[XFS] stop using uio in the readlink code
[linux-2.6/linux-2.6-openrd.git] / kernel / ptrace.c
blob3eca7a55f2ee8b13fb2fb6f3e09ea2d7a685acaf
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>
23 #include <asm/pgtable.h>
24 #include <asm/uaccess.h>
27 * ptrace a task: make the debugger its new parent and
28 * move it to the ptrace list.
30 * Must be called with the tasklist lock write-held.
32 void __ptrace_link(struct task_struct *child, struct task_struct *new_parent)
34 BUG_ON(!list_empty(&child->ptrace_list));
35 if (child->parent == new_parent)
36 return;
37 list_add(&child->ptrace_list, &child->parent->ptrace_children);
38 remove_parent(child);
39 child->parent = new_parent;
40 add_parent(child);
44 * Turn a tracing stop into a normal stop now, since with no tracer there
45 * would be no way to wake it up with SIGCONT or SIGKILL. If there was a
46 * signal sent that would resume the child, but didn't because it was in
47 * TASK_TRACED, resume it now.
48 * Requires that irqs be disabled.
50 void ptrace_untrace(struct task_struct *child)
52 spin_lock(&child->sighand->siglock);
53 if (child->state == TASK_TRACED) {
54 if (child->signal->flags & SIGNAL_STOP_STOPPED) {
55 child->state = TASK_STOPPED;
56 } else {
57 signal_wake_up(child, 1);
60 spin_unlock(&child->sighand->siglock);
64 * unptrace a task: move it back to its original parent and
65 * remove it from the ptrace list.
67 * Must be called with the tasklist lock write-held.
69 void __ptrace_unlink(struct task_struct *child)
71 BUG_ON(!child->ptrace);
73 child->ptrace = 0;
74 if (!list_empty(&child->ptrace_list)) {
75 list_del_init(&child->ptrace_list);
76 remove_parent(child);
77 child->parent = child->real_parent;
78 add_parent(child);
81 if (child->state == TASK_TRACED)
82 ptrace_untrace(child);
86 * Check that we have indeed attached to the thing..
88 int ptrace_check_attach(struct task_struct *child, int kill)
90 int ret = -ESRCH;
93 * We take the read lock around doing both checks to close a
94 * possible race where someone else was tracing our child and
95 * detached between these two checks. After this locked check,
96 * we are sure that this is our traced child and that can only
97 * be changed by us so it's not changing right after this.
99 read_lock(&tasklist_lock);
100 if ((child->ptrace & PT_PTRACED) && child->parent == current &&
101 (!(child->ptrace & PT_ATTACHED) || child->real_parent != current)
102 && child->signal != NULL) {
103 ret = 0;
104 spin_lock_irq(&child->sighand->siglock);
105 if (child->state == TASK_STOPPED) {
106 child->state = TASK_TRACED;
107 } else if (child->state != TASK_TRACED && !kill) {
108 ret = -ESRCH;
110 spin_unlock_irq(&child->sighand->siglock);
112 read_unlock(&tasklist_lock);
114 if (!ret && !kill) {
115 wait_task_inactive(child);
118 /* All systems go.. */
119 return ret;
122 static int may_attach(struct task_struct *task)
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.
132 int dumpable = 0;
133 /* Don't let security modules deny introspection */
134 if (task == current)
135 return 0;
136 if (((current->uid != task->euid) ||
137 (current->uid != task->suid) ||
138 (current->uid != task->uid) ||
139 (current->gid != task->egid) ||
140 (current->gid != task->sgid) ||
141 (current->gid != task->gid)) && !capable(CAP_SYS_PTRACE))
142 return -EPERM;
143 smp_rmb();
144 if (task->mm)
145 dumpable = get_dumpable(task->mm);
146 if (!dumpable && !capable(CAP_SYS_PTRACE))
147 return -EPERM;
149 return security_ptrace(current, task);
152 int ptrace_may_attach(struct task_struct *task)
154 int err;
155 task_lock(task);
156 err = may_attach(task);
157 task_unlock(task);
158 return !err;
161 int ptrace_attach(struct task_struct *task)
163 int retval;
164 unsigned long flags;
166 audit_ptrace(task);
168 retval = -EPERM;
169 if (task->pid <= 1)
170 goto out;
171 if (task->tgid == current->tgid)
172 goto out;
174 repeat:
176 * Nasty, nasty.
178 * We want to hold both the task-lock and the
179 * tasklist_lock for writing at the same time.
