pid namespaces: changes to show virtual ids to user
[GitHub/mt8127/android_kernel_alcatel_ttab.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>
a5cb013d 21#include <linux/audit.h>
b488893a 22#include <linux/pid_namespace.h>
1da177e4
LT
23
24#include <asm/pgtable.h>
25#include <asm/uaccess.h>
26
27/*
28 * ptrace a task: make the debugger its new parent and
29 * move it to the ptrace list.
30 *
31 * Must be called with the tasklist lock write-held.
32 */
36c8b586 33void __ptrace_link(struct task_struct *child, struct task_struct *new_parent)
1da177e4 34{
524223ca 35 BUG_ON(!list_empty(&child->ptrace_list));
1da177e4
LT
36 if (child->parent == new_parent)
37 return;
38 list_add(&child->ptrace_list, &child->parent->ptrace_children);
9b678ece 39 remove_parent(child);
1da177e4 40 child->parent = new_parent;
9b678ece 41 add_parent(child);
1da177e4
LT
42}
43
44/*
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.
50 */
36c8b586 51void ptrace_untrace(struct task_struct *child)
1da177e4
LT
52{
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;
57 } else {
58 signal_wake_up(child, 1);
59 }
60 }
61 spin_unlock(&child->sighand->siglock);
62}
63
64/*
65 * unptrace a task: move it back to its original parent and
66 * remove it from the ptrace list.
67 *
68 * Must be called with the tasklist lock write-held.
69 */
36c8b586 70void __ptrace_unlink(struct task_struct *child)
1da177e4 71{
5ecfbae0
ON
72 BUG_ON(!child->ptrace);
73
1da177e4
LT
74 child->ptrace = 0;
75 if (!list_empty(&child->ptrace_list)) {
76 list_del_init(&child->ptrace_list);
9b678ece 77 remove_parent(child);
1da177e4 78 child->parent = child->real_parent;
9b678ece 79 add_parent(child);
1da177e4
LT
80 }
81
e57a5059
RM
82 if (child->state == TASK_TRACED)
83 ptrace_untrace(child);
1da177e4
LT
84}
85
86/*
87 * Check that we have indeed attached to the thing..
88 */
89int ptrace_check_attach(struct task_struct *child, int kill)
90{
91 int ret = -ESRCH;
92
93 /*
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.
99 */
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) {
104 ret = 0;
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) {
109 ret = -ESRCH;
110 }
111 spin_unlock_irq(&child->sighand->siglock);
112 }
113 read_unlock(&tasklist_lock);
114
115 if (!ret && !kill) {
116 wait_task_inactive(child);
117 }
118
119 /* All systems go.. */
120 return ret;
121}
122
ab8d11be
MS
123static int may_attach(struct task_struct *task)
124{
df26c40e
EB
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.
128 *
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.
132 */
133 int dumpable = 0;
134 /* Don't let security modules deny introspection */
135 if (task == current)
136 return 0;
ab8d11be
MS
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))
143 return -EPERM;
144 smp_rmb();
df26c40e 145 if (task->mm)
6c5d5238 146 dumpable = get_dumpable(task->mm);
df26c40e 147 if (!dumpable && !capable(CAP_SYS_PTRACE))
ab8d11be
MS
148 return -EPERM;
149
150 return security_ptrace(current, task);
151}
152
153int ptrace_may_attach(struct task_struct *task)
154{
155 int err;
156 task_lock(task);
157 err = may_attach(task);
158 task_unlock(task);
159 return !err;
160}
161
1da177e4
LT
162int ptrace_attach(struct task_struct *task)
163{
164 int retval;
6175ecfe 165 unsigned long flags;
f5b40e36 166
a5cb013d
AV
167 audit_ptrace(task);
168
1da177e4
LT
169 retval = -EPERM;
170 if (task->pid <= 1)
f5b40e36 171 goto out;
28d838cc 172 if (task->tgid == current->tgid)
f5b40e36
LT
173 goto out;
174
f358166a
LT
175repeat:
176 /*
177 * Nasty, nasty.
178 *
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).
