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(struct task_struct *child, struct task_struct *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(struct task_struct *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(struct task_struct *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         /* May we inspect the given task?
 124          * This check is used both for attaching with ptrace
 125          * and for allowing access to sensitive information in /proc.
 126          *
 127          * ptrace_attach denies several cases that /proc allows
 128          * because setting up the necessary parent/child relationship
 129          * or halting the specified task is impossible.
 130          */
 131         int dumpable = 0;
 132         /* Don't let security modules deny introspection */
 133         if (task == current)
 134                 return 0;
 135         if (((current->uid != task->euid) ||
 136              (current->uid != task->suid) ||
 137              (current->uid != task->uid) ||
 138              (current->gid != task->egid) ||
 139              (current->gid != task->sgid) ||
 140              (current->gid != task->gid)) && !capable(CAP_SYS_PTRACE))
 141                 return -EPERM;
 142         smp_rmb();
 143         if (task->mm)
 144                 dumpable = task->mm->dumpable;
 145         if (!dumpable && !capable(CAP_SYS_PTRACE))
 146                 return -EPERM;
 147
 148         return security_ptrace(current, task);
 149 }
 150
 151 int ptrace_may_attach(struct task_struct *task)
 152 {
 153         int err;
 154         task_lock(task);
 155         err = may_attach(task);
 156         task_unlock(task);
 157         return !err;
 158 }
 159
 160 int ptrace_attach(struct task_struct *task)
 161 {
 162         int retval;
 163
 164         retval = -EPERM;
 165         if (task->pid <= 1)
 166                 goto out;
 167         if (task->tgid == current->tgid)
 168                 goto out;
 169
 170 repeat:
 171         /*
 172          * Nasty, nasty.
 173          *
 174          * We want to hold both the task-lock and the
 175          * tasklist_lock for writing at the same time.
 176          * But that's against the rules (tasklist_lock
 177          * is taken for reading by interrupts on other
 178          * cpu's that may have task_lock).
 179          */
 180         task_lock(task);
 181         local_irq_disable();
 182         if (!write_trylock(&tasklist_lock)) {
 183                 local_irq_enable();
 184                 task_unlock(task);
 185                 do {
 186                         cpu_relax();
 187                 } while (!write_can_lock(&tasklist_lock));
 188                 goto repeat;
 189         }
 190
 191         if (!task->mm)
 192                 goto bad;
 193         /* the same process cannot be attached many times */
 194         if (task->ptrace & PT_PTRACED)
 195                 goto bad;
 196         retval = may_attach(task);
 197         if (retval)
 198                 goto bad;
 199
 200         /* Go */
 201         task->ptrace |= PT_PTRACED | ((task->real_parent != current)
 202                                       ? PT_ATTACHED : 0);
 203         if (capable(CAP_SYS_PTRACE))
 204                 task->ptrace |= PT_PTRACE_CAP;
 205
 206         __ptrace_link(task, current);
 207
 208         force_sig_specific(SIGSTOP, task);
 209
 210 bad:
 211         write_unlock_irq(&tasklist_lock);
 212         task_unlock(task);
 213 out:
 214         return retval;
 215 }
 216
 217 static inline void __ptrace_detach(struct task_struct *child, unsigned int data)
 218 {
 219         child->exit_code = data;
 220         /* .. re-parent .. */
 221         __ptrace_unlink(child);
 222         /* .. and wake it up. */
 223         if (child->exit_state != EXIT_ZOMBIE)
 224                 wake_up_process(child);
 225 }
 226
 227 int ptrace_detach(struct task_struct *child, unsigned int data)
 228 {
 229         if (!valid_signal(data))
 230                 return -EIO;
 231
 232         /* Architecture-specific hardware disable .. */
 233         ptrace_disable(child);
 234
 235         write_lock_irq(&tasklist_lock);
 236         /* protect against de_thread()->release_task() */
 237         if (child->ptrace)
 238                 __ptrace_detach(child, data);
 239         write_unlock_irq(&tasklist_lock);
 240
 241         return 0;
 242 }
 243
 244 /*
 245  * Access another process' address space.
