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         BUG_ON(!list_empty(&child->ptrace_list));
  34         if (child->parent == new_parent)
  35                 return;
  36         list_add(&child->ptrace_list, &child->parent->ptrace_children);
  37         remove_parent(child);
  38         child->parent = new_parent;
  39         add_parent(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         spin_unlock(&child->sighand->siglock);
  60 }
  61
  62 /*
  63  * unptrace a task: move it back to its original parent and
  64  * remove it from the ptrace list.
  65  *
  66  * Must be called with the tasklist lock write-held.
  67  */
  68 void __ptrace_unlink(task_t *child)
  69 {
  70         BUG_ON(!child->ptrace);
  71
  72         child->ptrace = 0;
  73         if (!list_empty(&child->ptrace_list)) {
  74                 list_del_init(&child->ptrace_list);
  75                 remove_parent(child);
  76                 child->parent = child->real_parent;
  77                 add_parent(child);
  78         }
  79
  80         if (child->state == TASK_TRACED)
  81                 ptrace_untrace(child);
  82 }
  83
  84 /*
  85  * Check that we have indeed attached to the thing..
  86  */
  87 int ptrace_check_attach(struct task_struct *child, int kill)
  88 {
  89         int ret = -ESRCH;
  90
  91         /*
  92          * We take the read lock around doing both checks to close a
  93          * possible race where someone else was tracing our child and
  94          * detached between these two checks.  After this locked check,
  95          * we are sure that this is our traced child and that can only
  96          * be changed by us so it's not changing right after this.
  97          */
  98         read_lock(&tasklist_lock);
  99         if ((child->ptrace & PT_PTRACED) && child->parent == current &&
 100             (!(child->ptrace & PT_ATTACHED) || child->real_parent != current)
 101             && child->signal != NULL) {
 102                 ret = 0;
 103                 spin_lock_irq(&child->sighand->siglock);
 104                 if (child->state == TASK_STOPPED) {
 105                         child->state = TASK_TRACED;
 106                 } else if (child->state != TASK_TRACED && !kill) {
 107                         ret = -ESRCH;
 108                 }
 109                 spin_unlock_irq(&child->sighand->siglock);
 110         }
 111         read_unlock(&tasklist_lock);
 112
 113         if (!ret && !kill) {
 114                 wait_task_inactive(child);
 115         }
 116
 117         /* All systems go.. */
 118         return ret;
 119 }
 120
 121 static int may_attach(struct task_struct *task)
 122 {
 123         if (!task->mm)
 124                 return -EPERM;
 125         if (((current->uid != task->euid) ||
 126              (current->uid != task->suid) ||
 127              (current->uid != task->uid) ||
 128              (current->gid != task->egid) ||
 129              (current->gid != task->sgid) ||
 130              (current->gid != task->gid)) && !capable(CAP_SYS_PTRACE))
 131                 return -EPERM;
 132         smp_rmb();
 133         if (!task->mm->dumpable && !capable(CAP_SYS_PTRACE))
 134                 return -EPERM;
 135
 136         return security_ptrace(current, task);
 137 }
 138
 139 int ptrace_may_attach(struct task_struct *task)
 140 {
 141         int err;
 142         task_lock(task);
 143         err = may_attach(task);
 144         task_unlock(task);
 145         return !err;
 146 }
 147
 148 int ptrace_attach(struct task_struct *task)
 149 {
 150         int retval;
 151
 152         retval = -EPERM;
 153         if (task->pid <= 1)
 154                 goto out;
 155         if (task->tgid == current->tgid)
 156                 goto out;
 157
 158 repeat:
 159         /*
 160          * Nasty, nasty.
 161          *
 162          * We want to hold both the task-lock and the
 163          * tasklist_lock for writing at the same time.
 164          * But that's against the rules (tasklist_lock
 165          * is taken for reading by interrupts on other
 166          * cpu's that may have task_lock).
