blktrace: fix off-by-one bug
[linux-2.6/mini2440.git] / kernel / ptrace.c
blobc9cf48b21f052a620b34151cc9de6c4b48f4ad66
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)) {
63 if (child->signal->flags & SIGNAL_STOP_STOPPED) {
64 __set_task_state(child, TASK_STOPPED);
65 } else {
66 signal_wake_up(child, 1);
69 spin_unlock(&child->sighand->siglock);
73 * unptrace a task: move it back to its original parent and
74 * remove it from the ptrace list.
76 * Must be called with the tasklist lock write-held.
78 void __ptrace_unlink(struct task_struct *child)
80 BUG_ON(!child->ptrace);
82 child->ptrace = 0;
83 child->parent = child->real_parent;
84 list_del_init(&child->ptrace_entry);
86 arch_ptrace_untrace(child);
87 if (task_is_traced(child))
88 ptrace_untrace(child);
92 * Check that we have indeed attached to the thing..
94 int ptrace_check_attach(struct task_struct *child, int kill)
96 int ret = -ESRCH;
99 * We take the read lock around doing both checks to close a
100 * possible race where someone else was tracing our child and
101 * detached between these two checks. After this locked check,
102 * we are sure that this is our traced child and that can only
103 * be changed by us so it's not changing right after this.
105 read_lock(&tasklist_lock);
106 if ((child->ptrace & PT_PTRACED) && child->parent == current) {
107 ret = 0;
109 * child->sighand can't be NULL, release_task()
110 * does ptrace_unlink() before __exit_signal().
112 spin_lock_irq(&child->sighand->siglock);
113 if (task_is_stopped(child))
114 child->state = TASK_TRACED;
115 else if (!task_is_traced(child) && !kill)
116 ret = -ESRCH;
117 spin_unlock_irq(&child->sighand->siglock);
119 read_unlock(&tasklist_lock);
121 if (!ret && !kill)
122 ret = wait_task_inactive(child, TASK_TRACED) ? 0 : -ESRCH;
124 /* All systems go.. */
125 return ret;
128 int __ptrace_may_access(struct task_struct *task, unsigned int mode)
130 const struct cred *cred = current_cred(), *tcred;
132 /* May we inspect the given task?
133 * This check is used both for attaching with ptrace
134 * and for allowing access to sensitive information in /proc.
136 * ptrace_attach denies several cases that /proc allows
137 * because setting up the necessary parent/child relationship
138 * or halting the specified task is impossible.
140 int dumpable = 0;
141 /* Don't let security modules deny introspection */
142 if (task == current)
143 return 0;
144 rcu_read_lock();
145 tcred = __task_cred(task);
146 if ((cred->uid != tcred->euid ||
147 cred->uid != tcred->suid ||
148 cred->uid != tcred->uid ||
149 cred->gid != tcred->egid ||
150 cred->gid != tcred->sgid ||
151 cred->gid != tcred->gid) &&
152 !capable(CAP_SYS_PTRACE)) {
153 rcu_read_unlock();
154 return -EPERM;
156 rcu_read_unlock();
157 smp_rmb();
158 if (task->mm)
159 dumpable = get_dumpable(task->mm);
160 if (!dumpable && !capable(CAP_SYS_PTRACE))
161 return -EPERM;
163 return security_ptrace_may_access(task, mode);
166 bool ptrace_may_access(struct task_struct *task, unsigned int mode)
168 int err;
169 task_lock(task);
170 err = __ptrace_may_access(task, mode);
171 task_unlock(task);
172 return (!err ? true : false);
175 int ptrace_attach(struct task_struct *task)
177 int retval;
178 unsigned long flags;
180 audit_ptrace(task);
182 retval = -EPERM;
183 if (same_thread_group(task, current))
184 goto out;
186 /* Protect exec's credential calculations against our interference;
187 * SUID, SGID and LSM creds get determined differently under ptrace.
