PCI: pci_slot: grab refcount on slot's bus
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / kernel / ptrace.c
blobaaad0ec341948690d8d8cad7ad22ee0600035fad
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>
25 #include <asm/pgtable.h>
26 #include <asm/uaccess.h>
30 * Initialize a new task whose father had been ptraced.
32 * Called from copy_process().
34 void ptrace_fork(struct task_struct *child, unsigned long clone_flags)
36 arch_ptrace_fork(child, clone_flags);
40 * ptrace a task: make the debugger its new parent and
41 * move it to the ptrace list.
43 * Must be called with the tasklist lock write-held.
45 void __ptrace_link(struct task_struct *child, struct task_struct *new_parent)
47 BUG_ON(!list_empty(&child->ptrace_entry));
48 list_add(&child->ptrace_entry, &new_parent->ptraced);
49 child->parent = new_parent;
53 * Turn a tracing stop into a normal stop now, since with no tracer there
54 * would be no way to wake it up with SIGCONT or SIGKILL. If there was a
55 * signal sent that would resume the child, but didn't because it was in
56 * TASK_TRACED, resume it now.
57 * Requires that irqs be disabled.
59 static void ptrace_untrace(struct task_struct *child)
61 spin_lock(&child->sighand->siglock);
62 if (task_is_traced(child)) {
64 * If the group stop is completed or in progress,
65 * this thread was already counted as stopped.
67 if (child->signal->flags & SIGNAL_STOP_STOPPED ||
68 child->signal->group_stop_count)
69 __set_task_state(child, TASK_STOPPED);
70 else
71 signal_wake_up(child, 1);
73 spin_unlock(&child->sighand->siglock);
77 * unptrace a task: move it back to its original parent and
78 * remove it from the ptrace list.
80 * Must be called with the tasklist lock write-held.
82 void __ptrace_unlink(struct task_struct *child)
84 BUG_ON(!child->ptrace);
86 child->ptrace = 0;
87 child->parent = child->real_parent;
88 list_del_init(&child->ptrace_entry);
90 arch_ptrace_untrace(child);
91 if (task_is_traced(child))
92 ptrace_untrace(child);
96 * Check that we have indeed attached to the thing..
98 int ptrace_check_attach(struct task_struct *child, int kill)
100 int ret = -ESRCH;
103 * We take the read lock around doing both checks to close a
104 * possible race where someone else was tracing our child and
105 * detached between these two checks. After this locked check,
106 * we are sure that this is our traced child and that can only
107 * be changed by us so it's not changing right after this.
109 read_lock(&tasklist_lock);
110 if ((child->ptrace & PT_PTRACED) && child->parent == current) {
111 ret = 0;
113 * child->sighand can't be NULL, release_task()
114 * does ptrace_unlink() before __exit_signal().
116 spin_lock_irq(&child->sighand->siglock);
117 if (task_is_stopped(child))
118 child->state = TASK_TRACED;
119 else if (!task_is_traced(child) && !kill)
120 ret = -ESRCH;
121 spin_unlock_irq(&child->sighand->siglock);
123 read_unlock(&tasklist_lock);
125 if (!ret && !kill)
126 ret = wait_task_inactive(child, TASK_TRACED) ? 0 : -ESRCH;
128 /* All systems go.. */
129 return ret;
132 int __ptrace_may_access(struct task_struct *task, unsigned int mode)
134 const struct cred *cred = current_cred(), *tcred;
136 /* May we inspect the given task?
137 * This check is used both for attaching with ptrace
138 * and for allowing access to sensitive information in /proc.
140 * ptrace_attach denies several cases that /proc allows
141 * because setting up the necessary parent/child relationship
142 * or halting the specified task is impossible.
144 int dumpable = 0;
145 /* Don't let security modules deny introspection */
146 if (task == current)
147 return 0;
148 rcu_read_lock();
149 tcred = __task_cred(task);
150 if ((cred->uid != tcred->euid ||
151 cred->uid != tcred->suid ||
152 cred->uid != tcred->uid ||
153 cred->gid != tcred->egid ||
154 cred->gid != tcred->sgid ||
155 cred->gid != tcred->gid) &&
156 !capable(CAP_SYS_PTRACE)) {
157 rcu_read_unlock();
158 return -EPERM;
160 rcu_read_unlock();
161 smp_rmb();
162 if (task->mm)
163 dumpable = get_dumpable(task->mm);
164 if (!dumpable && !capable(CAP_SYS_PTRACE))
165 return -EPERM;
167 return security_ptrace_may_access(task, mode);
170 bool ptrace_may_access(struct task_struct *task, unsigned int mode)
172 int err;
173 task_lock(task);
174 err = __ptrace_may_access(task, mode);
175 task_unlock(task);
176 return (!err ? true : false);
179 int ptrace_attach(struct task_struct *task)
181 int retval;
182 unsigned long flags;
184 audit_ptrace(task);
186 retval = -EPERM;
187 if (same_thread_group(task, current))
188 goto out;
190 /* Protect exec's credential calculations against our interference;
191 * SUID, SGID and LSM creds get determined differently under ptrace.
