ptrace_get_task_struct: s/tasklist/rcu/, make it static
[linux-2.6/kvm.git] / kernel / ptrace.c
bloba64fe75a48ba32906145850829d42c82629bdff1
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;
171 audit_ptrace(task);
173 retval = -EPERM;
174 if (unlikely(task->flags & PF_KTHREAD))
175 goto out;
176 if (same_thread_group(task, current))
177 goto out;
180 * Protect exec's credential calculations against our interference;
181 * interference; SUID, SGID and LSM creds get determined differently
182 * under ptrace.
184 retval = mutex_lock_interruptible(&task->cred_guard_mutex);
185 if (retval < 0)
186 goto out;
188 task_lock(task);
189 retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH);
190 task_unlock(task);
191 if (retval)
192 goto unlock_creds;
194 write_lock_irq(&tasklist_lock);
195 retval = -EPERM;
196 if (unlikely(task->exit_state))
197 goto unlock_tasklist;
198 if (task->ptrace)
199 goto unlock_tasklist;
201 task->ptrace = PT_PTRACED;
202 if (capable(CAP_SYS_PTRACE))
203 task->ptrace |= PT_PTRACE_CAP;
205 __ptrace_link(task, current);
206 send_sig_info(SIGSTOP, SEND_SIG_FORCED, task);
208 retval = 0;
209 unlock_tasklist:
210 write_unlock_irq(&tasklist_lock);
211 unlock_creds:
212 mutex_unlock(&task->cred_guard_mutex);
213 out:
214 return retval;
218 * ptrace_traceme -- helper for PTRACE_TRACEME
220 * Performs checks and sets PT_PTRACED.
221 * Should be used by all ptrace implementations for PTRACE_TRACEME.
223 int ptrace_traceme(void)
225 int ret = -EPERM;
227 write_lock_irq(&tasklist_lock);
228 /* Are we already being traced? */
229 if (!current->ptrace) {
230 ret = security_ptrace_traceme(current->parent);
232 * Check PF_EXITING to ensure ->real_parent has not passed
233 * exit_ptrace(). Otherwise we don't report the error but
234 * pretend ->real_parent untraces us right after return.
236 if (!ret && !(current->real_parent->flags & PF_EXITING)) {
237 current->ptrace = PT_PTRACED;
238 __ptrace_link(current, current->real_parent);
241 write_unlock_irq(&tasklist_lock);
243 return ret;
247 * Called with irqs disabled, returns true if childs should reap themselves.
249 static int ignoring_children(struct sighand_struct *sigh)
251 int ret;
252 spin_lock(&sigh->siglock);
253 ret = (sigh->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) ||
254 (sigh->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT);
255 spin_unlock(&sigh->siglock);
256 return ret;
260 * Called with tasklist_lock held for writing.
261 * Unlink a traced task, and clean it up if it was a traced zombie.
262 * Return true if it needs to be reaped with release_task().
263 * (We can't call release_task() here because we already hold tasklist_lock.)
265 * If it's a zombie, our attachedness prevented normal parent notification
266 * or self-reaping. Do notification now if it would have happened earlier.
267 * If it should reap itself, return true.
269 * If it's our own child, there is no notification to do.
270 * But if our normal children self-reap, then this child
271 * was prevented by ptrace and we must reap it now.
273 static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
275 __ptrace_unlink(p);
277 if (p->exit_state == EXIT_ZOMBIE) {
278 if (!task_detached(p) && thread_group_empty(p)) {
279 if (!same_thread_group(p->real_parent, tracer))
280 do_notify_parent(p, p->exit_signal);
281 else if (ignoring_children(tracer->sighand))
282 p->exit_signal = -1;
284 if (task_detached(p)) {
285 /* Mark it as in the process of being reaped. */
286 p->exit_state = EXIT_DEAD;
287 return true;
291 return false;
294 int ptrace_detach(struct task_struct *child, unsigned int data)
296 bool dead = false;
298 if (!valid_signal(data))
299 return -EIO;
301 /* Architecture-specific hardware disable .. */
302 ptrace_disable(child);
303 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
305 write_lock_irq(&tasklist_lock);
307 * This child can be already killed. Make sure de_thread() or
308 * our sub-thread doing do_wait() didn't do release_task() yet.