 167          */
 168         task_lock(task);
 169         local_irq_disable();
 170         if (!write_trylock(&tasklist_lock)) {
 171                 local_irq_enable();
 172                 task_unlock(task);
 173                 do {
 174                         cpu_relax();
 175                 } while (!write_can_lock(&tasklist_lock));
 176                 goto repeat;
 177         }
 178
 179         /* the same process cannot be attached many times */
 180         if (task->ptrace & PT_PTRACED)
 181                 goto bad;
 182         retval = may_attach(task);
 183         if (retval)
 184                 goto bad;
 185
 186         /* Go */
 187         task->ptrace |= PT_PTRACED | ((task->real_parent != current)
 188                                       ? PT_ATTACHED : 0);
 189         if (capable(CAP_SYS_PTRACE))
 190                 task->ptrace |= PT_PTRACE_CAP;
 191
 192         __ptrace_link(task, current);
 193
 194         force_sig_specific(SIGSTOP, task);
 195
 196 bad:
 197         write_unlock_irq(&tasklist_lock);
 198         task_unlock(task);
 199 out:
 200         return retval;
 201 }
 202
 203 void __ptrace_detach(struct task_struct *child, unsigned int data)
 204 {
 205         child->exit_code = data;
 206         /* .. re-parent .. */
 207         __ptrace_unlink(child);
 208         /* .. and wake it up. */
 209         if (child->exit_state != EXIT_ZOMBIE)
 210                 wake_up_process(child);
 211 }
 212
 213 int ptrace_detach(struct task_struct *child, unsigned int data)
 214 {
 215         if (!valid_signal(data))
 216                 return -EIO;
 217
 218         /* Architecture-specific hardware disable .. */
 219         ptrace_disable(child);
 220
 221         write_lock_irq(&tasklist_lock);
 222         if (child->ptrace)
 223                 __ptrace_detach(child, data);
 224         write_unlock_irq(&tasklist_lock);
 225
 226         return 0;
 227 }
 228
 229 /*
 230  * Access another process' address space.
 231  * Source/target buffer must be kernel space,
 232  * Do not walk the page table directly, use get_user_pages
 233  */
 234
 235 int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write)
 236 {
 237         struct mm_struct *mm;
 238         struct vm_area_struct *vma;
 239         struct page *page;
 240         void *old_buf = buf;
 241
 242         mm = get_task_mm(tsk);
 243         if (!mm)
 244                 return 0;
 245
 246         down_read(&mm->mmap_sem);
 247         /* ignore errors, just check how much was sucessfully transfered */
 248         while (len) {
 249                 int bytes, ret, offset;
 250                 void *maddr;
 251
 252                 ret = get_user_pages(tsk, mm, addr, 1,
 253                                 write, 1, &page, &vma);
 254                 if (ret <= 0)
 255                         break;
 256
 257                 bytes = len;
 258                 offset = addr & (PAGE_SIZE-1);
 259                 if (bytes > PAGE_SIZE-offset)
 260                         bytes = PAGE_SIZE-offset;
 261
 262                 maddr = kmap(page);
 263                 if (write) {
 264                         copy_to_user_page(vma, page, addr,
 265                                           maddr + offset, buf, bytes);
 266                         set_page_dirty_lock(page);
 267                 } else {
 268                         copy_from_user_page(vma, page, addr,
 269                                             buf, maddr + offset, bytes);
 270                 }
 271                 kunmap(page);
 272                 page_cache_release(page);
 273                 len -= bytes;
 274                 buf += bytes;
 275                 addr += bytes;
 276         }
 277         up_read(&mm->mmap_sem);
 278         mmput(mm);
 279
 280         return buf - old_buf;
 281 }
 282
 283 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
 284 {
 285         int copied = 0;
 286
 287         while (len > 0) {
 288                 char buf[128];
 289                 int this_len, retval;
 290
 291                 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 292                 retval = access_process_vm(tsk, src, buf, this_len, 0);
 293                 if (!retval) {
 294                         if (copied)
 295                                 break;
 296                         return -EIO;
 297                 }
 298                 if (copy_to_user(dst, buf, retval))
 299                         return -EFAULT;
 300                 copied += retval;
 301                 src += retval;