193 retval = mutex_lock_interruptible(&current->cred_exec_mutex);
194 if (retval < 0)
195 goto out;
197 retval = -EPERM;
198 repeat:
200 * Nasty, nasty.
202 * We want to hold both the task-lock and the
203 * tasklist_lock for writing at the same time.
204 * But that's against the rules (tasklist_lock
205 * is taken for reading by interrupts on other
206 * cpu's that may have task_lock).
208 task_lock(task);
209 if (!write_trylock_irqsave(&tasklist_lock, flags)) {
210 task_unlock(task);
211 do {
212 cpu_relax();
213 } while (!write_can_lock(&tasklist_lock));
214 goto repeat;
217 if (!task->mm)
218 goto bad;
219 /* the same process cannot be attached many times */
220 if (task->ptrace & PT_PTRACED)
221 goto bad;
222 retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH);
223 if (retval)
224 goto bad;
226 /* Go */
227 task->ptrace |= PT_PTRACED;
228 if (capable(CAP_SYS_PTRACE))
229 task->ptrace |= PT_PTRACE_CAP;
231 __ptrace_link(task, current);
233 send_sig_info(SIGSTOP, SEND_SIG_FORCED, task);
234 bad:
235 write_unlock_irqrestore(&tasklist_lock, flags);
236 task_unlock(task);
237 mutex_unlock(&current->cred_exec_mutex);
238 out:
239 return retval;
243 * Called with irqs disabled, returns true if childs should reap themselves.
245 static int ignoring_children(struct sighand_struct *sigh)
247 int ret;
248 spin_lock(&sigh->siglock);
249 ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
250 (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
251 spin_unlock(&sigh->siglock);
252 return ret;
256 * Called with tasklist_lock held for writing.
257 * Unlink a traced task, and clean it up if it was a traced zombie.
258 * Return true if it needs to be reaped with release_task().
259 * (We can't call release_task() here because we already hold tasklist_lock.)
261 * If it's a zombie, our attachedness prevented normal parent notification
262 * or self-reaping. Do notification now if it would have happened earlier.
263 * If it should reap itself, return true.
265 * If it's our own child, there is no notification to do.
266 * But if our normal children self-reap, then this child
267 * was prevented by ptrace and we must reap it now.
269 static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
271 __ptrace_unlink(p);
273 if (p->exit_state == EXIT_ZOMBIE) {
274 if (!task_detached(p) && thread_group_empty(p)) {
275 if (!same_thread_group(p->real_parent, tracer))
276 do_notify_parent(p, p->exit_signal);
277 else if (ignoring_children(tracer->sighand))
278 p->exit_signal = -1;
280 if (task_detached(p)) {
281 /* Mark it as in the process of being reaped. */
282 p->exit_state = EXIT_DEAD;
283 return true;
287 return false;
290 int ptrace_detach(struct task_struct *child, unsigned int data)
292 bool dead = false;
294 if (!valid_signal(data))
295 return -EIO;
297 /* Architecture-specific hardware disable .. */
298 ptrace_disable(child);
299 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
301 write_lock_irq(&tasklist_lock);
303 * This child can be already killed. Make sure de_thread() or
304 * our sub-thread doing do_wait() didn't do release_task() yet.
306 if (child->ptrace) {
307 child->exit_code = data;
308 dead = __ptrace_detach(current, child);
310 write_unlock_irq(&tasklist_lock);
312 if (unlikely(dead))
313 release_task(child);
315 return 0;
319 * Detach all tasks we were using ptrace on.
