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->user->user_ns == tcred->user->user_ns &&
153 (cred->uid == tcred->euid &&
154 cred->uid == tcred->suid &&
155 cred->uid == tcred->uid &&
156 cred->gid == tcred->egid &&
157 cred->gid == tcred->sgid &&
158 cred->gid == tcred->gid))
160 if (ns_capable(tcred->user->user_ns, CAP_SYS_PTRACE))
168 dumpable = get_dumpable(task->mm);
169 if (!dumpable && !task_ns_capable(task, CAP_SYS_PTRACE))
172 return security_ptrace_access_check(task, mode);
175 bool ptrace_may_access(struct task_struct *task, unsigned int mode)
179 err = __ptrace_may_access(task, mode);
184 static int ptrace_attach(struct task_struct *task)
186 bool wait_trap = false;
192 if (unlikely(task->flags & PF_KTHREAD))
194 if (same_thread_group(task, current))
198 * Protect exec's credential calculations against our interference;
199 * interference; SUID, SGID and LSM creds get determined differently
202 retval = -ERESTARTNOINTR;
203 if (mutex_lock_interruptible(&task->signal->cred_guard_mutex))
207 retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH);
212 write_lock_irq(&tasklist_lock);
214 if (unlikely(task->exit_state))
215 goto unlock_tasklist;
217 goto unlock_tasklist;
219 task->ptrace = PT_PTRACED;
220 if (task_ns_capable(task, CAP_SYS_PTRACE))
221 task->ptrace |= PT_PTRACE_CAP;
223 __ptrace_link(task, current);
224 send_sig_info(SIGSTOP, SEND_SIG_FORCED, task);
226 spin_lock(&task->sighand->siglock);
229 * If the task is already STOPPED, set GROUP_STOP_PENDING and
230 * TRAPPING, and kick it so that it transits to TRACED. TRAPPING
231 * will be cleared if the child completes the transition or any
232 * event which clears the group stop states happens. We'll wait
233 * for the transition to complete before returning from this
236 * This hides STOPPED -> RUNNING -> TRACED transition from the
237 * attaching thread but a different thread in the same group can
238 * still observe the transient RUNNING state. IOW, if another
239 * thread's WNOHANG wait(2) on the stopped tracee races against
240 * ATTACH, the wait(2) may fail due to the transient RUNNING.
242 * The following task_is_stopped() test is safe as both transitions
243 * in and out of STOPPED are protected by siglock.
245 if (task_is_stopped(task)) {
246 task->group_stop |= GROUP_STOP_PENDING | GROUP_STOP_TRAPPING;
247 signal_wake_up(task, 1);
251 spin_unlock(&task->sighand->siglock);
255 write_unlock_irq(&tasklist_lock);
257 mutex_unlock(&task->signal->cred_guard_mutex);
260 wait_event(current->signal->wait_chldexit,
261 !(task->group_stop & GROUP_STOP_TRAPPING));
266 * ptrace_traceme -- helper for PTRACE_TRACEME
268 * Performs checks and sets PT_PTRACED.
269 * Should be used by all ptrace implementations for PTRACE_TRACEME.
271 static int ptrace_traceme(void)
275 write_lock_irq(&tasklist_lock);
276 /* Are we already being traced? */
277 if (!current->ptrace) {
278 ret = security_ptrace_traceme(current->parent);
280 * Check PF_EXITING to ensure ->real_parent has not passed
281 * exit_ptrace(). Otherwise we don't report the error but
282 * pretend ->real_parent untraces us right after return.
284 if (!ret && !(current->real_parent->flags & PF_EXITING)) {
285 current->ptrace = PT_PTRACED;
286 __ptrace_link(current, current->real_parent);
289 write_unlock_irq(&tasklist_lock);
295 * Called with irqs disabled, returns true if childs should reap themselves.
297 static int ignoring_children(struct sighand_struct *sigh)
300 spin_lock(&sigh->siglock);
301 ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
302 (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
303 spin_unlock(&sigh->siglock);
308 * Called with tasklist_lock held for writing.
309 * Unlink a traced task, and clean it up if it was a traced zombie.
310 * Return true if it needs to be reaped with release_task().
311 * (We can't call release_task() here because we already hold tasklist_lock.)
313 * If it's a zombie, our attachedness prevented normal parent notification
314 * or self-reaping. Do notification now if it would have happened earlier.
315 * If it should reap itself, return true.