189 retval = mutex_lock_interruptible(&current->cred_exec_mutex);
190 if (retval < 0)
191 goto out;
193 retval = -EPERM;
194 repeat:
196 * Nasty, nasty.
198 * We want to hold both the task-lock and the
199 * tasklist_lock for writing at the same time.
200 * But that's against the rules (tasklist_lock
201 * is taken for reading by interrupts on other
202 * cpu's that may have task_lock).
204 task_lock(task);
205 if (!write_trylock_irqsave(&tasklist_lock, flags)) {
206 task_unlock(task);
207 do {
208 cpu_relax();
209 } while (!write_can_lock(&tasklist_lock));
210 goto repeat;
213 if (!task->mm)
214 goto bad;
215 /* the same process cannot be attached many times */
216 if (task->ptrace & PT_PTRACED)
217 goto bad;
218 retval = __ptrace_may_access(task, PTRACE_MODE_ATTACH);
219 if (retval)
220 goto bad;
222 /* Go */
223 task->ptrace |= PT_PTRACED;
224 if (capable(CAP_SYS_PTRACE))
225 task->ptrace |= PT_PTRACE_CAP;
227 __ptrace_link(task, current);
229 send_sig_info(SIGSTOP, SEND_SIG_FORCED, task);
230 bad:
231 write_unlock_irqrestore(&tasklist_lock, flags);
232 task_unlock(task);
233 mutex_unlock(&current->cred_exec_mutex);
234 out:
235 return retval;
238 static inline void __ptrace_detach(struct task_struct *child, unsigned int data)
240 child->exit_code = data;
241 /* .. re-parent .. */
242 __ptrace_unlink(child);
243 /* .. and wake it up. */
244 if (child->exit_state != EXIT_ZOMBIE)
245 wake_up_process(child);
248 int ptrace_detach(struct task_struct *child, unsigned int data)
250 if (!valid_signal(data))
251 return -EIO;
253 /* Architecture-specific hardware disable .. */
254 ptrace_disable(child);
255 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
257 write_lock_irq(&tasklist_lock);
258 /* protect against de_thread()->release_task() */
259 if (child->ptrace)
260 __ptrace_detach(child, data);
261 write_unlock_irq(&tasklist_lock);
263 return 0;
266 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
268 int copied = 0;
270 while (len > 0) {
271 char buf[128];
272 int this_len, retval;
274 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
275 retval = access_process_vm(tsk, src, buf, this_len, 0);
276 if (!retval) {
277 if (copied)
278 break;
279 return -EIO;
281 if (copy_to_user(dst, buf, retval))
282 return -EFAULT;
283 copied += retval;
284 src += retval;
285 dst += retval;
286 len -= retval;
288 return copied;
291 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
293 int copied = 0;
295 while (len > 0) {
296 char buf[128];
297 int this_len, retval;
299 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
300 if (copy_from_user(buf, src, this_len))
301 return -EFAULT;
302 retval = access_process_vm(tsk, dst, buf, this_len, 1);
303 if (!retval) {
304 if (copied)
305 break;
306 return -EIO;
308 copied += retval;
309 src += retval;
310 dst += retval;
311 len -= retval;
313 return copied;
316 static int ptrace_setoptions(struct task_struct *child, long data)
318 child->ptrace &= ~PT_TRACE_MASK;
320 if (data & PTRACE_O_TRACESYSGOOD)
321 child->ptrace |= PT_TRACESYSGOOD;
323 if (data & PTRACE_O_TRACEFORK)
324 child->ptrace |= PT_TRACE_FORK;
326 if (data & PTRACE_O_TRACEVFORK)
327 child->ptrace |= PT_TRACE_VFORK;
329 if (data & PTRACE_O_TRACECLONE)
330 child->ptrace |= PT_TRACE_CLONE;
