fuse: don't use inode in fuse_file_poll
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / kernel / ptrace.c
blobdfcd83ceee3b246326cbec2a6eadb27abdba7823
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 * Initialize a new task whose father had been ptraced.
30 * Called from copy_process().
32 void ptrace_fork(struct task_struct *child, unsigned long clone_flags)
34 arch_ptrace_fork(child, clone_flags);
38 * ptrace a task: make the debugger its new parent and
39 * move it to the ptrace list.
41 * Must be called with the tasklist lock write-held.
43 void __ptrace_link(struct task_struct *child, struct task_struct *new_parent)
45 BUG_ON(!list_empty(&child->ptrace_entry));
46 list_add(&child->ptrace_entry, &new_parent->ptraced);
47 child->parent = new_parent;
51 * Turn a tracing stop into a normal stop now, since with no tracer there
52 * would be no way to wake it up with SIGCONT or SIGKILL. If there was a
53 * signal sent that would resume the child, but didn't because it was in
54 * TASK_TRACED, resume it now.
55 * Requires that irqs be disabled.
57 static void ptrace_untrace(struct task_struct *child)
59 spin_lock(&child->sighand->siglock);
60 if (task_is_traced(child)) {
62 * If the group stop is completed or in progress,
63 * this thread was already counted as stopped.
65 if (child->signal->flags & SIGNAL_STOP_STOPPED ||
66 child->signal->group_stop_count)
67 __set_task_state(child, TASK_STOPPED);
68 else
69 signal_wake_up(child, 1);
71 spin_unlock(&child->sighand->siglock);
75 * unptrace a task: move it back to its original parent and
76 * remove it from the ptrace list.
78 * Must be called with the tasklist lock write-held.
80 void __ptrace_unlink(struct task_struct *child)
82 BUG_ON(!child->ptrace);
84 child->ptrace = 0;
85 child->parent = child->real_parent;
86 list_del_init(&child->ptrace_entry);
88 arch_ptrace_untrace(child);
89 if (task_is_traced(child))
90 ptrace_untrace(child);
94 * Check that we have indeed attached to the thing..
96 int ptrace_check_attach(struct task_struct *child, int kill)
98 int ret = -ESRCH;
101 * We take the read lock around doing both checks to close a
102 * possible race where someone else was tracing our child and
103 * detached between these two checks. After this locked check,
104 * we are sure that this is our traced child and that can only
105 * be changed by us so it's not changing right after this.
107 read_lock(&tasklist_lock);
108 if ((child->ptrace & PT_PTRACED) && child->parent == current) {
109 ret = 0;
111 * child->sighand can't be NULL, release_task()
112 * does ptrace_unlink() before __exit_signal().
114 spin_lock_irq(&child->sighand->siglock);
115 if (task_is_stopped(child))
116 child->state = TASK_TRACED;
117 else if (!task_is_traced(child) && !kill)
118 ret = -ESRCH;
119 spin_unlock_irq(&child->sighand->siglock);
121 read_unlock(&tasklist_lock);
123 if (!ret && !kill)
124 ret = wait_task_inactive(child, TASK_TRACED) ? 0 : -ESRCH;
126 /* All systems go.. */
127 return ret;
130 int __ptrace_may_access(struct task_struct *task, unsigned int mode)
132 const struct cred *cred = current_cred(), *tcred;
134 /* May we inspect the given task?
135 * This check is used both for attaching with ptrace
136 * and for allowing access to sensitive information in /proc.
138 * ptrace_attach denies several cases that /proc allows
139 * because setting up the necessary parent/child relationship
140 * or halting the specified task is impossible.
142 int dumpable = 0;
143 /* Don't let security modules deny introspection */
144 if (task == current)
145 return 0;
146 rcu_read_lock();
147 tcred = __task_cred(task);
148 if ((cred->uid != tcred->euid ||
149 cred->uid != tcred->suid ||
150 cred->uid != tcred->uid ||
151 cred->gid != tcred->egid ||
152 cred->gid != tcred->sgid ||
153 cred->gid != tcred->gid) &&
154 !capable(CAP_SYS_PTRACE)) {
155 rcu_read_unlock();
156 return -EPERM;
158 rcu_read_unlock();
159 smp_rmb();
160 if (task->mm)
161 dumpable = get_dumpable(task->mm);
162 if (!dumpable && !capable(CAP_SYS_PTRACE))
163 return -EPERM;
165 return security_ptrace_may_access(task, mode);
168 bool ptrace_may_access(struct task_struct *task, unsigned int mode)
170 int err;
171 task_lock(task);
172 err = __ptrace_may_access(task, mode);
173 task_unlock(task);
174 return !err;
177 int ptrace_attach(struct task_struct *task)
179 int retval;
180 unsigned long flags;
182 audit_ptrace(task);
184 retval = -EPERM;
185 if (same_thread_group(task, current))
186 goto out;
188 /* Protect exec's credential calculations against our interference;
189 * SUID, SGID and LSM creds get determined differently under ptrace.
