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) {
116 * child->sighand can't be NULL, release_task()
117 * does ptrace_unlink() before __exit_signal().
119 spin_lock_irq(&child->sighand->siglock);
120 WARN_ON_ONCE(task_is_stopped(child));
121 if (task_is_traced(child) || kill)
123 spin_unlock_irq(&child->sighand->siglock);
125 read_unlock(&tasklist_lock);
128 ret = wait_task_inactive(child, TASK_TRACED) ? 0 : -ESRCH;
130 /* All systems go.. */
134 int __ptrace_may_access(struct task_struct *task, unsigned int mode)
136 const struct cred *cred = current_cred(), *tcred;
138 /* May we inspect the given task?
139 * This check is used both for attaching with ptrace
140 * and for allowing access to sensitive information in /proc.
142 * ptrace_attach denies several cases that /proc allows
143 * because setting up the necessary parent/child relationship
144 * or halting the specified task is impossible.
147 /* Don't let security modules deny introspection */
151 tcred = __task_cred(task);
152 if ((cred->uid != tcred->euid ||
153 cred->uid != tcred->suid ||
154 cred->uid != tcred->uid ||
155 cred->gid != tcred->egid ||
156 cred->gid != tcred->sgid ||
157 cred->gid != tcred->gid) &&
158 !capable(CAP_SYS_PTRACE)) {
165 dumpable = get_dumpable(task->mm);
166 if (!dumpable && !capable(CAP_SYS_PTRACE))
169 return security_ptrace_access_check(task, mode);
172 bool ptrace_may_access(struct task_struct *task, unsigned int mode)
176 err = __ptrace_may_access(task, mode);
181 static int ptrace_attach(struct task_struct *task)
183 bool wait_trap = false;
189 if (unlikely(task->flags & PF_KTHREAD))
191 if (same_thread_group(task, current))
195 * Protect exec's credential calculations against our interference;
196 * interference; SUID, SGID and LSM creds get determined differently
199 retval = -ERESTARTNOINTR;
200 if (mutex_lock_interruptible(&task->signal->cred_guard_mutex))
204 retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH);
209 write_lock_irq(&tasklist_lock);
211 if (unlikely(task->exit_state))
212 goto unlock_tasklist;
214 goto unlock_tasklist;
216 task->ptrace = PT_PTRACED;
217 if (capable(CAP_SYS_PTRACE))
218 task->ptrace |= PT_PTRACE_CAP;
220 __ptrace_link(task, current);
221 send_sig_info(SIGSTOP, SEND_SIG_FORCED, task);
223 spin_lock(&task->sighand->siglock);
226 * If the task is already STOPPED, set GROUP_STOP_PENDING and
227 * TRAPPING, and kick it so that it transits to TRACED. TRAPPING
228 * will be cleared if the child completes the transition or any
229 * event which clears the group stop states happens. We'll wait
230 * for the transition to complete before returning from this
233 * This hides STOPPED -> RUNNING -> TRACED transition from the
234 * attaching thread but a different thread in the same group can
235 * still observe the transient RUNNING state. IOW, if another
236 * thread's WNOHANG wait(2) on the stopped tracee races against
237 * ATTACH, the wait(2) may fail due to the transient RUNNING.
239 * The following task_is_stopped() test is safe as both transitions
240 * in and out of STOPPED are protected by siglock.
242 if (task_is_stopped(task)) {
243 task->group_stop |= GROUP_STOP_PENDING | GROUP_STOP_TRAPPING;
244 signal_wake_up(task, 1);
248 spin_unlock(&task->sighand->siglock);
252 write_unlock_irq(&tasklist_lock);
254 mutex_unlock(&task->signal->cred_guard_mutex);
257 wait_event(current->signal->wait_chldexit,
258 !(task->group_stop & GROUP_STOP_TRAPPING));
263 * ptrace_traceme -- helper for PTRACE_TRACEME
265 * Performs checks and sets PT_PTRACED.
266 * Should be used by all ptrace implementations for PTRACE_TRACEME.
268 static int ptrace_traceme(void)
272 write_lock_irq(&tasklist_lock);
273 /* Are we already being traced? */
274 if (!current->ptrace) {
275 ret = security_ptrace_traceme(current->parent);
277 * Check PF_EXITING to ensure ->real_parent has not passed
278 * exit_ptrace(). Otherwise we don't report the error but
279 * pretend ->real_parent untraces us right after return.
281 if (!ret && !(current->real_parent->flags & PF_EXITING)) {
282 current->ptrace = PT_PTRACED;
283 __ptrace_link(current, current->real_parent);
286 write_unlock_irq(&tasklist_lock);
292 * Called with irqs disabled, returns true if childs should reap themselves.
