2 * linux/kernel/ptrace.c
4 * (C) Copyright 1999 Linus Torvalds
6 * Common interfaces for "ptrace()" which we do not want
7 * to continually duplicate across every architecture.
10 #include <linux/capability.h>
11 #include <linux/module.h>
12 #include <linux/sched.h>
13 #include <linux/errno.h>
15 #include <linux/highmem.h>
16 #include <linux/pagemap.h>
17 #include <linux/ptrace.h>
18 #include <linux/security.h>
19 #include <linux/signal.h>
20 #include <linux/audit.h>
21 #include <linux/pid_namespace.h>
22 #include <linux/syscalls.h>
23 #include <linux/uaccess.h>
24 #include <linux/regset.h>
28 * ptrace a task: make the debugger its new parent and
29 * move it to the ptrace list.
31 * Must be called with the tasklist lock write-held.
33 void __ptrace_link(struct task_struct *child, struct task_struct *new_parent)
35 BUG_ON(!list_empty(&child->ptrace_entry));
36 list_add(&child->ptrace_entry, &new_parent->ptraced);
37 child->parent = new_parent;
41 * __ptrace_unlink - unlink ptracee and restore its execution state
42 * @child: ptracee to be unlinked
44 * Remove @child from the ptrace list, move it back to the original parent,
45 * and restore the execution state so that it conforms to the group stop
48 * Unlinking can happen via two paths - explicit PTRACE_DETACH or ptracer
49 * exiting. For PTRACE_DETACH, unless the ptracee has been killed between
50 * ptrace_check_attach() and here, it's guaranteed to be in TASK_TRACED.
51 * If the ptracer is exiting, the ptracee can be in any state.
53 * After detach, the ptracee should be in a state which conforms to the
54 * group stop. If the group is stopped or in the process of stopping, the
55 * ptracee should be put into TASK_STOPPED; otherwise, it should be woken
56 * up from TASK_TRACED.
58 * If the ptracee is in TASK_TRACED and needs to be moved to TASK_STOPPED,
59 * it goes through TRACED -> RUNNING -> STOPPED transition which is similar
60 * to but in the opposite direction of what happens while attaching to a
61 * stopped task. However, in this direction, the intermediate RUNNING
62 * state is not hidden even from the current ptracer and if it immediately
63 * re-attaches and performs a WNOHANG wait(2), it may fail.
66 * write_lock_irq(tasklist_lock)
68 void __ptrace_unlink(struct task_struct *child)
70 BUG_ON(!child->ptrace);
73 child->parent = child->real_parent;
74 list_del_init(&child->ptrace_entry);
76 spin_lock(&child->sighand->siglock);
79 * Reinstate GROUP_STOP_PENDING if group stop is in effect and
82 if (!(child->flags & PF_EXITING) &&
83 (child->signal->flags & SIGNAL_STOP_STOPPED ||
84 child->signal->group_stop_count))
85 child->group_stop |= GROUP_STOP_PENDING;
88 * If transition to TASK_STOPPED is pending or in TASK_TRACED, kick
89 * @child in the butt. Note that @resume should be used iff @child
90 * is in TASK_TRACED; otherwise, we might unduly disrupt
91 * TASK_KILLABLE sleeps.
93 if (child->group_stop & GROUP_STOP_PENDING || task_is_traced(child))
94 signal_wake_up(child, task_is_traced(child));
96 spin_unlock(&child->sighand->siglock);
100 * Check that we have indeed attached to the thing..
102 int ptrace_check_attach(struct task_struct *child, int kill)
107 * We take the read lock around doing both checks to close a
108 * possible race where someone else was tracing our child and
109 * detached between these two checks. After this locked check,
110 * we are sure that this is our traced child and that can only
111 * be changed by us so it's not changing right after this.
113 read_lock(&tasklist_lock);
114 if ((child->ptrace & PT_PTRACED) && child->parent == current) {
117 * child->sighand can't be NULL, release_task()
118 * does ptrace_unlink() before __exit_signal().
120 spin_lock_irq(&child->sighand->siglock);
121 if (task_is_stopped(child))
122 child->state = TASK_TRACED;
123 else if (!task_is_traced(child) && !kill)
125 spin_unlock_irq(&child->sighand->siglock);
127 read_unlock(&tasklist_lock);
130 ret = wait_task_inactive(child, TASK_TRACED) ? 0 : -ESRCH;
132 /* All systems go.. */
136 int __ptrace_may_access(struct task_struct *task, unsigned int mode)
138 const struct cred *cred = current_cred(), *tcred;
140 /* May we inspect the given task?
