group(task, current))
317 if (mode & PTRACE_MODE_FSCREDS) {
318 caller_uid = cred->fsuid;
319 caller_gid = cred->fsgid;
322 * Using the euid would make more sense here, but something
323 * in userland might rely on the old behavior, and this
324 * shouldn't be a security problem since
325 * PTRACE_MODE_REALCREDS implies that the caller explicitly
326 * used a syscall that requests access to another process
327 * (and not a filesystem syscall to procfs).
329 caller_uid = cred->uid;
330 caller_gid = cred->gid;
332 tcred = __task_cred(task);
333 if (uid_eq(caller_uid, tcred->euid) &&
334 uid_eq(caller_uid, tcred->suid) &&
335 uid_eq(caller_uid, tcred->uid) &&
336 gid_eq(caller_gid, tcred->egid) &&
337 gid_eq(caller_gid, tcred->sgid) &&
338 gid_eq(caller_gid, tcred->gid))
340 if (ptrace_has_cap(tcred->user_ns, mode))
347 * If a task drops privileges and becomes nondumpable (through a syscall
348 * like setresuid()) while we are trying to access it, we must ensure
349 * that the dumpability is read after the credentials; otherwise,
350 * we may be able to attach to a task that we shouldn't be able to
351 * attach to (as if the task had dropped privileges without becoming
353 * Pairs with a write barrier in commit_creds().
358 ((get_dumpable(mm) != SUID_DUMP_USER) &&
359 !ptrace_has_cap(mm->user_ns, mode)))
362 return security_ptrace_access_check(task, mode);
365 bool ptrace_may_access(struct task_struct *task, unsigned int mode)
369 err = __ptrace_may_access(task, mode);
374 static int ptrace_attach(struct task_struct *task, long request,
378 bool seize = (request == PTRACE_SEIZE);
385 if (flags & ~(unsigned long)PTRACE_O_MASK)
387 flags = PT_PTRACED | PT_SEIZED | (flags << PT_OPT_FLAG_SHIFT);
395 if (unlikely(task->flags & PF_KTHREAD))
397 if (same_thread_group(task, current))
401 * Protect exec's credential calculations against our interference;
402 * SUID, SGID and LSM creds get determined differently
405 retval = -ERESTARTNOINTR;
406 if (mutex_lock_interruptible(&task->signal->cred_guard_mutex))
410 retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH_REALCREDS);
415 write_lock_irq(&tasklist_lock);
417 if (unlikely(task->exit_state))
418 goto unlock_tasklist;
420 goto unlock_tasklist;
424 task->ptrace = flags;
426 ptrace_link(task, current);
428 /* SEIZE doesn't trap tracee on attach */
430 send_sig_info(SIGSTOP, SEND_SIG_PRIV, task);
432 spin_lock(&task->sighand->siglock);
435 * If the task is already STOPPED, set JOBCTL_TRAP_STOP and
436 * TRAPPING, and kick it so that it transits to TRACED. TRAPPING
437 * will be cleared if the child completes the transition or any
438 * event which clears the group stop states happens. We'll wait
439 * for the transition to complete before returning from this
442 * This hides STOPPED -> RUNNING -> TRACED transition from the
443 * attaching thread but a different thread in the same group can
444 * still observe the transient RUNNING state. IOW, if another
445 * thread's WNOHANG wait(2) on the stopped tracee races against
446 * ATTACH, the wait(2) may fail due to the transient RUNNING.
448 * The following task_is_stopped() test is safe as both transitions
449 * in and out of STOPPED are protected by siglock.
451 if (task_is_stopped(task) &&
452 task_set_jobctl_pending(task, JOBCTL_TRAP_STOP | JOBCTL_TRAPPING))
453 signal_wake_up_state(task, __TASK_STOPPED);
455 spin_unlock(&task->sighand->siglock);
459 write_unlock_irq(&tasklist_lock);
461 mutex_unlock(&task->signal->cred_guard_mutex);
465 * We do not bother to change retval or clear JOBCTL_TRAPPING
466 * if wait_on_bit() was interrupted by SIGKILL. The tracer will
467 * not return to user-mode, it will exit and clear this bit in
468 * __ptrace_unlink() if it wasn't already cleared by the tracee;
469 * and until then nobody can ptrace this task.