191 retval = mutex_lock_interruptible(&current->cred_exec_mutex);
192 if (retval < 0)
193 goto out;
195 retval = -EPERM;
196 repeat:
198 * Nasty, nasty.
200 * We want to hold both the task-lock and the
201 * tasklist_lock for writing at the same time.
202 * But that's against the rules (tasklist_lock
203 * is taken for reading by interrupts on other
204 * cpu's that may have task_lock).
206 task_lock(task);
207 if (!write_trylock_irqsave(&tasklist_lock, flags)) {
208 task_unlock(task);
209 do {
210 cpu_relax();
211 } while (!write_can_lock(&tasklist_lock));
212 goto repeat;
215 if (!task->mm)
216 goto bad;
217 /* the same process cannot be attached many times */
218 if (task->ptrace & PT_PTRACED)
219 goto bad;
220 retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH);
221 if (retval)
222 goto bad;
224 /* Go */
225 task->ptrace |= PT_PTRACED;
226 if (capable(CAP_SYS_PTRACE))
227 task->ptrace |= PT_PTRACE_CAP;
229 __ptrace_link(task, current);
231 send_sig_info(SIGSTOP, SEND_SIG_FORCED, task);
232 bad:
233 write_unlock_irqrestore(&tasklist_lock, flags);
234 task_unlock(task);
235 mutex_unlock(&current->cred_exec_mutex);
236 out:
237 return retval;
241 * Called with irqs disabled, returns true if childs should reap themselves.
243 static int ignoring_children(struct sighand_struct *sigh)
245 int ret;
246 spin_lock(&sigh->siglock);
247 ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
248 (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
249 spin_unlock(&sigh->siglock);
250 return ret;
254 * Called with tasklist_lock held for writing.
255 * Unlink a traced task, and clean it up if it was a traced zombie.
256 * Return true if it needs to be reaped with release_task().
257 * (We can't call release_task() here because we already hold tasklist_lock.)
259 * If it's a zombie, our attachedness prevented normal parent notification
260 * or self-reaping. Do notification now if it would have happened earlier.
261 * If it should reap itself, return true.
263 * If it's our own child, there is no notification to do.
264 * But if our normal children self-reap, then this child
265 * was prevented by ptrace and we must reap it now.
267 static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
269 __ptrace_unlink(p);
271 if (p->exit_state == EXIT_ZOMBIE) {
272 if (!task_detached(p) && thread_group_empty(p)) {
273 if (!same_thread_group(p->real_parent, tracer))
274 do_notify_parent(p, p->exit_signal);
275 else if (ignoring_children(tracer->sighand))
276 p->exit_signal = -1;
278 if (task_detached(p)) {
279 /* Mark it as in the process of being reaped. */
280 p->exit_state = EXIT_DEAD;
281 return true;
285 return false;
288 int ptrace_detach(struct task_struct *child, unsigned int data)
290 bool dead = false;
292 if (!valid_signal(data))
293 return -EIO;
295 /* Architecture-specific hardware disable .. */
296 ptrace_disable(child);
297 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
299 write_lock_irq(&tasklist_lock);
301 * This child can be already killed. Make sure de_thread() or
302 * our sub-thread doing do_wait() didn't do release_task() yet.
304 if (child->ptrace) {
305 child->exit_code = data;
306 dead = __ptrace_detach(current, child);
308 write_unlock_irq(&tasklist_lock);
310 if (unlikely(dead))
311 release_task(child);
313 return 0;
317 * Detach all tasks we were using ptrace on.