332 if (data & PTRACE_O_TRACEEXEC)
333 child->ptrace |= PT_TRACE_EXEC;
335 if (data & PTRACE_O_TRACEVFORKDONE)
336 child->ptrace |= PT_TRACE_VFORK_DONE;
338 if (data & PTRACE_O_TRACEEXIT)
339 child->ptrace |= PT_TRACE_EXIT;
341 return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
344 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t *info)
346 int error = -ESRCH;
348 read_lock(&tasklist_lock);
349 if (likely(child->sighand != NULL)) {
350 error = -EINVAL;
351 spin_lock_irq(&child->sighand->siglock);
352 if (likely(child->last_siginfo != NULL)) {
353 *info = *child->last_siginfo;
354 error = 0;
356 spin_unlock_irq(&child->sighand->siglock);
358 read_unlock(&tasklist_lock);
359 return error;
362 static int ptrace_setsiginfo(struct task_struct *child, const siginfo_t *info)
364 int error = -ESRCH;
366 read_lock(&tasklist_lock);
367 if (likely(child->sighand != NULL)) {
368 error = -EINVAL;
369 spin_lock_irq(&child->sighand->siglock);
370 if (likely(child->last_siginfo != NULL)) {
371 *child->last_siginfo = *info;
372 error = 0;
374 spin_unlock_irq(&child->sighand->siglock);
376 read_unlock(&tasklist_lock);
377 return error;
381 #ifdef PTRACE_SINGLESTEP
382 #define is_singlestep(request) ((request) == PTRACE_SINGLESTEP)
383 #else
384 #define is_singlestep(request) 0
385 #endif
387 #ifdef PTRACE_SINGLEBLOCK
388 #define is_singleblock(request) ((request) == PTRACE_SINGLEBLOCK)
389 #else
390 #define is_singleblock(request) 0
391 #endif
393 #ifdef PTRACE_SYSEMU
394 #define is_sysemu_singlestep(request) ((request) == PTRACE_SYSEMU_SINGLESTEP)
395 #else
396 #define is_sysemu_singlestep(request) 0
397 #endif
399 static int ptrace_resume(struct task_struct *child, long request, long data)
401 if (!valid_signal(data))
402 return -EIO;
404 if (request == PTRACE_SYSCALL)
405 set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
406 else
407 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
409 #ifdef TIF_SYSCALL_EMU
410 if (request == PTRACE_SYSEMU || request == PTRACE_SYSEMU_SINGLESTEP)
411 set_tsk_thread_flag(child, TIF_SYSCALL_EMU);
412 else
413 clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
414 #endif
416 if (is_singleblock(request)) {
417 if (unlikely(!arch_has_block_step()))
418 return -EIO;
419 user_enable_block_step(child);
420 } else if (is_singlestep(request) || is_sysemu_singlestep(request)) {
421 if (unlikely(!arch_has_single_step()))
422 return -EIO;
423 user_enable_single_step(child);
425 else
426 user_disable_single_step(child);
428 child->exit_code = data;
429 wake_up_process(child);
431 return 0;
434 int ptrace_request(struct task_struct *child, long request,
435 long addr, long data)
437 int ret = -EIO;
438 siginfo_t siginfo;
440 switch (request) {
441 case PTRACE_PEEKTEXT:
442 case PTRACE_PEEKDATA:
443 return generic_ptrace_peekdata(child, addr, data);
444 case PTRACE_POKETEXT:
445 case PTRACE_POKEDATA:
446 return generic_ptrace_pokedata(child, addr, data);
448 #ifdef PTRACE_OLDSETOPTIONS
449 case PTRACE_OLDSETOPTIONS:
450 #endif
451 case PTRACE_SETOPTIONS:
452 ret = ptrace_setoptions(child, data);
453 break;
454 case PTRACE_GETEVENTMSG:
455 ret = put_user(child->ptrace_message, (unsigned long __user *) data);
456 break;
