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
  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  */
  31 void __ptrace_link(task_t *child, task_t *new_parent)
  32 {
  33         if (!list_empty(&child->ptrace_list))
  34                 BUG();
  35         if (child->parent == new_parent)
  36                 return;
  37         list_add(&child->ptrace_list, &child->parent->ptrace_children);
  38         REMOVE_LINKS(child);
  39         child->parent = new_parent;
  40         SET_LINKS(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(task_t *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(task_t *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_LINKS(child);
  77                 child->parent = child->real_parent;
  78                 SET_LINKS(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         if (!task->mm)
 125                 return -EPERM;
 126         if (((current->uid != task->euid) ||
 127              (current->uid != task->suid) ||
 128              (current->uid != task->uid) ||
 129              (current->gid != task->egid) ||
 130              (current->gid != task->sgid) ||
 131              (current->gid != task->gid)) && !capable(CAP_SYS_PTRACE))
 132                 return -EPERM;
 133         smp_rmb();
 134         if (!task->mm->dumpable && !capable(CAP_SYS_PTRACE))
 135                 return -EPERM;
 136
 137         return security_ptrace(current, task);
 138 }
 139
 140 int ptrace_may_attach(struct task_struct *task)
 141 {
 142         int err;
 143         task_lock(task);
 144         err = may_attach(task);
 145         task_unlock(task);
 146         return !err;
 147 }
 148
 149 int ptrace_attach(struct task_struct *task)
 150 {
 151         int retval;
 152
 153         retval = -EPERM;
 154         if (task->pid <= 1)
 155                 goto out;
 156         if (task->tgid == current->tgid)
 157                 goto out;
 158
 159 repeat:
 160         /*
 161          * Nasty, nasty.
 162          *
 163          * We want to hold both the task-lock and the
 164          * tasklist_lock for writing at the same time.
 165          * But that's against the rules (tasklist_lock
 166          * is taken for reading by interrupts on other
 167          * cpu's that may have task_lock).
 168          */
 169         task_lock(task);
 170         local_irq_disable();
 171         if (!write_trylock(&tasklist_lock)) {
 172                 local_irq_enable();
 173                 task_unlock(task);
 174                 do {
 175                         cpu_relax();
 176                 } while (!write_can_lock(&tasklist_lock));
 177                 goto repeat;
 178         }
 179
 180         /* the same process cannot be attached many times */
 181         if (task->ptrace & PT_PTRACED)
 182                 goto bad;
 183         retval = may_attach(task);
 184         if (retval)
 185                 goto bad;
 186
 187         /* Go */
 188         task->ptrace |= PT_PTRACED | ((task->real_parent != current)
 189                                       ? PT_ATTACHED : 0);
 190         if (capable(CAP_SYS_PTRACE))
 191                 task->ptrace |= PT_PTRACE_CAP;
 192
 193         __ptrace_link(task, current);
 194
 195         force_sig_specific(SIGSTOP, task);
 196
 197 bad:
 198         write_unlock_irq(&tasklist_lock);
 199         task_unlock(task);
 200 out:
 201         return retval;
 202 }
 203
 204 void __ptrace_detach(struct task_struct *child, unsigned int data)
 205 {
 206         child->exit_code = data;
 207         /* .. re-parent .. */
 208         __ptrace_unlink(child);
 209         /* .. and wake it up. */
 210         if (child->exit_state != EXIT_ZOMBIE)
 211                 wake_up_process(child);
 212 }
 213
 214 int ptrace_detach(struct task_struct *child, unsigned int data)
 215 {
 216         if (!valid_signal(data))
 217                 return -EIO;
 218
 219         /* Architecture-specific hardware disable .. */
 220         ptrace_disable(child);
 221
 222         write_lock_irq(&tasklist_lock);
 223         if (child->ptrace)
 224                 __ptrace_detach(child, data);
 225         write_unlock_irq(&tasklist_lock);
 226
 227         return 0;
 228 }
 229
 230 /*
 231  * Access another process' address space.
