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>
24 #include <asm/pgtable.h>
25 #include <asm/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_list));
36 if (child->parent == new_parent)
38 list_add(&child->ptrace_list, &child->parent->ptrace_children);
40 child->parent = new_parent;
45 * Turn a tracing stop into a normal stop now, since with no tracer there
46 * would be no way to wake it up with SIGCONT or SIGKILL. If there was a
47 * signal sent that would resume the child, but didn't because it was in
48 * TASK_TRACED, resume it now.
49 * Requires that irqs be disabled.
51 void ptrace_untrace(struct task_struct *child)
53 spin_lock(&child->sighand->siglock);
54 if (child->state == TASK_TRACED) {
55 if (child->signal->flags & SIGNAL_STOP_STOPPED) {
56 child->state = TASK_STOPPED;
58 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 if (!list_empty(&child->ptrace_list)) {
76 list_del_init(&child->ptrace_list);
78 child->parent = child->real_parent;
82 if (child->state == TASK_TRACED)
83 ptrace_untrace(child);
87 * Check that we have indeed attached to the thing..
89 int ptrace_check_attach(struct task_struct *child, int kill)
94 * We take the read lock around doing both checks to close a
95 * possible race where someone else was tracing our child and
96 * detached between these two checks. After this locked check,
97 * we are sure that this is our traced child and that can only
98 * be changed by us so it's not changing right after this.
100 read_lock(&tasklist_lock);
101 if ((child->ptrace & PT_PTRACED) && child->parent == current &&
102 (!(child->ptrace & PT_ATTACHED) || child->real_parent != current)
103 && child->signal != NULL) {
105 spin_lock_irq(&child->sighand->siglock);
106 if (child->state == TASK_STOPPED) {
107 child->state = TASK_TRACED;
108 } else if (child->state != TASK_TRACED && !kill) {
111 spin_unlock_irq(&child->sighand->siglock);
113 read_unlock(&tasklist_lock);
116 wait_task_inactive(child);
119 /* All systems go.. */
123 static int may_attach(struct task_struct *task)
125 /* May we inspect the given task?
126 * This check is used both for attaching with ptrace
127 * and for allowing access to sensitive information in /proc.
129 * ptrace_attach denies several cases that /proc allows
130 * because setting up the necessary parent/child relationship
131 * or halting the specified task is impossible.
134 /* Don't let security modules deny introspection */
137 if (((current->uid != task->euid) ||
138 (current->uid != task->suid) ||
139 (current->uid != task->uid) ||
140 (current->gid != task->egid) ||
141 (current->gid != task->sgid) ||
142 (current->gid != task->gid)) && !capable(CAP_SYS_PTRACE))
146 dumpable = get_dumpable(task->mm);
147 if (!dumpable && !capable(CAP_SYS_PTRACE))
150 return security_ptrace(current, task);
153 int ptrace_may_attach(struct task_struct *task)
157 err = may_attach(task);
162 int ptrace_attach(struct task_struct *task)
172 if (task->tgid == current->tgid)
179 * We want to hold both the task-lock and the
180 * tasklist_lock for writing at the same time.
181 * But that's against the rules (tasklist_lock
182 * is taken for reading by interrupts on other
183 * cpu's that may have task_lock).
186 if (!write_trylock_irqsave(&tasklist_lock, flags)) {
190 } while (!write_can_lock(&tasklist_lock));
196 /* the same process cannot be attached many times */
197 if (task->ptrace & PT_PTRACED)
199 retval = may_attach(task);
204 task->ptrace |= PT_PTRACED | ((task->real_parent != current)
206 if (capable(CAP_SYS_PTRACE))
207 task->ptrace |= PT_PTRACE_CAP;
209 __ptrace_link(task, current);
211 force_sig_specific(SIGSTOP, task);
214 write_unlock_irqrestore(&tasklist_lock, flags);
220 static inline void __ptrace_detach(struct task_struct *child, unsigned int data)
222 child->exit_code = data;
223 /* .. re-parent .. */
224 __ptrace_unlink(child);
225 /* .. and wake it up. */
226 if (child->exit_state != EXIT_ZOMBIE)
227 wake_up_process(child);
230 int ptrace_detach(struct task_struct *child, unsigned int data)
232 if (!valid_signal(data))
235 /* Architecture-specific hardware disable .. */
236 ptrace_disable(child);
237 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
239 write_lock_irq(&tasklist_lock);
240 /* protect against de_thread()->release_task() */
242 __ptrace_detach(child, data);
243 write_unlock_irq(&tasklist_lock);
248 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
254 int this_len, retval;
256 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
257 retval = access_process_vm(tsk, src, buf, this_len, 0);
263 if (copy_to_user(dst, buf, retval))
273 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
279 int this_len, retval;
