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>
35 #include <linux/syscall_user_dispatch.h>
37 #include <asm/syscall.h> /* for syscall_get_* */
40 * Access another process' address space via ptrace.
41 * Source/target buffer must be kernel space,
42 * Do not walk the page table directly, use get_user_pages
44 int ptrace_access_vm(struct task_struct *tsk, unsigned long addr,
45 void *buf, int len, unsigned int gup_flags)
50 mm = get_task_mm(tsk);
55 (current != tsk->parent) ||
56 ((get_dumpable(mm) != SUID_DUMP_USER) &&
57 !ptracer_capable(tsk, mm->user_ns))) {
62 ret = __access_remote_vm(mm, addr, buf, len, gup_flags);
69 void __ptrace_link(struct task_struct *child, struct task_struct *new_parent,
70 const struct cred *ptracer_cred)
72 BUG_ON(!list_empty(&child->ptrace_entry));
73 list_add(&child->ptrace_entry, &new_parent->ptraced);
74 child->parent = new_parent;
75 child->ptracer_cred = get_cred(ptracer_cred);
79 * ptrace a task: make the debugger its new parent and
80 * move it to the ptrace list.
82 * Must be called with the tasklist lock write-held.
84 static void ptrace_link(struct task_struct *child, struct task_struct *new_parent)
86 __ptrace_link(child, new_parent, current_cred());
90 * __ptrace_unlink - unlink ptracee and restore its execution state
91 * @child: ptracee to be unlinked
93 * Remove @child from the ptrace list, move it back to the original parent,
94 * and restore the execution state so that it conforms to the group stop
97 * Unlinking can happen via two paths - explicit PTRACE_DETACH or ptracer
98 * exiting. For PTRACE_DETACH, unless the ptracee has been killed between
99 * ptrace_check_attach() and here, it's guaranteed to be in TASK_TRACED.
100 * If the ptracer is exiting, the ptracee can be in any state.
102 * After detach, the ptracee should be in a state which conforms to the
103 * group stop. If the group is stopped or in the process of stopping, the
104 * ptracee should be put into TASK_STOPPED; otherwise, it should be woken
105 * up from TASK_TRACED.
107 * If the ptracee is in TASK_TRACED and needs to be moved to TASK_STOPPED,
108 * it goes through TRACED -> RUNNING -> STOPPED transition which is similar
109 * to but in the opposite direction of what happens while attaching to a
110 * stopped task. However, in this direction, the intermediate RUNNING
111 * state is not hidden even from the current ptracer and if it immediately
112 * re-attaches and performs a WNOHANG wait(2), it may fail.
115 * write_lock_irq(tasklist_lock)
117 void __ptrace_unlink(struct task_struct *child)
119 const struct cred *old_cred;
120 BUG_ON(!child->ptrace);
122 clear_task_syscall_work(child, SYSCALL_TRACE);
123 #if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU)
124 clear_task_syscall_work(child, SYSCALL_EMU);
127 child->parent = child->real_parent;
128 list_del_init(&child->ptrace_entry);
129 old_cred = child->ptracer_cred;
130 child->ptracer_cred = NULL;
133 spin_lock(&child->sighand->siglock);
136 * Clear all pending traps and TRAPPING. TRAPPING should be
137 * cleared regardless of JOBCTL_STOP_PENDING. Do it explicitly.
139 task_clear_jobctl_pending(child, JOBCTL_TRAP_MASK);
140 task_clear_jobctl_trapping(child);
143 * Reinstate JOBCTL_STOP_PENDING if group stop is in effect and
146 if (!(child->flags & PF_EXITING) &&
147 (child->signal->flags & SIGNAL_STOP_STOPPED ||
148 child->signal->group_stop_count)) {
149 child->jobctl |= JOBCTL_STOP_PENDING;
152 * This is only possible if this thread was cloned by the
153 * traced task running in the stopped group, set the signal
154 * for the future reports.
155 * FIXME: we should change ptrace_init_task() to handle this
158 if (!(child->jobctl & JOBCTL_STOP_SIGMASK))
159 child->jobctl |= SIGSTOP;
163 * If transition to TASK_STOPPED is pending or in TASK_TRACED, kick
164 * @child in the butt. Note that @resume should be used iff @child
165 * is in TASK_TRACED; otherwise, we might unduly disrupt
166 * TASK_KILLABLE sleeps.