141 * This check is used both for attaching with ptrace
142 * and for allowing access to sensitive information in /proc.
144 * ptrace_attach denies several cases that /proc allows
145 * because setting up the necessary parent/child relationship
146 * or halting the specified task is impossible.
149 /* Don't let security modules deny introspection */
153 tcred = __task_cred(task);
154 if ((cred->uid != tcred->euid ||
155 cred->uid != tcred->suid ||
156 cred->uid != tcred->uid ||
157 cred->gid != tcred->egid ||
158 cred->gid != tcred->sgid ||
159 cred->gid != tcred->gid) &&
160 !capable(CAP_SYS_PTRACE)) {
167 dumpable = get_dumpable(task->mm);
168 if (!dumpable && !capable(CAP_SYS_PTRACE))
171 return security_ptrace_access_check(task, mode);
174 bool ptrace_may_access(struct task_struct *task, unsigned int mode)
178 err = __ptrace_may_access(task, mode);
183 static int ptrace_attach(struct task_struct *task)
185 bool wait_trap = false;
191 if (unlikely(task->flags & PF_KTHREAD))
193 if (same_thread_group(task, current))
197 * Protect exec's credential calculations against our interference;
198 * interference; SUID, SGID and LSM creds get determined differently
201 retval = -ERESTARTNOINTR;
202 if (mutex_lock_interruptible(&task->signal->cred_guard_mutex))
206 retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH);
211 write_lock_irq(&tasklist_lock);
213 if (unlikely(task->exit_state))
214 goto unlock_tasklist;
216 goto unlock_tasklist;
218 task->ptrace = PT_PTRACED;
219 if (capable(CAP_SYS_PTRACE))
220 task->ptrace |= PT_PTRACE_CAP;
222 __ptrace_link(task, current);
223 send_sig_info(SIGSTOP, SEND_SIG_FORCED, task);
225 spin_lock(&task->sighand->siglock);
228 * If the task is already STOPPED, set GROUP_STOP_PENDING and
229 * TRAPPING, and kick it so that it transits to TRACED. TRAPPING
230 * will be cleared if the child completes the transition or any
231 * event which clears the group stop states happens. We'll wait
232 * for the transition to complete before returning from this
235 * This hides STOPPED -> RUNNING -> TRACED transition from the
236 * attaching thread but a different thread in the same group can
237 * still observe the transient RUNNING state. IOW, if another
238 * thread's WNOHANG wait(2) on the stopped tracee races against
239 * ATTACH, the wait(2) may fail due to the transient RUNNING.
241 * The following task_is_stopped() test is safe as both transitions
242 * in and out of STOPPED are protected by siglock.
244 if (task_is_stopped(task)) {
245 task->group_stop |= GROUP_STOP_PENDING | GROUP_STOP_TRAPPING;
246 signal_wake_up(task, 1);
250 spin_unlock(&task->sighand->siglock);
254 write_unlock_irq(&tasklist_lock);
256 mutex_unlock(&task->signal->cred_guard_mutex);
259 wait_event(current->signal->wait_chldexit,
260 !(task->group_stop & GROUP_STOP_TRAPPING));
265 * ptrace_traceme -- helper for PTRACE_TRACEME
267 * Performs checks and sets PT_PTRACED.
268 * Should be used by all ptrace implementations for PTRACE_TRACEME.
270 static int ptrace_traceme(void)
274 write_lock_irq(&tasklist_lock);
275 /* Are we already being traced? */
276 if (!current->ptrace) {
277 ret = security_ptrace_traceme(current->parent);
279 * Check PF_EXITING to ensure ->real_parent has not passed
280 * exit_ptrace(). Otherwise we don't report the error but
281 * pretend ->real_parent untraces us right after return.
283 if (!ret && !(current->real_parent->flags & PF_EXITING)) {
284 current->ptrace = PT_PTRACED;
285 __ptrace_link(current, current->real_parent);
288 write_unlock_irq(&tasklist_lock);
294 * Called with irqs disabled, returns true if childs should reap themselves.
296 static int ignoring_children(struct sighand_struct *sigh)
299 spin_lock(&sigh->siglock);
300 ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
301 (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
302 spin_unlock(&sigh->siglock);
307 * Called with tasklist_lock held for writing.
308 * Unlink a traced task, and clean it up if it was a traced zombie.
309 * Return true if it needs to be reaped with release_task().
310 * (We can't call release_task() here because we already hold tasklist_lock.)