 232  * Source/target buffer must be kernel space,
 233  * Do not walk the page table directly, use get_user_pages
 234  */
 235
 236 int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write)
 237 {
 238         struct mm_struct *mm;
 239         struct vm_area_struct *vma;
 240         struct page *page;
 241         void *old_buf = buf;
 242
 243         mm = get_task_mm(tsk);
 244         if (!mm)
 245                 return 0;
 246
 247         down_read(&mm->mmap_sem);
 248         /* ignore errors, just check how much was sucessfully transfered */
 249         while (len) {
 250                 int bytes, ret, offset;
 251                 void *maddr;
 252
 253                 ret = get_user_pages(tsk, mm, addr, 1,
 254                                 write, 1, &page, &vma);
 255                 if (ret <= 0)
 256                         break;
 257
 258                 bytes = len;
 259                 offset = addr & (PAGE_SIZE-1);
 260                 if (bytes > PAGE_SIZE-offset)
 261                         bytes = PAGE_SIZE-offset;
 262
 263                 maddr = kmap(page);
 264                 if (write) {
 265                         copy_to_user_page(vma, page, addr,
 266                                           maddr + offset, buf, bytes);
 267                         set_page_dirty_lock(page);
 268                 } else {
 269                         copy_from_user_page(vma, page, addr,
 270                                             buf, maddr + offset, bytes);
 271                 }
 272                 kunmap(page);
 273                 page_cache_release(page);
 274                 len -= bytes;
 275                 buf += bytes;
 276                 addr += bytes;
 277         }
 278         up_read(&mm->mmap_sem);
 279         mmput(mm);
 280
 281         return buf - old_buf;
 282 }
 283
 284 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
 285 {
 286         int copied = 0;
 287
 288         while (len > 0) {
 289                 char buf[128];
 290                 int this_len, retval;
 291
 292                 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 293                 retval = access_process_vm(tsk, src, buf, this_len, 0);
 294                 if (!retval) {
 295                         if (copied)
 296                                 break;
 297                         return -EIO;
 298                 }
 299                 if (copy_to_user(dst, buf, retval))
 300                         return -EFAULT;
 301                 copied += retval;
 302                 src += retval;
 303                 dst += retval;
 304                 len -= retval;
 305         }
 306         return copied;
 307 }
 308
 309 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
 310 {
 311         int copied = 0;
 312
 313         while (len > 0) {
 314                 char buf[128];
 315                 int this_len, retval;
 316
 317                 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 318                 if (copy_from_user(buf, src, this_len))
 319                         return -EFAULT;
 320                 retval = access_process_vm(tsk, dst, buf, this_len, 1);
 321                 if (!retval) {
 322                         if (copied)
 323                                 break;
 324                         return -EIO;
 325                 }
 326                 copied += retval;
 327                 src += retval;
 328                 dst += retval;
 329                 len -= retval;
 330         }
 331         return copied;
 332 }
 333
 334 static int ptrace_setoptions(struct task_struct *child, long data)
 335 {
 336         child->ptrace &= ~PT_TRACE_MASK;
 337
 338         if (data & PTRACE_O_TRACESYSGOOD)
 339                 child->ptrace |= PT_TRACESYSGOOD;
 340
 341         if (data & PTRACE_O_TRACEFORK)
 342                 child->ptrace |= PT_TRACE_FORK;
 343
 344         if (data & PTRACE_O_TRACEVFORK)
 345                 child->ptrace |= PT_TRACE_VFORK;
 346
 347         if (data & PTRACE_O_TRACECLONE)
 348                 child->ptrace |= PT_TRACE_CLONE;
 349
 350         if (data & PTRACE_O_TRACEEXEC)
 351                 child->ptrace |= PT_TRACE_EXEC;
 352
 353         if (data & PTRACE_O_TRACEVFORKDONE)
 354                 child->ptrace |= PT_TRACE_VFORK_DONE;
 355
 356         if (data & PTRACE_O_TRACEEXIT)
 357                 child->ptrace |= PT_TRACE_EXIT;
 358
 359         return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
 360 }
 361
 362 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t __user * data)
 363 {
 364         siginfo_t lastinfo;
 365         int error = -ESRCH;
 366
 367         read_lock(&tasklist_lock);
 368         if (likely(child->sighand != NULL)) {
 369                 error = -EINVAL;
 370                 spin_lock_irq(&child->sighand->siglock);
 371                 if (likely(child->last_siginfo != NULL)) {
 372                         lastinfo = *child->last_siginfo;
 373                         error = 0;
 374                 }