168 if (child->jobctl & JOBCTL_STOP_PENDING || task_is_traced(child))
169 ptrace_signal_wake_up(child, true);
171 spin_unlock(&child->sighand->siglock);
174 static bool looks_like_a_spurious_pid(struct task_struct *task)
176 if (task->exit_code != ((PTRACE_EVENT_EXEC << 8) | SIGTRAP))
179 if (task_pid_vnr(task) == task->ptrace_message)
182 * The tracee changed its pid but the PTRACE_EVENT_EXEC event
183 * was not wait()'ed, most probably debugger targets the old
184 * leader which was destroyed in de_thread().
190 * Ensure that nothing can wake it up, even SIGKILL
192 * A task is switched to this state while a ptrace operation is in progress;
193 * such that the ptrace operation is uninterruptible.
195 static bool ptrace_freeze_traced(struct task_struct *task)
199 /* Lockless, nobody but us can set this flag */
200 if (task->jobctl & JOBCTL_LISTENING)
203 spin_lock_irq(&task->sighand->siglock);
204 if (task_is_traced(task) && !looks_like_a_spurious_pid(task) &&
205 !__fatal_signal_pending(task)) {
206 task->jobctl |= JOBCTL_PTRACE_FROZEN;
209 spin_unlock_irq(&task->sighand->siglock);
214 static void ptrace_unfreeze_traced(struct task_struct *task)
219 * The child may be awake and may have cleared
220 * JOBCTL_PTRACE_FROZEN (see ptrace_resume). The child will
221 * not set JOBCTL_PTRACE_FROZEN or enter __TASK_TRACED anew.
223 if (lock_task_sighand(task, &flags)) {
224 task->jobctl &= ~JOBCTL_PTRACE_FROZEN;
225 if (__fatal_signal_pending(task)) {
226 task->jobctl &= ~JOBCTL_TRACED;
227 wake_up_state(task, __TASK_TRACED);
229 unlock_task_sighand(task, &flags);
234 * ptrace_check_attach - check whether ptracee is ready for ptrace operation
235 * @child: ptracee to check for
236 * @ignore_state: don't check whether @child is currently %TASK_TRACED
238 * Check whether @child is being ptraced by %current and ready for further
239 * ptrace operations. If @ignore_state is %false, @child also should be in
240 * %TASK_TRACED state and on return the child is guaranteed to be traced
241 * and not executing. If @ignore_state is %true, @child can be in any
245 * Grabs and releases tasklist_lock and @child->sighand->siglock.
248 * 0 on success, -ESRCH if %child is not ready.
250 static int ptrace_check_attach(struct task_struct *child, bool ignore_state)
255 * We take the read lock around doing both checks to close a
256 * possible race where someone else was tracing our child and
257 * detached between these two checks. After this locked check,
258 * we are sure that this is our traced child and that can only
259 * be changed by us so it's not changing right after this.
261 read_lock(&tasklist_lock);
262 if (child->ptrace && child->parent == current) {
264 * child->sighand can't be NULL, release_task()
265 * does ptrace_unlink() before __exit_signal().
267 if (ignore_state || ptrace_freeze_traced(child))
270 read_unlock(&tasklist_lock);
272 if (!ret && !ignore_state &&
273 WARN_ON_ONCE(!wait_task_inactive(child, __TASK_TRACED|TASK_FROZEN)))
279 static bool ptrace_has_cap(struct user_namespace *ns, unsigned int mode)
281 if (mode & PTRACE_MODE_NOAUDIT)
282 return ns_capable_noaudit(ns, CAP_SYS_PTRACE);
283 return ns_capable(ns, CAP_SYS_PTRACE);
286 /* Returns 0 on success, -errno on denial. */
287 static int __ptrace_may_access(struct task_struct *task, unsigned int mode)
289 const struct cred *cred = current_cred(), *tcred;
290 struct mm_struct *mm;
294 if (!(mode & PTRACE_MODE_FSCREDS) == !(mode & PTRACE_MODE_REALCREDS)) {
295 WARN(1, "denying ptrace access check without PTRACE_MODE_*CREDS\n");
299 /* May we inspect the given task?