294 static int ignoring_children(struct sighand_struct *sigh)
297 spin_lock(&sigh->siglock);
298 ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
299 (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
300 spin_unlock(&sigh->siglock);
305 * Called with tasklist_lock held for writing.
306 * Unlink a traced task, and clean it up if it was a traced zombie.
307 * Return true if it needs to be reaped with release_task().
308 * (We can't call release_task() here because we already hold tasklist_lock.)
310 * If it's a zombie, our attachedness prevented normal parent notification
311 * or self-reaping. Do notification now if it would have happened earlier.
312 * If it should reap itself, return true.
314 * If it's our own child, there is no notification to do. But if our normal
315 * children self-reap, then this child was prevented by ptrace and we must
316 * reap it now, in that case we must also wake up sub-threads sleeping in
319 static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
323 if (p->exit_state == EXIT_ZOMBIE) {
324 if (!task_detached(p) && thread_group_empty(p)) {
325 if (!same_thread_group(p->real_parent, tracer))
326 do_notify_parent(p, p->exit_signal);
327 else if (ignoring_children(tracer->sighand)) {
328 __wake_up_parent(p, tracer);
332 if (task_detached(p)) {
333 /* Mark it as in the process of being reaped. */
334 p->exit_state = EXIT_DEAD;
342 static int ptrace_detach(struct task_struct *child, unsigned int data)
346 if (!valid_signal(data))
349 /* Architecture-specific hardware disable .. */
350 ptrace_disable(child);
351 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
353 write_lock_irq(&tasklist_lock);
355 * This child can be already killed. Make sure de_thread() or
356 * our sub-thread doing do_wait() didn't do release_task() yet.
359 child->exit_code = data;
360 dead = __ptrace_detach(current, child);
362 write_unlock_irq(&tasklist_lock);
371 * Detach all tasks we were using ptrace on. Called with tasklist held
372 * for writing, and returns with it held too. But note it can release
373 * and reacquire the lock.
375 void exit_ptrace(struct task_struct *tracer)
376 __releases(&tasklist_lock)
377 __acquires(&tasklist_lock)
379 struct task_struct *p, *n;
380 LIST_HEAD(ptrace_dead);
382 if (likely(list_empty(&tracer->ptraced)))
385 list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
386 if (__ptrace_detach(tracer, p))
387 list_add(&p->ptrace_entry, &ptrace_dead);
390 write_unlock_irq(&tasklist_lock);
391 BUG_ON(!list_empty(&tracer->ptraced));
393 list_for_each_entry_safe(p, n, &ptrace_dead, ptrace_entry) {
394 list_del_init(&p->ptrace_entry);
398 write_lock_irq(&tasklist_lock);
401 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
407 int this_len, retval;
409 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
410 retval = access_process_vm(tsk, src, buf, this_len, 0);
416 if (copy_to_user(dst, buf, retval))
426 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
432 int this_len, retval;
434 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
435 if (copy_from_user(buf, src, this_len))
437 retval = access_process_vm(tsk, dst, buf, this_len, 1);
451 static int ptrace_setoptions(struct task_struct *child, unsigned long data)
453 child->ptrace &= ~PT_TRACE_MASK;
455 if (data & PTRACE_O_TRACESYSGOOD)
456 child->ptrace |= PT_TRACESYSGOOD;
458 if (data & PTRACE_O_TRACEFORK)
459 child->ptrace |= PT_TRACE_FORK;
461 if (data & PTRACE_O_TRACEVFORK)
462 child->ptrace |= PT_TRACE_VFORK;
464 if (data & PTRACE_O_TRACECLONE)
465 child->ptrace |= PT_TRACE_CLONE;
467 if (data & PTRACE_O_TRACEEXEC)
468 child->ptrace |= PT_TRACE_EXEC;
470 if (data & PTRACE_O_TRACEVFORKDONE)
471 child->ptrace |= PT_TRACE_VFORK_DONE;
473 if (data & PTRACE_O_TRACEEXIT)
474 child->ptrace |= PT_TRACE_EXIT;
476 return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
479 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t *info)
484 if (lock_task_sighand(child, &flags)) {
486 if (likely(child->last_siginfo != NULL)) {