310 if (child->ptrace) {
311 child->exit_code = data;
312 dead = __ptrace_detach(current, child);
313 if (!child->exit_state)
314 wake_up_process(child);
316 write_unlock_irq(&tasklist_lock);
318 if (unlikely(dead))
319 release_task(child);
321 return 0;
325 * Detach all tasks we were using ptrace on.
327 void exit_ptrace(struct task_struct *tracer)
329 struct task_struct *p, *n;
330 LIST_HEAD(ptrace_dead);
332 write_lock_irq(&tasklist_lock);
333 list_for_each_entry_safe(p, n, &tracer->ptraced, ptrace_entry) {
334 if (__ptrace_detach(tracer, p))
335 list_add(&p->ptrace_entry, &ptrace_dead);
337 write_unlock_irq(&tasklist_lock);
339 BUG_ON(!list_empty(&tracer->ptraced));
341 list_for_each_entry_safe(p, n, &ptrace_dead, ptrace_entry) {
342 list_del_init(&p->ptrace_entry);
343 release_task(p);
347 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
349 int copied = 0;
351 while (len > 0) {
352 char buf[128];
353 int this_len, retval;
355 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
356 retval = access_process_vm(tsk, src, buf, this_len, 0);
357 if (!retval) {
358 if (copied)
359 break;
360 return -EIO;
362 if (copy_to_user(dst, buf, retval))
363 return -EFAULT;
364 copied += retval;
365 src += retval;
366 dst += retval;
367 len -= retval;
369 return copied;
372 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
374 int copied = 0;
376 while (len > 0) {
377 char buf[128];
378 int this_len, retval;
380 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
381 if (copy_from_user(buf, src, this_len))
382 return -EFAULT;
383 retval = access_process_vm(tsk, dst, buf, this_len, 1);
384 if (!retval) {
385 if (copied)
386 break;
387 return -EIO;
389 copied += retval;
390 src += retval;
391 dst += retval;
392 len -= retval;
394 return copied;
397 static int ptrace_setoptions(struct task_struct *child, long data)
399 child->ptrace &= ~PT_TRACE_MASK;
401 if (data & PTRACE_O_TRACESYSGOOD)
402 child->ptrace |= PT_TRACESYSGOOD;
404 if (data & PTRACE_O_TRACEFORK)
405 child->ptrace |= PT_TRACE_FORK;
407 if (data & PTRACE_O_TRACEVFORK)
408 child->ptrace |= PT_TRACE_VFORK;
410 if (data & PTRACE_O_TRACECLONE)
411 child->ptrace |= PT_TRACE_CLONE;
413 if (data & PTRACE_O_TRACEEXEC)
414 child->ptrace |= PT_TRACE_EXEC;
416 if (data & PTRACE_O_TRACEVFORKDONE)
417 child->ptrace |= PT_TRACE_VFORK_DONE;
419 if (data & PTRACE_O_TRACEEXIT)
420 child->ptrace |= PT_TRACE_EXIT;
422 return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
425 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t *info)
427 int error = -ESRCH;
429 read_lock(&tasklist_lock);
430 if (likely(child->sighand != NULL)) {
431 error = -EINVAL;
432 spin_lock_irq(&child->sighand->siglock);
433 if (likely(child->last_siginfo != NULL)) {
434 *info = *child->last_siginfo;
435 error = 0;
437 spin_unlock_irq(&child->sighand->siglock);
439 read_unlock(&tasklist_lock);
440 return error;
443 static int ptrace_setsiginfo(struct task_struct *child, const siginfo_t *info)
445 int error = -ESRCH;
447 read_lock(&tasklist_lock);
448 if (likely(child->sighand != NULL)) {
449 error = -EINVAL;
450 spin_lock_irq(&child->sighand->siglock);
451 if (likely(child->last_siginfo != NULL)) {
452 *child->last_siginfo = *info;
453 error = 0;
455 spin_unlock_irq(&child->sighand->siglock);
457 read_unlock(&tasklist_lock);
458 return error;
462 #ifdef PTRACE_SINGLESTEP
463 #define is_singlestep(request) ((request) == PTRACE_SINGLESTEP)
464 #else
465 #define is_singlestep(request) 0
466 #endif
468 #ifdef PTRACE_SINGLEBLOCK
469 #define is_singleblock(request) ((request) == PTRACE_SINGLEBLOCK)
470 #else
471 #define is_singleblock(request) 0
472 #endif