300 * This check is used both for attaching with ptrace
301 * and for allowing access to sensitive information in /proc.
303 * ptrace_attach denies several cases that /proc allows
304 * because setting up the necessary parent/child relationship
305 * or halting the specified task is impossible.
308 /* Don't let security modules deny introspection */
309 if (same_thread_group(task, current))
312 if (mode & PTRACE_MODE_FSCREDS) {
313 caller_uid = cred->fsuid;
314 caller_gid = cred->fsgid;
317 * Using the euid would make more sense here, but something
318 * in userland might rely on the old behavior, and this
319 * shouldn't be a security problem since
320 * PTRACE_MODE_REALCREDS implies that the caller explicitly
321 * used a syscall that requests access to another process
322 * (and not a filesystem syscall to procfs).
324 caller_uid = cred->uid;
325 caller_gid = cred->gid;
327 tcred = __task_cred(task);
328 if (uid_eq(caller_uid, tcred->euid) &&
329 uid_eq(caller_uid, tcred->suid) &&
330 uid_eq(caller_uid, tcred->uid) &&
331 gid_eq(caller_gid, tcred->egid) &&
332 gid_eq(caller_gid, tcred->sgid) &&
333 gid_eq(caller_gid, tcred->gid))
335 if (ptrace_has_cap(tcred->user_ns, mode))
342 * If a task drops privileges and becomes nondumpable (through a syscall
343 * like setresuid()) while we are trying to access it, we must ensure
344 * that the dumpability is read after the credentials; otherwise,
345 * we may be able to attach to a task that we shouldn't be able to
346 * attach to (as if the task had dropped privileges without becoming
348 * Pairs with a write barrier in commit_creds().
353 ((get_dumpable(mm) != SUID_DUMP_USER) &&
354 !ptrace_has_cap(mm->user_ns, mode)))
357 return security_ptrace_access_check(task, mode);
360 bool ptrace_may_access(struct task_struct *task, unsigned int mode)
364 err = __ptrace_may_access(task, mode);
369 static int check_ptrace_options(unsigned long data)
371 if (data & ~(unsigned long)PTRACE_O_MASK)
374 if (unlikely(data & PTRACE_O_SUSPEND_SECCOMP)) {
375 if (!IS_ENABLED(CONFIG_CHECKPOINT_RESTORE) ||
376 !IS_ENABLED(CONFIG_SECCOMP))
379 if (!capable(CAP_SYS_ADMIN))
382 if (seccomp_mode(¤t->seccomp) != SECCOMP_MODE_DISABLED ||
383 current->ptrace & PT_SUSPEND_SECCOMP)
389 static int ptrace_attach(struct task_struct *task, long request,
393 bool seize = (request == PTRACE_SEIZE);
401 * This duplicates the check in check_ptrace_options() because
402 * ptrace_attach() and ptrace_setoptions() have historically
403 * used different error codes for unknown ptrace options.
405 if (flags & ~(unsigned long)PTRACE_O_MASK)
407 retval = check_ptrace_options(flags);
410 flags = PT_PTRACED | PT_SEIZED | (flags << PT_OPT_FLAG_SHIFT);
418 if (unlikely(task->flags & PF_KTHREAD))
420 if (same_thread_group(task, current))
424 * Protect exec's credential calculations against our interference;
425 * SUID, SGID and LSM creds get determined differently
428 retval = -ERESTARTNOINTR;
429 if (mutex_lock_interruptible(&task->signal->cred_guard_mutex))
433 retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH_REALCREDS);
438 write_lock_irq(&tasklist_lock);
440 if (unlikely(task->exit_state))
441 goto unlock_tasklist;
443 goto unlock_tasklist;
445 task->ptrace = flags;
447 ptrace_link(task, current);
449 /* SEIZE doesn't trap tracee on attach */
451 send_sig_info(SIGSTOP, SEND_SIG_PRIV, task);
453 spin_lock(&task->sighand->siglock);
456 * If the task is already STOPPED, set JOBCTL_TRAP_STOP and
457 * TRAPPING, and kick it so that it transits to TRACED. TRAPPING
458 * will be cleared if the child completes the transition or any
459 * event which clears the group stop states happens. We'll wait
460 * for the transition to complete before returning from this
463 * This hides STOPPED -> RUNNING -> TRACED transition from the
464 * attaching thread but a different thread in the same group can
465 * still observe the transient RUNNING state. IOW, if another
466 * thread's WNOHANG wait(2) on the stopped tracee races against
467 * ATTACH, the wait(2) may fail due to the transient RUNNING.