487 *info = *child->last_siginfo;
490 unlock_task_sighand(child, &flags);
495 static int ptrace_setsiginfo(struct task_struct *child, const siginfo_t *info)
500 if (lock_task_sighand(child, &flags)) {
502 if (likely(child->last_siginfo != NULL)) {
503 *child->last_siginfo = *info;
506 unlock_task_sighand(child, &flags);
512 #ifdef PTRACE_SINGLESTEP
513 #define is_singlestep(request) ((request) == PTRACE_SINGLESTEP)
515 #define is_singlestep(request) 0
518 #ifdef PTRACE_SINGLEBLOCK
519 #define is_singleblock(request) ((request) == PTRACE_SINGLEBLOCK)
521 #define is_singleblock(request) 0
525 #define is_sysemu_singlestep(request) ((request) == PTRACE_SYSEMU_SINGLESTEP)
527 #define is_sysemu_singlestep(request) 0
530 static int ptrace_resume(struct task_struct *child, long request,
533 if (!valid_signal(data))
536 if (request == PTRACE_SYSCALL)
537 set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
539 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
541 #ifdef TIF_SYSCALL_EMU
542 if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
543 set_tsk_thread_flag(child, TIF_SYSCALL_EMU);
545 clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
548 if (is_singleblock(request)) {
549 if (unlikely(!arch_has_block_step()))
551 user_enable_block_step(child);
552 } else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
553 if (unlikely(!arch_has_single_step()))
555 user_enable_single_step(child);
557 user_disable_single_step(child);
560 child->exit_code = data;
561 wake_up_process(child);
566 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
568 static const struct user_regset *
569 find_regset(const struct user_regset_view *view, unsigned int type)
571 const struct user_regset *regset;
574 for (n = 0; n < view->n; ++n) {
575 regset = view->regsets + n;
576 if (regset->core_note_type == type)
583 static int ptrace_regset(struct task_struct *task, int req, unsigned int type,
586 const struct user_regset_view *view = task_user_regset_view(task);
587 const struct user_regset *regset = find_regset(view, type);
590 if (!regset || (kiov->iov_len % regset->size) != 0)
593 regset_no = regset - view->regsets;
594 kiov->iov_len = min(kiov->iov_len,
595 (__kernel_size_t) (regset->n * regset->size));
597 if (req == PTRACE_GETREGSET)
598 return copy_regset_to_user(task, view, regset_no, 0,
599 kiov->iov_len, kiov->iov_base);
601 return copy_regset_from_user(task, view, regset_no, 0,
602 kiov->iov_len, kiov->iov_base);
607 int ptrace_request(struct task_struct *child, long request,
608 unsigned long addr, unsigned long data)
612 void __user *datavp = (void __user *) data;
613 unsigned long __user *datalp = datavp;
616 case PTRACE_PEEKTEXT:
617 case PTRACE_PEEKDATA:
618 return generic_ptrace_peekdata(child, addr, data);
619 case PTRACE_POKETEXT:
620 case PTRACE_POKEDATA:
621 return generic_ptrace_pokedata(child, addr, data);
623 #ifdef PTRACE_OLDSETOPTIONS
624 case PTRACE_OLDSETOPTIONS:
626 case PTRACE_SETOPTIONS:
627 ret = ptrace_setoptions(child, data);
629 case PTRACE_GETEVENTMSG:
630 ret = put_user(child->ptrace_message, datalp);
633 case PTRACE_GETSIGINFO:
634 ret = ptrace_getsiginfo(child, &siginfo);
636 ret = copy_siginfo_to_user(datavp, &siginfo);
639 case PTRACE_SETSIGINFO:
640 if (copy_from_user(&siginfo, datavp, sizeof siginfo))
643 ret = ptrace_setsiginfo(child, &siginfo);
646 case PTRACE_DETACH: /* detach a process that was attached. */
647 ret = ptrace_detach(child, data);
650 #ifdef CONFIG_BINFMT_ELF_FDPIC
651 case PTRACE_GETFDPIC: {
652 struct mm_struct *mm = get_task_mm(child);
653 unsigned long tmp = 0;
660 case PTRACE_GETFDPIC_EXEC:
661 tmp = mm->context.exec_fdpic_loadmap;
663 case PTRACE_GETFDPIC_INTERP:
664 tmp = mm->context.interp_fdpic_loadmap;
671 ret = put_user(tmp, datalp);
676 #ifdef PTRACE_SINGLESTEP
677 case PTRACE_SINGLESTEP:
679 #ifdef PTRACE_SINGLEBLOCK
680 case PTRACE_SINGLEBLOCK:
684 case PTRACE_SYSEMU_SINGLESTEP:
688 return ptrace_resume(child, request, data);
691 if (child->exit_state) /* already dead */