471 wait_on_bit(&task->jobctl, JOBCTL_TRAPPING_BIT, TASK_KILLABLE);
472 proc_ptrace_connector(task, PTRACE_ATTACH);
479 * ptrace_traceme -- helper for PTRACE_TRACEME
481 * Performs checks and sets PT_PTRACED.
482 * Should be used by all ptrace implementations for PTRACE_TRACEME.
484 static int ptrace_traceme(void)
488 write_lock_irq(&tasklist_lock);
489 /* Are we already being traced? */
490 if (!current->ptrace) {
491 ret = security_ptrace_traceme(current->parent);
493 * Check PF_EXITING to ensure ->real_parent has not passed
494 * exit_ptrace(). Otherwise we don't report the error but
495 * pretend ->real_parent untraces us right after return.
497 if (!ret && !(current->real_parent->flags & PF_EXITING)) {
498 current->ptrace = PT_PTRACED;
499 ptrace_link(current, current->real_parent);
502 write_unlock_irq(&tasklist_lock);
508 * Called with irqs disabled, returns true if childs should reap themselves.
510 static int ignoring_children(struct sighand_struct *sigh)
513 spin_lock(&sigh->siglock);
514 ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
515 (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
516 spin_unlock(&sigh->siglock);
521 * Called with tasklist_lock held for writing.
522 * Unlink a traced task, and clean it up if it was a traced zombie.
523 * Return true if it needs to be reaped with release_task().
524 * (We can't call release_task() here because we already hold tasklist_lock.)
526 * If it's a zombie, our attachedness prevented normal parent notification
527 * or self-reaping. Do notification now if it would have happened earlier.
528 * If it should reap itself, return true.
530 * If it's our own child, there is no notification to do. But if our normal
531 * children self-reap, then this child was prevented by ptrace and we must
532 * reap it now, in that case we must also wake up sub-threads sleeping in
535 static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
541 if (p->exit_state != EXIT_ZOMBIE)
544 dead = !thread_group_leader(p);
546 if (!dead && thread_group_empty(p)) {
547 if (!same_thread_group(p->real_parent, tracer))
548 dead = do_notify_parent(p, p->exit_signal);
549 else if (ignoring_children(tracer->sighand)) {
550 __wake_up_parent(p, tracer);
554 /* Mark it as in the process of being reaped. */
556 p->exit_state = EXIT_DEAD;
560 static int ptrace_detach(struct task_struct *child, unsigned int data)
562 if (!valid_signal(data))
565 /* Architecture-specific hardware disable .. */
566 ptrace_disable(child);
568 write_lock_irq(&tasklist_lock);
570 * We rely on ptrace_freeze_traced(). It can't be killed and
571 * untraced by another thread, it can't be a zombie.
573 WARN_ON(!child->ptrace || child->exit_state);
575 * tasklist_lock avoids the race with wait_task_stopped(), see
576 * the comment in ptrace_resume().
578 child->exit_code = data;
579 __ptrace_detach(current, child);
580 write_unlock_irq(&tasklist_lock);
582 proc_ptrace_connector(child, PTRACE_DETACH);
588 * Detach all tasks we were using ptrace on. Called with tasklist held
591 void exit_ptrace(struct task_struct *tracer, struct list_head *dead)
593 struct task_struct *p, *n;
595 list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
596 if (unlikely(p->ptrace & PT_EXITKILL))
597 send_sig_info(SIGKILL, SEND_SIG_PRIV, p);
599 if (__ptrace_detach(tracer, p))
600 list_add(&p->ptrace_entry, dead);
604 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
610 int this_len, retval;
612 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
613 retval = ptrace_access_vm(tsk, src, buf, this_len, FOLL_FORCE);
620 if (copy_to_user(dst, buf, retval))
630 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
636 int this_len, retval;
638 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
639 if (copy_from_user(buf, src, this_len))
641 retval = ptrace_access_vm(tsk, dst, buf, this_len,
642 FOLL_FORCE | FOLL_WRITE);
656 static int ptrace_setoptions(struct task_struct *child, unsigned long data)
