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 (unlikely(task
->flags
& PF_KTHREAD
))
177 if (same_thread_group(task
, current
))
180 /* Protect the target's credential calculations against our
181 * interference; SUID, SGID and LSM creds get determined differently
184 retval
= mutex_lock_interruptible(&task
->cred_guard_mutex
);
191 * We want to hold both the task-lock and the
192 * tasklist_lock for writing at the same time.
193 * But that's against the rules (tasklist_lock
194 * is taken for reading by interrupts on other
195 * cpu's that may have task_lock).
198 if (!write_trylock_irqsave(&tasklist_lock
, flags
)) {
202 } while (!write_can_lock(&tasklist_lock
));
206 retval
= __ptrace_may_access(task
, PTRACE_MODE_ATTACH
);
211 if (unlikely(task
->exit_state
))
213 if (task
->ptrace
& PT_PTRACED
)
216 task
->ptrace
|= PT_PTRACED
;
217 if (capable(CAP_SYS_PTRACE
))
218 task
->ptrace
|= PT_PTRACE_CAP
;
220 __ptrace_link(task
, current
);
221 send_sig_info(SIGSTOP
, SEND_SIG_FORCED
, task
);
225 write_unlock_irqrestore(&tasklist_lock
, flags
);
227 mutex_unlock(&task
->cred_guard_mutex
);
233 * Called with irqs disabled, returns true if childs should reap themselves.
235 static int ignoring_children(struct sighand_struct
*sigh
)
238 spin_lock(&sigh
->siglock
);
239 ret
= (sigh
->action
[SIGCHLD
-1].sa
.sa_handler
== SIG_IGN
) ||
240 (sigh
->action
[SIGCHLD
-1].sa
.sa_flags
& SA_NOCLDWAIT
);
241 spin_unlock(&sigh
->siglock
);
246 * Called with tasklist_lock held for writing.
247 * Unlink a traced task, and clean it up if it was a traced zombie.
248 * Return true if it needs to be reaped with release_task().
249 * (We can't call release_task() here because we already hold tasklist_lock.)
251 * If it's a zombie, our attachedness prevented normal parent notification
252 * or self-reaping. Do notification now if it would have happened earlier.
253 * If it should reap itself, return true.
255 * If it's our own child, there is no notification to do.
256 * But if our normal children self-reap, then this child
257 * was prevented by ptrace and we must reap it now.
259 static bool __ptrace_detach(struct task_struct
*tracer
, struct task_struct
*p
)
263 if (p
->exit_state
== EXIT_ZOMBIE
) {
264 if (!task_detached(p
) && thread_group_empty(p
)) {
265 if (!same_thread_group(p
->real_parent
, tracer
))
266 do_notify_parent(p
, p
->exit_signal
);
267 else if (ignoring_children(tracer
->sighand
))
270 if (task_detached(p
)) {
271 /* Mark it as in the process of being reaped. */
272 p
->exit_state
= EXIT_DEAD
;
280 int ptrace_detach(struct task_struct
*child
, unsigned int data
)
284 if (!valid_signal(data
))
287 /* Architecture-specific hardware disable .. */
288 ptrace_disable(child
);
289 clear_tsk_thread_flag(child
, TIF_SYSCALL_TRACE
);
291 write_lock_irq(&tasklist_lock
);
293 * This child can be already killed. Make sure de_thread() or
294 * our sub-thread doing do_wait() didn't do release_task() yet.
297 child
->exit_code
= data
;
298 dead
= __ptrace_detach(current
, child
);
299 if (!child
->exit_state
)
300 wake_up_process(child
);
302 write_unlock_irq(&tasklist_lock
);
311 * Detach all tasks we were using ptrace on.
