ptrace: cleanup check/set of PT_PTRACED during attach
[linux-2.6/mini2440.git] / kernel / ptrace.c
blob12e21a949db14645a886fa4ef69963f9e451ca9d
1 /*
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.
8 */
10 #include <linux/capability.h>
11 #include <linux/module.h>
12 #include <linux/sched.h>
13 #include <linux/errno.h>
14 #include <linux/mm.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);
58 else
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);
74 child->ptrace = 0;
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)
88 int ret = -ESRCH;
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) {
99 ret = 0;
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)
108 ret = -ESRCH;
109 spin_unlock_irq(&child->sighand->siglock);
111 read_unlock(&tasklist_lock);
113 if (!ret && !kill)
114 ret = wait_task_inactive(child, TASK_TRACED) ? 0 : -ESRCH;
116 /* All systems go.. */
117 return ret;
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.
132 int dumpable = 0;
133 /* Don't let security modules deny introspection */
134 if (task == current)
135 return 0;
136 rcu_read_lock();
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)) {
145 rcu_read_unlock();
146 return -EPERM;
148 rcu_read_unlock();
149 smp_rmb();
150 if (task->mm)
151 dumpable = get_dumpable(task->mm);
152 if (!dumpable && !capable(CAP_SYS_PTRACE))
153 return -EPERM;
155 return security_ptrace_may_access(task, mode);
158 bool ptrace_may_access(struct task_struct *task, unsigned int mode)
160 int err;
161 task_lock(task);
162 err = __ptrace_may_access(task, mode);
163 task_unlock(task);
164 return !err;
167 int ptrace_attach(struct task_struct *task)
169 int retval;
170 unsigned long flags;
172 audit_ptrace(task);
174 retval = -EPERM;
175 if (unlikely(task->flags & PF_KTHREAD))
176 goto out;
177 if (same_thread_group(task, current))
178 goto out;
181 * Protect exec's credential calculations against our interference;
182 * interference; SUID, SGID and LSM creds get determined differently
183 * under ptrace.
185 retval = mutex_lock_interruptible(&task->cred_guard_mutex);
186 if (retval < 0)
187 goto out;
188 repeat:
190 * Nasty, nasty.
192 * We want to hold both the task-lock and the
193 * tasklist_lock for writing at the same time.
194 * But that's against the rules (tasklist_lock
195 * is taken for reading by interrupts on other
196 * cpu's that may have task_lock).
198 task_lock(task);
199 if (!write_trylock_irqsave(&tasklist_lock, flags)) {
200 task_unlock(task);
201 do {
202 cpu_relax();
203 } while (!write_can_lock(&tasklist_lock));
204 goto repeat;
207 retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH);
208 if (retval)
209 goto bad;
211 retval = -EPERM;
212 if (unlikely(task->exit_state))
213 goto bad;
214 if (task->ptrace)
215 goto bad;
217 task->ptrace = PT_PTRACED;
218 if (capable(CAP_SYS_PTRACE))
219 task->ptrace |= PT_PTRACE_CAP;
221 __ptrace_link(task, current);
222 send_sig_info(SIGSTOP, SEND_SIG_FORCED, task);
224 retval = 0;
225 bad:
226 write_unlock_irqrestore(&tasklist_lock, flags);
227 task_unlock(task);
228 mutex_unlock(&task->cred_guard_mutex);
229 out:
230 return retval;
234 * ptrace_traceme -- helper for PTRACE_TRACEME
236 * Performs checks and sets PT_PTRACED.
237 * Should be used by all ptrace implementations for PTRACE_TRACEME.
239 int ptrace_traceme(void)
241 int ret = -EPERM;
244 * Are we already being traced?
246 repeat:
247 task_lock(current);
248 if (!current->ptrace) {
250 * See ptrace_attach() comments about the locking here.
