Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland...
[linux-2.6-block.git] / kernel / ptrace.c
CommitLineData
1da177e4
LT
1/*
2 * linux/kernel/ptrace.c
3 *
4 * (C) Copyright 1999 Linus Torvalds
5 *
6 * Common interfaces for "ptrace()" which we do not want
7 * to continually duplicate across every architecture.
8 */
9
c59ede7b 10#include <linux/capability.h>
1da177e4
LT
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>
7ed20e1a 20#include <linux/signal.h>
1da177e4
LT
21
22#include <asm/pgtable.h>
23#include <asm/uaccess.h>
24
25/*
26 * ptrace a task: make the debugger its new parent and
27 * move it to the ptrace list.
28 *
29 * Must be called with the tasklist lock write-held.
30 */
31void __ptrace_link(task_t *child, task_t *new_parent)
32{
524223ca 33 BUG_ON(!list_empty(&child->ptrace_list));
1da177e4
LT
34 if (child->parent == new_parent)
35 return;
36 list_add(&child->ptrace_list, &child->parent->ptrace_children);
9b678ece 37 remove_parent(child);
1da177e4 38 child->parent = new_parent;
9b678ece 39 add_parent(child);
1da177e4
LT
40}
41
42/*
43 * Turn a tracing stop into a normal stop now, since with no tracer there
44 * would be no way to wake it up with SIGCONT or SIGKILL. If there was a
45 * signal sent that would resume the child, but didn't because it was in
46 * TASK_TRACED, resume it now.
47 * Requires that irqs be disabled.
48 */
49void ptrace_untrace(task_t *child)
50{
51 spin_lock(&child->sighand->siglock);
52 if (child->state == TASK_TRACED) {
53 if (child->signal->flags & SIGNAL_STOP_STOPPED) {
54 child->state = TASK_STOPPED;
55 } else {
56 signal_wake_up(child, 1);
57 }
58 }
59 spin_unlock(&child->sighand->siglock);
60}
61
62/*
63 * unptrace a task: move it back to its original parent and
64 * remove it from the ptrace list.
65 *
66 * Must be called with the tasklist lock write-held.
67 */
68void __ptrace_unlink(task_t *child)
69{
5ecfbae0
ON
70 BUG_ON(!child->ptrace);
71
1da177e4
LT
72 child->ptrace = 0;
73 if (!list_empty(&child->ptrace_list)) {
74 list_del_init(&child->ptrace_list);
9b678ece 75 remove_parent(child);
1da177e4 76 child->parent = child->real_parent;
9b678ece 77 add_parent(child);
1da177e4
LT
78 }
79
e57a5059
RM
80 if (child->state == TASK_TRACED)
81 ptrace_untrace(child);
1da177e4
LT
82}
83
84/*
85 * Check that we have indeed attached to the thing..
86 */
87int ptrace_check_attach(struct task_struct *child, int kill)
88{
89 int ret = -ESRCH;
90
91 /*
92 * We take the read lock around doing both checks to close a
93 * possible race where someone else was tracing our child and
94 * detached between these two checks. After this locked check,
95 * we are sure that this is our traced child and that can only
96 * be changed by us so it's not changing right after this.