312 * If it's a zombie, our attachedness prevented normal parent notification
313 * or self-reaping. Do notification now if it would have happened earlier.
314 * If it should reap itself, return true.
316 * If it's our own child, there is no notification to do. But if our normal
317 * children self-reap, then this child was prevented by ptrace and we must
318 * reap it now, in that case we must also wake up sub-threads sleeping in
321 static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
325 if (p->exit_state == EXIT_ZOMBIE) {
326 if (!task_detached(p) && thread_group_empty(p)) {
327 if (!same_thread_group(p->real_parent, tracer))
328 do_notify_parent(p, p->exit_signal);
329 else if (ignoring_children(tracer->sighand)) {
330 __wake_up_parent(p, tracer);
334 if (task_detached(p)) {
335 /* Mark it as in the process of being reaped. */
336 p->exit_state = EXIT_DEAD;
344 static int ptrace_detach(struct task_struct *child, unsigned int data)
348 if (!valid_signal(data))
351 /* Architecture-specific hardware disable .. */
352 ptrace_disable(child);
353 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
355 write_lock_irq(&tasklist_lock);
357 * This child can be already killed. Make sure de_thread() or
358 * our sub-thread doing do_wait() didn't do release_task() yet.
361 child->exit_code = data;
362 dead = __ptrace_detach(current, child);
364 write_unlock_irq(&tasklist_lock);
373 * Detach all tasks we were using ptrace on. Called with tasklist held
374 * for writing, and returns with it held too. But note it can release
375 * and reacquire the lock.
377 void exit_ptrace(struct task_struct *tracer)
378 __releases(&tasklist_lock)
379 __acquires(&tasklist_lock)
381 struct task_struct *p, *n;
382 LIST_HEAD(ptrace_dead);
384 if (likely(list_empty(&tracer->ptraced)))
387 list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
388 if (__ptrace_detach(tracer, p))
389 list_add(&p->ptrace_entry, &ptrace_dead);
392 write_unlock_irq(&tasklist_lock);
393 BUG_ON(!list_empty(&tracer->ptraced));
395 list_for_each_entry_safe(p, n, &ptrace_dead, ptrace_entry) {
396 list_del_init(&p->ptrace_entry);
400 write_lock_irq(&tasklist_lock);
403 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
409 int this_len, retval;
411 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
412 retval = access_process_vm(tsk, src, buf, this_len, 0);
418 if (copy_to_user(dst, buf, retval))
428 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
434 int this_len, retval;
436 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
437 if (copy_from_user(buf, src, this_len))
439 retval = access_process_vm(tsk, dst, buf, this_len, 1);
453 static int ptrace_setoptions(struct task_struct *child, unsigned long data)
455 child->ptrace &= ~PT_TRACE_MASK;
457 if (data & PTRACE_O_TRACESYSGOOD)
458 child->ptrace |= PT_TRACESYSGOOD;
460 if (data & PTRACE_O_TRACEFORK)
461 child->ptrace |= PT_TRACE_FORK;
463 if (data & PTRACE_O_TRACEVFORK)
464 child->ptrace |= PT_TRACE_VFORK;
466 if (data & PTRACE_O_TRACECLONE)
467 child->ptrace |= PT_TRACE_CLONE;
469 if (data & PTRACE_O_TRACEEXEC)
470 child->ptrace |= PT_TRACE_EXEC;
472 if (data & PTRACE_O_TRACEVFORKDONE)
473 child->ptrace |= PT_TRACE_VFORK_DONE;
475 if (data & PTRACE_O_TRACEEXIT)
476 child->ptrace |= PT_TRACE_EXIT;
478 return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
481 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t *info)
486 if (lock_task_sighand(child, &flags)) {
488 if (likely(child->last_siginfo != NULL)) {
489 *info = *child->last_siginfo;
492 unlock_task_sighand(child, &flags);
497 static int ptrace_setsiginfo(struct task_struct *child, const siginfo_t *info)
502 if (lock_task_sighand(child, &flags)) {
504 if (likely(child->last_siginfo != NULL)) {
505 *child->last_siginfo = *info;
508 unlock_task_sighand(child, &flags);
