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/module.h>
  11 #include <linux/sched.h>
  12 #include <linux/errno.h>
  13 #include <linux/mm.h>
  14 #include <linux/highmem.h>
  15 #include <linux/pagemap.h>
  16 #include <linux/smp_lock.h>
  17 #include <linux/ptrace.h>
  18 #include <linux/security.h>
  19 #include <linux/signal.h>
  20
  21 #include <asm/pgtable.h>
  22 #include <asm/uaccess.h>
  23
  24 /*
  25  * ptrace a task: make the debugger its new parent and
  26  * move it to the ptrace list.
  27  *
  28  * Must be called with the tasklist lock write-held.
  29  */
  30 void __ptrace_link(task_t *child, task_t *new_parent)
  31 {
  32         if (!list_empty(&child->ptrace_list))
  33                 BUG();
  34         if (child->parent == new_parent)
  35                 return;
  36         list_add(&child->ptrace_list, &child->parent->ptrace_children);
  37         REMOVE_LINKS(child);
  38         child->parent = new_parent;
  39         SET_LINKS(child);
  40 }
  41
  42 /*
  43  * Turn a tracing stop into a normal stop now, since with no tracer there
  44  * would be no way to wake it up with SIGCONT or SIGKILL.  If there was a
  45  * signal sent that would resume the child, but didn't because it was in
  46  * TASK_TRACED, resume it now.
  47  * Requires that irqs be disabled.
  48  */
  49 void ptrace_untrace(task_t *child)
  50 {
  51         spin_lock(&child->sighand->siglock);
  52         if (child->state == TASK_TRACED) {
  53                 if (child->signal->flags & SIGNAL_STOP_STOPPED) {
  54                         child->state = TASK_STOPPED;
  55                 } else {
  56                         signal_wake_up(child, 1);
  57                 }
  58         }
  59         if (child->signal->flags & SIGNAL_GROUP_EXIT) {
  60                 sigaddset(&child->pending.signal, SIGKILL);
  61                 signal_wake_up(child, 1);
  62         }
  63         spin_unlock(&child->sighand->siglock);
  64 }
  65
  66 /*
  67  * unptrace a task: move it back to its original parent and
  68  * remove it from the ptrace list.
  69  *
  70  * Must be called with the tasklist lock write-held.
  71  */
  72 void __ptrace_unlink(task_t *child)
  73 {
  74         if (!child->ptrace)
  75                 BUG();
  76         child->ptrace = 0;
  77         if (!list_empty(&child->ptrace_list)) {
  78                 list_del_init(&child->ptrace_list);
  79                 REMOVE_LINKS(child);
  80                 child->parent = child->real_parent;
  81                 SET_LINKS(child);
  82         }
  83
  84         ptrace_untrace(child);
  85 }
  86
  87 /*
  88  * Check that we have indeed attached to the thing..
  89  */
  90 int ptrace_check_attach(struct task_struct *child, int kill)
  91 {
  92         int ret = -ESRCH;
  93
  94         /*
  95          * We take the read lock around doing both checks to close a
  96          * possible race where someone else was tracing our child and
  97          * detached between these two checks.  After this locked check,
  98          * we are sure that this is our traced child and that can only
  99          * be changed by us so it's not changing right after this.