474 #ifdef PTRACE_SYSEMU
475 #define is_sysemu_singlestep(request) ((request) == PTRACE_SYSEMU_SINGLESTEP)
476 #else
477 #define is_sysemu_singlestep(request) 0
478 #endif
480 static int ptrace_resume(struct task_struct *child, long request, long data)
482 if (!valid_signal(data))
483 return -EIO;
485 if (request == PTRACE_SYSCALL)
486 set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
487 else
488 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
490 #ifdef TIF_SYSCALL_EMU
491 if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
492 set_tsk_thread_flag(child, TIF_SYSCALL_EMU);
493 else
494 clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
495 #endif
497 if (is_singleblock(request)) {
498 if (unlikely(!arch_has_block_step()))
499 return -EIO;
500 user_enable_block_step(child);
501 } else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
502 if (unlikely(!arch_has_single_step()))
503 return -EIO;
504 user_enable_single_step(child);
505 } else {
506 user_disable_single_step(child);
509 child->exit_code = data;
510 wake_up_process(child);
512 return 0;
515 int ptrace_request(struct task_struct *child, long request,
516 long addr, long data)
518 int ret = -EIO;
519 siginfo_t siginfo;
521 switch (request) {
522 case PTRACE_PEEKTEXT:
523 case PTRACE_PEEKDATA:
524 return generic_ptrace_peekdata(child, addr, data);
525 case PTRACE_POKETEXT:
526 case PTRACE_POKEDATA:
527 return generic_ptrace_pokedata(child, addr, data);
529 #ifdef PTRACE_OLDSETOPTIONS
530 case PTRACE_OLDSETOPTIONS:
531 #endif
532 case PTRACE_SETOPTIONS:
533 ret = ptrace_setoptions(child, data);
534 break;
535 case PTRACE_GETEVENTMSG:
536 ret = put_user(child->ptrace_message, (unsigned long __user *) data);
537 break;
539 case PTRACE_GETSIGINFO:
540 ret = ptrace_getsiginfo(child, &siginfo);
541 if (!ret)
542 ret = copy_siginfo_to_user((siginfo_t __user *) data,
543 &siginfo);
544 break;
546 case PTRACE_SETSIGINFO:
547 if (copy_from_user(&siginfo, (siginfo_t __user *) data,
548 sizeof siginfo))
549 ret = -EFAULT;
550 else
551 ret = ptrace_setsiginfo(child, &siginfo);
552 break;
554 case PTRACE_DETACH: /* detach a process that was attached. */
555 ret = ptrace_detach(child, data);
556 break;
558 #ifdef PTRACE_SINGLESTEP
559 case PTRACE_SINGLESTEP:
560 #endif
561 #ifdef PTRACE_SINGLEBLOCK
562 case PTRACE_SINGLEBLOCK:
563 #endif
564 #ifdef PTRACE_SYSEMU
565 case PTRACE_SYSEMU:
566 case PTRACE_SYSEMU_SINGLESTEP:
567 #endif
568 case PTRACE_SYSCALL:
569 case PTRACE_CONT:
570 return ptrace_resume(child, request, data);
572 case PTRACE_KILL:
573 if (child->exit_state) /* already dead */
574 return 0;
575 return ptrace_resume(child, request, SIGKILL);
577 default:
578 break;
581 return ret;
584 static struct task_struct *ptrace_get_task_struct(pid_t pid)
586 struct task_struct *child;
588 rcu_read_lock();
589 child = find_task_by_vpid(pid);
590 if (child)
591 get_task_struct(child);
592 rcu_read_unlock();
594 if (!child)
595 return ERR_PTR(-ESRCH);
596 return child;
599 #ifndef arch_ptrace_attach
600 #define arch_ptrace_attach(child) do { } while (0)
601 #endif
603 SYSCALL_DEFINE4(ptrace, long, request, long, pid, long, addr, long, data)
605 struct task_struct *child;
606 long ret;
609 * This lock_kernel fixes a subtle race with suid exec
611 lock_kernel();
612 if (request == PTRACE_TRACEME) {
613 ret = ptrace_traceme();
614 if (!ret)
615 arch_ptrace_attach(current);
616 goto out;
619 child = ptrace_get_task_struct(pid);
620 if (IS_ERR(child)) {
621 ret = PTR_ERR(child);
622 goto out;
625 if (request == PTRACE_ATTACH) {
626 ret = ptrace_attach(child);
628 * Some architectures need to do book-keeping after
629 * a ptrace attach.