252 unsigned long flags;
253 if (!write_trylock_irqsave(&tasklist_lock, flags)) {
254 task_unlock(current);
255 do {
256 cpu_relax();
257 } while (!write_can_lock(&tasklist_lock));
258 goto repeat;
261 ret = security_ptrace_traceme(current->parent);
264 * Check PF_EXITING to ensure ->real_parent has not passed
265 * exit_ptrace(). Otherwise we don't report the error but
266 * pretend ->real_parent untraces us right after return.
268 if (!ret && !(current->real_parent->flags & PF_EXITING)) {
269 current->ptrace = PT_PTRACED;
270 __ptrace_link(current, current->real_parent);
273 write_unlock_irqrestore(&tasklist_lock, flags);
275 task_unlock(current);
276 return ret;
280 * Called with irqs disabled, returns true if childs should reap themselves.
282 static int ignoring_children(struct sighand_struct *sigh)
284 int ret;
285 spin_lock(&sigh->siglock);
286 ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
287 (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
288 spin_unlock(&sigh->siglock);
289 return ret;
293 * Called with tasklist_lock held for writing.
294 * Unlink a traced task, and clean it up if it was a traced zombie.
295 * Return true if it needs to be reaped with release_task().
296 * (We can't call release_task() here because we already hold tasklist_lock.)
298 * If it's a zombie, our attachedness prevented normal parent notification
299 * or self-reaping. Do notification now if it would have happened earlier.
300 * If it should reap itself, return true.
302 * If it's our own child, there is no notification to do.
303 * But if our normal children self-reap, then this child
304 * was prevented by ptrace and we must reap it now.
306 static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
308 __ptrace_unlink(p);
310 if (p->exit_state == EXIT_ZOMBIE) {
311 if (!task_detached(p) && thread_group_empty(p)) {
312 if (!same_thread_group(p->real_parent, tracer))
313 do_notify_parent(p, p->exit_signal);
314 else if (ignoring_children(tracer->sighand))
315 p->exit_signal = -1;
317 if (task_detached(p)) {
318 /* Mark it as in the process of being reaped. */
319 p->exit_state = EXIT_DEAD;
320 return true;
324 return false;
327 int ptrace_detach(struct task_struct *child, unsigned int data)
329 bool dead = false;
331 if (!valid_signal(data))
332 return -EIO;
334 /* Architecture-specific hardware disable .. */
335 ptrace_disable(child);
336 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
338 write_lock_irq(&tasklist_lock);
340 * This child can be already killed. Make sure de_thread() or
341 * our sub-thread doing do_wait() didn't do release_task() yet.
343 if (child->ptrace) {
344 child->exit_code = data;
345 dead = __ptrace_detach(current, child);
346 if (!child->exit_state)
347 wake_up_process(child);
349 write_unlock_irq(&tasklist_lock);
351 if (unlikely(dead))
352 release_task(child);
354 return 0;
358 * Detach all tasks we were using ptrace on.
