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