 375                 spin_unlock_irq(&child->sighand->siglock);
 376         }
 377         read_unlock(&tasklist_lock);
 378         if (!error)
 379                 return copy_siginfo_to_user(data, &lastinfo);
 380         return error;
 381 }
 382
 383 static int ptrace_setsiginfo(struct task_struct *child, siginfo_t __user * data)
 384 {
 385         siginfo_t newinfo;
 386         int error = -ESRCH;
 387
 388         if (copy_from_user(&newinfo, data, sizeof (siginfo_t)))
 389                 return -EFAULT;
 390
 391         read_lock(&tasklist_lock);
 392         if (likely(child->sighand != NULL)) {
 393                 error = -EINVAL;
 394                 spin_lock_irq(&child->sighand->siglock);
 395                 if (likely(child->last_siginfo != NULL)) {
 396                         *child->last_siginfo = newinfo;
 397                         error = 0;
 398                 }
 399                 spin_unlock_irq(&child->sighand->siglock);
 400         }
 401         read_unlock(&tasklist_lock);
 402         return error;
 403 }
 404
 405 int ptrace_request(struct task_struct *child, long request,
 406                    long addr, long data)
 407 {
 408         int ret = -EIO;
 409
 410         switch (request) {
 411 #ifdef PTRACE_OLDSETOPTIONS
 412         case PTRACE_OLDSETOPTIONS:
 413 #endif
 414         case PTRACE_SETOPTIONS:
 415                 ret = ptrace_setoptions(child, data);
 416                 break;
 417         case PTRACE_GETEVENTMSG:
 418                 ret = put_user(child->ptrace_message, (unsigned long __user *) data);
 419                 break;
 420         case PTRACE_GETSIGINFO:
 421                 ret = ptrace_getsiginfo(child, (siginfo_t __user *) data);
 422                 break;
 423         case PTRACE_SETSIGINFO:
 424                 ret = ptrace_setsiginfo(child, (siginfo_t __user *) data);
 425                 break;
 426         default:
 427                 break;
 428         }
 429
 430         return ret;
 431 }
 432
 433 /**
 434  * ptrace_traceme  --  helper for PTRACE_TRACEME
 435  *
 436  * Performs checks and sets PT_PTRACED.
 437  * Should be used by all ptrace implementations for PTRACE_TRACEME.
 438  */
 439 int ptrace_traceme(void)
 440 {
 441         int ret = -EPERM;
 442
 443         /*
 444          * Are we already being traced?
 445          */
 446         task_lock(current);
 447         if (!(current->ptrace & PT_PTRACED)) {
 448                 ret = security_ptrace(current->parent, current);
 449                 /*
 450                  * Set the ptrace bit in the process ptrace flags.
 451                  */
 452                 if (!ret)
 453                         current->ptrace |= PT_PTRACED;
 454         }
 455         task_unlock(current);
 456         return ret;
 457 }
 458
 459 /**
 460  * ptrace_get_task_struct  --  grab a task struct reference for ptrace
 461  * @pid:       process id to grab a task_struct reference of
 462  *
 463  * This function is a helper for ptrace implementations.  It checks
 464  * permissions and then grabs a task struct for use of the actual
 465  * ptrace implementation.
 466  *
 467  * Returns the task_struct for @pid or an ERR_PTR() on failure.
 468  */
 469 struct task_struct *ptrace_get_task_struct(pid_t pid)
 470 {
 471         struct task_struct *child;
 472
 473         /*
 474          * Tracing init is not allowed.
 475          */
 476         if (pid == 1)
 477                 return ERR_PTR(-EPERM);
 478
 479         read_lock(&tasklist_lock);
 480         child = find_task_by_pid(pid);
 481         if (child)
 482                 get_task_struct(child);
 483         read_unlock(&tasklist_lock);
 484         if (!child)
 485                 return ERR_PTR(-ESRCH);
 486         return child;
 487 }
 488
 489 #ifndef __ARCH_SYS_PTRACE
 490 asmlinkage long sys_ptrace(long request, long pid, long addr, long data)
 491 {
 492         struct task_struct *child;
 493         long ret;
 494
 495         /*
 496          * This lock_kernel fixes a subtle race with suid exec
 497          */
 498         lock_kernel();
 499         if (request == PTRACE_TRACEME) {
 500                 ret = ptrace_traceme();
 501                 goto out;
 502         }
 503
 504         child = ptrace_get_task_struct(pid);
 505         if (IS_ERR(child)) {
 506                 ret = PTR_ERR(child);
 507                 goto out;
 508         }
 509
 510         if (request == PTRACE_ATTACH) {
 511                 ret = ptrace_attach(child);
 512                 goto out_put_task_struct;
 513         }
 514
 515         ret = ptrace_check_attach(child, request == PTRACE_KILL);
 516         if (ret < 0)
 517                 goto out_put_task_struct;
 518
 519         ret = arch_ptrace(child, request, addr, data);
 520         if (ret < 0)
 521                 goto out_put_task_struct;
 522
 523  out_put_task_struct:
 524         put_task_struct(child);
 525  out:
 526         unlock_kernel();
 527         return ret;
 528 }
 529 #endif /* __ARCH_SYS_PTRACE */