184 */
f5b40e36 185 task_lock(task);
6175ecfe 186 if (!write_trylock_irqsave(&tasklist_lock, flags)) {
f358166a
LT
187 task_unlock(task);
188 do {
189 cpu_relax();
190 } while (!write_can_lock(&tasklist_lock));
191 goto repeat;
192 }
f5b40e36 193
df26c40e
EB
194 if (!task->mm)
195 goto bad;
1da177e4
LT
196 /* the same process cannot be attached many times */
197 if (task->ptrace & PT_PTRACED)
198 goto bad;
ab8d11be 199 retval = may_attach(task);
1da177e4
LT
200 if (retval)
201 goto bad;
202
203 /* Go */
204 task->ptrace |= PT_PTRACED | ((task->real_parent != current)
205 ? PT_ATTACHED : 0);
206 if (capable(CAP_SYS_PTRACE))
207 task->ptrace |= PT_PTRACE_CAP;
1da177e4 208
1da177e4 209 __ptrace_link(task, current);
1da177e4
LT
210
211 force_sig_specific(SIGSTOP, task);
1da177e4
LT
212
213bad:
6175ecfe 214 write_unlock_irqrestore(&tasklist_lock, flags);
1da177e4 215 task_unlock(task);
f5b40e36 216out:
1da177e4
LT
217 return retval;
218}
219
d5f70c00 220static inline void __ptrace_detach(struct task_struct *child, unsigned int data)
5ecfbae0
ON
221{
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);
228}
229
1da177e4
LT
230int ptrace_detach(struct task_struct *child, unsigned int data)
231{
7ed20e1a 232 if (!valid_signal(data))
5ecfbae0 233 return -EIO;
1da177e4
LT
234
235 /* Architecture-specific hardware disable .. */
236 ptrace_disable(child);
7d941432 237 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
1da177e4 238
1da177e4 239 write_lock_irq(&tasklist_lock);
d5f70c00 240 /* protect against de_thread()->release_task() */
5ecfbae0
ON
241 if (child->ptrace)
242 __ptrace_detach(child, data);
1da177e4
LT
243 write_unlock_irq(&tasklist_lock);
244
245 return 0;
246}
247
1da177e4
LT
248int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
249{
250 int copied = 0;
251
252 while (len > 0) {
253 char buf[128];
254 int this_len, retval;
255
256 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
257 retval = access_process_vm(tsk, src, buf, this_len, 0);
258 if (!retval) {
259 if (copied)
260 break;
261 return -EIO;
262 }
263 if (copy_to_user(dst, buf, retval))
264 return -EFAULT;
265 copied += retval;
266 src += retval;
267 dst += retval;
268 len -= retval;
269 }
270 return copied;
271}
272
273int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
274{
275 int copied = 0;
276
277 while (len > 0) {
278 char buf[128];
279 int this_len, retval;
280
281 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
282 if (copy_from_user(buf, src, this_len))
283 return -EFAULT;
284 retval = access_process_vm(tsk, dst, buf, this_len, 1);
285 if (!retval) {
286 if (copied)
287 break;
288 return -EIO;
289 }
290 copied += retval;
291 src += retval;
292 dst += retval;
293 len -= retval;
294 }
295 return copied;
296}
297
298static int ptrace_setoptions(struct task_struct *child, long data)
299{
300 child->ptrace &= ~PT_TRACE_MASK;
301
302 if (data & PTRACE_O_TRACESYSGOOD)
303 child->ptrace |= PT_TRACESYSGOOD;
304
305 if (data & PTRACE_O_TRACEFORK)
306 child->ptrace |= PT_TRACE_FORK;
307
308 if (data & PTRACE_O_TRACEVFORK)
309 child->ptrace |= PT_TRACE_VFORK;
310
311 if (data & PTRACE_O_TRACECLONE)
312 child->ptrace |= PT_TRACE_CLONE;
313
314 if (data & PTRACE_O_TRACEEXEC)
315 child->ptrace |= PT_TRACE_EXEC;
316
317 if (data & PTRACE_O_TRACEVFORKDONE)
318 child->ptrace |= PT_TRACE_VFORK_DONE;
319
320 if (data & PTRACE_O_TRACEEXIT)
321 child->ptrace |= PT_TRACE_EXIT;
322
323 return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
324}
325
326static int ptrace_getsiginfo(struct task_struct *child, siginfo_t __user * data)
327{
328 siginfo_t lastinfo;
329 int error = -ESRCH;
330
331 read_lock(&tasklist_lock);
332 if (likely(child->sighand != NULL)) {
333 error = -EINVAL;
334 spin_lock_irq(&child->sighand->siglock);
335 if (likely(child->last_siginfo != NULL)) {
336 lastinfo = *child->last_siginfo;
337 error = 0;
338 }
339 spin_unlock_irq(&child->sighand->siglock);
340 }
341 read_unlock(&tasklist_lock);
342 if (!error)
343 return copy_siginfo_to_user(data, &lastinfo);
344 return error;
345}
346
347static int ptrace_setsiginfo(struct task_struct *child, siginfo_t __user * data)
348{
349 siginfo_t newinfo;
350 int error = -ESRCH;
351
352 if (copy_from_user(&newinfo, data, sizeof (siginfo_t)))
353 return -EFAULT;
354
355 read_lock(&tasklist_lock);
356 if (likely(child->sighand != NULL)) {
357 error = -EINVAL;
358 spin_lock_irq(&child->sighand->siglock);
359 if (likely(child->last_siginfo != NULL)) {
360 *child->last_siginfo = newinfo;
361 error = 0;
362 }
363 spin_unlock_irq(&child->sighand->siglock);
364 }
365 read_unlock(&tasklist_lock);
366 return error;
367}
368
369int ptrace_request(struct task_struct *child, long request,
370 long addr, long data)
371{
372 int ret = -EIO;
373
374 switch (request) {
375#ifdef PTRACE_OLDSETOPTIONS
376 case PTRACE_OLDSETOPTIONS:
377#endif
378 case PTRACE_SETOPTIONS:
379 ret = ptrace_setoptions(child, data);
380 break;
381 case PTRACE_GETEVENTMSG:
382 ret = put_user(child->ptrace_message, (unsigned long __user *) data);
383 break;
384 case PTRACE_GETSIGINFO:
385 ret = ptrace_getsiginfo(child, (siginfo_t __user *) data);
386 break;
387 case PTRACE_SETSIGINFO:
388 ret = ptrace_setsiginfo(child, (siginfo_t __user *) data);
389 break;
1bcf5482
AD
390 case PTRACE_DETACH: /* detach a process that was attached. */
391 ret = ptrace_detach(child, data);
392 break;
1da177e4
LT
393 default:
394 break;
395 }
396
397 return ret;
398}
481bed45 399
6b9c7ed8
CH
400/**
401 * ptrace_traceme -- helper for PTRACE_TRACEME
402 *
403 * Performs checks and sets PT_PTRACED.