360 void exit_ptrace(struct task_struct *tracer)
362 struct task_struct *p, *n;
363 LIST_HEAD(ptrace_dead);
365 write_lock_irq(&tasklist_lock);
366 list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
367 if (__ptrace_detach(tracer, p))
368 list_add(&p->ptrace_entry, &ptrace_dead);
370 write_unlock_irq(&tasklist_lock);
372 BUG_ON(!list_empty(&tracer->ptraced));
374 list_for_each_entry_safe(p, n, &ptrace_dead, ptrace_entry) {
375 list_del_init(&p->ptrace_entry);
376 release_task(p);
380 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
382 int copied = 0;
384 while (len > 0) {
385 char buf[128];
386 int this_len, retval;
388 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
389 retval = access_process_vm(tsk, src, buf, this_len, 0);
390 if (!retval) {
391 if (copied)
392 break;
393 return -EIO;
395 if (copy_to_user(dst, buf, retval))
396 return -EFAULT;
397 copied += retval;
398 src += retval;
399 dst += retval;
400 len -= retval;
402 return copied;
405 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
407 int copied = 0;
409 while (len > 0) {
410 char buf[128];
411 int this_len, retval;
413 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
414 if (copy_from_user(buf, src, this_len))
415 return -EFAULT;
416 retval = access_process_vm(tsk, dst, buf, this_len, 1);
417 if (!retval) {
418 if (copied)
419 break;
420 return -EIO;
422 copied += retval;
423 src += retval;
424 dst += retval;
425 len -= retval;
427 return copied;
430 static int ptrace_setoptions(struct task_struct *child, long data)
432 child->ptrace &= ~PT_TRACE_MASK;
434 if (data & PTRACE_O_TRACESYSGOOD)
435 child->ptrace |= PT_TRACESYSGOOD;
437 if (data & PTRACE_O_TRACEFORK)
438 child->ptrace |= PT_TRACE_FORK;
440 if (data & PTRACE_O_TRACEVFORK)
441 child->ptrace |= PT_TRACE_VFORK;
443 if (data & PTRACE_O_TRACECLONE)
444 child->ptrace |= PT_TRACE_CLONE;
446 if (data & PTRACE_O_TRACEEXEC)
447 child->ptrace |= PT_TRACE_EXEC;
449 if (data & PTRACE_O_TRACEVFORKDONE)
450 child->ptrace |= PT_TRACE_VFORK_DONE;
452 if (data & PTRACE_O_TRACEEXIT)
453 child->ptrace |= PT_TRACE_EXIT;
455 return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
458 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t *info)
460 int error = -ESRCH;
462 read_lock(&tasklist_lock);
463 if (likely(child->sighand != NULL)) {
464 error = -EINVAL;
465 spin_lock_irq(&child->sighand->siglock);
466 if (likely(child->last_siginfo != NULL)) {
467 *info = *child->last_siginfo;
468 error = 0;
470 spin_unlock_irq(&child->sighand->siglock);
472 read_unlock(&tasklist_lock);
473 return error;
476 static int ptrace_setsiginfo(struct task_struct *child, const siginfo_t *info)
478 int error = -ESRCH;
480 read_lock(&tasklist_lock);
481 if (likely(child->sighand != NULL)) {
482 error = -EINVAL;
483 spin_lock_irq(&child->sighand->siglock);
484 if (likely(child->last_siginfo != NULL)) {
485 *child->last_siginfo = *info;
486 error = 0;
488 spin_unlock_irq(&child->sighand->siglock);
490 read_unlock(&tasklist_lock);
491 return error;
495 #ifdef PTRACE_SINGLESTEP
496 #define is_singlestep(request) ((request) == PTRACE_SINGLESTEP)
497 #else
498 #define is_singlestep(request) 0
499 #endif
501 #ifdef PTRACE_SINGLEBLOCK
502 #define is_singleblock(request) ((request) == PTRACE_SINGLEBLOCK)
503 #else
504 #define is_singleblock(request) 0
505 #endif
507 #ifdef PTRACE_SYSEMU
508 #define is_sysemu_singlestep(request) ((request) == PTRACE_SYSEMU_SINGLESTEP)
509 #else
510 #define is_sysemu_singlestep(request) 0
511 #endif
513 static int ptrace_resume(struct task_struct *child, long request, long data)
515 if (!valid_signal(data))
516 return -EIO;
518 if (request == PTRACE_SYSCALL)