321 void exit_ptrace(struct task_struct *tracer)
323 struct task_struct *p, *n;
324 LIST_HEAD(ptrace_dead);
326 write_lock_irq(&tasklist_lock);
327 list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
328 if (__ptrace_detach(tracer, p))
329 list_add(&p->ptrace_entry, &ptrace_dead);
331 write_unlock_irq(&tasklist_lock);
333 BUG_ON(!list_empty(&tracer->ptraced));
335 list_for_each_entry_safe(p, n, &ptrace_dead, ptrace_entry) {
336 list_del_init(&p->ptrace_entry);
337 release_task(p);
341 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
343 int copied = 0;
345 while (len > 0) {
346 char buf[128];
347 int this_len, retval;
349 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
350 retval = access_process_vm(tsk, src, buf, this_len, 0);
351 if (!retval) {
352 if (copied)
353 break;
354 return -EIO;
356 if (copy_to_user(dst, buf, retval))
357 return -EFAULT;
358 copied += retval;
359 src += retval;
360 dst += retval;
361 len -= retval;
363 return copied;
366 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
368 int copied = 0;
370 while (len > 0) {
371 char buf[128];
372 int this_len, retval;
374 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
375 if (copy_from_user(buf, src, this_len))
376 return -EFAULT;
377 retval = access_process_vm(tsk, dst, buf, this_len, 1);
378 if (!retval) {
379 if (copied)
380 break;
381 return -EIO;
383 copied += retval;
384 src += retval;
385 dst += retval;
386 len -= retval;
388 return copied;
391 static int ptrace_setoptions(struct task_struct *child, long data)
393 child->ptrace &= ~PT_TRACE_MASK;
395 if (data & PTRACE_O_TRACESYSGOOD)
396 child->ptrace |= PT_TRACESYSGOOD;
398 if (data & PTRACE_O_TRACEFORK)
399 child->ptrace |= PT_TRACE_FORK;
401 if (data & PTRACE_O_TRACEVFORK)
402 child->ptrace |= PT_TRACE_VFORK;
404 if (data & PTRACE_O_TRACECLONE)
405 child->ptrace |= PT_TRACE_CLONE;
407 if (data & PTRACE_O_TRACEEXEC)
408 child->ptrace |= PT_TRACE_EXEC;
410 if (data & PTRACE_O_TRACEVFORKDONE)
411 child->ptrace |= PT_TRACE_VFORK_DONE;
413 if (data & PTRACE_O_TRACEEXIT)
414 child->ptrace |= PT_TRACE_EXIT;
416 return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
419 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t *info)
421 int error = -ESRCH;
423 read_lock(&tasklist_lock);
424 if (likely(child->sighand != NULL)) {
425 error = -EINVAL;
426 spin_lock_irq(&child->sighand->siglock);
427 if (likely(child->last_siginfo != NULL)) {
428 *info = *child->last_siginfo;
429 error = 0;
431 spin_unlock_irq(&child->sighand->siglock);
433 read_unlock(&tasklist_lock);
434 return error;
437 static int ptrace_setsiginfo(struct task_struct *child, const siginfo_t *info)
439 int error = -ESRCH;
441 read_lock(&tasklist_lock);
442 if (likely(child->sighand != NULL)) {
443 error = -EINVAL;
444 spin_lock_irq(&child->sighand->siglock);
445 if (likely(child->last_siginfo != NULL)) {
446 *child->last_siginfo = *info;
447 error = 0;
449 spin_unlock_irq(&child->sighand->siglock);
451 read_unlock(&tasklist_lock);
452 return error;
456 #ifdef PTRACE_SINGLESTEP
457 #define is_singlestep(request) ((request) == PTRACE_SINGLESTEP)
458 #else
459 #define is_singlestep(request) 0
460 #endif
462 #ifdef PTRACE_SINGLEBLOCK
463 #define is_singleblock(request) ((request) == PTRACE_SINGLEBLOCK)
464 #else
465 #define is_singleblock(request) 0
466 #endif
468 #ifdef PTRACE_SYSEMU
469 #define is_sysemu_singlestep(request) ((request) == PTRACE_SYSEMU_SINGLESTEP)
470 #else
471 #define is_sysemu_singlestep(request) 0
472 #endif
474 static int ptrace_resume(struct task_struct *child, long request, long data)
476 if (!valid_signal(data))
477 return -EIO;
479 if (request == PTRACE_SYSCALL)
480 set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
481 else
482 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
484 #ifdef TIF_SYSCALL_EMU
485 if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
486 set_tsk_thread_flag(child, TIF_SYSCALL_EMU);
487 else
488 clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
489 #endif
491 if (is_singleblock(request)) {
492 if (unlikely(!arch_has_block_step()))
493 return -EIO;
494 user_enable_block_step(child);