660 if (data & ~(unsigned long)PTRACE_O_MASK)
663 if (unlikely(data & PTRACE_O_SUSPEND_SECCOMP)) {
664 if (!IS_ENABLED(CONFIG_CHECKPOINT_RESTORE) ||
665 !IS_ENABLED(CONFIG_SECCOMP))
668 if (!capable(CAP_SYS_ADMIN))
671 if (seccomp_mode(¤t->seccomp) != SECCOMP_MODE_DISABLED ||
672 current->ptrace & PT_SUSPEND_SECCOMP)
676 /* Avoid intermediate state when all opts are cleared */
677 flags = child->ptrace;
678 flags &= ~(PTRACE_O_MASK << PT_OPT_FLAG_SHIFT);
679 flags |= (data << PT_OPT_FLAG_SHIFT);
680 child->ptrace = flags;
685 static int ptrace_getsiginfo(struct task_struct *child, kernel_siginfo_t *info)
690 if (lock_task_sighand(child, &flags)) {
692 if (likely(child->last_siginfo != NULL)) {
693 copy_siginfo(info, child->last_siginfo);
696 unlock_task_sighand(child, &flags);
701 static int ptrace_setsiginfo(struct task_struct *child, const kernel_siginfo_t *info)
706 if (lock_task_sighand(child, &flags)) {
708 if (likely(child->last_siginfo != NULL)) {
709 copy_siginfo(child->last_siginfo, info);
712 unlock_task_sighand(child, &flags);
717 static int ptrace_peek_siginfo(struct task_struct *child,
721 struct ptrace_peeksiginfo_args arg;
722 struct sigpending *pending;
726 ret = copy_from_user(&arg, (void __user *) addr,
727 sizeof(struct ptrace_peeksiginfo_args));
731 if (arg.flags & ~PTRACE_PEEKSIGINFO_SHARED)
732 return -EINVAL; /* unknown flags */
737 /* Ensure arg.off fits in an unsigned long */
738 if (arg.off > ULONG_MAX)
741 if (arg.flags & PTRACE_PEEKSIGINFO_SHARED)
742 pending = &child->signal->shared_pending;
744 pending = &child->pending;
746 for (i = 0; i < arg.nr; ) {
747 kernel_siginfo_t info;
748 unsigned long off = arg.off + i;
751 spin_lock_irq(&child->sighand->siglock);
752 list_for_each_entry(q, &pending->list, list) {
755 copy_siginfo(&info, &q->info);
759 spin_unlock_irq(&child->sighand->siglock);
761 if (!found) /* beyond the end of the list */
765 if (unlikely(in_compat_syscall())) {
766 compat_siginfo_t __user *uinfo = compat_ptr(data);
768 if (copy_siginfo_to_user32(uinfo, &info)) {
776 siginfo_t __user *uinfo = (siginfo_t __user *) data;
778 if (copy_siginfo_to_user(uinfo, &info)) {
784 data += sizeof(siginfo_t);
787 if (signal_pending(current))
800 static long ptrace_get_rseq_configuration(struct task_struct *task,
801 unsigned long size, void __user *data)
803 struct ptrace_rseq_configuration conf = {
804 .rseq_abi_pointer = (u64)(uintptr_t)task->rseq,
805 .rseq_abi_size = sizeof(*task->rseq),
806 .signature = task->rseq_sig,
810 size = min_t(unsigned long, size, sizeof(conf));
811 if (copy_to_user(data, &conf, size))
817 #ifdef PTRACE_SINGLESTEP
818 #define is_singlestep(request) ((request) == PTRACE_SINGLESTEP)
820 #define is_singlestep(request) 0
823 #ifdef PTRACE_SINGLEBLOCK
824 #define is_singleblock(request) ((request) == PTRACE_SINGLEBLOCK)
826 #define is_singleblock(request) 0
830 #define is_sysemu_singlestep(request) ((request) == PTRACE_SYSEMU_SINGLESTEP)
832 #define is_sysemu_singlestep(request) 0
835 static int ptrace_resume(struct task_struct *child, long request,
840 if (!valid_signal(data))
843 if (request == PTRACE_SYSCALL)
844 set_task_syscall_work(child, SYSCALL_TRACE);
846 clear_task_syscall_work(child, SYSCALL_TRACE);
848 #if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU)
849 if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
850 set_task_syscall_work(child, SYSCALL_EMU);
852 clear_task_syscall_work(child, SYSCALL_EMU);
855 if (is_singleblock(request)) {
856 if (unlikely(!arch_has_block_step()))
858 user_enable_block_step(child);
859 } else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
860 if (unlikely(!arch_has_single_step()))
862 user_enable_single_step(child);
864 user_disable_single_step(child);
868 * Change ->exit_code and ->state under siglock to avoid the race
869 * with wait_task_stopped() in between; a non-zero ->exit_code will
870 * wrongly look like another report from tracee.