519 set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
520 else
521 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
523 #ifdef TIF_SYSCALL_EMU
524 if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
525 set_tsk_thread_flag(child, TIF_SYSCALL_EMU);
526 else
527 clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
528 #endif
530 if (is_singleblock(request)) {
531 if (unlikely(!arch_has_block_step()))
532 return -EIO;
533 user_enable_block_step(child);
534 } else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
535 if (unlikely(!arch_has_single_step()))
536 return -EIO;
537 user_enable_single_step(child);
538 } else {
539 user_disable_single_step(child);
542 child->exit_code = data;
543 wake_up_process(child);
545 return 0;
548 int ptrace_request(struct task_struct *child, long request,
549 long addr, long data)
551 int ret = -EIO;
552 siginfo_t siginfo;
554 switch (request) {
555 case PTRACE_PEEKTEXT:
556 case PTRACE_PEEKDATA:
557 return generic_ptrace_peekdata(child, addr, data);
558 case PTRACE_POKETEXT:
559 case PTRACE_POKEDATA:
560 return generic_ptrace_pokedata(child, addr, data);
562 #ifdef PTRACE_OLDSETOPTIONS
563 case PTRACE_OLDSETOPTIONS:
564 #endif
565 case PTRACE_SETOPTIONS:
566 ret = ptrace_setoptions(child, data);
567 break;
568 case PTRACE_GETEVENTMSG:
569 ret = put_user(child->ptrace_message, (unsigned long __user *) data);
570 break;
572 case PTRACE_GETSIGINFO:
573 ret = ptrace_getsiginfo(child, &siginfo);
574 if (!ret)
575 ret = copy_siginfo_to_user((siginfo_t __user *) data,
576 &siginfo);
577 break;
579 case PTRACE_SETSIGINFO:
580 if (copy_from_user(&siginfo, (siginfo_t __user *) data,
581 sizeof siginfo))
582 ret = -EFAULT;
583 else
584 ret = ptrace_setsiginfo(child, &siginfo);
585 break;
587 case PTRACE_DETACH: /* detach a process that was attached. */
588 ret = ptrace_detach(child, data);
589 break;
591 #ifdef PTRACE_SINGLESTEP
592 case PTRACE_SINGLESTEP:
593 #endif
594 #ifdef PTRACE_SINGLEBLOCK
595 case PTRACE_SINGLEBLOCK:
596 #endif
597 #ifdef PTRACE_SYSEMU
598 case PTRACE_SYSEMU:
599 case PTRACE_SYSEMU_SINGLESTEP:
600 #endif
601 case PTRACE_SYSCALL:
602 case PTRACE_CONT:
603 return ptrace_resume(child, request, data);
605 case PTRACE_KILL:
606 if (child->exit_state) /* already dead */
607 return 0;
608 return ptrace_resume(child, request, SIGKILL);
610 default:
611 break;
614 return ret;
618 * ptrace_get_task_struct -- grab a task struct reference for ptrace
619 * @pid: process id to grab a task_struct reference of
621 * This function is a helper for ptrace implementations. It checks
622 * permissions and then grabs a task struct for use of the actual
623 * ptrace implementation.
625 * Returns the task_struct for @pid or an ERR_PTR() on failure.
627 struct task_struct *ptrace_get_task_struct(pid_t pid)
629 struct task_struct *child;
631 read_lock(&tasklist_lock);
632 child = find_task_by_vpid(pid);
633 if (child)
634 get_task_struct(child);
636 read_unlock(&tasklist_lock);
637 if (!child)
638 return ERR_PTR(-ESRCH);
639 return child;
642 #ifndef arch_ptrace_attach
643 #define arch_ptrace_attach(child) do { } while (0)
644 #endif
646 SYSCALL_DEFINE4(ptrace, long, request, long, pid, long, addr, long, data)
648 struct task_struct *child;
649 long ret;
652 * This lock_kernel fixes a subtle race with suid exec
654 lock_kernel();
655 if (request == PTRACE_TRACEME) {
656 ret = ptrace_traceme();
657 if (!ret)
658 arch_ptrace_attach(current);
659 goto out;
662 child = ptrace_get_task_struct(pid);
663 if (IS_ERR(child)) {
664 ret = PTR_ERR(child);
665 goto out;
668 if (request == PTRACE_ATTACH) {
669 ret = ptrace_attach(child);
671 * Some architectures need to do book-keeping after
672 * a ptrace attach.