313 void exit_ptrace(struct task_struct
*tracer
)
315 struct task_struct
*p
, *n
;
316 LIST_HEAD(ptrace_dead
);
318 write_lock_irq(&tasklist_lock
);
319 list_for_each_entry_safe(p
, n
, &tracer
->ptraced
, ptrace_entry
) {
320 if (__ptrace_detach(tracer
, p
))
321 list_add(&p
->ptrace_entry
, &ptrace_dead
);
323 write_unlock_irq(&tasklist_lock
);
325 BUG_ON(!list_empty(&tracer
->ptraced
));
327 list_for_each_entry_safe(p
, n
, &ptrace_dead
, ptrace_entry
) {
328 list_del_init(&p
->ptrace_entry
);
333 int ptrace_readdata(struct task_struct
*tsk
, unsigned long src
, char __user
*dst
, int len
)
339 int this_len
, retval
;
341 this_len
= (len
> sizeof(buf
)) ? sizeof(buf
) : len
;
342 retval
= access_process_vm(tsk
, src
, buf
, this_len
, 0);
348 if (copy_to_user(dst
, buf
, retval
))
358 int ptrace_writedata(struct task_struct
*tsk
, char __user
*src
, unsigned long dst
, int len
)
364 int this_len
, retval
;
366 this_len
= (len
> sizeof(buf
)) ? sizeof(buf
) : len
;
367 if (copy_from_user(buf
, src
, this_len
))
369 retval
= access_process_vm(tsk
, dst
, buf
, this_len
, 1);
383 static int ptrace_setoptions(struct task_struct
*child
, long data
)
385 child
->ptrace
&= ~PT_TRACE_MASK
;
387 if (data
& PTRACE_O_TRACESYSGOOD
)
388 child
->ptrace
|= PT_TRACESYSGOOD
;
390 if (data
& PTRACE_O_TRACEFORK
)
391 child
->ptrace
|= PT_TRACE_FORK
;
393 if (data
& PTRACE_O_TRACEVFORK
)
394 child
->ptrace
|= PT_TRACE_VFORK
;
396 if (data
& PTRACE_O_TRACECLONE
)
397 child
->ptrace
|= PT_TRACE_CLONE
;
399 if (data
& PTRACE_O_TRACEEXEC
)
400 child
->ptrace
|= PT_TRACE_EXEC
;
402 if (data
& PTRACE_O_TRACEVFORKDONE
)
403 child
->ptrace
|= PT_TRACE_VFORK_DONE
;
405 if (data
& PTRACE_O_TRACEEXIT
)
406 child
->ptrace
|= PT_TRACE_EXIT
;
408 return (data
& ~PTRACE_O_MASK
) ? -EINVAL
: 0;
411 static int ptrace_getsiginfo(struct task_struct
*child
, siginfo_t
*info
)
415 read_lock(&tasklist_lock
);
416 if (likely(child
->sighand
!= NULL
)) {
418 spin_lock_irq(&child
->sighand
->siglock
);
419 if (likely(child
->last_siginfo
!= NULL
)) {
420 *info
= *child
->last_siginfo
;
423 spin_unlock_irq(&child
->sighand
->siglock
);
425 read_unlock(&tasklist_lock
);
429 static int ptrace_setsiginfo(struct task_struct
*child
, const siginfo_t
*info
)
433 read_lock(&tasklist_lock
);
434 if (likely(child
->sighand
!= NULL
)) {
436 spin_lock_irq(&child
->sighand
->siglock
);
437 if (likely(child
->last_siginfo
!= NULL
)) {
438 *child
->last_siginfo
= *info
;
441 spin_unlock_irq(&child
->sighand
->siglock
);
443 read_unlock(&tasklist_lock
);
448 #ifdef PTRACE_SINGLESTEP
449 #define is_singlestep(request) ((request) == PTRACE_SINGLESTEP)
451 #define is_singlestep(request) 0
454 #ifdef PTRACE_SINGLEBLOCK
455 #define is_singleblock(request) ((request) == PTRACE_SINGLEBLOCK)
457 #define is_singleblock(request) 0
461 #define is_sysemu_singlestep(request) ((request) == PTRACE_SYSEMU_SINGLESTEP)
463 #define is_sysemu_singlestep(request) 0
466 static int ptrace_resume(struct task_struct
*child
, long request
, long data
)
468 if (!valid_signal(data
))
471 if (request
== PTRACE_SYSCALL
)
472 set_tsk_thread_flag(child
, TIF_SYSCALL_TRACE
);
474 clear_tsk_thread_flag(child