872 * Note that we need siglock even if ->exit_code == data and/or this
873 * status was not reported yet, the new status must not be cleared by
874 * wait_task_stopped() after resume.
876 * If data == 0 we do not care if wait_task_stopped() reports the old
877 * status and clears the code too; this can't race with the tracee, it
878 * takes siglock after resume.
880 need_siglock = data && !thread_group_empty(current);
882 spin_lock_irq(&child->sighand->siglock);
883 child->exit_code = data;
884 wake_up_state(child, __TASK_TRACED);
886 spin_unlock_irq(&child->sighand->siglock);
891 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
893 static const struct user_regset *
894 find_regset(const struct user_regset_view *view, unsigned int type)
896 const struct user_regset *regset;
899 for (n = 0; n < view->n; ++n) {
900 regset = view->regsets + n;
901 if (regset->core_note_type == type)
908 static int ptrace_regset(struct task_struct *task, int req, unsigned int type,
911 const struct user_regset_view *view = task_user_regset_view(task);
912 const struct user_regset *regset = find_regset(view, type);
915 if (!regset || (kiov->iov_len % regset->size) != 0)
918 regset_no = regset - view->regsets;
919 kiov->iov_len = min(kiov->iov_len,
920 (__kernel_size_t) (regset->n * regset->size));
922 if (req == PTRACE_GETREGSET)
923 return copy_regset_to_user(task, view, regset_no, 0,
924 kiov->iov_len, kiov->iov_base);
926 return copy_regset_from_user(task, view, regset_no, 0,
927 kiov->iov_len, kiov->iov_base);
931 * This is declared in linux/regset.h and defined in machine-dependent
932 * code. We put the export here, near the primary machine-neutral use,
933 * to ensure no machine forgets it.
935 EXPORT_SYMBOL_GPL(task_user_regset_view);
938 ptrace_get_syscall_info_entry(struct task_struct *child, struct pt_regs *regs,
939 struct ptrace_syscall_info *info)
941 unsigned long args[ARRAY_SIZE(info->entry.args)];
944 info->op = PTRACE_SYSCALL_INFO_ENTRY;
945 info->entry.nr = syscall_get_nr(child, regs);
946 syscall_get_arguments(child, regs, args);
947 for (i = 0; i < ARRAY_SIZE(args); i++)
948 info->entry.args[i] = args[i];
950 /* args is the last field in struct ptrace_syscall_info.entry */
951 return offsetofend(struct ptrace_syscall_info, entry.args);
955 ptrace_get_syscall_info_seccomp(struct task_struct *child, struct pt_regs *regs,
956 struct ptrace_syscall_info *info)
959 * As struct ptrace_syscall_info.entry is currently a subset
960 * of struct ptrace_syscall_info.seccomp, it makes sense to
961 * initialize that subset using ptrace_get_syscall_info_entry().
962 * This can be reconsidered in the future if these structures
963 * diverge significantly enough.
965 ptrace_get_syscall_info_entry(child, regs, info);
966 info->op = PTRACE_SYSCALL_INFO_SECCOMP;
967 info->seccomp.ret_data = child->ptrace_message;
969 /* ret_data is the last field in struct ptrace_syscall_info.seccomp */
970 return offsetofend(struct ptrace_syscall_info, seccomp.ret_data);
974 ptrace_get_syscall_info_exit(struct task_struct *child, struct pt_regs *regs,
975 struct ptrace_syscall_info *info)
977 info->op = PTRACE_SYSCALL_INFO_EXIT;
978 info->exit.rval = syscall_get_error(child, regs);
979 info->exit.is_error = !!info->exit.rval;
980 if (!info->exit.is_error)
981 info->exit.rval = syscall_get_return_value(child, regs);
983 /* is_error is the last field in struct ptrace_syscall_info.exit */
984 return offsetofend(struct ptrace_syscall_info, exit.is_error);
988 ptrace_get_syscall_info(struct task_struct *child, unsigned long user_size,
991 struct pt_regs *regs = task_pt_regs(child);
992 struct ptrace_syscall_info info = {
993 .op = PTRACE_SYSCALL_INFO_NONE,
994 .arch = syscall_get_arch(child),
995 .instruction_pointer = instruction_pointer(regs),
996 .stack_pointer = user_stack_pointer(regs),
998 unsigned long actual_size = offsetof(struct ptrace_syscall_info, entry);
999 unsigned long write_size;
1002 * This does not need lock_task_sighand() to access
1003 * child->last_siginfo because ptrace_freeze_traced()
1004 * called earlier by ptrace_check_attach() ensures that
1005 * the tracee cannot go away and clear its last_siginfo.