319 void exit_ptrace(struct task_struct *tracer)
321 struct task_struct *p, *n;
322 LIST_HEAD(ptrace_dead);
324 write_lock_irq(&tasklist_lock);
325 list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
326 if (__ptrace_detach(tracer, p))
327 list_add(&p->ptrace_entry, &ptrace_dead);
329 write_unlock_irq(&tasklist_lock);
331 BUG_ON(!list_empty(&tracer->ptraced));
333 list_for_each_entry_safe(p, n, &ptrace_dead, ptrace_entry) {
334 list_del_init(&p->ptrace_entry);
335 release_task(p);
339 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
341 int copied = 0;
343 while (len > 0) {
344 char buf[128];
345 int this_len, retval;
347 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
348 retval = access_process_vm(tsk, src, buf, this_len, 0);
349 if (!retval) {
350 if (copied)
351 break;
352 return -EIO;
354 if (copy_to_user(dst, buf, retval))
355 return -EFAULT;
356 copied += retval;
357 src += retval;
358 dst += retval;
359 len -= retval;
361 return copied;
364 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
366 int copied = 0;
368 while (len > 0) {
369 char buf[128];
370 int this_len, retval;
372 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
373 if (copy_from_user(buf, src, this_len))
374 return -EFAULT;
375 retval = access_process_vm(tsk, dst, buf, this_len, 1);
376 if (!retval) {
377 if (copied)
378 break;
379 return -EIO;
381 copied += retval;
382 src += retval;
383 dst += retval;
384 len -= retval;
386 return copied;
389 static int ptrace_setoptions(struct task_struct *child, long data)
391 child->ptrace &= ~PT_TRACE_MASK;
393 if (data & PTRACE_O_TRACESYSGOOD)
394 child->ptrace |= PT_TRACESYSGOOD;
396 if (data & PTRACE_O_TRACEFORK)
397 child->ptrace |= PT_TRACE_FORK;
399 if (data & PTRACE_O_TRACEVFORK)
400 child->ptrace |= PT_TRACE_VFORK;
402 if (data & PTRACE_O_TRACECLONE)
403 child->ptrace |= PT_TRACE_CLONE;
405 if (data & PTRACE_O_TRACEEXEC)
406 child->ptrace |= PT_TRACE_EXEC;
408 if (data & PTRACE_O_TRACEVFORKDONE)
409 child->ptrace |= PT_TRACE_VFORK_DONE;
411 if (data & PTRACE_O_TRACEEXIT)
412 child->ptrace |= PT_TRACE_EXIT;
414 return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
417 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t *info)
419 int error = -ESRCH;
421 read_lock(&tasklist_lock);
422 if (likely(child->sighand != NULL)) {
423 error = -EINVAL;
424 spin_lock_irq(&child->sighand->siglock);
425 if (likely(child->last_siginfo != NULL)) {
426 *info = *child->last_siginfo;
427 error = 0;
429 spin_unlock_irq(&child->sighand->siglock);
431 read_unlock(&tasklist_lock);
432 return error;
435 static int ptrace_setsiginfo(struct task_struct *child, const siginfo_t *info)
437 int error = -ESRCH;
439 read_lock(&tasklist_lock);
440 if (likely(child->sighand != NULL)) {
441 error = -EINVAL;
442 spin_lock_irq(&child->sighand->siglock);
443 if (likely(child->last_siginfo != NULL)) {
444 *child->last_siginfo = *info;
445 error = 0;
447 spin_unlock_irq(&child->sighand->siglock);
449 read_unlock(&tasklist_lock);
450 return error;
454 #ifdef PTRACE_SINGLESTEP
455 #define is_singlestep(request) ((request) == PTRACE_SINGLESTEP)
456 #else
457 #define is_singlestep(request) 0
458 #endif
460 #ifdef PTRACE_SINGLEBLOCK
461 #define is_singleblock(request) ((request) == PTRACE_SINGLEBLOCK)
462 #else
463 #define is_singleblock(request) 0
464 #endif
466 #ifdef PTRACE_SYSEMU
467 #define is_sysemu_singlestep(request) ((request) == PTRACE_SYSEMU_SINGLESTEP)
468 #else
469 #define is_sysemu_singlestep(request) 0
470 #endif
472 static int ptrace_resume(struct task_struct *child, long request, long data)
474 if (!valid_signal(data))
475 return -EIO;
477 if (request == PTRACE_SYSCALL)
478 set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
479 else
480 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
482 #ifdef TIF_SYSCALL_EMU
483 if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
484 set_tsk_thread_flag(child, TIF_SYSCALL_EMU);
485 else
486 clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
487 #endif
489 if (is_singleblock(request)) {
490 if (unlikely(!arch_has_block_step()))
491 return -EIO;
492 user_enable_block_step(child);
493 } else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