631 if (!ret)
632 arch_ptrace_attach(child);
633 goto out_put_task_struct;
636 ret = ptrace_check_attach(child, request == PTRACE_KILL);
637 if (ret < 0)
638 goto out_put_task_struct;
640 ret = arch_ptrace(child, request, addr, data);
642 out_put_task_struct:
643 put_task_struct(child);
644 out:
645 unlock_kernel();
646 return ret;
649 int generic_ptrace_peekdata(struct task_struct *tsk, long addr, long data)
651 unsigned long tmp;
652 int copied;
654 copied = access_process_vm(tsk, addr, &tmp, sizeof(tmp), 0);
655 if (copied != sizeof(tmp))
656 return -EIO;
657 return put_user(tmp, (unsigned long __user *)data);
660 int generic_ptrace_pokedata(struct task_struct *tsk, long addr, long data)
662 int copied;
664 copied = access_process_vm(tsk, addr, &data, sizeof(data), 1);
665 return (copied == sizeof(data)) ? 0 : -EIO;
668 #if defined CONFIG_COMPAT
669 #include <linux/compat.h>
671 int compat_ptrace_request(struct task_struct *child, compat_long_t request,
672 compat_ulong_t addr, compat_ulong_t data)
674 compat_ulong_t __user *datap = compat_ptr(data);
675 compat_ulong_t word;
676 siginfo_t siginfo;
677 int ret;
679 switch (request) {
680 case PTRACE_PEEKTEXT:
681 case PTRACE_PEEKDATA:
682 ret = access_process_vm(child, addr, &word, sizeof(word), 0);
683 if (ret != sizeof(word))
684 ret = -EIO;
685 else
686 ret = put_user(word, datap);
687 break;
689 case PTRACE_POKETEXT:
690 case PTRACE_POKEDATA:
691 ret = access_process_vm(child, addr, &data, sizeof(data), 1);
692 ret = (ret != sizeof(data) ? -EIO : 0);
693 break;
695 case PTRACE_GETEVENTMSG:
696 ret = put_user((compat_ulong_t) child->ptrace_message, datap);
697 break;
699 case PTRACE_GETSIGINFO:
700 ret = ptrace_getsiginfo(child, &siginfo);
701 if (!ret)
702 ret = copy_siginfo_to_user32(
703 (struct compat_siginfo __user *) datap,
704 &siginfo);
705 break;
707 case PTRACE_SETSIGINFO:
708 memset(&siginfo, 0, sizeof siginfo);
709 if (copy_siginfo_from_user32(
710 &siginfo, (struct compat_siginfo __user *) datap))
711 ret = -EFAULT;
712 else
713 ret = ptrace_setsiginfo(child, &siginfo);
714 break;
716 default:
717 ret = ptrace_request(child, request, addr, data);
720 return ret;
723 asmlinkage long compat_sys_ptrace(compat_long_t request, compat_long_t pid,
724 compat_long_t addr, compat_long_t data)
726 struct task_struct *child;
727 long ret;
730 * This lock_kernel fixes a subtle race with suid exec
732 lock_kernel();
733 if (request == PTRACE_TRACEME) {
734 ret = ptrace_traceme();
735 goto out;
738 child = ptrace_get_task_struct(pid);
739 if (IS_ERR(child)) {
740 ret = PTR_ERR(child);
741 goto out;
744 if (request == PTRACE_ATTACH) {
745 ret = ptrace_attach(child);
747 * Some architectures need to do book-keeping after
748 * a ptrace attach.
750 if (!ret)
751 arch_ptrace_attach(child);
752 goto out_put_task_struct;
755 ret = ptrace_check_attach(child, request == PTRACE_KILL);
756 if (!ret)
757 ret = compat_arch_ptrace(child, request, addr, data);
759 out_put_task_struct:
760 put_task_struct(child);
761 out:
762 unlock_kernel();
763 return ret;
765 #endif /* CONFIG_COMPAT */