469 * The following task_is_stopped() test is safe as both transitions
470 * in and out of STOPPED are protected by siglock.
472 if (task_is_stopped(task) &&
473 task_set_jobctl_pending(task, JOBCTL_TRAP_STOP | JOBCTL_TRAPPING)) {
474 task->jobctl &= ~JOBCTL_STOPPED;
475 signal_wake_up_state(task, __TASK_STOPPED);
478 spin_unlock(&task->sighand->siglock);
482 write_unlock_irq(&tasklist_lock);
484 mutex_unlock(&task->signal->cred_guard_mutex);
488 * We do not bother to change retval or clear JOBCTL_TRAPPING
489 * if wait_on_bit() was interrupted by SIGKILL. The tracer will
490 * not return to user-mode, it will exit and clear this bit in
491 * __ptrace_unlink() if it wasn't already cleared by the tracee;
492 * and until then nobody can ptrace this task.
494 wait_on_bit(&task->jobctl, JOBCTL_TRAPPING_BIT, TASK_KILLABLE);
495 proc_ptrace_connector(task, PTRACE_ATTACH);
502 * ptrace_traceme -- helper for PTRACE_TRACEME
504 * Performs checks and sets PT_PTRACED.
505 * Should be used by all ptrace implementations for PTRACE_TRACEME.
507 static int ptrace_traceme(void)
511 write_lock_irq(&tasklist_lock);
512 /* Are we already being traced? */
513 if (!current->ptrace) {
514 ret = security_ptrace_traceme(current->parent);
516 * Check PF_EXITING to ensure ->real_parent has not passed
517 * exit_ptrace(). Otherwise we don't report the error but
518 * pretend ->real_parent untraces us right after return.
520 if (!ret && !(current->real_parent->flags & PF_EXITING)) {
521 current->ptrace = PT_PTRACED;
522 ptrace_link(current, current->real_parent);
525 write_unlock_irq(&tasklist_lock);
531 * Called with irqs disabled, returns true if childs should reap themselves.
533 static int ignoring_children(struct sighand_struct *sigh)
536 spin_lock(&sigh->siglock);
537 ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
538 (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
539 spin_unlock(&sigh->siglock);
544 * Called with tasklist_lock held for writing.
545 * Unlink a traced task, and clean it up if it was a traced zombie.
546 * Return true if it needs to be reaped with release_task().
547 * (We can't call release_task() here because we already hold tasklist_lock.)
549 * If it's a zombie, our attachedness prevented normal parent notification
550 * or self-reaping. Do notification now if it would have happened earlier.
551 * If it should reap itself, return true.
553 * If it's our own child, there is no notification to do. But if our normal
554 * children self-reap, then this child was prevented by ptrace and we must
555 * reap it now, in that case we must also wake up sub-threads sleeping in
558 static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
564 if (p->exit_state != EXIT_ZOMBIE)
567 dead = !thread_group_leader(p);
569 if (!dead && thread_group_empty(p)) {
570 if (!same_thread_group(p->real_parent, tracer))
571 dead = do_notify_parent(p, p->exit_signal);
572 else if (ignoring_children(tracer->sighand)) {
573 __wake_up_parent(p, tracer);
577 /* Mark it as in the process of being reaped. */
579 p->exit_state = EXIT_DEAD;
583 static int ptrace_detach(struct task_struct *child, unsigned int data)
585 if (!valid_signal(data))
588 /* Architecture-specific hardware disable .. */
589 ptrace_disable(child);
591 write_lock_irq(&tasklist_lock);
593 * We rely on ptrace_freeze_traced(). It can't be killed and
594 * untraced by another thread, it can't be a zombie.