97 */
98 read_lock(&tasklist_lock);
99 if ((child->ptrace & PT_PTRACED) && child->parent == current &&
100 (!(child->ptrace & PT_ATTACHED) || child->real_parent != current)
101 && child->signal != NULL) {
102 ret = 0;
103 spin_lock_irq(&child->sighand->siglock);
104 if (child->state == TASK_STOPPED) {
105 child->state = TASK_TRACED;
106 } else if (child->state != TASK_TRACED && !kill) {
107 ret = -ESRCH;
108 }
109 spin_unlock_irq(&child->sighand->siglock);
110 }
111 read_unlock(&tasklist_lock);
112
113 if (!ret && !kill) {
114 wait_task_inactive(child);
115 }
116
117 /* All systems go.. */
118 return ret;
119}
120
ab8d11be
MS
121static int may_attach(struct task_struct *task)
122{
123 if (!task->mm)
124 return -EPERM;
125 if (((current->uid != task->euid) ||
126 (current->uid != task->suid) ||
127 (current->uid != task->uid) ||
128 (current->gid != task->egid) ||
129 (current->gid != task->sgid) ||
130 (current->gid != task->gid)) && !capable(CAP_SYS_PTRACE))
131 return -EPERM;
132 smp_rmb();
133 if (!task->mm->dumpable && !capable(CAP_SYS_PTRACE))
134 return -EPERM;
135
136 return security_ptrace(current, task);
137}
138
139int ptrace_may_attach(struct task_struct *task)
140{
141 int err;
142 task_lock(task);
143 err = may_attach(task);
144 task_unlock(task);
145 return !err;
146}
147
1da177e4
LT
148int ptrace_attach(struct task_struct *task)
149{
150 int retval;
f5b40e36 151
1da177e4
LT
152 retval = -EPERM;
153 if (task->pid <= 1)
f5b40e36 154 goto out;
28d838cc 155 if (task->tgid == current->tgid)
f5b40e36
LT
156 goto out;
157
158 write_lock_irq(&tasklist_lock);
159 task_lock(task);
160
1da177e4
LT
161 /* the same process cannot be attached many times */
162 if (task->ptrace & PT_PTRACED)
163 goto bad;
ab8d11be 164 retval = may_attach(task);
1da177e4
LT
165 if (retval)
166 goto bad;
167
168 /* Go */
169 task->ptrace |= PT_PTRACED | ((task->real_parent != current)
170 ? PT_ATTACHED : 0);
171 if (capable(CAP_SYS_PTRACE))
172 task->ptrace |= PT_PTRACE_CAP;
1da177e4 173
1da177e4 174 __ptrace_link(task, current);
1da177e4
LT
175
176 force_sig_specific(SIGSTOP, task);
1da177e4
LT
177
178bad:
f5b40e36 179 write_unlock_irq(&tasklist_lock);
1da177e4 180 task_unlock(task);
f5b40e36 181out:
1da177e4
LT
182 return retval;
183}
184
5ecfbae0
ON
185void __ptrace_detach(struct task_struct *child, unsigned int data)
186{
187 child->exit_code = data;
188 /* .. re-parent .. */
189 __ptrace_unlink(child);
190 /* .. and wake it up. */
191 if (child->exit_state != EXIT_ZOMBIE)
192 wake_up_process(child);
193}
194
1da177e4
LT
195int ptrace_detach(struct task_struct *child, unsigned int data)
196{
7ed20e1a 197 if (!valid_signal(data))
5ecfbae0 198 return -EIO;
1da177e4
LT
199
200 /* Architecture-specific hardware disable .. */
201 ptrace_disable(child);
202
1da177e4 203 write_lock_irq(&tasklist_lock);
5ecfbae0
ON
204 if (child->ptrace)
205 __ptrace_detach(child, data);
1da177e4
LT
206 write_unlock_irq(&tasklist_lock);
207
208 return 0;
209}
210
211/*
212 * Access another process' address space.