494 if (unlikely(!arch_has_single_step()))
495 return -EIO;
496 user_enable_single_step(child);
497 } else {
498 user_disable_single_step(child);
501 child->exit_code = data;
502 wake_up_process(child);
504 return 0;
507 int ptrace_request(struct task_struct *child, long request,
508 long addr, long data)
510 int ret = -EIO;
511 siginfo_t siginfo;
513 switch (request) {
514 case PTRACE_PEEKTEXT:
515 case PTRACE_PEEKDATA:
516 return generic_ptrace_peekdata(child, addr, data);
517 case PTRACE_POKETEXT:
518 case PTRACE_POKEDATA:
519 return generic_ptrace_pokedata(child, addr, data);
521 #ifdef PTRACE_OLDSETOPTIONS
522 case PTRACE_OLDSETOPTIONS:
523 #endif
524 case PTRACE_SETOPTIONS:
525 ret = ptrace_setoptions(child, data);
526 break;
527 case PTRACE_GETEVENTMSG:
528 ret = put_user(child->ptrace_message, (unsigned long __user *) data);
529 break;
531 case PTRACE_GETSIGINFO:
532 ret = ptrace_getsiginfo(child, &siginfo);
533 if (!ret)
534 ret = copy_siginfo_to_user((siginfo_t __user *) data,
535 &siginfo);
536 break;
538 case PTRACE_SETSIGINFO:
539 if (copy_from_user(&siginfo, (siginfo_t __user *) data,
540 sizeof siginfo))
541 ret = -EFAULT;
542 else
543 ret = ptrace_setsiginfo(child, &siginfo);
544 break;
546 case PTRACE_DETACH: /* detach a process that was attached. */
547 ret = ptrace_detach(child, data);
548 break;
550 #ifdef PTRACE_SINGLESTEP
551 case PTRACE_SINGLESTEP:
552 #endif
553 #ifdef PTRACE_SINGLEBLOCK
554 case PTRACE_SINGLEBLOCK:
555 #endif
556 #ifdef PTRACE_SYSEMU
557 case PTRACE_SYSEMU:
558 case PTRACE_SYSEMU_SINGLESTEP:
559 #endif
560 case PTRACE_SYSCALL:
561 case PTRACE_CONT:
562 return ptrace_resume(child, request, data);
564 case PTRACE_KILL:
565 if (child->exit_state) /* already dead */
566 return 0;
567 return ptrace_resume(child, request, SIGKILL);
569 default:
570 break;
573 return ret;
577 * ptrace_traceme -- helper for PTRACE_TRACEME
579 * Performs checks and sets PT_PTRACED.
580 * Should be used by all ptrace implementations for PTRACE_TRACEME.
582 int ptrace_traceme(void)
584 int ret = -EPERM;
587 * Are we already being traced?
589 repeat:
590 task_lock(current);
591 if (!(current->ptrace & PT_PTRACED)) {
593 * See ptrace_attach() comments about the locking here.
595 unsigned long flags;
596 if (!write_trylock_irqsave(&tasklist_lock, flags)) {
597 task_unlock(current);
598 do {
599 cpu_relax();
600 } while (!write_can_lock(&tasklist_lock));
601 goto repeat;
604 ret = security_ptrace_traceme(current->parent);
607 * Check PF_EXITING to ensure ->real_parent has not passed
608 * exit_ptrace(). Otherwise we don't report the error but
609 * pretend ->real_parent untraces us right after return.
611 if (!ret && !(current->real_parent->flags & PF_EXITING)) {
612 current->ptrace |= PT_PTRACED;
613 __ptrace_link(current, current->real_parent);
616 write_unlock_irqrestore(&tasklist_lock, flags);
618 task_unlock(current);
619 return ret;
623 * ptrace_get_task_struct -- grab a task struct reference for ptrace
624 * @pid: process id to grab a task_struct reference of
626 * This function is a helper for ptrace implementations. It checks
627 * permissions and then grabs a task struct for use of the actual
628 * ptrace implementation.
630 * Returns the task_struct for @pid or an ERR_PTR() on failure.
632 struct task_struct *ptrace_get_task_struct(pid_t pid)
634 struct task_struct *child;
636 read_lock(&tasklist_lock);
637 child = find_task_by_vpid(pid);
638 if (child)
639 get_task_struct(child);
641 read_unlock(&tasklist_lock);
642 if (!child)
643 return ERR_PTR(-ESRCH);
644 return child;
647 #ifndef arch_ptrace_attach
648 #define arch_ptrace_attach(child) do { } while (0)
649 #endif
651 SYSCALL_DEFINE4(ptrace, long, request, long, pid, long, addr, long, data)
653 struct task_struct *child;
654 long ret;
657 * This lock_kernel fixes a subtle race with suid exec
659 lock_kernel();
660 if (request == PTRACE_TRACEME) {
661 ret = ptrace_traceme();
662 if (!ret)
663 arch_ptrace_attach(current);
664 goto out;
667 child = ptrace_get_task_struct(pid);
668 if (IS_ERR(child)) {
669 ret = PTR_ERR(child);
670 goto out;
673 if (request == PTRACE_ATTACH) {
674 ret = ptrace_attach(child);
676 * Some architectures need to do book-keeping after
677 * a ptrace attach.