281 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
282 if (copy_from_user(buf, src, this_len))
284 retval = access_process_vm(tsk, dst, buf, this_len, 1);
298 static int ptrace_setoptions(struct task_struct *child, long data)
300 child->ptrace &= ~PT_TRACE_MASK;
302 if (data & PTRACE_O_TRACESYSGOOD)
303 child->ptrace |= PT_TRACESYSGOOD;
305 if (data & PTRACE_O_TRACEFORK)
306 child->ptrace |= PT_TRACE_FORK;
308 if (data & PTRACE_O_TRACEVFORK)
309 child->ptrace |= PT_TRACE_VFORK;
311 if (data & PTRACE_O_TRACECLONE)
312 child->ptrace |= PT_TRACE_CLONE;
314 if (data & PTRACE_O_TRACEEXEC)
315 child->ptrace |= PT_TRACE_EXEC;
317 if (data & PTRACE_O_TRACEVFORKDONE)
318 child->ptrace |= PT_TRACE_VFORK_DONE;
320 if (data & PTRACE_O_TRACEEXIT)
321 child->ptrace |= PT_TRACE_EXIT;
323 return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
326 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t __user * data)
331 read_lock(&tasklist_lock);
332 if (likely(child->sighand != NULL)) {
334 spin_lock_irq(&child->sighand->siglock);
335 if (likely(child->last_siginfo != NULL)) {
336 lastinfo = *child->last_siginfo;
339 spin_unlock_irq(&child->sighand->siglock);
341 read_unlock(&tasklist_lock);
343 return copy_siginfo_to_user(data, &lastinfo);
347 static int ptrace_setsiginfo(struct task_struct *child, siginfo_t __user * data)
352 if (copy_from_user(&newinfo, data, sizeof (siginfo_t)))
355 read_lock(&tasklist_lock);
356 if (likely(child->sighand != NULL)) {
358 spin_lock_irq(&child->sighand->siglock);
359 if (likely(child->last_siginfo != NULL)) {
360 *child->last_siginfo = newinfo;
363 spin_unlock_irq(&child->sighand->siglock);
365 read_unlock(&tasklist_lock);
369 int ptrace_request(struct task_struct *child, long request,
370 long addr, long data)
375 #ifdef PTRACE_OLDSETOPTIONS
376 case PTRACE_OLDSETOPTIONS:
378 case PTRACE_SETOPTIONS:
379 ret = ptrace_setoptions(child, data);
381 case PTRACE_GETEVENTMSG:
382 ret = put_user(child->ptrace_message, (unsigned long __user *) data);
384 case PTRACE_GETSIGINFO:
385 ret = ptrace_getsiginfo(child, (siginfo_t __user *) data);
387 case PTRACE_SETSIGINFO:
388 ret = ptrace_setsiginfo(child, (siginfo_t __user *) data);
390 case PTRACE_DETACH: /* detach a process that was attached. */
391 ret = ptrace_detach(child, data);
401 * ptrace_traceme -- helper for PTRACE_TRACEME
403 * Performs checks and sets PT_PTRACED.
404 * Should be used by all ptrace implementations for PTRACE_TRACEME.
406 int ptrace_traceme(void)
411 * Are we already being traced?
414 if (!(current->ptrace & PT_PTRACED)) {
415 ret = security_ptrace(current->parent, current);
417 * Set the ptrace bit in the process ptrace flags.
420 current->ptrace |= PT_PTRACED;
422 task_unlock(current);
427 * ptrace_get_task_struct -- grab a task struct reference for ptrace
428 * @pid: process id to grab a task_struct reference of
430 * This function is a helper for ptrace implementations. It checks
431 * permissions and then grabs a task struct for use of the actual
432 * ptrace implementation.
434 * Returns the task_struct for @pid or an ERR_PTR() on failure.
436 struct task_struct *ptrace_get_task_struct(pid_t pid)
438 struct task_struct *child;
441 * Tracing init is not allowed.
444 return ERR_PTR(-EPERM);
446 read_lock(&tasklist_lock);
447 child = find_task_by_vpid(pid);
449 get_task_struct(child);
451 read_unlock(&tasklist_lock);
453 return ERR_PTR(-ESRCH);
457 #ifndef arch_ptrace_attach
458 #define arch_ptrace_attach(child) do { } while (0)
461 #ifndef __ARCH_SYS_PTRACE
462 asmlinkage long sys_ptrace(long request, long pid, long addr, long data)
464 struct task_struct *child;
468 * This lock_kernel fixes a subtle race with suid exec
471 if (request == PTRACE_TRACEME) {
472 ret = ptrace_traceme();
476 child = ptrace_get_task_struct(pid);
478 ret = PTR_ERR(child);
482 if (request == PTRACE_ATTACH) {
483 ret = ptrace_attach(child);
485 * Some architectures need to do book-keeping after
489 arch_ptrace_attach(child);
490 goto out_put_task_struct;
493 ret = ptrace_check_attach(child, request == PTRACE_KILL);
495 goto out_put_task_struct;
497 ret = arch_ptrace(child, request, addr, data);
499 goto out_put_task_struct;
502 put_task_struct(child);
507 #endif /* __ARCH_SYS_PTRACE */
509 int generic_ptrace_peekdata(struct task_struct *tsk, long addr, long data)
514 copied = access_process_vm(tsk, addr, &tmp, sizeof(tmp), 0);
515 if (copied != sizeof(tmp))
517 return put_user(tmp, (unsigned long __user *)data);
520 int generic_ptrace_pokedata(struct task_struct *tsk, long addr, long data)
524 copied = access_process_vm(tsk, addr, &data, sizeof(data), 1);
525 return (copied == sizeof(data)) ? 0 : -EIO;