213 * Source/target buffer must be kernel space,
214 * Do not walk the page table directly, use get_user_pages
215 */
216
217int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write)
218{
219 struct mm_struct *mm;
220 struct vm_area_struct *vma;
221 struct page *page;
222 void *old_buf = buf;
223
224 mm = get_task_mm(tsk);
225 if (!mm)
226 return 0;
227
228 down_read(&mm->mmap_sem);
229 /* ignore errors, just check how much was sucessfully transfered */
230 while (len) {
231 int bytes, ret, offset;
232 void *maddr;
233
234 ret = get_user_pages(tsk, mm, addr, 1,
235 write, 1, &page, &vma);
236 if (ret <= 0)
237 break;
238
239 bytes = len;
240 offset = addr & (PAGE_SIZE-1);
241 if (bytes > PAGE_SIZE-offset)
242 bytes = PAGE_SIZE-offset;
243
244 maddr = kmap(page);
245 if (write) {
246 copy_to_user_page(vma, page, addr,
247 maddr + offset, buf, bytes);
16bf1348 248 set_page_dirty_lock(page);
1da177e4
LT
249 } else {
250 copy_from_user_page(vma, page, addr,
251 buf, maddr + offset, bytes);
252 }
253 kunmap(page);
254 page_cache_release(page);
255 len -= bytes;
256 buf += bytes;
257 addr += bytes;
258 }
259 up_read(&mm->mmap_sem);
260 mmput(mm);
261
262 return buf - old_buf;
263}
264
265int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
266{
267 int copied = 0;
268
269 while (len > 0) {
270 char buf[128];
271 int this_len, retval;
272
273 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
274 retval = access_process_vm(tsk, src, buf, this_len, 0);
275 if (!retval) {
276 if (copied)
277 break;
278 return -EIO;
279 }
280 if (copy_to_user(dst, buf, retval))
281 return -EFAULT;
282 copied += retval;
283 src += retval;
284 dst += retval;
285 len -= retval;
286 }
287 return copied;
288}
289
290int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
291{
292 int copied = 0;
293
294 while (len > 0) {
295 char buf[128];
296 int this_len, retval;
297
298 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
299 if (copy_from_user(buf, src, this_len))
300 return -EFAULT;
301 retval = access_process_vm(tsk, dst, buf, this_len, 1);
302 if (!retval) {
303 if (copied)
304 break;
305 return -EIO;
306 }
307 copied += retval;
308 src += retval;
309 dst += retval;
310 len -= retval;
311 }
312 return copied;
313}
314
315static int ptrace_setoptions(struct task_struct *child, long data)
316{
317 child->ptrace &= ~PT_TRACE_MASK;
318
319 if (data & PTRACE_O_TRACESYSGOOD)
320 child->ptrace |= PT_TRACESYSGOOD;
321
322 if (data & PTRACE_O_TRACEFORK)
323 child->ptrace |= PT_TRACE_FORK;
324
325 if (data & PTRACE_O_TRACEVFORK)
326 child->ptrace |= PT_TRACE_VFORK;
327
328 if (data & PTRACE_O_TRACECLONE)
329 child->ptrace |= PT_TRACE_CLONE;
330
331 if (data & PTRACE_O_TRACEEXEC)
332 child->ptrace |= PT_TRACE_EXEC;
333
334 if (data & PTRACE_O_TRACEVFORKDONE)
335 child->ptrace |= PT_TRACE_VFORK_DONE;
336
337 if (data & PTRACE_O_TRACEEXIT)
338 child->ptrace |= PT_TRACE_EXIT;
339
340 return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
341}
342
343static int ptrace_getsiginfo(struct task_struct *child, siginfo_t __user * data)
344{
345 siginfo_t lastinfo;
346 int error = -ESRCH;
347
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 lastinfo = *child->last_siginfo;
354 error = 0;
355 }
356 spin_unlock_irq(&child->sighand->siglock);
357 }
358 read_unlock(&tasklist_lock);
359 if (!error)
360 return copy_siginfo_to_user(data, &lastinfo);
361 return error;
362}
363
364static int ptrace_setsiginfo(struct task_struct *child, siginfo_t __user * data)
365{
366 siginfo_t newinfo;
367 int error = -ESRCH;
368
369 if (copy_from_user(&newinfo, data, sizeof (siginfo_t)))
370 return -EFAULT;
371
372 read_lock(&tasklist_lock);
373 if (likely(child->sighand != NULL)) {
374 error = -EINVAL;
375 spin_lock_irq(&child->sighand->siglock);
376 if (likely(child->last_siginfo != NULL)) {
377 *child->last_siginfo = newinfo;
378 error = 0;
379 }
380 spin_unlock_irq(&child->sighand->siglock);
381 }
382 read_unlock(&tasklist_lock);
383 return error;
384}
385
386int ptrace_request(struct task_struct *child, long request,
387 long addr, long data)
388{
389 int ret = -EIO;
390
391 switch (request) {
392#ifdef PTRACE_OLDSETOPTIONS
393 case PTRACE_OLDSETOPTIONS:
394#endif
395 case PTRACE_SETOPTIONS:
396 ret = ptrace_setoptions(child, data);
397 break;
398 case PTRACE_GETEVENTMSG:
399 ret = put_user(child->ptrace_message, (unsigned long __user *) data);
400 break;
401 case PTRACE_GETSIGINFO:
402 ret = ptrace_getsiginfo(child, (siginfo_t __user *) data);
403 break;
404 case PTRACE_SETSIGINFO:
405 ret = ptrace_setsiginfo(child, (siginfo_t __user *) data);
406 break;
407 default:
408 break;
409 }
410
411 return ret;
412}
481bed45 413
6b9c7ed8
CH
414/**
415 * ptrace_traceme -- helper for PTRACE_TRACEME
416 *
417 * Performs checks and sets PT_PTRACED.
