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         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         BUG_ON(!child->ptrace);
  75
  76         child->ptrace = 0;
  77         if (!list_empty(&child->ptrace_list)) {
  78                 list_del_init(&child->ptrace_list);
  79                 remove_parent(child);
  80                 child->parent = child->real_parent;
  81                 add_parent(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 void __ptrace_detach(struct task_struct *child, unsigned int data)
 187 {
 188         child->exit_code = data;
 189         /* .. re-parent .. */
 190         __ptrace_unlink(child);
 191         /* .. and wake it up. */
 192         if (child->exit_state != EXIT_ZOMBIE)
 193                 wake_up_process(child);
 194 }
 195
 196 int ptrace_detach(struct task_struct *child, unsigned int data)
 197 {
 198         if (!valid_signal(data))
 199                 return -EIO;
 200
 201         /* Architecture-specific hardware disable .. */
 202         ptrace_disable(child);
 203
 204         write_lock_irq(&tasklist_lock);
 205         if (child->ptrace)
 206                 __ptrace_detach(child, data);
 207         write_unlock_irq(&tasklist_lock);
 208
 209         return 0;
 210 }
 211
 212 /*
 213  * Access another process' address space.
 214  * Source/target buffer must be kernel space,
 215  * Do not walk the page table directly, use get_user_pages
 216  */
 217
 218 int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write)
 219 {
 220         struct mm_struct *mm;
 221         struct vm_area_struct *vma;
 222         struct page *page;
 223         void *old_buf = buf;
 224
 225         mm = get_task_mm(tsk);
 226         if (!mm)
 227                 return 0;
 228
 229         down_read(&mm->mmap_sem);
 230         /* ignore errors, just check how much was sucessfully transfered */
 231         while (len) {
 232                 int bytes, ret, offset;
 233                 void *maddr;
 234
 235                 ret = get_user_pages(tsk, mm, addr, 1,
 236                                 write, 1, &page, &vma);
 237                 if (ret <= 0)
 238                         break;
 239
 240                 bytes = len;
 241                 offset = addr & (PAGE_SIZE-1);
 242                 if (bytes > PAGE_SIZE-offset)
 243                         bytes = PAGE_SIZE-offset;
 244
 245                 maddr = kmap(page);
 246                 if (write) {
 247                         copy_to_user_page(vma, page, addr,
 248                                           maddr + offset, buf, bytes);
 249                         set_page_dirty_lock(page);
 250                 } else {
 251                         copy_from_user_page(vma, page, addr,
 252                                             buf, maddr + offset, bytes);
 253                 }
 254                 kunmap(page);
 255                 page_cache_release(page);
 256                 len -= bytes;
 257                 buf += bytes;
 258                 addr += bytes;
 259         }
 260         up_read(&mm->mmap_sem);
 261         mmput(mm);
 262
 263         return buf - old_buf;
 264 }
 265
 266 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
 267 {
 268         int copied = 0;
 269
 270         while (len > 0) {
 271                 char buf[128];
 272                 int this_len, retval;
 273
 274                 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 275                 retval = access_process_vm(tsk, src, buf, this_len, 0);
 276                 if (!retval) {
 277                         if (copied)
 278                                 break;
 279                         return -EIO;
 280                 }
 281                 if (copy_to_user(dst, buf, retval))
 282                         return -EFAULT;
 283                 copied += retval;
 284                 src += retval;
 285                 dst += retval;
 286                 len -= retval;
 287         }
 288         return copied;
 289 }
 290
 291 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
 292 {
 293         int copied = 0;
 294
 295         while (len > 0) {
 296                 char buf[128];
 297                 int this_len, retval;
 298
 299                 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 300                 if (copy_from_user(buf, src, this_len))
 301                         return -EFAULT;
 302                 retval = access_process_vm(tsk, dst, buf, this_len, 1);
 303                 if (!retval) {
 304                         if (copied)
 305                                 break;
 306                         return -EIO;
 307                 }
 308                 copied += retval;
 309                 src += retval;
 310                 dst += retval;
 311                 len -= retval;
 312         }
 313         return copied;
 314 }
 315
 316 static int ptrace_setoptions(struct task_struct *child, long data)
 317 {
 318         child->ptrace &= ~PT_TRACE_MASK;
 319
 320         if (data & PTRACE_O_TRACESYSGOOD)
 321                 child->ptrace |= PT_TRACESYSGOOD;
 322
 323         if (data & PTRACE_O_TRACEFORK)
 324                 child->ptrace |= PT_TRACE_FORK;
 325
 326         if (data & PTRACE_O_TRACEVFORK)
 327                 child->ptrace |= PT_TRACE_VFORK;
 328
 329         if (data & PTRACE_O_TRACECLONE)
 330                 child->ptrace |= PT_TRACE_CLONE;
 331
 332         if (data & PTRACE_O_TRACEEXEC)
 333                 child->ptrace |= PT_TRACE_EXEC;
 334
 335         if (data & PTRACE_O_TRACEVFORKDONE)
 336                 child->ptrace |= PT_TRACE_VFORK_DONE;
 337