 302                 dst += retval;
 303                 len -= retval;
 304         }
 305         return copied;
 306 }
 307
 308 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
 309 {
 310         int copied = 0;
 311
 312         while (len > 0) {
 313                 char buf[128];
 314                 int this_len, retval;
 315
 316                 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 317                 if (copy_from_user(buf, src, this_len))
 318                         return -EFAULT;
 319                 retval = access_process_vm(tsk, dst, buf, this_len, 1);
 320                 if (!retval) {
 321                         if (copied)
 322                                 break;
 323                         return -EIO;
 324                 }
 325                 copied += retval;
 326                 src += retval;
 327                 dst += retval;
 328                 len -= retval;
 329         }
 330         return copied;
 331 }
 332
 333 static int ptrace_setoptions(struct task_struct *child, long data)
 334 {
 335         child->ptrace &= ~PT_TRACE_MASK;
 336
 337         if (data & PTRACE_O_TRACESYSGOOD)
 338                 child->ptrace |= PT_TRACESYSGOOD;
 339
 340         if (data & PTRACE_O_TRACEFORK)
 341                 child->ptrace |= PT_TRACE_FORK;
 342
 343         if (data & PTRACE_O_TRACEVFORK)
 344                 child->ptrace |= PT_TRACE_VFORK;
 345
 346         if (data & PTRACE_O_TRACECLONE)
 347                 child->ptrace |= PT_TRACE_CLONE;
 348
 349         if (data & PTRACE_O_TRACEEXEC)
 350                 child->ptrace |= PT_TRACE_EXEC;
 351
 352         if (data & PTRACE_O_TRACEVFORKDONE)
 353                 child->ptrace |= PT_TRACE_VFORK_DONE;
 354
 355         if (data & PTRACE_O_TRACEEXIT)
 356                 child->ptrace |= PT_TRACE_EXIT;
 357
 358         return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
 359 }
 360
 361 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t __user * data)
 362 {
 363         siginfo_t lastinfo;
 364         int error = -ESRCH;
 365
 366         read_lock(&tasklist_lock);
 367         if (likely(child->sighand != NULL)) {
 368                 error = -EINVAL;
 369                 spin_lock_irq(&child->sighand->siglock);
 370                 if (likely(child->last_siginfo != NULL)) {
 371                         lastinfo = *child->last_siginfo;
 372                         error = 0;
 373                 }
 374                 spin_unlock_irq(&child->sighand->siglock);
 375         }
 376         read_unlock(&tasklist_lock);
 377         if (!error)
 378                 return copy_siginfo_to_user(data, &lastinfo);
 379         return error;
 380 }
 381
 382 static int ptrace_setsiginfo(struct task_struct *child, siginfo_t __user * data)
 383 {
 384         siginfo_t newinfo;
 385         int error = -ESRCH;
 386
 387         if (copy_from_user(&newinfo, data, sizeof (siginfo_t)))
 388                 return -EFAULT;
 389
 390         read_lock(&tasklist_lock);
 391         if (likely(child->sighand != NULL)) {
 392                 error = -EINVAL;
 393                 spin_lock_irq(&child->sighand->siglock);
 394                 if (likely(child->last_siginfo != NULL)) {
 395                         *child->last_siginfo = newinfo;
 396                         error = 0;
 397                 }
 398                 spin_unlock_irq(&child->sighand->siglock);
 399         }
 400         read_unlock(&tasklist_lock);
 401         return error;
 402 }
 403
 404 int ptrace_request(struct task_struct *child, long request,
 405                    long addr, long data)
 406 {
 407         int ret = -EIO;
 408
 409         switch (request) {
 410 #ifdef PTRACE_OLDSETOPTIONS
 411         case PTRACE_OLDSETOPTIONS:
 412 #endif
 413         case PTRACE_SETOPTIONS:
 414                 ret = ptrace_setoptions(child, data);
 415                 break;
 416         case PTRACE_GETEVENTMSG:
 417                 ret = put_user(child->ptrace_message, (unsigned long __user *) data);
 418                 break;
 419         case PTRACE_GETSIGINFO:
 420                 ret = ptrace_getsiginfo(child, (siginfo_t __user *) data);
 421                 break;
 422         case PTRACE_SETSIGINFO:
 423                 ret = ptrace_setsiginfo(child, (siginfo_t __user *) data);
 424                 break;
 425         default:
 426                 break;
 427         }
 428
 429         return ret;
 430 }
 431
 432 /**
 433  * ptrace_traceme  --  helper for PTRACE_TRACEME
 434  *
 435  * Performs checks and sets PT_PTRACED.