458 case PTRACE_GETSIGINFO:
459 ret = ptrace_getsiginfo(child, &siginfo);
460 if (!ret)
461 ret = copy_siginfo_to_user((siginfo_t __user *) data,
462 &siginfo);
463 break;
465 case PTRACE_SETSIGINFO:
466 if (copy_from_user(&siginfo, (siginfo_t __user *) data,
467 sizeof siginfo))
468 ret = -EFAULT;
469 else
470 ret = ptrace_setsiginfo(child, &siginfo);
471 break;
473 case PTRACE_DETACH: /* detach a process that was attached. */
474 ret = ptrace_detach(child, data);
475 break;
477 #ifdef PTRACE_SINGLESTEP
478 case PTRACE_SINGLESTEP:
479 #endif
480 #ifdef PTRACE_SINGLEBLOCK
481 case PTRACE_SINGLEBLOCK:
482 #endif
483 #ifdef PTRACE_SYSEMU
484 case PTRACE_SYSEMU:
485 case PTRACE_SYSEMU_SINGLESTEP:
486 #endif
487 case PTRACE_SYSCALL:
488 case PTRACE_CONT:
489 return ptrace_resume(child, request, data);
491 case PTRACE_KILL:
492 if (child->exit_state) /* already dead */
493 return 0;
494 return ptrace_resume(child, request, SIGKILL);
496 default:
497 break;
500 return ret;
504 * ptrace_traceme -- helper for PTRACE_TRACEME
506 * Performs checks and sets PT_PTRACED.
507 * Should be used by all ptrace implementations for PTRACE_TRACEME.
509 int ptrace_traceme(void)
511 int ret = -EPERM;
514 * Are we already being traced?
516 repeat:
517 task_lock(current);
518 if (!(current->ptrace & PT_PTRACED)) {
520 * See ptrace_attach() comments about the locking here.
522 unsigned long flags;
523 if (!write_trylock_irqsave(&tasklist_lock, flags)) {
524 task_unlock(current);
525 do {
526 cpu_relax();
527 } while (!write_can_lock(&tasklist_lock));
528 goto repeat;
531 ret = security_ptrace_traceme(current->parent);
534 * Set the ptrace bit in the process ptrace flags.
535 * Then link us on our parent's ptraced list.
537 if (!ret) {
538 current->ptrace |= PT_PTRACED;
539 __ptrace_link(current, current->real_parent);
542 write_unlock_irqrestore(&tasklist_lock, flags);
544 task_unlock(current);
545 return ret;
549 * ptrace_get_task_struct -- grab a task struct reference for ptrace
550 * @pid: process id to grab a task_struct reference of
552 * This function is a helper for ptrace implementations. It checks
553 * permissions and then grabs a task struct for use of the actual
554 * ptrace implementation.
556 * Returns the task_struct for @pid or an ERR_PTR() on failure.
558 struct task_struct *ptrace_get_task_struct(pid_t pid)
560 struct task_struct *child;
562 read_lock(&tasklist_lock);
563 child = find_task_by_vpid(pid);
564 if (child)
565 get_task_struct(child);
567 read_unlock(&tasklist_lock);
568 if (!child)
569 return ERR_PTR(-ESRCH);
570 return child;
573 #ifndef arch_ptrace_attach
574 #define arch_ptrace_attach(child) do { } while (0)
575 #endif
577 SYSCALL_DEFINE4(ptrace, long, request, long, pid, long, addr, long, data)
579 struct task_struct *child;
580 long ret;
583 * This lock_kernel fixes a subtle race with suid exec
585 lock_kernel();
586 if (request == PTRACE_TRACEME) {
587 ret = ptrace_traceme();
588 if (!ret)
589 arch_ptrace_attach(current);
590 goto out;
593 child = ptrace_get_task_struct(pid);
594 if (IS_ERR(child)) {
595 ret = PTR_ERR(child);
596 goto out;
599 if (request == PTRACE_ATTACH) {
600 ret = ptrace_attach(child);
602 * Some architectures need to do book-keeping after
603 * a ptrace attach.