495 } else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
496 if (unlikely(!arch_has_single_step()))
497 return -EIO;
498 user_enable_single_step(child);
500 else
501 user_disable_single_step(child);
503 child->exit_code = data;
504 wake_up_process(child);
506 return 0;
509 int ptrace_request(struct task_struct *child, long request,
510 long addr, long data)
512 int ret = -EIO;
513 siginfo_t siginfo;
515 switch (request) {
516 case PTRACE_PEEKTEXT:
517 case PTRACE_PEEKDATA:
518 return generic_ptrace_peekdata(child, addr, data);
519 case PTRACE_POKETEXT:
520 case PTRACE_POKEDATA:
521 return generic_ptrace_pokedata(child, addr, data);
523 #ifdef PTRACE_OLDSETOPTIONS
524 case PTRACE_OLDSETOPTIONS:
525 #endif
526 case PTRACE_SETOPTIONS:
527 ret = ptrace_setoptions(child, data);
528 break;
529 case PTRACE_GETEVENTMSG:
530 ret = put_user(child->ptrace_message, (unsigned long __user *) data);
531 break;
533 case PTRACE_GETSIGINFO:
534 ret = ptrace_getsiginfo(child, &siginfo);
535 if (!ret)
536 ret = copy_siginfo_to_user((siginfo_t __user *) data,
537 &siginfo);
538 break;
540 case PTRACE_SETSIGINFO:
541 if (copy_from_user(&siginfo, (siginfo_t __user *) data,
542 sizeof siginfo))
543 ret = -EFAULT;
544 else
545 ret = ptrace_setsiginfo(child, &siginfo);
546 break;
548 case PTRACE_DETACH: /* detach a process that was attached. */
549 ret = ptrace_detach(child, data);
550 break;
552 #ifdef PTRACE_SINGLESTEP
553 case PTRACE_SINGLESTEP:
554 #endif
555 #ifdef PTRACE_SINGLEBLOCK
556 case PTRACE_SINGLEBLOCK:
557 #endif
558 #ifdef PTRACE_SYSEMU
559 case PTRACE_SYSEMU:
560 case PTRACE_SYSEMU_SINGLESTEP:
561 #endif
562 case PTRACE_SYSCALL:
563 case PTRACE_CONT:
564 return ptrace_resume(child, request, data);
566 case PTRACE_KILL:
567 if (child->exit_state) /* already dead */
568 return 0;
569 return ptrace_resume(child, request, SIGKILL);
571 default:
572 break;
575 return ret;
579 * ptrace_traceme -- helper for PTRACE_TRACEME
581 * Performs checks and sets PT_PTRACED.
582 * Should be used by all ptrace implementations for PTRACE_TRACEME.
584 int ptrace_traceme(void)
586 int ret = -EPERM;
589 * Are we already being traced?
591 repeat:
592 task_lock(current);
593 if (!(current->ptrace & PT_PTRACED)) {
595 * See ptrace_attach() comments about the locking here.
597 unsigned long flags;
598 if (!write_trylock_irqsave(&tasklist_lock, flags)) {
599 task_unlock(current);
600 do {
601 cpu_relax();
602 } while (!write_can_lock(&tasklist_lock));
603 goto repeat;
606 ret = security_ptrace_traceme(current->parent);
609 * Set the ptrace bit in the process ptrace flags.
610 * Then link us on our parent's ptraced list.
612 if (!ret) {
613 current->ptrace |= PT_PTRACED;
614 __ptrace_link(current, current->real_parent);
617 write_unlock_irqrestore(&tasklist_lock, flags);
619 task_unlock(current);
620 return ret;
624 * ptrace_get_task_struct -- grab a task struct reference for ptrace
625 * @pid: process id to grab a task_struct reference of
627 * This function is a helper for ptrace implementations. It checks
628 * permissions and then grabs a task struct for use of the actual
629 * ptrace implementation.
631 * Returns the task_struct for @pid or an ERR_PTR() on failure.
633 struct task_struct *ptrace_get_task_struct(pid_t pid)
635 struct task_struct *child;
637 read_lock(&tasklist_lock);
638 child = find_task_by_vpid(pid);
639 if (child)
640 get_task_struct(child);
642 read_unlock(&tasklist_lock);
643 if (!child)
644 return ERR_PTR(-ESRCH);
645 return child;
648 #ifndef arch_ptrace_attach
649 #define arch_ptrace_attach(child) do { } while (0)
650 #endif
652 SYSCALL_DEFINE4(ptrace, long, request, long, pid, long, addr, long, data)
654 struct task_struct *child;
655 long ret;
658 * This lock_kernel fixes a subtle race with suid exec
660 lock_kernel();
661 if (request == PTRACE_TRACEME) {
662 ret = ptrace_traceme();
663 if (!ret)
664 arch_ptrace_attach(current);
665 goto out;
668 child = ptrace_get_task_struct(pid);
669 if (IS_ERR(child)) {
670 ret = PTR_ERR(child);
671 goto out;
674 if (request == PTRACE_ATTACH) {
675 ret = ptrace_attach(child);
677 * Some architectures need to do book-keeping after
678 * a ptrace attach.
