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/smp_lock.h>
18 #include <linux/ptrace.h>
19 #include <linux/security.h>
20 #include <linux/signal.h>
21 #include <linux/audit.h>
22 #include <linux/pid_namespace.h>
23 #include <linux/syscalls.h>
24 #include <linux/uaccess.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 * Turn a tracing stop into a normal stop now, since with no tracer there
42 * would be no way to wake it up with SIGCONT or SIGKILL. If there was a
43 * signal sent that would resume the child, but didn't because it was in
44 * TASK_TRACED, resume it now.
45 * Requires that irqs be disabled.
47 static void ptrace_untrace(struct task_struct *child)
49 spin_lock(&child->sighand->siglock);
50 if (task_is_traced(child)) {
52 * If the group stop is completed or in progress,
53 * this thread was already counted as stopped.
55 if (child->signal->flags & SIGNAL_STOP_STOPPED ||
56 child->signal->group_stop_count)
57 __set_task_state(child, TASK_STOPPED);
59 signal_wake_up(child, 1);
61 spin_unlock(&child->sighand->siglock);
65 * unptrace a task: move it back to its original parent and
66 * remove it from the ptrace list.
68 * Must be called with the tasklist lock write-held.
70 void __ptrace_unlink(struct task_struct *child)
72 BUG_ON(!child->ptrace);
75 child->parent = child->real_parent;
76 list_del_init(&child->ptrace_entry);
78 arch_ptrace_untrace(child);
79 if (task_is_traced(child))
80 ptrace_untrace(child);
84 * Check that we have indeed attached to the thing..
86 int ptrace_check_attach(struct task_struct *child, int kill)
91 * We take the read lock around doing both checks to close a
92 * possible race where someone else was tracing our child and
93 * detached between these two checks. After this locked check,
94 * we are sure that this is our traced child and that can only
95 * be changed by us so it's not changing right after this.
97 read_lock(&tasklist_lock);
98 if ((child->ptrace & PT_PTRACED) && child->parent == current) {
101 * child->sighand can't be NULL, release_task()
102 * does ptrace_unlink() before __exit_signal().
104 spin_lock_irq(&child->sighand->siglock);
105 if (task_is_stopped(child))
106 child->state = TASK_TRACED;
107 else if (!task_is_traced(child) && !kill)
109 spin_unlock_irq(&child->sighand->siglock);
111 read_unlock(&tasklist_lock);
114 ret = wait_task_inactive(child, TASK_TRACED) ? 0 : -ESRCH;
116 /* All systems go.. */
120 int __ptrace_may_access(struct task_struct *task, unsigned int mode)
122 const struct cred *cred = current_cred(), *tcred;
124 /* May we inspect the given task?
125 * This check is used both for attaching with ptrace
126 * and for allowing access to sensitive information in /proc.
128 * ptrace_attach denies several cases that /proc allows
129 * because setting up the necessary parent/child relationship
130 * or halting the specified task is impossible.
133 /* Don't let security modules deny introspection */
137 tcred = __task_cred(task);
138 if ((cred->uid != tcred->euid ||
139 cred->uid != tcred->suid ||
140 cred->uid != tcred->uid ||
141 cred->gid != tcred->egid ||
142 cred->gid != tcred->sgid ||
143 cred->gid != tcred->gid) &&
144 !capable(CAP_SYS_PTRACE)) {
151 dumpable = get_dumpable(task->mm);
152 if (!dumpable && !capable(CAP_SYS_PTRACE))
155 return security_ptrace_may_access(task, mode);
158 bool ptrace_may_access(struct task_struct *task, unsigned int mode)
162 err = __ptrace_may_access(task, mode);
167 int ptrace_attach(struct task_struct *task)
175 if (same_thread_group(task, current))
178 /* Protect exec's credential calculations against our interference;
179 * SUID, SGID and LSM creds get determined differently under ptrace.
181 retval = mutex_lock_interruptible(&task->cred_exec_mutex);
190 * We want to hold both the task-lock and the
191 * tasklist_lock for writing at the same time.
192 * But that's against the rules (tasklist_lock
193 * is taken for reading by interrupts on other
194 * cpu's that may have task_lock).