 246  * Source/target buffer must be kernel space,
 247  * Do not walk the page table directly, use get_user_pages
 248  */
 249
 250 int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write)
 251 {
 252         struct mm_struct *mm;
 253         struct vm_area_struct *vma;
 254         struct page *page;
 255         void *old_buf = buf;
 256
 257         mm = get_task_mm(tsk);
 258         if (!mm)
 259                 return 0;
 260
 261         down_read(&mm->mmap_sem);
 262         /* ignore errors, just check how much was sucessfully transfered */
 263         while (len) {
 264                 int bytes, ret, offset;
 265                 void *maddr;
 266
 267                 ret = get_user_pages(tsk, mm, addr, 1,
 268                                 write, 1, &page, &vma);
 269                 if (ret <= 0)
 270                         break;
 271
 272                 bytes = len;
 273                 offset = addr & (PAGE_SIZE-1);
 274                 if (bytes > PAGE_SIZE-offset)
 275                         bytes = PAGE_SIZE-offset;
 276
 277                 maddr = kmap(page);
 278                 if (write) {
 279                         copy_to_user_page(vma, page, addr,
 280                                           maddr + offset, buf, bytes);
 281                         set_page_dirty_lock(page);
 282                 } else {
 283                         copy_from_user_page(vma, page, addr,
 284                                             buf, maddr + offset, bytes);
 285                 }
 286                 kunmap(page);
 287                 page_cache_release(page);
 288                 len -= bytes;
 289                 buf += bytes;
 290                 addr += bytes;
 291         }
 292         up_read(&mm->mmap_sem);
 293         mmput(mm);
 294
 295         return buf - old_buf;
 296 }
 297
 298 int ptrace_readdata(struct task_struct *tsk, unsigned long src, char __user *dst, int len)
 299 {
 300         int copied = 0;
 301
 302         while (len > 0) {
 303                 char buf[128];
 304                 int this_len, retval;
 305
 306                 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 307                 retval = access_process_vm(tsk, src, buf, this_len, 0);
 308                 if (!retval) {
 309                         if (copied)
 310                                 break;
 311                         return -EIO;
 312                 }
 313                 if (copy_to_user(dst, buf, retval))
 314                         return -EFAULT;
 315                 copied += retval;
 316                 src += retval;
 317                 dst += retval;
 318                 len -= retval;
 319         }
 320         return copied;
 321 }
 322
 323 int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned long dst, int len)
 324 {
 325         int copied = 0;
 326
 327         while (len > 0) {
 328                 char buf[128];
 329                 int this_len, retval;
 330
 331                 this_len = (len > sizeof(buf)) ? sizeof(buf) : len;
 332                 if (copy_from_user(buf, src, this_len))
 333                         return -EFAULT;
 334                 retval = access_process_vm(tsk, dst, buf, this_len, 1);
 335                 if (!retval) {
 336                         if (copied)
 337                                 break;
 338                         return -EIO;
 339                 }
 340                 copied += retval;
 341                 src += retval;
 342                 dst += retval;
 343                 len -= retval;
 344         }
 345         return copied;
 346 }
 347
 348 static int ptrace_setoptions(struct task_struct *child, long data)
 349 {
 350         child->ptrace &= ~PT_TRACE_MASK;
 351
 352         if (data & PTRACE_O_TRACESYSGOOD)
 353                 child->ptrace |= PT_TRACESYSGOOD;
 354
 355         if (data & PTRACE_O_TRACEFORK)
 356                 child->ptrace |= PT_TRACE_FORK;
 357
 358         if (data & PTRACE_O_TRACEVFORK)