 338         if (data & PTRACE_O_TRACEEXIT)
 339                 child->ptrace |= PT_TRACE_EXIT;
 340
 341         return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
 342 }
 343
 344 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t __user * data)
 345 {
 346         siginfo_t lastinfo;
 347         int error = -ESRCH;
 348
 349         read_lock(&tasklist_lock);
 350         if (likely(child->sighand != NULL)) {
 351                 error = -EINVAL;
 352                 spin_lock_irq(&child->sighand->siglock);
 353                 if (likely(child->last_siginfo != NULL)) {
 354                         lastinfo = *child->last_siginfo;
 355                         error = 0;
 356                 }
 357                 spin_unlock_irq(&child->sighand->siglock);
 358         }
 359         read_unlock(&tasklist_lock);
 360         if (!error)
 361                 return copy_siginfo_to_user(data, &lastinfo);
 362         return error;
 363 }
 364
 365 static int ptrace_setsiginfo(struct task_struct *child, siginfo_t __user * data)
 366 {
 367         siginfo_t newinfo;
 368         int error = -ESRCH;
 369
 370         if (copy_from_user(&newinfo, data, sizeof (siginfo_t)))
 371                 return -EFAULT;
 372
 373         read_lock(&tasklist_lock);
 374         if (likely(child->sighand != NULL)) {
 375                 error = -EINVAL;
 376                 spin_lock_irq(&child->sighand->siglock);
 377                 if (likely(child->last_siginfo != NULL)) {
 378                         *child->last_siginfo = newinfo;
 379                         error = 0;
 380                 }
 381                 spin_unlock_irq(&child->sighand->siglock);
 382         }
 383         read_unlock(&tasklist_lock);
 384         return error;
 385 }
 386
 387 int ptrace_request(struct task_struct *child, long request,
 388                    long addr, long data)
 389 {
 390         int ret = -EIO;
 391
 392         switch (request) {
 393 #ifdef PTRACE_OLDSETOPTIONS
 394         case PTRACE_OLDSETOPTIONS:
 395 #endif
 396         case PTRACE_SETOPTIONS:
 397                 ret = ptrace_setoptions(child, data);
 398                 break;
 399         case PTRACE_GETEVENTMSG:
 400                 ret = put_user(child->ptrace_message, (unsigned long __user *) data);
 401                 break;
 402         case PTRACE_GETSIGINFO:
 403                 ret = ptrace_getsiginfo(child, (siginfo_t __user *) data);
 404                 break;
 405         case PTRACE_SETSIGINFO:
 406                 ret = ptrace_setsiginfo(child, (siginfo_t __user *) data);
 407                 break;
 408         default:
 409                 break;
 410         }
 411
 412         return ret;
 413 }
 414
 415 /**
 416  * ptrace_traceme  --  helper for PTRACE_TRACEME
 417  *
 418  * Performs checks and sets PT_PTRACED.
 419  * Should be used by all ptrace implementations for PTRACE_TRACEME.
 420  */
 421 int ptrace_traceme(void)
 422 {
 423         int ret;
 424
 425         /*
 426          * Are we already being traced?
 427          */
 428         if (current->ptrace & PT_PTRACED)
 429                 return -EPERM;
 430         ret = security_ptrace(current->parent, current);
 431         if (ret)
 432                 return -EPERM;
 433         /*
 434          * Set the ptrace bit in the process ptrace flags.
 435          */
 436         current->ptrace |= PT_PTRACED;
 437         return 0;
 438 }
 439
 440 /**
 441  * ptrace_get_task_struct  --  grab a task struct reference for ptrace
 442  * @pid:       process id to grab a task_struct reference of
 443  *
 444  * This function is a helper for ptrace implementations.  It checks
 445  * permissions and then grabs a task struct for use of the actual
 446  * ptrace implementation.
 447  *
 448  * Returns the task_struct for @pid or an ERR_PTR() on failure.
 449  */
 450 struct task_struct *ptrace_get_task_struct(pid_t pid)
 451 {
 452         struct task_struct *child;
 453
 454         /*
 455          * Tracing init is not allowed.
 456          */
 457         if (pid == 1)
 458                 return ERR_PTR(-EPERM);
 459
 460         read_lock(&tasklist_lock);
 461         child = find_task_by_pid(pid);
 462         if (child)
 463                 get_task_struct(child);
 464         read_unlock(&tasklist_lock);
 465         if (!child)
 466                 return ERR_PTR(-ESRCH);
 467         return child;
 468 }
 469
 470 #ifndef __ARCH_SYS_PTRACE
 471 asmlinkage long sys_ptrace(long request, long pid, long addr, long data)
 472 {
 473         struct task_struct *child;
 474         long ret;
 475
 476         /*
 477          * This lock_kernel fixes a subtle race with suid exec
 478          */
 479         lock_kernel();
 480         if (request == PTRACE_TRACEME) {
 481                 ret = ptrace_traceme();
 482                 goto out;
 483         }
 484
 485         child = ptrace_get_task_struct(pid);
 486         if (IS_ERR(child)) {
 487                 ret = PTR_ERR(child);
 488                 goto out;
 489         }
 490
 491         if (request == PTRACE_ATTACH) {
 492                 ret = ptrace_attach(child);
 493                 goto out_put_task_struct;
 494         }
 495
 496         ret = ptrace_check_attach(child, request == PTRACE_KILL);
 497         if (ret < 0)
 498                 goto out_put_task_struct;
 499
 500         ret = arch_ptrace(child, request, addr, data);
 501         if (ret < 0)
 502                 goto out_put_task_struct;
 503
 504  out_put_task_struct:
 505         put_task_struct(child);
 506  out:
 507         unlock_kernel();
 508         return ret;
 509 }
 510 #endif /* __ARCH_SYS_PTRACE */