197 if (!write_trylock_irqsave(&tasklist_lock, flags)) {
201 } while (!write_can_lock(&tasklist_lock));
207 /* the same process cannot be attached many times */
208 if (task->ptrace & PT_PTRACED)
210 retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH);
215 task->ptrace |= PT_PTRACED;
216 if (capable(CAP_SYS_PTRACE))
217 task->ptrace |= PT_PTRACE_CAP;
219 __ptrace_link(task, current);
221 send_sig_info(SIGSTOP, SEND_SIG_FORCED, task);
223 write_unlock_irqrestore(&tasklist_lock, flags);
225 mutex_unlock(&task->cred_exec_mutex);
231 * Called with irqs disabled, returns true if childs should reap themselves.
233 static int ignoring_children(struct sighand_struct *sigh)
236 spin_lock(&sigh->siglock);
237 ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
238 (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
239 spin_unlock(&sigh->siglock);
244 * Called with tasklist_lock held for writing.
245 * Unlink a traced task, and clean it up if it was a traced zombie.
246 * Return true if it needs to be reaped with release_task().
247 * (We can't call release_task() here because we already hold tasklist_lock.)
249 * If it's a zombie, our attachedness prevented normal parent notification
250 * or self-reaping. Do notification now if it would have happened earlier.
251 * If it should reap itself, return true.
253 * If it's our own child, there is no notification to do.
254 * But if our normal children self-reap, then this child
255 * was prevented by ptrace and we must reap it now.
257 static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
261 if (p->exit_state == EXIT_ZOMBIE) {
262 if (!task_detached(p) && thread_group_empty(p)) {
263 if (!same_thread_group(p->real_parent, tracer))
264 do_notify_parent(p, p->exit_signal);
265 else if (ignoring_children(tracer->sighand))
268 if (task_detached(p)) {
269 /* Mark it as in the process of being reaped. */
270 p->exit_state = EXIT_DEAD;
278 int ptrace_detach(struct task_struct *child, unsigned int data)
282 if (!valid_signal(data))
285 /* Architecture-specific hardware disable .. */
286 ptrace_disable(child);
287 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
289 write_lock_irq(&tasklist_lock);
291 * This child can be already killed. Make sure de_thread() or
292 * our sub-thread doing do_wait() didn't do release_task() yet.
295 child->exit_code = data;
296 dead = __ptrace_detach(current, child);
298 write_unlock_irq(&tasklist_lock);
307 * Detach all tasks we were using ptrace on.
309 void exit_ptrace(struct task_struct *tracer)
311 struct task_struct *p, *n;
312 LIST_HEAD(ptrace_dead);
314 write_lock_irq(&tasklist_lock);
315 list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
316 if (__ptrace_detach(tracer, p))
317 list_add(&p->ptrace_entry, &ptrace_dead);
319 write_unlock_irq(&tasklist_lock);
321 BUG_ON(!list_empty(&tracer->ptraced));
323 list_for_each_entry_safe(p, n, &ptrace_dead, ptrace_entry) {
324 list_del_init(&p->ptrace_entry);
329 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
335 int this_len, retval;
337 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
338 retval = access_process_vm(tsk, src, buf, this_len, 0);
344 if (copy_to_user(dst, buf, retval))
354 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
360 int this_len, retval;
362 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
363 if (copy_from_user(buf, src, this_len))
365 retval = access_process_vm(tsk, dst, buf, this_len, 1);
379 static int ptrace_setoptions(struct task_struct *child, long data)
381 child->ptrace &= ~PT_TRACE_MASK;
383 if (data & PTRACE_O_TRACESYSGOOD)
384 child->ptrace |= PT_TRACESYSGOOD;
386 if (data & PTRACE_O_TRACEFORK)
387 child->ptrace |= PT_TRACE_FORK;
389 if (data & PTRACE_O_TRACEVFORK)
390 child->ptrace |= PT_TRACE_VFORK;
392 if (data & PTRACE_O_TRACECLONE)
393 child->ptrace |= PT_TRACE_CLONE;
395 if (data & PTRACE_O_TRACEEXEC)
396 child->ptrace |= PT_TRACE_EXEC;
398 if (data & PTRACE_O_TRACEVFORKDONE)
399 child->ptrace |= PT_TRACE_VFORK_DONE;
