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         unsigned long flags;
 165
 166         audit_ptrace(task);
 167
 168         retval = -EPERM;
 169         if (task->pid <= 1)
 170                 goto out;
 171         if (task->tgid == current->tgid)
 172                 goto out;
 173
 174 repeat:
 175         /*
 176          * Nasty, nasty.
 177          *
 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).
 183          */
 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;
 191         }
 192
 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;
 201
 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;
 207
 208         __ptrace_link(task, current);
 209
 210         force_sig_specific(SIGSTOP, task);
 211
 212 bad:
 213         write_unlock_irqrestore(&tasklist_lock, flags);
 214         task_unlock(task);
 215 out:
 216         return retval;
 217 }
 218
 219 static inline void __ptrace_detach(struct task_struct *child, unsigned int data)
 220 {
 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);
 227 }
 228
 229 int ptrace_detach(struct task_struct *child, unsigned int data)
 230 {
 231         if (!valid_signal(data))
 232                 return -EIO;
 233
 234         /* Architecture-specific hardware disable .. */
 235         ptrace_disable(child);
 236
 237         write_lock_irq(&tasklist_lock);
 238         /* protect against de_thread()->release_task() */
 239         if (child->ptrace)
 240                 __ptrace_detach(child, data);
 241         write_unlock_irq(&tasklist_lock);
 242
 243         return 0;
 244 }
 245
 246 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
 247 {
 248         int copied = 0;
 249
 250         while (len > 0) {
 251                 char buf[128];
 252                 int this_len, retval;
 253
 254                 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 255                 retval = access_process_vm(tsk, src, buf, this_len, 0);
 256                 if (!retval) {
 257                         if (copied)
 258                                 break;
 259                         return -EIO;
 260                 }
 261                 if (copy_to_user(dst, buf, retval))
 262                         return -EFAULT;
 263                 copied += retval;
 264                 src += retval;
 265                 dst += retval;
 266                 len -= retval;
 267         }
 268         return copied;
 269 }
 270
 271 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
 272 {
 273         int copied = 0;
 274
 275         while (len > 0) {
 276                 char buf[128];
 277                 int this_len, retval;
 278
 279                 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 280                 if (copy_from_user(buf, src, this_len))
 281                         return -EFAULT;
 282                 retval = access_process_vm(tsk, dst, buf, this_len, 1);
 283                 if (!retval) {
 284                         if (copied)
 285                                 break;
 286                         return -EIO;
 287                 }
 288                 copied += retval;
 289                 src += retval;
 290                 dst += retval;
 291                 len -= retval;
 292         }
 293         return copied;
 294 }
 295
 296 static int ptrace_setoptions(struct task_struct *child, long data)
 297 {
 298         child->ptrace &= ~PT_TRACE_MASK;
 299
 300         if (data & PTRACE_O_TRACESYSGOOD)
 301                 child->ptrace |= PT_TRACESYSGOOD;
 302
 303         if (data & PTRACE_O_TRACEFORK)
 304                 child->ptrace |= PT_TRACE_FORK;
 305
 306         if (data & PTRACE_O_TRACEVFORK)
 307                 child->ptrace |= PT_TRACE_VFORK;
 308
 309         if (data & PTRACE_O_TRACECLONE)
 310                 child->ptrace |= PT_TRACE_CLONE;
 311
 312         if (data & PTRACE_O_TRACEEXEC)
 313                 child->ptrace |= PT_TRACE_EXEC;
 314
 315         if (data & PTRACE_O_TRACEVFORKDONE)
 316                 child->ptrace |= PT_TRACE_VFORK_DONE;
 317
 318         if (data & PTRACE_O_TRACEEXIT)
 319                 child->ptrace |= PT_TRACE_EXIT;
 320
 321         return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
 322 }
 323
 324 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t __user * data)
 325 {
 326         siginfo_t lastinfo;
 327         int error = -ESRCH;
 328
 329         read_lock(&tasklist_lock);
 330         if (likely(child->sighand != NULL)) {
 331                 error = -EINVAL;
 332                 spin_lock_irq(&child->sighand->siglock);
 333                 if (likely(child->last_siginfo != NULL)) {
 334                         lastinfo = *child->last_siginfo;
 335                         error = 0;
 336                 }
 337                 spin_unlock_irq(&child->sighand->siglock);
 338         }
 339         read_unlock(&tasklist_lock);