1007 switch (child->last_siginfo ? child->last_siginfo->si_code : 0) {
1008 case SIGTRAP | 0x80:
1009 switch (child->ptrace_message) {
1010 case PTRACE_EVENTMSG_SYSCALL_ENTRY:
1011 actual_size = ptrace_get_syscall_info_entry(child, regs,
1014 case PTRACE_EVENTMSG_SYSCALL_EXIT:
1015 actual_size = ptrace_get_syscall_info_exit(child, regs,
1020 case SIGTRAP | (PTRACE_EVENT_SECCOMP << 8):
1021 actual_size = ptrace_get_syscall_info_seccomp(child, regs,
1026 write_size = min(actual_size, user_size);
1027 return copy_to_user(datavp, &info, write_size) ? -EFAULT : actual_size;
1029 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
1031 int ptrace_request(struct task_struct *child, long request,
1032 unsigned long addr, unsigned long data)
1034 bool seized = child->ptrace & PT_SEIZED;
1036 kernel_siginfo_t siginfo, *si;
1037 void __user *datavp = (void __user *) data;
1038 unsigned long __user *datalp = datavp;
1039 unsigned long flags;
1042 case PTRACE_PEEKTEXT:
1043 case PTRACE_PEEKDATA:
1044 return generic_ptrace_peekdata(child, addr, data);
1045 case PTRACE_POKETEXT:
1046 case PTRACE_POKEDATA:
1047 return generic_ptrace_pokedata(child, addr, data);
1049 #ifdef PTRACE_OLDSETOPTIONS
1050 case PTRACE_OLDSETOPTIONS:
1052 case PTRACE_SETOPTIONS:
1053 ret = ptrace_setoptions(child, data);
1055 case PTRACE_GETEVENTMSG:
1056 ret = put_user(child->ptrace_message, datalp);
1059 case PTRACE_PEEKSIGINFO:
1060 ret = ptrace_peek_siginfo(child, addr, data);
1063 case PTRACE_GETSIGINFO:
1064 ret = ptrace_getsiginfo(child, &siginfo);
1066 ret = copy_siginfo_to_user(datavp, &siginfo);
1069 case PTRACE_SETSIGINFO:
1070 ret = copy_siginfo_from_user(&siginfo, datavp);
1072 ret = ptrace_setsiginfo(child, &siginfo);
1075 case PTRACE_GETSIGMASK: {
1078 if (addr != sizeof(sigset_t)) {
1083 if (test_tsk_restore_sigmask(child))
1084 mask = &child->saved_sigmask;
1086 mask = &child->blocked;
1088 if (copy_to_user(datavp, mask, sizeof(sigset_t)))
1096 case PTRACE_SETSIGMASK: {
1099 if (addr != sizeof(sigset_t)) {
1104 if (copy_from_user(&new_set, datavp, sizeof(sigset_t))) {
1109 sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
1112 * Every thread does recalc_sigpending() after resume, so
1113 * retarget_shared_pending() and recalc_sigpending() are not
1116 spin_lock_irq(&child->sighand->siglock);
1117 child->blocked = new_set;
1118 spin_unlock_irq(&child->sighand->siglock);
1120 clear_tsk_restore_sigmask(child);
1126 case PTRACE_INTERRUPT:
1128 * Stop tracee without any side-effect on signal or job
1129 * control. At least one trap is guaranteed to happen
1130 * after this request. If @child is already trapped, the
1131 * current trap is not disturbed and another trap will
1132 * happen after the current trap is ended with PTRACE_CONT.