180 * But that's against the rules (tasklist_lock
181 * is taken for reading by interrupts on other
182 * cpu's that may have task_lock).
184 task_lock(task);
185 if (!write_trylock_irqsave(&tasklist_lock, flags)) {
186 task_unlock(task);
187 do {
188 cpu_relax();
189 } while (!write_can_lock(&tasklist_lock));
190 goto repeat;
193 if (!task->mm)
194 goto bad;
195 /* the same process cannot be attached many times */
196 if (task->ptrace & PT_PTRACED)
197 goto bad;
198 retval = may_attach(task);
199 if (retval)
200 goto bad;
202 /* Go */
203 task->ptrace |= PT_PTRACED | ((task->real_parent != current)
204 ? PT_ATTACHED : 0);
205 if (capable(CAP_SYS_PTRACE))
206 task->ptrace |= PT_PTRACE_CAP;
208 __ptrace_link(task, current);
210 force_sig_specific(SIGSTOP, task);
212 bad:
213 write_unlock_irqrestore(&tasklist_lock, flags);
214 task_unlock(task);
215 out:
216 return retval;
219 static inline void __ptrace_detach(struct task_struct *child, unsigned int data)
221 child->exit_code = data;
222 /* .. re-parent .. */
223 __ptrace_unlink(child);
224 /* .. and wake it up. */
225 if (child->exit_state != EXIT_ZOMBIE)
226 wake_up_process(child);
229 int ptrace_detach(struct task_struct *child, unsigned int data)
231 if (!valid_signal(data))
232 return -EIO;
234 /* Architecture-specific hardware disable .. */
235 ptrace_disable(child);
236 clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
238 write_lock_irq(&tasklist_lock);
239 /* protect against de_thread()->release_task() */
240 if (child->ptrace)
241 __ptrace_detach(child, data);
242 write_unlock_irq(&tasklist_lock);
244 return 0;
247 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
249 int copied = 0;
251 while (len > 0) {
252 char buf[128];
253 int this_len, retval;
255 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
256 retval = access_process_vm(tsk, src, buf, this_len, 0);
257 if (!retval) {
258 if (copied)
259 break;
260 return -EIO;
262 if (copy_to_user(dst, buf, retval))
263 return -EFAULT;
264 copied += retval;
265 src += retval;
266 dst += retval;
267 len -= retval;
269 return copied;
272 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
274 int copied = 0;
276 while (len > 0) {
277 char buf[128];
278 int this_len, retval;
280 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
281 if (copy_from_user(buf, src, this_len))
282 return -EFAULT;
283 retval = access_process_vm(tsk, dst, buf, this_len, 1);
284 if (!retval) {
285 if (copied)
286 break;
287 return -EIO;
289 copied += retval;
290 src += retval;
291 dst += retval;
292 len -= retval;
294 return copied;
297 static int ptrace_setoptions(struct task_struct *child, long data)
299 child->ptrace &= ~PT_TRACE_MASK;
301 if (data & PTRACE_O_TRACESYSGOOD)
302 child->ptrace |= PT_TRACESYSGOOD;
304 if (data & PTRACE_O_TRACEFORK)
305 child->ptrace |= PT_TRACE_FORK;
307 if (data & PTRACE_O_TRACEVFORK)
308 child->ptrace |= PT_TRACE_VFORK;
310 if (data & PTRACE_O_TRACECLONE)
311 child->ptrace |= PT_TRACE_CLONE;
313 if (data & PTRACE_O_TRACEEXEC)
314 child->ptrace |= PT_TRACE_EXEC;
316 if (data & PTRACE_O_TRACEVFORKDONE)
317 child->ptrace |= PT_TRACE_VFORK_DONE;
319 if (data & PTRACE_O_TRACEEXIT)
320 child->ptrace |= PT_TRACE_EXIT;
322 return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
325 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t __user * data)
327 siginfo_t lastinfo;
328 int error = -ESRCH;
330 read_lock(&tasklist_lock);
331 if (likely(child->sighand != NULL)) {
332 error = -EINVAL;
333 spin_lock_irq(&child->sighand->siglock);
334 if (likely(child->last_siginfo != NULL)) {
335 lastinfo = *child->last_siginfo;
336 error = 0;
338 spin_unlock_irq(&child->sighand->siglock);
340 read_unlock(&tasklist_lock);
341 if (!error)
342 return copy_siginfo_to_user(data, &lastinfo);
343 return error;