317 * If it's our own child, there is no notification to do. But if our normal
318 * children self-reap, then this child was prevented by ptrace and we must
319 * reap it now, in that case we must also wake up sub-threads sleeping in
322 static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
326 if (p->exit_state == EXIT_ZOMBIE) {
327 if (!task_detached(p) && thread_group_empty(p)) {
328 if (!same_thread_group(p->real_parent, tracer))
329 do_notify_parent(p, p->exit_signal);
330 else if (ignoring_children(tracer->sighand)) {
331 __wake_up_parent(p, tracer);
335 if (task_detached(p)) {
336 /* Mark it as in the process of being reaped. */
337 p->exit_state = EXIT_DEAD;
345 static int ptrace_detach(struct task_struct *child, unsigned int data)
349 if (!valid_signal(data))
352 /* Architecture-specific hardware disable .. */
353 ptrace_disable(child);
354 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
356 write_lock_irq(&tasklist_lock);
358 * This child can be already killed. Make sure de_thread() or
359 * our sub-thread doing do_wait() didn't do release_task() yet.
362 child->exit_code = data;
363 dead = __ptrace_detach(current, child);
365 write_unlock_irq(&tasklist_lock);
374 * Detach all tasks we were using ptrace on. Called with tasklist held
375 * for writing, and returns with it held too. But note it can release
376 * and reacquire the lock.
378 void exit_ptrace(struct task_struct *tracer)
379 __releases(&tasklist_lock)
380 __acquires(&tasklist_lock)
382 struct task_struct *p, *n;
383 LIST_HEAD(ptrace_dead);
385 if (likely(list_empty(&tracer->ptraced)))
388 list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
389 if (__ptrace_detach(tracer, p))
390 list_add(&p->ptrace_entry, &ptrace_dead);
393 write_unlock_irq(&tasklist_lock);
394 BUG_ON(!list_empty(&tracer->ptraced));
396 list_for_each_entry_safe(p, n, &ptrace_dead, ptrace_entry) {
397 list_del_init(&p->ptrace_entry);
401 write_lock_irq(&tasklist_lock);
404 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
410 int this_len, retval;
412 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
413 retval = access_process_vm(tsk, src, buf, this_len, 0);
419 if (copy_to_user(dst, buf, retval))
429 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
435 int this_len, retval;
437 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
438 if (copy_from_user(buf, src, this_len))
440 retval = access_process_vm(tsk, dst, buf, this_len, 1);
454 static int ptrace_setoptions(struct task_struct *child, unsigned long data)
456 child->ptrace &= ~PT_TRACE_MASK;
458 if (data & PTRACE_O_TRACESYSGOOD)
459 child->ptrace |= PT_TRACESYSGOOD;
461 if (data & PTRACE_O_TRACEFORK)
462 child->ptrace |= PT_TRACE_FORK;
464 if (data & PTRACE_O_TRACEVFORK)
465 child->ptrace |= PT_TRACE_VFORK;
467 if (data & PTRACE_O_TRACECLONE)
468 child->ptrace |= PT_TRACE_CLONE;
470 if (data & PTRACE_O_TRACEEXEC)
471 child->ptrace |= PT_TRACE_EXEC;
473 if (data & PTRACE_O_TRACEVFORKDONE)
474 child->ptrace |= PT_TRACE_VFORK_DONE;
476 if (data & PTRACE_O_TRACEEXIT)
477 child->ptrace |= PT_TRACE_EXIT;
479 return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
482 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t *info)
487 if (lock_task_sighand(child, &flags)) {
489 if (likely(child->last_siginfo != NULL)) {
490 *info = *child->last_siginfo;
493 unlock_task_sighand(child, &flags);
498 static int ptrace_setsiginfo(struct task_struct *child, const siginfo_t *info)
503 if (lock_task_sighand(child, &flags)) {
505 if (likely(child->last_siginfo != NULL)) {
506 *child->last_siginfo = *info;
509 unlock_task_sighand(child, &flags);
515 #ifdef PTRACE_SINGLESTEP
516 #define is_singlestep(request) ((request) == PTRACE_SINGLESTEP)
518 #define is_singlestep(request) 0