679 if (!ret)
680 arch_ptrace_attach(child);
681 goto out_put_task_struct;
684 ret = ptrace_check_attach(child, request == PTRACE_KILL);
685 if (ret < 0)
686 goto out_put_task_struct;
688 ret = arch_ptrace(child, request, addr, data);
690 out_put_task_struct:
691 put_task_struct(child);
692 out:
693 unlock_kernel();
694 return ret;
697 int generic_ptrace_peekdata(struct task_struct *tsk, long addr, long data)
699 unsigned long tmp;
700 int copied;
702 copied = access_process_vm(tsk, addr, &tmp, sizeof(tmp), 0);
703 if (copied != sizeof(tmp))
704 return -EIO;
705 return put_user(tmp, (unsigned long __user *)data);
708 int generic_ptrace_pokedata(struct task_struct *tsk, long addr, long data)
710 int copied;
712 copied = access_process_vm(tsk, addr, &data, sizeof(data), 1);
713 return (copied == sizeof(data)) ? 0 : -EIO;
716 #if defined CONFIG_COMPAT
717 #include <linux/compat.h>
719 int compat_ptrace_request(struct task_struct *child, compat_long_t request,
720 compat_ulong_t addr, compat_ulong_t data)
722 compat_ulong_t __user *datap = compat_ptr(data);
723 compat_ulong_t word;
724 siginfo_t siginfo;
725 int ret;
727 switch (request) {
728 case PTRACE_PEEKTEXT:
729 case PTRACE_PEEKDATA:
730 ret = access_process_vm(child, addr, &word, sizeof(word), 0);
731 if (ret != sizeof(word))
732 ret = -EIO;
733 else
734 ret = put_user(word, datap);
735 break;
737 case PTRACE_POKETEXT:
738 case PTRACE_POKEDATA:
739 ret = access_process_vm(child, addr, &data, sizeof(data), 1);
740 ret = (ret != sizeof(data) ? -EIO : 0);
741 break;
743 case PTRACE_GETEVENTMSG:
744 ret = put_user((compat_ulong_t) child->ptrace_message, datap);
745 break;
747 case PTRACE_GETSIGINFO:
748 ret = ptrace_getsiginfo(child, &siginfo);
749 if (!ret)
750 ret = copy_siginfo_to_user32(
751 (struct compat_siginfo __user *) datap,
752 &siginfo);
753 break;
755 case PTRACE_SETSIGINFO:
756 memset(&siginfo, 0, sizeof siginfo);
757 if (copy_siginfo_from_user32(
758 &siginfo, (struct compat_siginfo __user *) datap))
759 ret = -EFAULT;
760 else
761 ret = ptrace_setsiginfo(child, &siginfo);
762 break;
764 default:
765 ret = ptrace_request(child, request, addr, data);
768 return ret;
771 asmlinkage long compat_sys_ptrace(compat_long_t request, compat_long_t pid,
772 compat_long_t addr, compat_long_t data)
774 struct task_struct *child;
775 long ret;
778 * This lock_kernel fixes a subtle race with suid exec
780 lock_kernel();
781 if (request == PTRACE_TRACEME) {
782 ret = ptrace_traceme();
783 goto out;
786 child = ptrace_get_task_struct(pid);
787 if (IS_ERR(child)) {
788 ret = PTR_ERR(child);
789 goto out;
792 if (request == PTRACE_ATTACH) {
793 ret = ptrace_attach(child);
795 * Some architectures need to do book-keeping after
796 * a ptrace attach.
798 if (!ret)
799 arch_ptrace_attach(child);
800 goto out_put_task_struct;
803 ret = ptrace_check_attach(child, request == PTRACE_KILL);
804 if (!ret)
805 ret = compat_arch_ptrace(child, request, addr, data);
807 out_put_task_struct:
808 put_task_struct(child);
809 out:
810 unlock_kernel();
811 return ret;
813 #endif /* CONFIG_COMPAT */