680 if (!ret)
681 arch_ptrace_attach(child);
682 goto out_put_task_struct;
685 ret = ptrace_check_attach(child, request == PTRACE_KILL);
686 if (ret < 0)
687 goto out_put_task_struct;
689 ret = arch_ptrace(child, request, addr, data);
691 out_put_task_struct:
692 put_task_struct(child);
693 out:
694 unlock_kernel();
695 return ret;
698 int generic_ptrace_peekdata(struct task_struct *tsk, long addr, long data)
700 unsigned long tmp;
701 int copied;
703 copied = access_process_vm(tsk, addr, &tmp, sizeof(tmp), 0);
704 if (copied != sizeof(tmp))
705 return -EIO;
706 return put_user(tmp, (unsigned long __user *)data);
709 int generic_ptrace_pokedata(struct task_struct *tsk, long addr, long data)
711 int copied;
713 copied = access_process_vm(tsk, addr, &data, sizeof(data), 1);
714 return (copied == sizeof(data)) ? 0 : -EIO;
717 #if defined CONFIG_COMPAT
718 #include <linux/compat.h>
720 int compat_ptrace_request(struct task_struct *child, compat_long_t request,
721 compat_ulong_t addr, compat_ulong_t data)
723 compat_ulong_t __user *datap = compat_ptr(data);
724 compat_ulong_t word;
725 siginfo_t siginfo;
726 int ret;
728 switch (request) {
729 case PTRACE_PEEKTEXT:
730 case PTRACE_PEEKDATA:
731 ret = access_process_vm(child, addr, &word, sizeof(word), 0);
732 if (ret != sizeof(word))
733 ret = -EIO;
734 else
735 ret = put_user(word, datap);
736 break;
738 case PTRACE_POKETEXT:
739 case PTRACE_POKEDATA:
740 ret = access_process_vm(child, addr, &data, sizeof(data), 1);
741 ret = (ret != sizeof(data) ? -EIO : 0);
742 break;
744 case PTRACE_GETEVENTMSG:
745 ret = put_user((compat_ulong_t) child->ptrace_message, datap);
746 break;
748 case PTRACE_GETSIGINFO:
749 ret = ptrace_getsiginfo(child, &siginfo);
750 if (!ret)
751 ret = copy_siginfo_to_user32(
752 (struct compat_siginfo __user *) datap,
753 &siginfo);
754 break;
756 case PTRACE_SETSIGINFO:
757 memset(&siginfo, 0, sizeof siginfo);
758 if (copy_siginfo_from_user32(
759 &siginfo, (struct compat_siginfo __user *) datap))
760 ret = -EFAULT;
761 else
762 ret = ptrace_setsiginfo(child, &siginfo);
763 break;
765 default:
766 ret = ptrace_request(child, request, addr, data);
769 return ret;
772 asmlinkage long compat_sys_ptrace(compat_long_t request, compat_long_t pid,
773 compat_long_t addr, compat_long_t data)
775 struct task_struct *child;
776 long ret;
779 * This lock_kernel fixes a subtle race with suid exec
781 lock_kernel();
782 if (request == PTRACE_TRACEME) {
783 ret = ptrace_traceme();
784 goto out;
787 child = ptrace_get_task_struct(pid);
788 if (IS_ERR(child)) {
789 ret = PTR_ERR(child);
790 goto out;
793 if (request == PTRACE_ATTACH) {
794 ret = ptrace_attach(child);
796 * Some architectures need to do book-keeping after
797 * a ptrace attach.
799 if (!ret)
800 arch_ptrace_attach(child);
801 goto out_put_task_struct;
804 ret = ptrace_check_attach(child, request == PTRACE_KILL);
805 if (!ret)
806 ret = compat_arch_ptrace(child, request, addr, data);
808 out_put_task_struct:
809 put_task_struct(child);
810 out:
811 unlock_kernel();
812 return ret;
814 #endif /* CONFIG_COMPAT */