521 #ifdef PTRACE_SINGLEBLOCK
522 #define is_singleblock(request) ((request) == PTRACE_SINGLEBLOCK)
524 #define is_singleblock(request) 0
528 #define is_sysemu_singlestep(request) ((request) == PTRACE_SYSEMU_SINGLESTEP)
530 #define is_sysemu_singlestep(request) 0
533 static int ptrace_resume(struct task_struct *child, long request,
536 if (!valid_signal(data))
539 if (request == PTRACE_SYSCALL)
540 set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
542 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
544 #ifdef TIF_SYSCALL_EMU
545 if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
546 set_tsk_thread_flag(child, TIF_SYSCALL_EMU);
548 clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
551 if (is_singleblock(request)) {
552 if (unlikely(!arch_has_block_step()))
554 user_enable_block_step(child);
555 } else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
556 if (unlikely(!arch_has_single_step()))
558 user_enable_single_step(child);
560 user_disable_single_step(child);
563 child->exit_code = data;
564 wake_up_process(child);
569 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
571 static const struct user_regset *
572 find_regset(const struct user_regset_view *view, unsigned int type)
574 const struct user_regset *regset;
577 for (n = 0; n < view->n; ++n) {
578 regset = view->regsets + n;
579 if (regset->core_note_type == type)
586 static int ptrace_regset(struct task_struct *task, int req, unsigned int type,
589 const struct user_regset_view *view = task_user_regset_view(task);
590 const struct user_regset *regset = find_regset(view, type);
593 if (!regset || (kiov->iov_len % regset->size) != 0)
596 regset_no = regset - view->regsets;
597 kiov->iov_len = min(kiov->iov_len,
598 (__kernel_size_t) (regset->n * regset->size));
600 if (req == PTRACE_GETREGSET)
601 return copy_regset_to_user(task, view, regset_no, 0,
602 kiov->iov_len, kiov->iov_base);
604 return copy_regset_from_user(task, view, regset_no, 0,
605 kiov->iov_len, kiov->iov_base);
610 int ptrace_request(struct task_struct *child, long request,
611 unsigned long addr, unsigned long data)
615 void __user *datavp = (void __user *) data;
616 unsigned long __user *datalp = datavp;
619 case PTRACE_PEEKTEXT:
620 case PTRACE_PEEKDATA:
621 return generic_ptrace_peekdata(child, addr, data);
622 case PTRACE_POKETEXT:
623 case PTRACE_POKEDATA:
624 return generic_ptrace_pokedata(child, addr, data);
626 #ifdef PTRACE_OLDSETOPTIONS
627 case PTRACE_OLDSETOPTIONS:
629 case PTRACE_SETOPTIONS:
630 ret = ptrace_setoptions(child, data);
632 case PTRACE_GETEVENTMSG:
633 ret = put_user(child->ptrace_message, datalp);
636 case PTRACE_GETSIGINFO:
637 ret = ptrace_getsiginfo(child, &siginfo);
639 ret = copy_siginfo_to_user(datavp, &siginfo);
642 case PTRACE_SETSIGINFO:
643 if (copy_from_user(&siginfo, datavp, sizeof siginfo))
646 ret = ptrace_setsiginfo(child, &siginfo);
649 case PTRACE_DETACH: /* detach a process that was attached. */
650 ret = ptrace_detach(child, data);
653 #ifdef CONFIG_BINFMT_ELF_FDPIC
654 case PTRACE_GETFDPIC: {
655 struct mm_struct *mm = get_task_mm(child);
656 unsigned long tmp = 0;
663 case PTRACE_GETFDPIC_EXEC:
664 tmp = mm->context.exec_fdpic_loadmap;
666 case PTRACE_GETFDPIC_INTERP:
667 tmp = mm->context.interp_fdpic_loadmap;
674 ret = put_user(tmp, datalp);
679 #ifdef PTRACE_SINGLESTEP
680 case PTRACE_SINGLESTEP:
682 #ifdef PTRACE_SINGLEBLOCK
683 case PTRACE_SINGLEBLOCK:
687 case PTRACE_SYSEMU_SINGLESTEP:
691 return ptrace_resume(child, request, data);
694 if (child->exit_state) /* already dead */
696 return ptrace_resume(child, request, SIGKILL);
698 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
699 case PTRACE_GETREGSET:
700 case PTRACE_SETREGSET:
703 struct iovec __user *uiov = datavp;