596 WARN_ON(!child->ptrace || child->exit_state);
598 * tasklist_lock avoids the race with wait_task_stopped(), see
599 * the comment in ptrace_resume().
601 child->exit_code = data;
602 __ptrace_detach(current, child);
603 write_unlock_irq(&tasklist_lock);
605 proc_ptrace_connector(child, PTRACE_DETACH);
611 * Detach all tasks we were using ptrace on. Called with tasklist held
614 void exit_ptrace(struct task_struct *tracer, struct list_head *dead)
616 struct task_struct *p, *n;
618 list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
619 if (unlikely(p->ptrace & PT_EXITKILL))
620 send_sig_info(SIGKILL, SEND_SIG_PRIV, p);
622 if (__ptrace_detach(tracer, p))
623 list_add(&p->ptrace_entry, dead);
627 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
633 int this_len, retval;
635 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
636 retval = ptrace_access_vm(tsk, src, buf, this_len, FOLL_FORCE);
643 if (copy_to_user(dst, buf, retval))
653 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
659 int this_len, retval;
661 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
662 if (copy_from_user(buf, src, this_len))
664 retval = ptrace_access_vm(tsk, dst, buf, this_len,
665 FOLL_FORCE | FOLL_WRITE);
679 static int ptrace_setoptions(struct task_struct *child, unsigned long data)
684 ret = check_ptrace_options(data);
688 /* Avoid intermediate state when all opts are cleared */
689 flags = child->ptrace;
690 flags &= ~(PTRACE_O_MASK << PT_OPT_FLAG_SHIFT);
691 flags |= (data << PT_OPT_FLAG_SHIFT);
692 child->ptrace = flags;
697 static int ptrace_getsiginfo(struct task_struct *child, kernel_siginfo_t *info)
702 if (lock_task_sighand(child, &flags)) {
704 if (likely(child->last_siginfo != NULL)) {
705 copy_siginfo(info, child->last_siginfo);
708 unlock_task_sighand(child, &flags);
713 static int ptrace_setsiginfo(struct task_struct *child, const kernel_siginfo_t *info)
718 if (lock_task_sighand(child, &flags)) {
720 if (likely(child->last_siginfo != NULL)) {
721 copy_siginfo(child->last_siginfo, info);
724 unlock_task_sighand(child, &flags);
729 static int ptrace_peek_siginfo(struct task_struct *child,
733 struct ptrace_peeksiginfo_args arg;
734 struct sigpending *pending;
738 ret = copy_from_user(&arg, (void __user *) addr,
739 sizeof(struct ptrace_peeksiginfo_args));
743 if (arg.flags & ~PTRACE_PEEKSIGINFO_SHARED)
744 return -EINVAL; /* unknown flags */
749 /* Ensure arg.off fits in an unsigned long */
750 if (arg.off > ULONG_MAX)
753 if (arg.flags & PTRACE_PEEKSIGINFO_SHARED)
754 pending = &child->signal->shared_pending;
756 pending = &child->pending;
758 for (i = 0; i < arg.nr; ) {
759 kernel_siginfo_t info;
760 unsigned long off = arg.off + i;
763 spin_lock_irq(&child->sighand->siglock);
764 list_for_each_entry(q, &pending->list, list) {
767 copy_siginfo(&info, &q->info);
771 spin_unlock_irq(&child->sighand->siglock);
773 if (!found) /* beyond the end of the list */
777 if (unlikely(in_compat_syscall())) {
778 compat_siginfo_t __user *uinfo = compat_ptr(data);
780 if (copy_siginfo_to_user32(uinfo, &info)) {
788 siginfo_t __user *uinfo = (siginfo_t __user *) data;
790 if (copy_siginfo_to_user(uinfo, &info)) {
796 data += sizeof(siginfo_t);
799 if (signal_pending(current))
812 static long ptrace_get_rseq_configuration(struct task_struct *task,
813 unsigned long size, void __user *data)