, TIF_SYSCALL_TRACE
);
476 #ifdef TIF_SYSCALL_EMU
477 if (request
== PTRACE_SYSEMU
|| request
== PTRACE_SYSEMU_SINGLESTEP
)
478 set_tsk_thread_flag(child
, TIF_SYSCALL_EMU
);
480 clear_tsk_thread_flag(child
, TIF_SYSCALL_EMU
);
483 if (is_singleblock(request
)) {
484 if (unlikely(!arch_has_block_step()))
486 user_enable_block_step(child
);
487 } else if (is_singlestep(request
) || is_sysemu_singlestep(request
)) {
488 if (unlikely(!arch_has_single_step()))
490 user_enable_single_step(child
);
492 user_disable_single_step(child
);
495 child
->exit_code
= data
;
496 wake_up_process(child
);
501 int ptrace_request(struct task_struct
*child
, long request
,
502 long addr
, long data
)
508 case PTRACE_PEEKTEXT
:
509 case PTRACE_PEEKDATA
:
510 return generic_ptrace_peekdata(child
, addr
, data
);
511 case PTRACE_POKETEXT
:
512 case PTRACE_POKEDATA
:
513 return generic_ptrace_pokedata(child
, addr
, data
);
515 #ifdef PTRACE_OLDSETOPTIONS
516 case PTRACE_OLDSETOPTIONS
:
518 case PTRACE_SETOPTIONS
:
519 ret
= ptrace_setoptions(child
, data
);
521 case PTRACE_GETEVENTMSG
:
522 ret
= put_user(child
->ptrace_message
, (unsigned long __user
*) data
);
525 case PTRACE_GETSIGINFO
:
526 ret
= ptrace_getsiginfo(child
, &siginfo
);
528 ret
= copy_siginfo_to_user((siginfo_t __user
*) data
,
532 case PTRACE_SETSIGINFO
:
533 if (copy_from_user(&siginfo
, (siginfo_t __user
*) data
,
537 ret
= ptrace_setsiginfo(child
, &siginfo
);
540 case PTRACE_DETACH
: /* detach a process that was attached. */
541 ret
= ptrace_detach(child
, data
);
544 #ifdef PTRACE_SINGLESTEP
545 case PTRACE_SINGLESTEP
:
547 #ifdef PTRACE_SINGLEBLOCK
548 case PTRACE_SINGLEBLOCK
:
552 case PTRACE_SYSEMU_SINGLESTEP
:
556 return ptrace_resume(child
, request
, data
);
559 if (child
->exit_state
) /* already dead */
561 return ptrace_resume(child
, request
, SIGKILL
);
571 * ptrace_traceme -- helper for PTRACE_TRACEME
573 * Performs checks and sets PT_PTRACED.
574 * Should be used by all ptrace implementations for PTRACE_TRACEME.
576 int ptrace_traceme(void)
581 * Are we already being traced?
585 if (!(current
->ptrace
& PT_PTRACED
)) {
587 * See ptrace_attach() comments about the locking here.
590 if (!write_trylock_irqsave(&tasklist_lock
, flags
)) {
591 task_unlock(current
);
594 } while (!write_can_lock(&tasklist_lock
));
598 ret
= security_ptrace_traceme(current
->parent
);
601 * Check PF_EXITING to ensure ->real_parent has not passed
602 * exit_ptrace(). Otherwise we don't report the error but
603 * pretend ->real_parent untraces us right after return.
605 if (!ret
&& !(current
->real_parent
->flags
& PF_EXITING
)) {
606 current
->ptrace
|= PT_PTRACED
;
607 __ptrace_link(current
, current
->real_parent
);
610 write_unlock_irqrestore(&tasklist_lock
, flags
);
612 task_unlock(current
);
617 * ptrace_get_task_struct -- grab a task struct reference for ptrace
618 * @pid: process id to grab a task_struct reference of
620 * This function is a helper for ptrace implementations. It checks
621 * permissions and then grabs a task struct for use of the actual
622 * ptrace implementation.
624 * Returns the task_struct for @pid or an ERR_PTR() on failure.