346 static int ptrace_setsiginfo(struct task_struct *child, siginfo_t __user * data)
348 siginfo_t newinfo;
349 int error = -ESRCH;
351 if (copy_from_user(&newinfo, data, sizeof (siginfo_t)))
352 return -EFAULT;
354 read_lock(&tasklist_lock);
355 if (likely(child->sighand != NULL)) {
356 error = -EINVAL;
357 spin_lock_irq(&child->sighand->siglock);
358 if (likely(child->last_siginfo != NULL)) {
359 *child->last_siginfo = newinfo;
360 error = 0;
362 spin_unlock_irq(&child->sighand->siglock);
364 read_unlock(&tasklist_lock);
365 return error;
368 int ptrace_request(struct task_struct *child, long request,
369 long addr, long data)
371 int ret = -EIO;
373 switch (request) {
374 #ifdef PTRACE_OLDSETOPTIONS
375 case PTRACE_OLDSETOPTIONS:
376 #endif
377 case PTRACE_SETOPTIONS:
378 ret = ptrace_setoptions(child, data);
379 break;
380 case PTRACE_GETEVENTMSG:
381 ret = put_user(child->ptrace_message, (unsigned long __user *) data);
382 break;
383 case PTRACE_GETSIGINFO:
384 ret = ptrace_getsiginfo(child, (siginfo_t __user *) data);
385 break;
386 case PTRACE_SETSIGINFO:
387 ret = ptrace_setsiginfo(child, (siginfo_t __user *) data);
388 break;
389 default:
390 break;
393 return ret;
397 * ptrace_traceme -- helper for PTRACE_TRACEME
399 * Performs checks and sets PT_PTRACED.
400 * Should be used by all ptrace implementations for PTRACE_TRACEME.
402 int ptrace_traceme(void)
404 int ret = -EPERM;
407 * Are we already being traced?
409 task_lock(current);
410 if (!(current->ptrace & PT_PTRACED)) {
411 ret = security_ptrace(current->parent, current);
413 * Set the ptrace bit in the process ptrace flags.
415 if (!ret)
416 current->ptrace |= PT_PTRACED;
418 task_unlock(current);
419 return ret;
423 * ptrace_get_task_struct -- grab a task struct reference for ptrace
424 * @pid: process id to grab a task_struct reference of
426 * This function is a helper for ptrace implementations. It checks
427 * permissions and then grabs a task struct for use of the actual
428 * ptrace implementation.
430 * Returns the task_struct for @pid or an ERR_PTR() on failure.
432 struct task_struct *ptrace_get_task_struct(pid_t pid)
434 struct task_struct *child;
437 * Tracing init is not allowed.
439 if (pid == 1)
440 return ERR_PTR(-EPERM);
442 read_lock(&tasklist_lock);
443 child = find_task_by_pid(pid);
444 if (child)
445 get_task_struct(child);
447 read_unlock(&tasklist_lock);
448 if (!child)
449 return ERR_PTR(-ESRCH);
450 return child;
453 #ifndef __ARCH_SYS_PTRACE
454 asmlinkage long sys_ptrace(long request, long pid, long addr, long data)
456 struct task_struct *child;
457 long ret;
460 * This lock_kernel fixes a subtle race with suid exec
462 lock_kernel();
463 if (request == PTRACE_TRACEME) {
464 ret = ptrace_traceme();
465 goto out;
468 child = ptrace_get_task_struct(pid);
469 if (IS_ERR(child)) {
470 ret = PTR_ERR(child);
471 goto out;
474 if (request == PTRACE_ATTACH) {
475 ret = ptrace_attach(child);
476 goto out_put_task_struct;
479 ret = ptrace_check_attach(child, request == PTRACE_KILL);
480 if (ret < 0)
481 goto out_put_task_struct;
483 ret = arch_ptrace(child, request, addr, data);
484 if (ret < 0)
485 goto out_put_task_struct;
487 out_put_task_struct:
488 put_task_struct(child);
489 out:
490 unlock_kernel();
491 return ret;
493 #endif /* __ARCH_SYS_PTRACE */
495 int generic_ptrace_peekdata(struct task_struct *tsk, long addr, long data)
497 unsigned long tmp;
498 int copied;
500 copied = access_process_vm(tsk, addr, &tmp, sizeof(tmp), 0);
501 if (copied != sizeof(tmp))
502 return -EIO;
503 return put_user(tmp, (unsigned long __user *)data);
506 int generic_ptrace_pokedata(struct task_struct *tsk, long addr, long data)
508 int copied;
510 copied = access_process_vm(tsk, addr, &data, sizeof(data), 1);
511 return (copied == sizeof(data)) ? 0 : -EIO;