 100          */
 101         read_lock(&tasklist_lock);
 102         if ((child->ptrace & PT_PTRACED) && child->parent == current &&
 103             (!(child->ptrace & PT_ATTACHED) || child->real_parent != current)
 104             && child->signal != NULL) {
 105                 ret = 0;
 106                 spin_lock_irq(&child->sighand->siglock);
 107                 if (child->state == TASK_STOPPED) {
 108                         child->state = TASK_TRACED;
 109                 } else if (child->state != TASK_TRACED && !kill) {
 110                         ret = -ESRCH;
 111                 }
 112                 spin_unlock_irq(&child->sighand->siglock);
 113         }
 114         read_unlock(&tasklist_lock);
 115
 116         if (!ret && !kill) {
 117                 wait_task_inactive(child);
 118         }
 119
 120         /* All systems go.. */
 121         return ret;
 122 }
 123
 124 static int may_attach(struct task_struct *task)
 125 {
 126         if (!task->mm)
 127                 return -EPERM;
 128         if (((current->uid != task->euid) ||
 129              (current->uid != task->suid) ||
 130              (current->uid != task->uid) ||
 131              (current->gid != task->egid) ||
 132              (current->gid != task->sgid) ||
 133              (current->gid != task->gid)) && !capable(CAP_SYS_PTRACE))
 134                 return -EPERM;
 135         smp_rmb();
 136         if (!task->mm->dumpable && !capable(CAP_SYS_PTRACE))
 137                 return -EPERM;
 138
 139         return security_ptrace(current, task);
 140 }
 141
 142 int ptrace_may_attach(struct task_struct *task)
 143 {
 144         int err;
 145         task_lock(task);
 146         err = may_attach(task);
 147         task_unlock(task);
 148         return !err;
 149 }
 150
 151 int ptrace_attach(struct task_struct *task)
 152 {
 153         int retval;
 154         task_lock(task);
 155         retval = -EPERM;
 156         if (task->pid <= 1)
 157                 goto bad;
 158         if (task->tgid == current->tgid)
 159                 goto bad;
 160         /* the same process cannot be attached many times */
 161         if (task->ptrace & PT_PTRACED)
 162                 goto bad;
 163         retval = may_attach(task);
 164         if (retval)
 165                 goto bad;
 166
 167         /* Go */
 168         task->ptrace |= PT_PTRACED | ((task->real_parent != current)
 169                                       ? PT_ATTACHED : 0);
 170         if (capable(CAP_SYS_PTRACE))
 171                 task->ptrace |= PT_PTRACE_CAP;
 172         task_unlock(task);
 173
 174         write_lock_irq(&tasklist_lock);
 175         __ptrace_link(task, current);
 176         write_unlock_irq(&tasklist_lock);
 177
 178         force_sig_specific(SIGSTOP, task);
 179         return 0;
 180
 181 bad:
 182         task_unlock(task);
 183         return retval;
 184 }
 185
 186 int ptrace_detach(struct task_struct *child, unsigned int data)
 187 {
 188         if (!valid_signal(data))
 189                 return  -EIO;
 190
 191         /* Architecture-specific hardware disable .. */
 192         ptrace_disable(child);
 193
 194         /* .. re-parent .. */
 195         child->exit_code = data;
 196
 197         write_lock_irq(&tasklist_lock);
 198         __ptrace_unlink(child);
 199         /* .. and wake it up. */
 200         if (child->exit_state != EXIT_ZOMBIE)
 201                 wake_up_process(child);
 202         write_unlock_irq(&tasklist_lock);
 203
 204         return 0;
 205 }
 206
 207 /*
 208  * Access another process' address space.