693 return ptrace_resume(child, request, SIGKILL);
695 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
696 case PTRACE_GETREGSET:
697 case PTRACE_SETREGSET:
700 struct iovec __user *uiov = datavp;
702 if (!access_ok(VERIFY_WRITE, uiov, sizeof(*uiov)))
705 if (__get_user(kiov.iov_base, &uiov->iov_base) ||
706 __get_user(kiov.iov_len, &uiov->iov_len))
709 ret = ptrace_regset(child, request, addr, &kiov);
711 ret = __put_user(kiov.iov_len, &uiov->iov_len);
722 static struct task_struct *ptrace_get_task_struct(pid_t pid)
724 struct task_struct *child;
727 child = find_task_by_vpid(pid);
729 get_task_struct(child);
733 return ERR_PTR(-ESRCH);
737 #ifndef arch_ptrace_attach
738 #define arch_ptrace_attach(child) do { } while (0)
741 SYSCALL_DEFINE4(ptrace, long, request, long, pid, unsigned long, addr,
744 struct task_struct *child;
747 if (request == PTRACE_TRACEME) {
748 ret = ptrace_traceme();
750 arch_ptrace_attach(current);
754 child = ptrace_get_task_struct(pid);
756 ret = PTR_ERR(child);
760 if (request == PTRACE_ATTACH) {
761 ret = ptrace_attach(child);
763 * Some architectures need to do book-keeping after
767 arch_ptrace_attach(child);
768 goto out_put_task_struct;
771 ret = ptrace_check_attach(child, request == PTRACE_KILL);
773 goto out_put_task_struct;
775 ret = arch_ptrace(child, request, addr, data);
778 put_task_struct(child);
783 int generic_ptrace_peekdata(struct task_struct *tsk, unsigned long addr,
789 copied = access_process_vm(tsk, addr, &tmp, sizeof(tmp), 0);
790 if (copied != sizeof(tmp))
792 return put_user(tmp, (unsigned long __user *)data);
795 int generic_ptrace_pokedata(struct task_struct *tsk, unsigned long addr,
800 copied = access_process_vm(tsk, addr, &data, sizeof(data), 1);
801 return (copied == sizeof(data)) ? 0 : -EIO;
804 #if defined CONFIG_COMPAT
805 #include <linux/compat.h>
807 int compat_ptrace_request(struct task_struct *child, compat_long_t request,
808 compat_ulong_t addr, compat_ulong_t data)
810 compat_ulong_t __user *datap = compat_ptr(data);
816 case PTRACE_PEEKTEXT:
817 case PTRACE_PEEKDATA:
818 ret = access_process_vm(child, addr, &word, sizeof(word), 0);
819 if (ret != sizeof(word))
822 ret = put_user(word, datap);
825 case PTRACE_POKETEXT:
826 case PTRACE_POKEDATA:
827 ret = access_process_vm(child, addr, &data, sizeof(data), 1);
828 ret = (ret != sizeof(data) ? -EIO : 0);
831 case PTRACE_GETEVENTMSG:
832 ret = put_user((compat_ulong_t) child->ptrace_message, datap);
835 case PTRACE_GETSIGINFO:
836 ret = ptrace_getsiginfo(child, &siginfo);
838 ret = copy_siginfo_to_user32(
839 (struct compat_siginfo __user *) datap,
843 case PTRACE_SETSIGINFO:
844 memset(&siginfo, 0, sizeof siginfo);
845 if (copy_siginfo_from_user32(
846 &siginfo, (struct compat_siginfo __user *) datap))
849 ret = ptrace_setsiginfo(child, &siginfo);
851 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
852 case PTRACE_GETREGSET:
853 case PTRACE_SETREGSET:
856 struct compat_iovec __user *uiov =
857 (struct compat_iovec __user *) datap;
861 if (!access_ok(VERIFY_WRITE, uiov, sizeof(*uiov)))
864 if (__get_user(ptr, &uiov->iov_base) ||
865 __get_user(len, &uiov->iov_len))
868 kiov.iov_base = compat_ptr(ptr);
871 ret = ptrace_regset(child, request, addr, &kiov);
873 ret = __put_user(kiov.iov_len, &uiov->iov_len);
879 ret = ptrace_request(child, request, addr, data);
885 asmlinkage long compat_sys_ptrace(compat_long_t request, compat_long_t pid,
886 compat_long_t addr, compat_long_t data)
888 struct task_struct *child;
891 if (request == PTRACE_TRACEME) {
892 ret = ptrace_traceme();
896 child = ptrace_get_task_struct(pid);
898 ret = PTR_ERR(child);
902 if (request == PTRACE_ATTACH) {
903 ret = ptrace_attach(child);
905 * Some architectures need to do book-keeping after
909 arch_ptrace_attach(child);
910 goto out_put_task_struct;
913 ret = ptrace_check_attach(child, request == PTRACE_KILL);
915 ret = compat_arch_ptrace(child, request, addr, data);
918 put_task_struct(child);
922 #endif /* CONFIG_COMPAT */