404 * Should be used by all ptrace implementations for PTRACE_TRACEME.
405 */
406int ptrace_traceme(void)
481bed45 407{
f5b40e36 408 int ret = -EPERM;
481bed45
CH
409
410 /*
6b9c7ed8
CH
411 * Are we already being traced?
412 */
f5b40e36
LT
413 task_lock(current);
414 if (!(current->ptrace & PT_PTRACED)) {
415 ret = security_ptrace(current->parent, current);
416 /*
417 * Set the ptrace bit in the process ptrace flags.
418 */
419 if (!ret)
420 current->ptrace |= PT_PTRACED;
421 }
422 task_unlock(current);
423 return ret;
6b9c7ed8 424}
481bed45 425
6b9c7ed8
CH
426/**
427 * ptrace_get_task_struct -- grab a task struct reference for ptrace
428 * @pid: process id to grab a task_struct reference of
429 *
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.
433 *
434 * Returns the task_struct for @pid or an ERR_PTR() on failure.
435 */
436struct task_struct *ptrace_get_task_struct(pid_t pid)
437{
438 struct task_struct *child;
481bed45
CH
439
440 /*
6b9c7ed8 441 * Tracing init is not allowed.
481bed45
CH
442 */
443 if (pid == 1)
6b9c7ed8 444 return ERR_PTR(-EPERM);
481bed45 445
481bed45 446 read_lock(&tasklist_lock);
b488893a
PE
447 child = find_task_by_pid_ns(pid,
448 current->nsproxy->pid_ns);
481bed45
CH
449 if (child)
450 get_task_struct(child);
f400e198 451
481bed45
CH
452 read_unlock(&tasklist_lock);
453 if (!child)
6b9c7ed8
CH
454 return ERR_PTR(-ESRCH);
455 return child;
481bed45
CH
456}
457
0ac15559
CH
458#ifndef arch_ptrace_attach
459#define arch_ptrace_attach(child) do { } while (0)
460#endif
461
6b9c7ed8 462#ifndef __ARCH_SYS_PTRACE
481bed45
CH
463asmlinkage long sys_ptrace(long request, long pid, long addr, long data)
464{
465 struct task_struct *child;
466 long ret;
467
468 /*
469 * This lock_kernel fixes a subtle race with suid exec
470 */
471 lock_kernel();
6b9c7ed8
CH
472 if (request == PTRACE_TRACEME) {
473 ret = ptrace_traceme();
481bed45 474 goto out;
6b9c7ed8
CH
475 }
476
477 child = ptrace_get_task_struct(pid);
478 if (IS_ERR(child)) {
479 ret = PTR_ERR(child);
480 goto out;
481 }
481bed45
CH
482
483 if (request == PTRACE_ATTACH) {
484 ret = ptrace_attach(child);
0ac15559
CH
485 /*
486 * Some architectures need to do book-keeping after
487 * a ptrace attach.
488 */
489 if (!ret)
490 arch_ptrace_attach(child);
005f18df 491 goto out_put_task_struct;
481bed45
CH
492 }
493
494 ret = ptrace_check_attach(child, request == PTRACE_KILL);
495 if (ret < 0)
496 goto out_put_task_struct;
497
498 ret = arch_ptrace(child, request, addr, data);
499 if (ret < 0)
500 goto out_put_task_struct;
501
502 out_put_task_struct:
503 put_task_struct(child);
504 out:
505 unlock_kernel();
506 return ret;
507}
508#endif /* __ARCH_SYS_PTRACE */
76647323
AD
509
510int generic_ptrace_peekdata(struct task_struct *tsk, long addr, long data)
511{
512 unsigned long tmp;
513 int copied;
514
515 copied = access_process_vm(tsk, addr, &tmp, sizeof(tmp), 0);
516 if (copied != sizeof(tmp))
517 return -EIO;
518 return put_user(tmp, (unsigned long __user *)data);
519}
f284ce72
AD
520
521int generic_ptrace_pokedata(struct task_struct *tsk, long addr, long data)
522{
523 int copied;
524
525 copied = access_process_vm(tsk, addr, &data, sizeof(data), 1);
526 return (copied == sizeof(data)) ? 0 : -EIO;
527}