815 struct ptrace_rseq_configuration conf = {
816 .rseq_abi_pointer = (u64)(uintptr_t)task->rseq,
817 .rseq_abi_size = task->rseq_len,
818 .signature = task->rseq_sig,
822 size = min_t(unsigned long, size, sizeof(conf));
823 if (copy_to_user(data, &conf, size))
829 #define is_singlestep(request) ((request) == PTRACE_SINGLESTEP)
831 #ifdef PTRACE_SINGLEBLOCK
832 #define is_singleblock(request) ((request) == PTRACE_SINGLEBLOCK)
834 #define is_singleblock(request) 0
838 #define is_sysemu_singlestep(request) ((request) == PTRACE_SYSEMU_SINGLESTEP)
840 #define is_sysemu_singlestep(request) 0
843 static int ptrace_resume(struct task_struct *child, long request,
846 if (!valid_signal(data))
849 if (request == PTRACE_SYSCALL)
850 set_task_syscall_work(child, SYSCALL_TRACE);
852 clear_task_syscall_work(child, SYSCALL_TRACE);
854 #if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU)
855 if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
856 set_task_syscall_work(child, SYSCALL_EMU);
858 clear_task_syscall_work(child, SYSCALL_EMU);
861 if (is_singleblock(request)) {
862 if (unlikely(!arch_has_block_step()))
864 user_enable_block_step(child);
865 } else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
866 if (unlikely(!arch_has_single_step()))
868 user_enable_single_step(child);
870 user_disable_single_step(child);
874 * Change ->exit_code and ->state under siglock to avoid the race
875 * with wait_task_stopped() in between; a non-zero ->exit_code will
876 * wrongly look like another report from tracee.
878 * Note that we need siglock even if ->exit_code == data and/or this
879 * status was not reported yet, the new status must not be cleared by
880 * wait_task_stopped() after resume.
882 spin_lock_irq(&child->sighand->siglock);
883 child->exit_code = data;
884 child->jobctl &= ~JOBCTL_TRACED;
885 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 case PTRACE_SINGLESTEP:
1210 #ifdef PTRACE_SINGLEBLOCK
1211 case PTRACE_SINGLEBLOCK:
1213 #ifdef PTRACE_SYSEMU
1215 case PTRACE_SYSEMU_SINGLESTEP:
1217 case PTRACE_SYSCALL:
1219 return ptrace_resume(child, request, data);
1222 send_sig_info(SIGKILL, SEND_SIG_NOINFO, child);
1225 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
1226 case PTRACE_GETREGSET:
1227 case PTRACE_SETREGSET: {
1229 struct iovec __user *uiov = datavp;
1231 if (!access_ok(uiov, sizeof(*uiov)))
1234 if (__get_user(kiov.iov_base, &uiov->iov_base) ||
1235 __get_user(kiov.iov_len, &uiov->iov_len))
1238 ret = ptrace_regset(child, request, addr, &kiov);
1240 ret = __put_user(kiov.iov_len, &uiov->iov_len);
1244 case PTRACE_GET_SYSCALL_INFO:
1245 ret = ptrace_get_syscall_info(child, addr, datavp);
1249 case PTRACE_SECCOMP_GET_FILTER:
1250 ret = seccomp_get_filter(child, addr, datavp);
1253 case PTRACE_SECCOMP_GET_METADATA:
1254 ret = seccomp_get_metadata(child, addr, datavp);
1258 case PTRACE_GET_RSEQ_CONFIGURATION:
1259 ret = ptrace_get_rseq_configuration(child, addr, datavp);
1263 case PTRACE_SET_SYSCALL_USER_DISPATCH_CONFIG:
1264 ret = syscall_user_dispatch_set_config(child, addr, datavp);
1267 case PTRACE_GET_SYSCALL_USER_DISPATCH_CONFIG:
1268 ret = syscall_user_dispatch_get_config(child, addr, datavp);
1278 SYSCALL_DEFINE4(ptrace, long, request, long, pid, unsigned long, addr,
1279 unsigned long, data)
1281 struct task_struct *child;
