kernel/ptrace.c

   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
  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
  23 #include <asm/pgtable.h>
  24 #include <asm/uaccess.h>
  25
  26 /*
  27  * ptrace a task: make the debugger its new parent and
  28  * move it to the ptrace list.
  29  *
  30  * Must be called with the tasklist lock write-held.
  31  */
  32 void __ptrace_link(struct task_struct *child, struct task_struct *new_parent)
  33 {
  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);
  41 }
  42
  43 /*
  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.
  49  */
  50 void ptrace_untrace(struct task_struct *child)
  51 {
  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);
  58                 }
  59         }
  60         spin_unlock(&child->sighand->siglock);
  61 }
  62
  63 /*
  64  * unptrace a task: move it back to its original parent and
  65  * remove it from the ptrace list.
  66  *
  67  * Must be called with the tasklist lock write-held.
  68  */
  69 void __ptrace_unlink(struct task_struct *child)
  70 {
  71         BUG_ON(!child->ptrace);
  72
  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);
  79         }
  80
  81         if (child->state == TASK_TRACED)
  82                 ptrace_untrace(child);
  83 }
  84
  85 /*
  86  * Check that we have indeed attached to the thing..
  87  */
  88 int ptrace_check_attach(struct task_struct *child, int kill)
  89 {
  90         int ret = -ESRCH;
  91
  92         /*
  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.
  98          */
  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;
 109                 }
 110                 spin_unlock_irq(&child->sighand->siglock);
 111         }
 112         read_unlock(&tasklist_lock);
 113
 114         if (!ret && !kill) {
 115                 wait_task_inactive(child);
 116         }
 117
 118         /* All systems go.. */
 119         return ret;
 120 }
 121
 122 static int may_attach(struct task_struct *task)
 123 {
 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.
 127          *
 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.
 131          */
 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 = task->mm->dumpable;
 146         if (!dumpable && !capable(CAP_SYS_PTRACE))
 147                 return -EPERM;
 148
 149         return security_ptrace(current, task);
 150 }
 151
 152 int ptrace_may_attach(struct task_struct *task)
 153 {
 154         int err;
 155         task_lock(task);
 156         err = may_attach(task);
 157         task_unlock(task);
 158         return !err;
 159 }
 160
 161 int ptrace_attach(struct task_struct *task)
 162 {
 163         int retval;
 164
 165         audit_ptrace(task);
 166
 167         retval = -EPERM;
 168         if (task->pid <= 1)
 169                 goto out;
 170         if (task->tgid == current->tgid)
 171                 goto out;
 172
 173 repeat:
 174         /*
 175          * Nasty, nasty.
 176          *
 177          * We want to hold both the task-lock and the
 178          * tasklist_lock for writing at the same time.
 179          * But that's against the rules (tasklist_lock
 180          * is taken for reading by interrupts on other
 181          * cpu's that may have task_lock).
 182          */
 183         task_lock(task);
 184         local_irq_disable();
 185         if (!write_trylock(&tasklist_lock)) {
 186                 local_irq_enable();
 187                 task_unlock(task);
 188                 do {
 189                         cpu_relax();
 190                 } while (!write_can_lock(&tasklist_lock));
 191                 goto repeat;
 192         }
 193
 194         if (!task->mm)
 195                 goto bad;
 196         /* the same process cannot be attached many times */
 197         if (task->ptrace & PT_PTRACED)
 198                 goto bad;
 199         retval = may_attach(task);
 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;
 208
 209         __ptrace_link(task, current);
 210
 211         force_sig_specific(SIGSTOP, task);
 212
 213 bad:
 214         write_unlock_irq(&tasklist_lock);
 215         task_unlock(task);
 216 out:
 217         return retval;
 218 }
 219
 220 static inline void __ptrace_detach(struct task_struct *child, unsigned int data)
 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
 230 int ptrace_detach(struct task_struct *child, unsigned int data)
 231 {
 232         if (!valid_signal(data))
 233                 return -EIO;
 234
 235         /* Architecture-specific hardware disable .. */
 236         ptrace_disable(child);
 237
 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);
 243
 244         return 0;
 245 }
 246
 247 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
 248 {
 249         int copied = 0;
 250
 251         while (len > 0) {
 252                 char buf[128];
 253                 int this_len, retval;
 254
 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;
 261                 }
 262                 if (copy_to_user(dst, buf, retval))
 263                         return -EFAULT;
 264                 copied += retval;
 265                 src += retval;
 266                 dst += retval;
 267                 len -= retval;
 268         }
 269         return copied;
 270 }
 271
 272 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
 273 {
 274         int copied = 0;
 275
 276         while (len > 0) {
 277                 char buf[128];