 209  * Source/target buffer must be kernel space,
 210  * Do not walk the page table directly, use get_user_pages
 211  */
 212
 213 int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write)
 214 {
 215         struct mm_struct *mm;
 216         struct vm_area_struct *vma;
 217         struct page *page;
 218         void *old_buf = buf;
 219
 220         mm = get_task_mm(tsk);
 221         if (!mm)
 222                 return 0;
 223
 224         down_read(&mm->mmap_sem);
 225         /* ignore errors, just check how much was sucessfully transfered */
 226         while (len) {
 227                 int bytes, ret, offset;
 228                 void *maddr;
 229
 230                 ret = get_user_pages(tsk, mm, addr, 1,
 231                                 write, 1, &page, &vma);
 232                 if (ret <= 0)
 233                         break;
 234
 235                 bytes = len;
 236                 offset = addr & (PAGE_SIZE-1);
 237                 if (bytes > PAGE_SIZE-offset)
 238                         bytes = PAGE_SIZE-offset;
 239
 240                 maddr = kmap(page);
 241                 if (write) {
 242                         copy_to_user_page(vma, page, addr,
 243                                           maddr + offset, buf, bytes);
 244                         set_page_dirty_lock(page);
 245                 } else {
 246                         copy_from_user_page(vma, page, addr,
 247                                             buf, maddr + offset, bytes);
 248                 }
 249                 kunmap(page);
 250                 page_cache_release(page);
 251                 len -= bytes;
 252                 buf += bytes;
 253                 addr += bytes;
 254         }
 255         up_read(&mm->mmap_sem);
 256         mmput(mm);
 257
 258         return buf - old_buf;
 259 }
 260
 261 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
 262 {
 263         int copied = 0;
 264
 265         while (len > 0) {
 266                 char buf[128];
 267                 int this_len, retval;
 268
 269                 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 270                 retval = access_process_vm(tsk, src, buf, this_len, 0);
 271                 if (!retval) {
 272                         if (copied)
 273                                 break;
 274                         return -EIO;
 275                 }
 276                 if (copy_to_user(dst, buf, retval))
 277                         return -EFAULT;
 278                 copied += retval;
 279                 src += retval;
 280                 dst += retval;
 281                 len -= retval;
 282         }
 283         return copied;
 284 }
 285
 286 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
 287 {
 288         int copied = 0;
 289
 290         while (len > 0) {
 291                 char buf[128];
 292                 int this_len, retval;
 293
 294                 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 295                 if (copy_from_user(buf, src, this_len))
 296                         return -EFAULT;
 297                 retval = access_process_vm(tsk, dst, buf, this_len, 1);
 298                 if (!retval) {
 299                         if (copied)
 300                                 break;
 301                         return -EIO;
 302                 }
 303                 copied += retval;
 304                 src += retval;
 305                 dst += retval;
 306                 len -= retval;
 307         }
 308         return copied;
 309 }
 310
 311 static int ptrace_setoptions(struct task_struct *child, long data)
 312 {
 313         child->ptrace &= ~PT_TRACE_MASK;
 314
 315         if (data & PTRACE_O_TRACESYSGOOD)
 316                 child->ptrace |= PT_TRACESYSGOOD;
 317
 318         if (data & PTRACE_O_TRACEFORK)
 319                 child->ptrace |= PT_TRACE_FORK;
 320
 321         if (data & PTRACE_O_TRACEVFORK)
 322                 child->ptrace |= PT_TRACE_VFORK;
 323
 324         if (data & PTRACE_O_TRACECLONE)
 325                 child->ptrace |= PT_TRACE_CLONE;
 326
 327         if (data & PTRACE_O_TRACEEXEC)
 328                 child->ptrace |= PT_TRACE_EXEC;
 329
 330         if (data & PTRACE_O_TRACEVFORKDONE)
 331                 child->ptrace |= PT_TRACE_VFORK_DONE;
 332
 333         if (data & PTRACE_O_TRACEEXIT)
 334                 child->ptrace |= PT_TRACE_EXIT;
 335
 336         return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
 337 }
 338
 339 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t __user * data)
 340 {
 341         siginfo_t lastinfo;
 342         int error = -ESRCH;
 343
 344         read_lock(&tasklist_lock);
 345         if (likely(child->sighand != NULL)) {
 346                 error = -EINVAL;