626 struct task_struct
*ptrace_get_task_struct(pid_t pid
)
628 struct task_struct
*child
;
630 read_lock(&tasklist_lock
);
631 child
= find_task_by_vpid(pid
);
633 get_task_struct(child
);
635 read_unlock(&tasklist_lock
);
637 return ERR_PTR(-ESRCH
);
641 #ifndef arch_ptrace_attach
642 #define arch_ptrace_attach(child) do { } while (0)
645 SYSCALL_DEFINE4(ptrace
, long, request
, long, pid
, long, addr
, long, data
)
647 struct task_struct
*child
;
651 * This lock_kernel fixes a subtle race with suid exec
654 if (request
== PTRACE_TRACEME
) {
655 ret
= ptrace_traceme();
657 arch_ptrace_attach(current
);
661 child
= ptrace_get_task_struct(pid
);
663 ret
= PTR_ERR(child
);
667 if (request
== PTRACE_ATTACH
) {
668 ret
= ptrace_attach(child
);
670 * Some architectures need to do book-keeping after
674 arch_ptrace_attach(child
);
675 goto out_put_task_struct
;
678 ret
= ptrace_check_attach(child
, request
== PTRACE_KILL
);
680 goto out_put_task_struct
;
682 ret
= arch_ptrace(child
, request
, addr
, data
);
685 put_task_struct(child
);
691 int generic_ptrace_peekdata(struct task_struct
*tsk
, long addr
, long data
)
696 copied
= access_process_vm(tsk
, addr
, &tmp
, sizeof(tmp
), 0);
697 if (copied
!= sizeof(tmp
))
699 return put_user(tmp
, (unsigned long __user
*)data
);
702 int generic_ptrace_pokedata(struct task_struct
*tsk
, long addr
, long data
)
706 copied
= access_process_vm(tsk
, addr
, &data
, sizeof(data
), 1);
707 return (copied
== sizeof(data
)) ? 0 : -EIO
;
710 #if defined CONFIG_COMPAT
711 #include <linux/compat.h>
713 int compat_ptrace_request(struct task_struct
*child
, compat_long_t request
,
714 compat_ulong_t addr
, compat_ulong_t data
)
716 compat_ulong_t __user
*datap
= compat_ptr(data
);
722 case PTRACE_PEEKTEXT
:
723 case PTRACE_PEEKDATA
:
724 ret
= access_process_vm(child
, addr
, &word
, sizeof(word
), 0);
725 if (ret
!= sizeof(word
))
728 ret
= put_user(word
, datap
);
731 case PTRACE_POKETEXT
:
732 case PTRACE_POKEDATA
:
733 ret
= access_process_vm(child
, addr
, &data
, sizeof(data
), 1);
734 ret
= (ret
!= sizeof(data
) ? -EIO
: 0);
737 case PTRACE_GETEVENTMSG
:
738 ret
= put_user((compat_ulong_t
) child
->ptrace_message
, datap
);
741 case PTRACE_GETSIGINFO
:
742 ret
= ptrace_getsiginfo(child
, &siginfo
);
744 ret
= copy_siginfo_to_user32(
745 (struct compat_siginfo __user
*) datap
,
749 case PTRACE_SETSIGINFO
:
750 memset(&siginfo
, 0, sizeof siginfo
);
751 if (copy_siginfo_from_user32(
752 &siginfo
, (struct compat_siginfo __user
*) datap
))
755 ret
= ptrace_setsiginfo(child
, &siginfo
);
759 ret
= ptrace_request(child
, request
, addr
, data
);
765 asmlinkage
long compat_sys_ptrace(compat_long_t request
, compat_long_t pid
,
766 compat_long_t addr
, compat_long_t data
)
768 struct task_struct
*child
;
772 * This lock_kernel fixes a subtle race with suid exec
775 if (request
== PTRACE_TRACEME
) {
776 ret
= ptrace_traceme();
780 child
= ptrace_get_task_struct(pid
);
782 ret
= PTR_ERR(child
);
786 if (request
== PTRACE_ATTACH
) {
787 ret
= ptrace_attach(child
);
789 * Some architectures need to do book-keeping after
793 arch_ptrace_attach(child
);
794 goto out_put_task_struct
;
797 ret
= ptrace_check_attach(child
, request
== PTRACE_KILL
);
799 ret
= compat_arch_ptrace(child
, request
, addr
, data
);
802 put_task_struct(child
);
807 #endif /* CONFIG_COMPAT */