605 if (!ret)
606 arch_ptrace_attach(child);
607 goto out_put_task_struct;
610 ret = ptrace_check_attach(child, request == PTRACE_KILL);
611 if (ret < 0)
612 goto out_put_task_struct;
614 ret = arch_ptrace(child, request, addr, data);
615 if (ret < 0)
616 goto out_put_task_struct;
618 out_put_task_struct:
619 put_task_struct(child);
620 out:
621 unlock_kernel();
622 return ret;
625 int generic_ptrace_peekdata(struct task_struct *tsk, long addr, long data)
627 unsigned long tmp;
628 int copied;
630 copied = access_process_vm(tsk, addr, &tmp, sizeof(tmp), 0);
631 if (copied != sizeof(tmp))
632 return -EIO;
633 return put_user(tmp, (unsigned long __user *)data);
636 int generic_ptrace_pokedata(struct task_struct *tsk, long addr, long data)
638 int copied;
640 copied = access_process_vm(tsk, addr, &data, sizeof(data), 1);
641 return (copied == sizeof(data)) ? 0 : -EIO;
644 #if defined CONFIG_COMPAT
645 #include <linux/compat.h>
647 int compat_ptrace_request(struct task_struct *child, compat_long_t request,
648 compat_ulong_t addr, compat_ulong_t data)
650 compat_ulong_t __user *datap = compat_ptr(data);
651 compat_ulong_t word;
652 siginfo_t siginfo;
653 int ret;
655 switch (request) {
656 case PTRACE_PEEKTEXT:
657 case PTRACE_PEEKDATA:
658 ret = access_process_vm(child, addr, &word, sizeof(word), 0);
659 if (ret != sizeof(word))
660 ret = -EIO;
661 else
662 ret = put_user(word, datap);
663 break;
665 case PTRACE_POKETEXT:
666 case PTRACE_POKEDATA:
667 ret = access_process_vm(child, addr, &data, sizeof(data), 1);
668 ret = (ret != sizeof(data) ? -EIO : 0);
669 break;
671 case PTRACE_GETEVENTMSG:
672 ret = put_user((compat_ulong_t) child->ptrace_message, datap);
673 break;
675 case PTRACE_GETSIGINFO:
676 ret = ptrace_getsiginfo(child, &siginfo);
677 if (!ret)
678 ret = copy_siginfo_to_user32(
679 (struct compat_siginfo __user *) datap,
680 &siginfo);
681 break;
683 case PTRACE_SETSIGINFO:
684 memset(&siginfo, 0, sizeof siginfo);
685 if (copy_siginfo_from_user32(
686 &siginfo, (struct compat_siginfo __user *) datap))
687 ret = -EFAULT;
688 else
689 ret = ptrace_setsiginfo(child, &siginfo);
690 break;
692 default:
693 ret = ptrace_request(child, request, addr, data);
696 return ret;
699 asmlinkage long compat_sys_ptrace(compat_long_t request, compat_long_t pid,
700 compat_long_t addr, compat_long_t data)
702 struct task_struct *child;
703 long ret;
706 * This lock_kernel fixes a subtle race with suid exec
708 lock_kernel();
709 if (request == PTRACE_TRACEME) {
710 ret = ptrace_traceme();
711 goto out;
714 child = ptrace_get_task_struct(pid);
715 if (IS_ERR(child)) {
716 ret = PTR_ERR(child);
717 goto out;
720 if (request == PTRACE_ATTACH) {
721 ret = ptrace_attach(child);
723 * Some architectures need to do book-keeping after
724 * a ptrace attach.
726 if (!ret)
727 arch_ptrace_attach(child);
728 goto out_put_task_struct;
731 ret = ptrace_check_attach(child, request == PTRACE_KILL);
732 if (!ret)
733 ret = compat_arch_ptrace(child, request, addr, data);
735 out_put_task_struct:
736 put_task_struct(child);
737 out:
738 unlock_kernel();
739 return ret;
741 #endif /* CONFIG_COMPAT */