 278                 int this_len, retval;
 279
 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;
 288                 }
 289                 copied += retval;
 290                 src += retval;
 291                 dst += retval;
 292                 len -= retval;
 293         }
 294         return copied;
 295 }
 296
 297 static int ptrace_setoptions(struct task_struct *child, long data)
 298 {
 299         child->ptrace &= ~PT_TRACE_MASK;
 300
 301         if (data & PTRACE_O_TRACESYSGOOD)
 302                 child->ptrace |= PT_TRACESYSGOOD;
 303
 304         if (data & PTRACE_O_TRACEFORK)
 305                 child->ptrace |= PT_TRACE_FORK;
 306
 307         if (data & PTRACE_O_TRACEVFORK)
 308                 child->ptrace |= PT_TRACE_VFORK;
 309
 310         if (data & PTRACE_O_TRACECLONE)
 311                 child->ptrace |= PT_TRACE_CLONE;
 312
 313         if (data & PTRACE_O_TRACEEXEC)
 314                 child->ptrace |= PT_TRACE_EXEC;
 315
 316         if (data & PTRACE_O_TRACEVFORKDONE)
 317                 child->ptrace |= PT_TRACE_VFORK_DONE;
 318
 319         if (data & PTRACE_O_TRACEEXIT)
 320                 child->ptrace |= PT_TRACE_EXIT;
 321
 322         return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
 323 }
 324
 325 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t __user * data)
 326 {
 327         siginfo_t lastinfo;
 328         int error = -ESRCH;
 329
 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;
 337                 }
 338                 spin_unlock_irq(&child->sighand->siglock);
 339         }
 340         read_unlock(&tasklist_lock);
 341         if (!error)
 342                 return copy_siginfo_to_user(data, &lastinfo);
 343         return error;
 344 }
 345
 346 static int ptrace_setsiginfo(struct task_struct *child, siginfo_t __user * data)
 347 {
 348         siginfo_t newinfo;
 349         int error = -ESRCH;
 350
 351         if (copy_from_user(&newinfo, data, sizeof (siginfo_t)))
 352                 return -EFAULT;
 353
 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;
 361                 }
 362                 spin_unlock_irq(&child->sighand->siglock);
 363         }
 364         read_unlock(&tasklist_lock);
 365         return error;
 366 }
 367
 368 int ptrace_request(struct task_struct *child, long request,
 369                    long addr, long data)
 370 {
 371         int ret = -EIO;
 372
 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;
 391         }
 392
 393         return ret;
 394 }
 395
 396 /**
 397  * ptrace_traceme  --  helper for PTRACE_TRACEME
 398  *
 399  * Performs checks and sets PT_PTRACED.
 400  * Should be used by all ptrace implementations for PTRACE_TRACEME.
 401  */
 402 int ptrace_traceme(void)
 403 {
 404         int ret = -EPERM;
 405
 406         /*
 407          * Are we already being traced?
 408          */
 409         task_lock(current);
 410         if (!(current->ptrace & PT_PTRACED)) {
 411                 ret = security_ptrace(current->parent, current);
 412                 /*
 413                  * Set the ptrace bit in the process ptrace flags.
 414                  */
 415                 if (!ret)
 416                         current->ptrace |= PT_PTRACED;
 417         }
 418         task_unlock(current);
 419         return ret;
 420 }
 421
 422 /**
 423  * ptrace_get_task_struct  --  grab a task struct reference for ptrace
 424  * @pid:       process id to grab a task_struct reference of
 425  *
 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.
 429  *
 430  * Returns the task_struct for @pid or an ERR_PTR() on failure.
 431  */
 432 struct task_struct *ptrace_get_task_struct(pid_t pid)
 433 {
 434         struct task_struct *child;
 435
 436         /*
 437          * Tracing init is not allowed.
 438          */
 439         if (pid == 1)
 440                 return ERR_PTR(-EPERM);
 441
 442         read_lock(&tasklist_lock);
 443         child = find_task_by_pid(pid);
 444         if (child)
 445                 get_task_struct(child);
 446
 447         read_unlock(&tasklist_lock);
 448         if (!child)
 449                 return ERR_PTR(-ESRCH);
 450         return child;
 451 }
 452
 453 #ifndef __ARCH_SYS_PTRACE
 454 asmlinkage long sys_ptrace(long request, long pid, long addr, long data)
 455 {
 456         struct task_struct *child;
 457         long ret;
 458
 459         /*
 460          * This lock_kernel fixes a subtle race with suid exec
 461          */
 462         lock_kernel();
 463         if (request == PTRACE_TRACEME) {
 464                 ret = ptrace_traceme();
 465                 goto out;
 466         }
 467
 468         child = ptrace_get_task_struct(pid);
 469         if (IS_ERR(child)) {
 470                 ret = PTR_ERR(child);
 471                 goto out;
 472         }
 473
 474         if (request == PTRACE_ATTACH) {
 475                 ret = ptrace_attach(child);
 476                 goto out_put_task_struct;
 477         }
 478
 479         ret = ptrace_check_attach(child, request == PTRACE_KILL);
 480         if (ret < 0)
 481                 goto out_put_task_struct;
 482
 483         ret = arch_ptrace(child, request, addr, data);
 484         if (ret < 0)
 485                 goto out_put_task_struct;
 486
 487  out_put_task_struct:
 488         put_task_struct(child);
 489  out:
 490         unlock_kernel();
 491         return ret;
 492 }
 493 #endif /* __ARCH_SYS_PTRACE */