514 #ifdef PTRACE_SINGLESTEP
515 #define is_singlestep(request) ((request) == PTRACE_SINGLESTEP)
517 #define is_singlestep(request) 0
520 #ifdef PTRACE_SINGLEBLOCK
521 #define is_singleblock(request) ((request) == PTRACE_SINGLEBLOCK)
523 #define is_singleblock(request) 0
527 #define is_sysemu_singlestep(request) ((request) == PTRACE_SYSEMU_SINGLESTEP)
529 #define is_sysemu_singlestep(request) 0
532 static int ptrace_resume(struct task_struct *child, long request,
535 if (!valid_signal(data))
538 if (request == PTRACE_SYSCALL)
539 set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
541 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
543 #ifdef TIF_SYSCALL_EMU
544 if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
545 set_tsk_thread_flag(child, TIF_SYSCALL_EMU);
547 clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
550 if (is_singleblock(request)) {
551 if (unlikely(!arch_has_block_step()))
553 user_enable_block_step(child);
554 } else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
555 if (unlikely(!arch_has_single_step()))
557 user_enable_single_step(child);
559 user_disable_single_step(child);
562 child->exit_code = data;
563 wake_up_process(child);
568 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
570 static const struct user_regset *
571 find_regset(const struct user_regset_view *view, unsigned int type)
573 const struct user_regset *regset;
576 for (n = 0; n < view->n; ++n) {
577 regset = view->regsets + n;
578 if (regset->core_note_type == type)
585 static int ptrace_regset(struct task_struct *task, int req, unsigned int type,
588 const struct user_regset_view *view = task_user_regset_view(task);
589 const struct user_regset *regset = find_regset(view, type);
592 if (!regset || (kiov->iov_len % regset->size) != 0)
595 regset_no = regset - view->regsets;
596 kiov->iov_len = min(kiov->iov_len,
597 (__kernel_size_t) (regset->n * regset->size));
599 if (req == PTRACE_GETREGSET)
600 return copy_regset_to_user(task, view, regset_no, 0,
601 kiov->iov_len, kiov->iov_base);
603 return copy_regset_from_user(task, view, regset_no, 0,
604 kiov->iov_len, kiov->iov_base);
609 int ptrace_request(struct task_struct *child, long request,
610 unsigned long addr, unsigned long data)
614 void __user *datavp = (void __user *) data;
615 unsigned long __user *datalp = datavp;
618 case PTRACE_PEEKTEXT:
619 case PTRACE_PEEKDATA:
620 return generic_ptrace_peekdata(child, addr, data);
621 case PTRACE_POKETEXT:
622 case PTRACE_POKEDATA:
623 return generic_ptrace_pokedata(child, addr, data);
625 #ifdef PTRACE_OLDSETOPTIONS
626 case PTRACE_OLDSETOPTIONS:
628 case PTRACE_SETOPTIONS:
629 ret = ptrace_setoptions(child, data);
631 case PTRACE_GETEVENTMSG:
632 ret = put_user(child->ptrace_message, datalp);
635 case PTRACE_GETSIGINFO:
636 ret = ptrace_getsiginfo(child, &siginfo);
638 ret = copy_siginfo_to_user(datavp, &siginfo);
641 case PTRACE_SETSIGINFO:
642 if (copy_from_user(&siginfo, datavp, sizeof siginfo))
645 ret = ptrace_setsiginfo(child, &siginfo);
648 case PTRACE_DETACH: /* detach a process that was attached. */
649 ret = ptrace_detach(child, data);
652 #ifdef CONFIG_BINFMT_ELF_FDPIC
653 case PTRACE_GETFDPIC: {
654 struct mm_struct *mm = get_task_mm(child);
655 unsigned long tmp = 0;
662 case PTRACE_GETFDPIC_EXEC:
663 tmp = mm->context.exec_fdpic_loadmap;
665 case PTRACE_GETFDPIC_INTERP:
666 tmp = mm->context.interp_fdpic_loadmap;
673 ret = put_user(tmp, datalp);
678 #ifdef PTRACE_SINGLESTEP
679 case PTRACE_SINGLESTEP:
681 #ifdef PTRACE_SINGLEBLOCK
682 case PTRACE_SINGLEBLOCK:
686 case PTRACE_SYSEMU_SINGLESTEP:
690 return ptrace_resume(child, request, data);
693 if (child->exit_state) /* already dead */
695 return ptrace_resume(child, request, SIGKILL);
697 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
698 case PTRACE_GETREGSET:
699 case PTRACE_SETREGSET:
702 struct iovec __user *uiov = datavp;
704 if (!access_ok(VERIFY_WRITE, uiov, sizeof(*uiov)))
707 if (__get_user(kiov.iov_base, &uiov->iov_base) ||
708 __get_user(kiov.iov_len, &uiov->iov_len))
711 ret = ptrace_regset(child, request, addr, &kiov);
713 ret = __put_user(kiov.iov_len, &uiov->iov_len);
724 static struct task_struct *ptrace_get_task_struct(pid_t pid)
726 struct task_struct *child;
729 child = find_task_by_vpid(pid);
731 get_task_struct(child);
735 return ERR_PTR(-ESRCH);
739 #ifndef arch_ptrace_attach
740 #define arch_ptrace_attach(child) do { } while (0)
743 SYSCALL_DEFINE4(ptrace, long, request, long, pid, unsigned long, addr,
746 struct task_struct *child;
749 if (request == PTRACE_TRACEME) {
750 ret = ptrace_traceme();
752 arch_ptrace_attach(current);
756 child = ptrace_get_task_struct(pid);
758 ret = PTR_ERR(child);
762 if (request == PTRACE_ATTACH) {
763 ret = ptrace_attach(child);
765 * Some architectures need to do book-keeping after
769 arch_ptrace_attach(child);
770 goto out_put_task_struct;
773 ret = ptrace_check_attach(child, request == PTRACE_KILL);
775 goto out_put_task_struct;
777 ret = arch_ptrace(child, request, addr, data);
780 put_task_struct(child);
785 int generic_ptrace_peekdata(struct task_struct *tsk, unsigned long addr,
791 copied = access_process_vm(tsk, addr, &tmp, sizeof(tmp), 0);
792 if (copied != sizeof(tmp))
794 return put_user(tmp, (unsigned long __user *)data);
797 int generic_ptrace_pokedata(struct task_struct *tsk, unsigned long addr,
802 copied = access_process_vm(tsk, addr, &data, sizeof(data), 1);
803 return (copied == sizeof(data)) ? 0 : -EIO;
806 #if defined CONFIG_COMPAT
807 #include <linux/compat.h>
809 int compat_ptrace_request(struct task_struct *child, compat_long_t request,
810 compat_ulong_t addr, compat_ulong_t data)
812 compat_ulong_t __user *datap = compat_ptr(data);
818 case PTRACE_PEEKTEXT:
819 case PTRACE_PEEKDATA:
820 ret = access_process_vm(child, addr, &word, sizeof(word), 0);
821 if (ret != sizeof(word))
824 ret = put_user(word, datap);
827 case PTRACE_POKETEXT:
828 case PTRACE_POKEDATA:
829 ret = access_process_vm(child, addr, &data, sizeof(data), 1);
830 ret = (ret != sizeof(data) ? -EIO : 0);
833 case PTRACE_GETEVENTMSG:
834 ret = put_user((compat_ulong_t) child->ptrace_message, datap);
837 case PTRACE_GETSIGINFO:
838 ret = ptrace_getsiginfo(child, &siginfo);
840 ret = copy_siginfo_to_user32(
841 (struct compat_siginfo __user *) datap,
845 case PTRACE_SETSIGINFO:
846 memset(&siginfo, 0, sizeof siginfo);
847 if (copy_siginfo_from_user32(
848 &siginfo, (struct compat_siginfo __user *) datap))
851 ret = ptrace_setsiginfo(child, &siginfo);
853 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
854 case PTRACE_GETREGSET:
855 case PTRACE_SETREGSET:
858 struct compat_iovec __user *uiov =
859 (struct compat_iovec __user *) datap;
863 if (!access_ok(VERIFY_WRITE, uiov, sizeof(*uiov)))
866 if (__get_user(ptr, &uiov->iov_base) ||
867 __get_user(len, &uiov->iov_len))
870 kiov.iov_base = compat_ptr(ptr);
873 ret = ptrace_regset(child, request, addr, &kiov);
875 ret = __put_user(kiov.iov_len, &uiov->iov_len);
881 ret = ptrace_request(child, request, addr, data);
887 asmlinkage long compat_sys_ptrace(compat_long_t request, compat_long_t pid,
888 compat_long_t addr, compat_long_t data)
890 struct task_struct *child;
893 if (request == PTRACE_TRACEME) {
894 ret = ptrace_traceme();
898 child = ptrace_get_task_struct(pid);
900 ret = PTR_ERR(child);
904 if (request == PTRACE_ATTACH) {
905 ret = ptrace_attach(child);
907 * Some architectures need to do book-keeping after
911 arch_ptrace_attach(child);
912 goto out_put_task_struct;
915 ret = ptrace_check_attach(child, request == PTRACE_KILL);
917 ret = compat_arch_ptrace(child, request, addr, data);
920 put_task_struct(child);
924 #endif /* CONFIG_COMPAT */