674 if (!ret)
675 arch_ptrace_attach(child);
676 goto out_put_task_struct;
679 ret = ptrace_check_attach(child, request == PTRACE_KILL);
680 if (ret < 0)
681 goto out_put_task_struct;
683 ret = arch_ptrace(child, request, addr, data);
685 out_put_task_struct:
686 put_task_struct(child);
687 out:
688 unlock_kernel();
689 return ret;
692 int generic_ptrace_peekdata(struct task_struct *tsk, long addr, long data)
694 unsigned long tmp;
695 int copied;
697 copied = access_process_vm(tsk, addr, &tmp, sizeof(tmp), 0);
698 if (copied != sizeof(tmp))
699 return -EIO;
700 return put_user(tmp, (unsigned long __user *)data);
703 int generic_ptrace_pokedata(struct task_struct *tsk, long addr, long data)
705 int copied;
707 copied = access_process_vm(tsk, addr, &data, sizeof(data), 1);
708 return (copied == sizeof(data)) ? 0 : -EIO;
711 #if defined CONFIG_COMPAT
712 #include <linux/compat.h>
714 int compat_ptrace_request(struct task_struct *child, compat_long_t request,
715 compat_ulong_t addr, compat_ulong_t data)
717 compat_ulong_t __user *datap = compat_ptr(data);
718 compat_ulong_t word;
719 siginfo_t siginfo;
720 int ret;
722 switch (request) {
723 case PTRACE_PEEKTEXT:
724 case PTRACE_PEEKDATA:
725 ret = access_process_vm(child, addr, &word, sizeof(word), 0);
726 if (ret != sizeof(word))
727 ret = -EIO;
728 else
729 ret = put_user(word, datap);
730 break;
732 case PTRACE_POKETEXT:
733 case PTRACE_POKEDATA:
734 ret = access_process_vm(child, addr, &data, sizeof(data), 1);
735 ret = (ret != sizeof(data) ? -EIO : 0);
736 break;
738 case PTRACE_GETEVENTMSG:
739 ret = put_user((compat_ulong_t) child->ptrace_message, datap);
740 break;
742 case PTRACE_GETSIGINFO:
743 ret = ptrace_getsiginfo(child, &siginfo);
744 if (!ret)
745 ret = copy_siginfo_to_user32(
746 (struct compat_siginfo __user *) datap,
747 &siginfo);
748 break;
750 case PTRACE_SETSIGINFO:
751 memset(&siginfo, 0, sizeof siginfo);
752 if (copy_siginfo_from_user32(
753 &siginfo, (struct compat_siginfo __user *) datap))
754 ret = -EFAULT;
755 else
756 ret = ptrace_setsiginfo(child, &siginfo);
757 break;
759 default:
760 ret = ptrace_request(child, request, addr, data);
763 return ret;
766 asmlinkage long compat_sys_ptrace(compat_long_t request, compat_long_t pid,
767 compat_long_t addr, compat_long_t data)
769 struct task_struct *child;
770 long ret;
773 * This lock_kernel fixes a subtle race with suid exec
775 lock_kernel();
776 if (request == PTRACE_TRACEME) {
777 ret = ptrace_traceme();
778 goto out;
781 child = ptrace_get_task_struct(pid);
782 if (IS_ERR(child)) {
783 ret = PTR_ERR(child);
784 goto out;
787 if (request == PTRACE_ATTACH) {
788 ret = ptrace_attach(child);
790 * Some architectures need to do book-keeping after
791 * a ptrace attach.
793 if (!ret)
794 arch_ptrace_attach(child);
795 goto out_put_task_struct;
798 ret = ptrace_check_attach(child, request == PTRACE_KILL);
799 if (!ret)
800 ret = compat_arch_ptrace(child, request, addr, data);
802 out_put_task_struct:
803 put_task_struct(child);
804 out:
805 unlock_kernel();
806 return ret;
808 #endif /* CONFIG_COMPAT */