401 if (data & PTRACE_O_TRACEEXIT)
402 child->ptrace |= PT_TRACE_EXIT;
404 return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
407 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t *info)
411 read_lock(&tasklist_lock);
412 if (likely(child->sighand != NULL)) {
414 spin_lock_irq(&child->sighand->siglock);
415 if (likely(child->last_siginfo != NULL)) {
416 *info = *child->last_siginfo;
419 spin_unlock_irq(&child->sighand->siglock);
421 read_unlock(&tasklist_lock);
425 static int ptrace_setsiginfo(struct task_struct *child, const siginfo_t *info)
429 read_lock(&tasklist_lock);
430 if (likely(child->sighand != NULL)) {
432 spin_lock_irq(&child->sighand->siglock);
433 if (likely(child->last_siginfo != NULL)) {
434 *child->last_siginfo = *info;
437 spin_unlock_irq(&child->sighand->siglock);
439 read_unlock(&tasklist_lock);
444 #ifdef PTRACE_SINGLESTEP
445 #define is_singlestep(request) ((request) == PTRACE_SINGLESTEP)
447 #define is_singlestep(request) 0
450 #ifdef PTRACE_SINGLEBLOCK
451 #define is_singleblock(request) ((request) == PTRACE_SINGLEBLOCK)
453 #define is_singleblock(request) 0
457 #define is_sysemu_singlestep(request) ((request) == PTRACE_SYSEMU_SINGLESTEP)
459 #define is_sysemu_singlestep(request) 0
462 static int ptrace_resume(struct task_struct *child, long request, long data)
464 if (!valid_signal(data))
467 if (request == PTRACE_SYSCALL)
468 set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
470 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
472 #ifdef TIF_SYSCALL_EMU
473 if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
474 set_tsk_thread_flag(child, TIF_SYSCALL_EMU);
476 clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
479 if (is_singleblock(request)) {
480 if (unlikely(!arch_has_block_step()))
482 user_enable_block_step(child);
483 } else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
484 if (unlikely(!arch_has_single_step()))
486 user_enable_single_step(child);
488 user_disable_single_step(child);
491 child->exit_code = data;
492 wake_up_process(child);
497 int ptrace_request(struct task_struct *child, long request,
498 long addr, long data)
504 case PTRACE_PEEKTEXT:
505 case PTRACE_PEEKDATA:
506 return generic_ptrace_peekdata(child, addr, data);
507 case PTRACE_POKETEXT:
508 case PTRACE_POKEDATA:
509 return generic_ptrace_pokedata(child, addr, data);
511 #ifdef PTRACE_OLDSETOPTIONS
512 case PTRACE_OLDSETOPTIONS:
514 case PTRACE_SETOPTIONS:
515 ret = ptrace_setoptions(child, data);
517 case PTRACE_GETEVENTMSG:
518 ret = put_user(child->ptrace_message, (unsigned long __user *) data);
521 case PTRACE_GETSIGINFO:
522 ret = ptrace_getsiginfo(child, &siginfo);
524 ret = copy_siginfo_to_user((siginfo_t __user *) data,
528 case PTRACE_SETSIGINFO:
529 if (copy_from_user(&siginfo, (siginfo_t __user *) data,
533 ret = ptrace_setsiginfo(child, &siginfo);
536 case PTRACE_DETACH: /* detach a process that was attached. */
537 ret = ptrace_detach(child, data);
540 #ifdef PTRACE_SINGLESTEP
541 case PTRACE_SINGLESTEP:
543 #ifdef PTRACE_SINGLEBLOCK
544 case PTRACE_SINGLEBLOCK:
548 case PTRACE_SYSEMU_SINGLESTEP:
552 return ptrace_resume(child, request, data);
555 if (child->exit_state) /* already dead */
557 return ptrace_resume(child, request, SIGKILL);
567 * ptrace_traceme -- helper for PTRACE_TRACEME
569 * Performs checks and sets PT_PTRACED.
570 * Should be used by all ptrace implementations for PTRACE_TRACEME.
572 int ptrace_traceme(void)
577 * Are we already being traced?
581 if (!(current->ptrace & PT_PTRACED)) {
583 * See ptrace_attach() comments about the locking here.
586 if (!write_trylock_irqsave(&tasklist_lock, flags)) {
587 task_unlock(current);
590 } while (!write_can_lock(&tasklist_lock));
594 ret = security_ptrace_traceme(current->parent);
597 * Check PF_EXITING to ensure ->real_parent has not passed
598 * exit_ptrace(). Otherwise we don't report the error but
599 * pretend ->real_parent untraces us right after return.