418 * Should be used by all ptrace implementations for PTRACE_TRACEME.
419 */
420int ptrace_traceme(void)
481bed45 421{
f5b40e36 422 int ret = -EPERM;
481bed45
CH
423
424 /*
6b9c7ed8
CH
425 * Are we already being traced?
426 */
f5b40e36
LT
427 task_lock(current);
428 if (!(current->ptrace & PT_PTRACED)) {
429 ret = security_ptrace(current->parent, current);
430 /*
431 * Set the ptrace bit in the process ptrace flags.
432 */
433 if (!ret)
434 current->ptrace |= PT_PTRACED;
435 }
436 task_unlock(current);
437 return ret;
6b9c7ed8 438}
481bed45 439
6b9c7ed8
CH
440/**
441 * ptrace_get_task_struct -- grab a task struct reference for ptrace
442 * @pid: process id to grab a task_struct reference of
443 *
444 * This function is a helper for ptrace implementations. It checks
445 * permissions and then grabs a task struct for use of the actual
446 * ptrace implementation.
447 *
448 * Returns the task_struct for @pid or an ERR_PTR() on failure.
449 */
450struct task_struct *ptrace_get_task_struct(pid_t pid)
451{
452 struct task_struct *child;
481bed45
CH
453
454 /*
6b9c7ed8 455 * Tracing init is not allowed.
481bed45
CH
456 */
457 if (pid == 1)
6b9c7ed8 458 return ERR_PTR(-EPERM);
481bed45 459
481bed45
CH
460 read_lock(&tasklist_lock);
461 child = find_task_by_pid(pid);
462 if (child)
463 get_task_struct(child);
464 read_unlock(&tasklist_lock);
465 if (!child)
6b9c7ed8
CH
466 return ERR_PTR(-ESRCH);
467 return child;
481bed45
CH
468}
469
6b9c7ed8 470#ifndef __ARCH_SYS_PTRACE
481bed45
CH
471asmlinkage long sys_ptrace(long request, long pid, long addr, long data)
472{
473 struct task_struct *child;
474 long ret;
475
476 /*
477 * This lock_kernel fixes a subtle race with suid exec
478 */
479 lock_kernel();
6b9c7ed8
CH
480 if (request == PTRACE_TRACEME) {
481 ret = ptrace_traceme();
481bed45 482 goto out;
6b9c7ed8
CH
483 }
484
485 child = ptrace_get_task_struct(pid);
486 if (IS_ERR(child)) {
487 ret = PTR_ERR(child);
488 goto out;
489 }
481bed45
CH
490
491 if (request == PTRACE_ATTACH) {
492 ret = ptrace_attach(child);
005f18df 493 goto out_put_task_struct;
481bed45
CH
494 }
495
496 ret = ptrace_check_attach(child, request == PTRACE_KILL);
497 if (ret < 0)
498 goto out_put_task_struct;
499
500 ret = arch_ptrace(child, request, addr, data);
501 if (ret < 0)
502 goto out_put_task_struct;
503
504 out_put_task_struct:
505 put_task_struct(child);
506 out:
507 unlock_kernel();
508 return ret;
509}
510#endif /* __ARCH_SYS_PTRACE */