1134 * The actual trap might not be PTRACE_EVENT_STOP trap but
1135 * the pending condition is cleared regardless.
1137 if (unlikely(!seized || !lock_task_sighand(child, &flags)))
1141 * INTERRUPT doesn't disturb existing trap sans one
1142 * exception. If ptracer issued LISTEN for the current
1143 * STOP, this INTERRUPT should clear LISTEN and re-trap
1146 if (likely(task_set_jobctl_pending(child, JOBCTL_TRAP_STOP)))
1147 ptrace_signal_wake_up(child, child->jobctl & JOBCTL_LISTENING);
1149 unlock_task_sighand(child, &flags);
1155 * Listen for events. Tracee must be in STOP. It's not
1156 * resumed per-se but is not considered to be in TRACED by
1157 * wait(2) or ptrace(2). If an async event (e.g. group
1158 * stop state change) happens, tracee will enter STOP trap
1159 * again. Alternatively, ptracer can issue INTERRUPT to
1160 * finish listening and re-trap tracee into STOP.
1162 if (unlikely(!seized || !lock_task_sighand(child, &flags)))
1165 si = child->last_siginfo;
1166 if (likely(si && (si->si_code >> 8) == PTRACE_EVENT_STOP)) {
1167 child->jobctl |= JOBCTL_LISTENING;
1169 * If NOTIFY is set, it means event happened between
1170 * start of this trap and now. Trigger re-trap.
1172 if (child->jobctl & JOBCTL_TRAP_NOTIFY)
1173 ptrace_signal_wake_up(child, true);
1176 unlock_task_sighand(child, &flags);
1179 case PTRACE_DETACH: /* detach a process that was attached. */
1180 ret = ptrace_detach(child, data);
1183 #ifdef CONFIG_BINFMT_ELF_FDPIC
1184 case PTRACE_GETFDPIC: {
1185 struct mm_struct *mm = get_task_mm(child);
1186 unsigned long tmp = 0;
1193 case PTRACE_GETFDPIC_EXEC:
1194 tmp = mm->context.exec_fdpic_loadmap;
1196 case PTRACE_GETFDPIC_INTERP:
1197 tmp = mm->context.interp_fdpic_loadmap;
1204 ret = put_user(tmp, datalp);
1209 #ifdef PTRACE_SINGLESTEP
1210 case PTRACE_SINGLESTEP:
1212 #ifdef PTRACE_SINGLEBLOCK
1213 case PTRACE_SINGLEBLOCK:
1215 #ifdef PTRACE_SYSEMU
1217 case PTRACE_SYSEMU_SINGLESTEP:
1219 case PTRACE_SYSCALL:
1221 return ptrace_resume(child, request, data);
1224 if (child->exit_state) /* already dead */
1226 return ptrace_resume(child, request, SIGKILL);
1228 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
1229 case PTRACE_GETREGSET:
1230 case PTRACE_SETREGSET: {
1232 struct iovec __user *uiov = datavp;
1234 if (!access_ok(uiov, sizeof(*uiov)))
1237 if (__get_user(kiov.iov_base, &uiov->iov_base) ||
1238 __get_user(kiov.iov_len, &uiov->iov_len))
1241 ret = ptrace_regset(child, request, addr, &kiov);
1243 ret = __put_user(kiov.iov_len, &uiov->iov_len);
1247 case PTRACE_GET_SYSCALL_INFO:
1248 ret = ptrace_get_syscall_info(child, addr, datavp);
1252 case PTRACE_SECCOMP_GET_FILTER:
1253 ret = seccomp_get_filter(child, addr, datavp);
1256 case PTRACE_SECCOMP_GET_METADATA:
1257 ret = seccomp_get_metadata(child, addr, datavp);
1261 case PTRACE_GET_RSEQ_CONFIGURATION:
1262 ret = ptrace_get_rseq_configuration(child, addr, datavp);
1273 #ifndef arch_ptrace_attach
1274 #define arch_ptrace_attach(child) do { } while (0)
1277 SYSCALL_DEFINE4(ptrace, long, request, long, pid, unsigned long, addr,
1278 unsigned long, data)
1280 struct task_struct *child;
1283 if (request == PTRACE_TRACEME) {
1284 ret = ptrace_traceme();
1286 arch_ptrace_attach(current);
1290 child = find_get_task_by_vpid(pid);
1296 if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1297 ret = ptrace_attach(child, request, addr, data);