 347                 spin_lock_irq(&child->sighand->siglock);
 348                 if (likely(child->last_siginfo != NULL)) {
 349                         lastinfo = *child->last_siginfo;
 350                         error = 0;
 351                 }
 352                 spin_unlock_irq(&child->sighand->siglock);
 353         }
 354         read_unlock(&tasklist_lock);
 355         if (!error)
 356                 return copy_siginfo_to_user(data, &lastinfo);
 357         return error;
 358 }
 359
 360 static int ptrace_setsiginfo(struct task_struct *child, siginfo_t __user * data)
 361 {
 362         siginfo_t newinfo;
 363         int error = -ESRCH;
 364
 365         if (copy_from_user(&newinfo, data, sizeof (siginfo_t)))
 366                 return -EFAULT;
 367
 368         read_lock(&tasklist_lock);
 369         if (likely(child->sighand != NULL)) {
 370                 error = -EINVAL;
 371                 spin_lock_irq(&child->sighand->siglock);
 372                 if (likely(child->last_siginfo != NULL)) {
 373                         *child->last_siginfo = newinfo;
 374                         error = 0;
 375                 }
 376                 spin_unlock_irq(&child->sighand->siglock);
 377         }
 378         read_unlock(&tasklist_lock);
 379         return error;
 380 }
 381
 382 int ptrace_request(struct task_struct *child, long request,
 383                    long addr, long data)
 384 {
 385         int ret = -EIO;
 386
 387         switch (request) {
 388 #ifdef PTRACE_OLDSETOPTIONS
 389         case PTRACE_OLDSETOPTIONS:
 390 #endif
 391         case PTRACE_SETOPTIONS:
 392                 ret = ptrace_setoptions(child, data);
 393                 break;
 394         case PTRACE_GETEVENTMSG:
 395                 ret = put_user(child->ptrace_message, (unsigned long __user *) data);
 396                 break;
 397         case PTRACE_GETSIGINFO:
 398                 ret = ptrace_getsiginfo(child, (siginfo_t __user *) data);
 399                 break;
 400         case PTRACE_SETSIGINFO:
 401                 ret = ptrace_setsiginfo(child, (siginfo_t __user *) data);
 402                 break;
 403         default:
 404                 break;
 405         }
 406
 407         return ret;
 408 }
 409
 410 #ifndef __ARCH_SYS_PTRACE
 411 static int ptrace_get_task_struct(long request, long pid,
 412                 struct task_struct **childp)
 413 {
 414         struct task_struct *child;
 415         int ret;
 416
 417         /*
 418          * Callers use child == NULL as an indication to exit early even
 419          * when the return value is 0, so make sure it is non-NULL here.
 420          */
 421         *childp = NULL;
 422
 423         if (request == PTRACE_TRACEME) {
 424                 /*
 425                  * Are we already being traced?
 426                  */
 427                 if (current->ptrace & PT_PTRACED)
 428                         return -EPERM;
 429                 ret = security_ptrace(current->parent, current);
 430                 if (ret)
 431                         return -EPERM;
 432                 /*
 433                  * Set the ptrace bit in the process ptrace flags.
 434                  */
 435                 current->ptrace |= PT_PTRACED;
 436                 return 0;
 437         }
 438
 439         /*
 440          * You may not mess with init
 441          */
 442         if (pid == 1)
 443                 return -EPERM;
 444
 445         ret = -ESRCH;
 446         read_lock(&tasklist_lock);
 447         child = find_task_by_pid(pid);
 448         if (child)
 449                 get_task_struct(child);
 450         read_unlock(&tasklist_lock);
 451         if (!child)
 452                 return -ESRCH;
 453
 454         *childp = child;
 455         return 0;
 456 }
 457
 458 asmlinkage long sys_ptrace(long request, long pid, long addr, long data)
 459 {
 460         struct task_struct *child;
 461         long ret;
 462
 463         /*
 464          * This lock_kernel fixes a subtle race with suid exec
 465          */
 466         lock_kernel();
 467         ret = ptrace_get_task_struct(request, pid, &child);
 468         if (!child)
 469                 goto out;
 470
 471         if (request == PTRACE_ATTACH) {
 472                 ret = ptrace_attach(child);
 473                 goto out;
 474         }
 475
 476         ret = ptrace_check_attach(child, request == PTRACE_KILL);
 477         if (ret < 0)
 478                 goto out_put_task_struct;
 479
 480         ret = arch_ptrace(child, request, addr, data);
 481         if (ret < 0)
 482                 goto out_put_task_struct;
 483
 484  out_put_task_struct:
 485         put_task_struct(child);
 486  out:
 487         unlock_kernel();
 488         return ret;
 489 }
 490 #endif /* __ARCH_SYS_PTRACE */