 340         if (!error)
 341                 return copy_siginfo_to_user(data, &lastinfo);
 342         return error;
 343 }
 344
 345 static int ptrace_setsiginfo(struct task_struct *child, siginfo_t __user * data)
 346 {
 347         siginfo_t newinfo;
 348         int error = -ESRCH;
 349
 350         if (copy_from_user(&newinfo, data, sizeof (siginfo_t)))
 351                 return -EFAULT;
 352
 353         read_lock(&tasklist_lock);
 354         if (likely(child->sighand != NULL)) {
 355                 error = -EINVAL;
 356                 spin_lock_irq(&child->sighand->siglock);
 357                 if (likely(child->last_siginfo != NULL)) {
 358                         *child->last_siginfo = newinfo;
 359                         error = 0;
 360                 }
 361                 spin_unlock_irq(&child->sighand->siglock);
 362         }
 363         read_unlock(&tasklist_lock);
 364         return error;
 365 }
 366
 367 int ptrace_request(struct task_struct *child, long request,
 368                    long addr, long data)
 369 {
 370         int ret = -EIO;
 371
 372         switch (request) {
 373 #ifdef PTRACE_OLDSETOPTIONS
 374         case PTRACE_OLDSETOPTIONS:
 375 #endif
 376         case PTRACE_SETOPTIONS:
 377                 ret = ptrace_setoptions(child, data);
 378                 break;
 379         case PTRACE_GETEVENTMSG:
 380                 ret = put_user(child->ptrace_message, (unsigned long __user *) data);
 381                 break;
 382         case PTRACE_GETSIGINFO:
 383                 ret = ptrace_getsiginfo(child, (siginfo_t __user *) data);
 384                 break;
 385         case PTRACE_SETSIGINFO:
 386                 ret = ptrace_setsiginfo(child, (siginfo_t __user *) data);
 387                 break;
 388         default:
 389                 break;
 390         }
 391
 392         return ret;
 393 }
 394
 395 /**
 396  * ptrace_traceme  --  helper for PTRACE_TRACEME
 397  *
 398  * Performs checks and sets PT_PTRACED.
 399  * Should be used by all ptrace implementations for PTRACE_TRACEME.
 400  */
 401 int ptrace_traceme(void)
 402 {
 403         int ret = -EPERM;
 404
 405         /*
 406          * Are we already being traced?
 407          */
 408         task_lock(current);
 409         if (!(current->ptrace & PT_PTRACED)) {
 410                 ret = security_ptrace(current->parent, current);
 411                 /*
 412                  * Set the ptrace bit in the process ptrace flags.
 413                  */
 414                 if (!ret)
 415                         current->ptrace |= PT_PTRACED;
 416         }
 417         task_unlock(current);
 418         return ret;
 419 }
 420
 421 /**
 422  * ptrace_get_task_struct  --  grab a task struct reference for ptrace
 423  * @pid:       process id to grab a task_struct reference of
 424  *
 425  * This function is a helper for ptrace implementations.  It checks
 426  * permissions and then grabs a task struct for use of the actual
 427  * ptrace implementation.
 428  *
 429  * Returns the task_struct for @pid or an ERR_PTR() on failure.
 430  */
 431 struct task_struct *ptrace_get_task_struct(pid_t pid)
 432 {
 433         struct task_struct *child;
 434
 435         /*
 436          * Tracing init is not allowed.
 437          */
 438         if (pid == 1)
 439                 return ERR_PTR(-EPERM);
 440
 441         read_lock(&tasklist_lock);
 442         child = find_task_by_pid(pid);
 443         if (child)
 444                 get_task_struct(child);
 445
 446         read_unlock(&tasklist_lock);
 447         if (!child)
 448                 return ERR_PTR(-ESRCH);
 449         return child;
 450 }
 451
 452 #ifndef __ARCH_SYS_PTRACE
 453 asmlinkage long sys_ptrace(long request, long pid, long addr, long data)
 454 {
 455         struct task_struct *child;
 456         long ret;
 457
 458         /*
 459          * This lock_kernel fixes a subtle race with suid exec
 460          */
 461         lock_kernel();
 462         if (request == PTRACE_TRACEME) {
 463                 ret = ptrace_traceme();
 464                 goto out;
 465         }
 466
 467         child = ptrace_get_task_struct(pid);
 468         if (IS_ERR(child)) {
 469                 ret = PTR_ERR(child);
 470                 goto out;
 471         }
 472
 473         if (request == PTRACE_ATTACH) {
 474                 ret = ptrace_attach(child);
 475                 goto out_put_task_struct;
 476         }
 477
 478         ret = ptrace_check_attach(child, request == PTRACE_KILL);
 479         if (ret < 0)
 480                 goto out_put_task_struct;
 481
 482         ret = arch_ptrace(child, request, addr, data);
 483         if (ret < 0)
 484                 goto out_put_task_struct;
 485
 486  out_put_task_struct:
 487         put_task_struct(child);
 488  out:
 489         unlock_kernel();
 490         return ret;
 491 }
 492 #endif /* __ARCH_SYS_PTRACE */