1284 if (request == PTRACE_TRACEME) {
1285 ret = ptrace_traceme();
1289 child = find_get_task_by_vpid(pid);
1295 if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1296 ret = ptrace_attach(child, request, addr, data);
1297 goto out_put_task_struct;
1300 ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1301 request == PTRACE_INTERRUPT);
1303 goto out_put_task_struct;
1305 ret = arch_ptrace(child, request, addr, data);
1306 if (ret || request != PTRACE_DETACH)
1307 ptrace_unfreeze_traced(child);
1309 out_put_task_struct:
1310 put_task_struct(child);
1315 int generic_ptrace_peekdata(struct task_struct *tsk, unsigned long addr,
1321 copied = ptrace_access_vm(tsk, addr, &tmp, sizeof(tmp), FOLL_FORCE);
1322 if (copied != sizeof(tmp))
1324 return put_user(tmp, (unsigned long __user *)data);
1327 int generic_ptrace_pokedata(struct task_struct *tsk, unsigned long addr,
1332 copied = ptrace_access_vm(tsk, addr, &data, sizeof(data),
1333 FOLL_FORCE | FOLL_WRITE);
1334 return (copied == sizeof(data)) ? 0 : -EIO;
1337 #if defined CONFIG_COMPAT
1339 int compat_ptrace_request(struct task_struct *child, compat_long_t request,
1340 compat_ulong_t addr, compat_ulong_t data)
1342 compat_ulong_t __user *datap = compat_ptr(data);
1343 compat_ulong_t word;
1344 kernel_siginfo_t siginfo;
1348 case PTRACE_PEEKTEXT:
1349 case PTRACE_PEEKDATA:
1350 ret = ptrace_access_vm(child, addr, &word, sizeof(word),
1352 if (ret != sizeof(word))
1355 ret = put_user(word, datap);
1358 case PTRACE_POKETEXT:
1359 case PTRACE_POKEDATA:
1360 ret = ptrace_access_vm(child, addr, &data, sizeof(data),
1361 FOLL_FORCE | FOLL_WRITE);
1362 ret = (ret != sizeof(data) ? -EIO : 0);
1365 case PTRACE_GETEVENTMSG:
1366 ret = put_user((compat_ulong_t) child->ptrace_message, datap);
1369 case PTRACE_GETSIGINFO:
1370 ret = ptrace_getsiginfo(child, &siginfo);
1372 ret = copy_siginfo_to_user32(
1373 (struct compat_siginfo __user *) datap,
1377 case PTRACE_SETSIGINFO:
1378 ret = copy_siginfo_from_user32(
1379 &siginfo, (struct compat_siginfo __user *) datap);
1381 ret = ptrace_setsiginfo(child, &siginfo);
1383 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
1384 case PTRACE_GETREGSET:
1385 case PTRACE_SETREGSET:
1388 struct compat_iovec __user *uiov =
1389 (struct compat_iovec __user *) datap;
1393 if (!access_ok(uiov, sizeof(*uiov)))
1396 if (__get_user(ptr, &uiov->iov_base) ||
1397 __get_user(len, &uiov->iov_len))
1400 kiov.iov_base = compat_ptr(ptr);
1403 ret = ptrace_regset(child, request, addr, &kiov);
1405 ret = __put_user(kiov.iov_len, &uiov->iov_len);
1411 ret = ptrace_request(child, request, addr, data);
1417 COMPAT_SYSCALL_DEFINE4(ptrace, compat_long_t, request, compat_long_t, pid,
1418 compat_long_t, addr, compat_long_t, data)
1420 struct task_struct *child;
1423 if (request == PTRACE_TRACEME) {
1424 ret = ptrace_traceme();
1428 child = find_get_task_by_vpid(pid);
1434 if (request == PTRACE_ATTACH || request == PTRACE_SEIZE) {
1435 ret = ptrace_attach(child, request, addr, data);
1436 goto out_put_task_struct;
1439 ret = ptrace_check_attach(child, request == PTRACE_KILL ||
1440 request == PTRACE_INTERRUPT);
1442 ret = compat_arch_ptrace(child, request, addr, data);
1443 if (ret || request != PTRACE_DETACH)
1444 ptrace_unfreeze_traced(child);
1447 out_put_task_struct:
1448 put_task_struct(child);
1452 #endif /* CONFIG_COMPAT */