601 if (!ret && !(current->real_parent->flags & PF_EXITING)) {
602 current->ptrace |= PT_PTRACED;
603 __ptrace_link(current, current->real_parent);
606 write_unlock_irqrestore(&tasklist_lock, flags);
608 task_unlock(current);
613 * ptrace_get_task_struct -- grab a task struct reference for ptrace
614 * @pid: process id to grab a task_struct reference of
616 * This function is a helper for ptrace implementations. It checks
617 * permissions and then grabs a task struct for use of the actual
618 * ptrace implementation.
620 * Returns the task_struct for @pid or an ERR_PTR() on failure.
622 struct task_struct *ptrace_get_task_struct(pid_t pid)
624 struct task_struct *child;
626 read_lock(&tasklist_lock);
627 child = find_task_by_vpid(pid);
629 get_task_struct(child);
631 read_unlock(&tasklist_lock);
633 return ERR_PTR(-ESRCH);
637 #ifndef arch_ptrace_attach
638 #define arch_ptrace_attach(child) do { } while (0)
641 SYSCALL_DEFINE4(ptrace, long, request, long, pid, long, addr, long, data)
643 struct task_struct *child;
647 * This lock_kernel fixes a subtle race with suid exec
650 if (request == PTRACE_TRACEME) {
651 ret = ptrace_traceme();
653 arch_ptrace_attach(current);
657 child = ptrace_get_task_struct(pid);
659 ret = PTR_ERR(child);
663 if (request == PTRACE_ATTACH) {
664 ret = ptrace_attach(child);
666 * Some architectures need to do book-keeping after
670 arch_ptrace_attach(child);
671 goto out_put_task_struct;
674 ret = ptrace_check_attach(child, request == PTRACE_KILL);
676 goto out_put_task_struct;
678 ret = arch_ptrace(child, request, addr, data);
681 put_task_struct(child);
687 int generic_ptrace_peekdata(struct task_struct *tsk, long addr, long data)
692 copied = access_process_vm(tsk, addr, &tmp, sizeof(tmp), 0);
693 if (copied != sizeof(tmp))
695 return put_user(tmp, (unsigned long __user *)data);
698 int generic_ptrace_pokedata(struct task_struct *tsk, long addr, long data)
702 copied = access_process_vm(tsk, addr, &data, sizeof(data), 1);
703 return (copied == sizeof(data)) ? 0 : -EIO;
706 #if defined CONFIG_COMPAT
707 #include <linux/compat.h>
709 int compat_ptrace_request(struct task_struct *child, compat_long_t request,
710 compat_ulong_t addr, compat_ulong_t data)
712 compat_ulong_t __user *datap = compat_ptr(data);
718 case PTRACE_PEEKTEXT:
719 case PTRACE_PEEKDATA:
720 ret = access_process_vm(child, addr, &word, sizeof(word), 0);
721 if (ret != sizeof(word))
724 ret = put_user(word, datap);
727 case PTRACE_POKETEXT:
728 case PTRACE_POKEDATA:
729 ret = access_process_vm(child, addr, &data, sizeof(data), 1);
730 ret = (ret != sizeof(data) ? -EIO : 0);
733 case PTRACE_GETEVENTMSG:
734 ret = put_user((compat_ulong_t) child->ptrace_message, datap);
737 case PTRACE_GETSIGINFO:
738 ret = ptrace_getsiginfo(child, &siginfo);
740 ret = copy_siginfo_to_user32(
741 (struct compat_siginfo __user *) datap,
745 case PTRACE_SETSIGINFO:
746 memset(&siginfo, 0, sizeof siginfo);
747 if (copy_siginfo_from_user32(
748 &siginfo, (struct compat_siginfo __user *) datap))
751 ret = ptrace_setsiginfo(child, &siginfo);
755 ret = ptrace_request(child, request, addr, data);
761 asmlinkage long compat_sys_ptrace(compat_long_t request, compat_long_t pid,
762 compat_long_t addr, compat_long_t data)
764 struct task_struct *child;
768 * This lock_kernel fixes a subtle race with suid exec
771 if (request == PTRACE_TRACEME) {
772 ret = ptrace_traceme();
776 child = ptrace_get_task_struct(pid);
778 ret = PTR_ERR(child);
782 if (request == PTRACE_ATTACH) {
783 ret = ptrace_attach(child);
785 * Some architectures need to do book-keeping after
789 arch_ptrace_attach(child);
790 goto out_put_task_struct;
793 ret = ptrace_check_attach(child, request == PTRACE_KILL);
795 ret = compat_arch_ptrace(child, request, addr, data);
798 put_task_struct(child);
803 #endif /* CONFIG_COMPAT */