 436  * Should be used by all ptrace implementations for PTRACE_TRACEME.
 437  */
 438 int ptrace_traceme(void)
 439 {
 440         int ret = -EPERM;
 441
 442         /*
 443          * Are we already being traced?
 444          */
 445         task_lock(current);
 446         if (!(current->ptrace & PT_PTRACED)) {
 447                 ret = security_ptrace(current->parent, current);
 448                 /*
 449                  * Set the ptrace bit in the process ptrace flags.
 450                  */
 451                 if (!ret)
 452                         current->ptrace |= PT_PTRACED;
 453         }
 454         task_unlock(current);
 455         return ret;
 456 }
 457
 458 /**
 459  * ptrace_get_task_struct  --  grab a task struct reference for ptrace
 460  * @pid:       process id to grab a task_struct reference of
 461  *
 462  * This function is a helper for ptrace implementations.  It checks
 463  * permissions and then grabs a task struct for use of the actual
 464  * ptrace implementation.
 465  *
 466  * Returns the task_struct for @pid or an ERR_PTR() on failure.
 467  */
 468 struct task_struct *ptrace_get_task_struct(pid_t pid)
 469 {
 470         struct task_struct *child;
 471
 472         /*
 473          * Tracing init is not allowed.
 474          */
 475         if (pid == 1)
 476                 return ERR_PTR(-EPERM);
 477
 478         read_lock(&tasklist_lock);
 479         child = find_task_by_pid(pid);
 480         if (child)
 481                 get_task_struct(child);
 482         read_unlock(&tasklist_lock);
 483         if (!child)
 484                 return ERR_PTR(-ESRCH);
 485         return child;
 486 }
 487
 488 #ifndef __ARCH_SYS_PTRACE
 489 asmlinkage long sys_ptrace(long request, long pid, long addr, long data)
 490 {
 491         struct task_struct *child;
 492         long ret;
 493
 494         /*
 495          * This lock_kernel fixes a subtle race with suid exec
 496          */
 497         lock_kernel();
 498         if (request == PTRACE_TRACEME) {
 499                 ret = ptrace_traceme();
 500                 goto out;
 501         }
 502
 503         child = ptrace_get_task_struct(pid);
 504         if (IS_ERR(child)) {
 505                 ret = PTR_ERR(child);
 506                 goto out;
 507         }
 508
 509         if (request == PTRACE_ATTACH) {
 510                 ret = ptrace_attach(child);
 511                 goto out_put_task_struct;
 512         }
 513
 514         ret = ptrace_check_attach(child, request == PTRACE_KILL);
 515         if (ret < 0)
 516                 goto out_put_task_struct;
 517
 518         ret = arch_ptrace(child, request, addr, data);
 519         if (ret < 0)
 520                 goto out_put_task_struct;
 521
 522  out_put_task_struct:
 523         put_task_struct(child);
 524  out:
 525         unlock_kernel();
 526         return ret;
 527 }
 528 #endif /* __ARCH_SYS_PTRACE */