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