705 if (!access_ok(VERIFY_WRITE, uiov, sizeof(*uiov)))
708 if (__get_user(kiov.iov_base, &uiov->iov_base) ||
709 __get_user(kiov.iov_len, &uiov->iov_len))
712 ret = ptrace_regset(child, request, addr, &kiov);
714 ret = __put_user(kiov.iov_len, &uiov->iov_len);
725 static struct task_struct *ptrace_get_task_struct(pid_t pid)
727 struct task_struct *child;
730 child = find_task_by_vpid(pid);
732 get_task_struct(child);
736 return ERR_PTR(-ESRCH);
740 #ifndef arch_ptrace_attach
741 #define arch_ptrace_attach(child) do { } while (0)
744 SYSCALL_DEFINE4(ptrace, long, request, long, pid, unsigned long, addr,
747 struct task_struct *child;
750 if (request == PTRACE_TRACEME) {
751 ret = ptrace_traceme();
753 arch_ptrace_attach(current);
757 child = ptrace_get_task_struct(pid);
759 ret = PTR_ERR(child);
763 if (request == PTRACE_ATTACH) {
764 ret = ptrace_attach(child);
766 * Some architectures need to do book-keeping after
770 arch_ptrace_attach(child);
771 goto out_put_task_struct;
774 ret = ptrace_check_attach(child, request == PTRACE_KILL);
776 goto out_put_task_struct;
778 ret = arch_ptrace(child, request, addr, data);
781 put_task_struct(child);
786 int generic_ptrace_peekdata(struct task_struct *tsk, unsigned long addr,
792 copied = access_process_vm(tsk, addr, &tmp, sizeof(tmp), 0);
793 if (copied != sizeof(tmp))
795 return put_user(tmp, (unsigned long __user *)data);
798 int generic_ptrace_pokedata(struct task_struct *tsk, unsigned long addr,
803 copied = access_process_vm(tsk, addr, &data, sizeof(data), 1);
804 return (copied == sizeof(data)) ? 0 : -EIO;
807 #if defined CONFIG_COMPAT
808 #include <linux/compat.h>
810 int compat_ptrace_request(struct task_struct *child, compat_long_t request,
811 compat_ulong_t addr, compat_ulong_t data)
813 compat_ulong_t __user *datap = compat_ptr(data);
819 case PTRACE_PEEKTEXT:
820 case PTRACE_PEEKDATA:
821 ret = access_process_vm(child, addr, &word, sizeof(word), 0);
822 if (ret != sizeof(word))
825 ret = put_user(word, datap);
828 case PTRACE_POKETEXT:
829 case PTRACE_POKEDATA:
830 ret = access_process_vm(child, addr, &data, sizeof(data), 1);
831 ret = (ret != sizeof(data) ? -EIO : 0);
834 case PTRACE_GETEVENTMSG:
835 ret = put_user((compat_ulong_t) child->ptrace_message, datap);
838 case PTRACE_GETSIGINFO:
839 ret = ptrace_getsiginfo(child, &siginfo);
841 ret = copy_siginfo_to_user32(
842 (struct compat_siginfo __user *) datap,
846 case PTRACE_SETSIGINFO:
847 memset(&siginfo, 0, sizeof siginfo);
848 if (copy_siginfo_from_user32(
849 &siginfo, (struct compat_siginfo __user *) datap))
852 ret = ptrace_setsiginfo(child, &siginfo);
854 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
855 case PTRACE_GETREGSET:
856 case PTRACE_SETREGSET:
859 struct compat_iovec __user *uiov =
860 (struct compat_iovec __user *) datap;
864 if (!access_ok(VERIFY_WRITE, uiov, sizeof(*uiov)))
867 if (__get_user(ptr, &uiov->iov_base) ||
868 __get_user(len, &uiov->iov_len))
871 kiov.iov_base = compat_ptr(ptr);
874 ret = ptrace_regset(child, request, addr, &kiov);
876 ret = __put_user(kiov.iov_len, &uiov->iov_len);
882 ret = ptrace_request(child, request, addr, data);
888 asmlinkage long compat_sys_ptrace(compat_long_t request, compat_long_t pid,
889 compat_long_t addr, compat_long_t data)
891 struct task_struct *child;
894 if (request == PTRACE_TRACEME) {
895 ret = ptrace_traceme();
899 child = ptrace_get_task_struct(pid);
901 ret = PTR_ERR(child);
905 if (request == PTRACE_ATTACH) {
906 ret = ptrace_attach(child);
908 * Some architectures need to do book-keeping after
912 arch_ptrace_attach(child);
913 goto out_put_task_struct;
916 ret = ptrace_check_attach(child, request == PTRACE_KILL);
918 ret = compat_arch_ptrace(child, request, addr, data);
921 put_task_struct(child);
925 #endif /* CONFIG_COMPAT */