1299 * Some architectures need to do book-keeping after
1303 arch_ptrace_attach(child);
1304 goto out_put_task_struct;
1307 ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1308 request == PTRACE_INTERRUPT);
1310 goto out_put_task_struct;
1312 ret = arch_ptrace(child, request, addr, data);
1313 if (ret || request != PTRACE_DETACH)
1314 ptrace_unfreeze_traced(child);
1316 out_put_task_struct:
1317 put_task_struct(child);
1322 int generic_ptrace_peekdata(struct task_struct *tsk, unsigned long addr,
1328 copied = ptrace_access_vm(tsk, addr, &tmp, sizeof(tmp), FOLL_FORCE);
1329 if (copied != sizeof(tmp))
1331 return put_user(tmp, (unsigned long __user *)data);
1334 int generic_ptrace_pokedata(struct task_struct *tsk, unsigned long addr,
1339 copied = ptrace_access_vm(tsk, addr, &data, sizeof(data),
1340 FOLL_FORCE | FOLL_WRITE);
1341 return (copied == sizeof(data)) ? 0 : -EIO;
1344 #if defined CONFIG_COMPAT
1346 int compat_ptrace_request(struct task_struct *child, compat_long_t request,
1347 compat_ulong_t addr, compat_ulong_t data)
1349 compat_ulong_t __user *datap = compat_ptr(data);
1350 compat_ulong_t word;
1351 kernel_siginfo_t siginfo;
1355 case PTRACE_PEEKTEXT:
1356 case PTRACE_PEEKDATA:
1357 ret = ptrace_access_vm(child, addr, &word, sizeof(word),
1359 if (ret != sizeof(word))
1362 ret = put_user(word, datap);
1365 case PTRACE_POKETEXT:
1366 case PTRACE_POKEDATA:
1367 ret = ptrace_access_vm(child, addr, &data, sizeof(data),
1368 FOLL_FORCE | FOLL_WRITE);
1369 ret = (ret != sizeof(data) ? -EIO : 0);
1372 case PTRACE_GETEVENTMSG:
1373 ret = put_user((compat_ulong_t) child->ptrace_message, datap);
1376 case PTRACE_GETSIGINFO:
1377 ret = ptrace_getsiginfo(child, &siginfo);
1379 ret = copy_siginfo_to_user32(
1380 (struct compat_siginfo __user *) datap,
1384 case PTRACE_SETSIGINFO:
1385 ret = copy_siginfo_from_user32(
1386 &siginfo, (struct compat_siginfo __user *) datap);
1388 ret = ptrace_setsiginfo(child, &siginfo);
1390 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
1391 case PTRACE_GETREGSET:
1392 case PTRACE_SETREGSET:
1395 struct compat_iovec __user *uiov =
1396 (struct compat_iovec __user *) datap;
1400 if (!access_ok(uiov, sizeof(*uiov)))
1403 if (__get_user(ptr, &uiov->iov_base) ||
1404 __get_user(len, &uiov->iov_len))
1407 kiov.iov_base = compat_ptr(ptr);
1410 ret = ptrace_regset(child, request, addr, &kiov);
1412 ret = __put_user(kiov.iov_len, &uiov->iov_len);
1418 ret = ptrace_request(child, request, addr, data);
1424 COMPAT_SYSCALL_DEFINE4(ptrace, compat_long_t, request, compat_long_t, pid,
1425 compat_long_t, addr, compat_long_t, data)
1427 struct task_struct *child;
1430 if (request == PTRACE_TRACEME) {
1431 ret = ptrace_traceme();
1435 child = find_get_task_by_vpid(pid);
1441 if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1442 ret = ptrace_attach(child, request, addr, data);
1444 * Some architectures need to do book-keeping after
1448 arch_ptrace_attach(child);
1449 goto out_put_task_struct;
1452 ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1453 request == PTRACE_INTERRUPT);
1455 ret = compat_arch_ptrace(child, request, addr, data);
1456 if (ret || request != PTRACE_DETACH)
1457 ptrace_unfreeze_traced(child);
1460 out_put_task_struct:
1461 put_task_struct(child);
1465 #endif /* CONFIG_COMPAT */