 359                 child->ptrace |= PT_TRACE_VFORK;
 360
 361         if (data & PTRACE_O_TRACECLONE)
 362                 child->ptrace |= PT_TRACE_CLONE;
 363
 364         if (data & PTRACE_O_TRACEEXEC)
 365                 child->ptrace |= PT_TRACE_EXEC;
 366
 367         if (data & PTRACE_O_TRACEVFORKDONE)
 368                 child->ptrace |= PT_TRACE_VFORK_DONE;
 369
 370         if (data & PTRACE_O_TRACEEXIT)
 371                 child->ptrace |= PT_TRACE_EXIT;
 372
 373         return (data & ~PTRACE_O_MASK) ? -EINVAL : 0;
 374 }
 375
 376 static int ptrace_getsiginfo(struct task_struct *child, siginfo_t __user * data)
 377 {
 378         siginfo_t lastinfo;
 379         int error = -ESRCH;
 380
 381         read_lock(&tasklist_lock);
 382         if (likely(child->sighand != NULL)) {
 383                 error = -EINVAL;
 384                 spin_lock_irq(&child->sighand->siglock);
 385                 if (likely(child->last_siginfo != NULL)) {
 386                         lastinfo = *child->last_siginfo;
 387                         error = 0;
 388                 }
 389                 spin_unlock_irq(&child->sighand->siglock);
 390         }
 391         read_unlock(&tasklist_lock);
 392         if (!error)
 393                 return copy_siginfo_to_user(data, &lastinfo);
 394         return error;
 395 }
 396
 397 static int ptrace_setsiginfo(struct task_struct *child, siginfo_t __user * data)
 398 {
 399         siginfo_t newinfo;
 400         int error = -ESRCH;
 401
 402         if (copy_from_user(&newinfo, data, sizeof (siginfo_t)))
 403                 return -EFAULT;
 404
 405         read_lock(&tasklist_lock);
 406         if (likely(child->sighand != NULL)) {
 407                 error = -EINVAL;
 408                 spin_lock_irq(&child->sighand->siglock);
 409                 if (likely(child->last_siginfo != NULL)) {
 410                         *child->last_siginfo = newinfo;
 411                         error = 0;
 412                 }
 413                 spin_unlock_irq(&child->sighand->siglock);
 414         }
 415         read_unlock(&tasklist_lock);
 416         return error;
 417 }
 418
 419 int ptrace_request(struct task_struct *child, long request,
 420                    long addr, long data)
 421 {
 422         int ret = -EIO;
 423
 424         switch (request) {
 425 #ifdef PTRACE_OLDSETOPTIONS
 426         case PTRACE_OLDSETOPTIONS:
 427 #endif
 428         case PTRACE_SETOPTIONS:
 429                 ret = ptrace_setoptions(child, data);
 430                 break;
 431         case PTRACE_GETEVENTMSG:
 432                 ret = put_user(child->ptrace_message, (unsigned long __user *) data);
 433                 break;
 434         case PTRACE_GETSIGINFO:
 435                 ret = ptrace_getsiginfo(child, (siginfo_t __user *) data);
 436                 break;
 437         case PTRACE_SETSIGINFO:
 438                 ret = ptrace_setsiginfo(child, (siginfo_t __user *) data);
 439                 break;
 440         default:
 441                 break;
 442         }
 443
 444         return ret;
 445 }
 446
 447 /**
 448  * ptrace_traceme  --  helper for PTRACE_TRACEME
 449  *
 450  * Performs checks and sets PT_PTRACED.
 451  * Should be used by all ptrace implementations for PTRACE_TRACEME.
 452  */
 453 int ptrace_traceme(void)
 454 {
 455         int ret = -EPERM;
 456
 457         /*
 458          * Are we already being traced?
 459          */
 460         task_lock(current);
 461         if (!(current->ptrace & PT_PTRACED)) {
 462                 ret = security_ptrace(current->parent, current);
 463                 /*
 464                  * Set the ptrace bit in the process ptrace flags.
 465                  */
 466                 if (!ret)
 467                         current->ptrace |= PT_PTRACED;
 468         }
 469         task_unlock(current);
 470         return ret;
 471 }
 472
 473 /**
 474  * ptrace_get_task_struct  --  grab a task struct reference for ptrace
 475  * @pid:       process id to grab a task_struct reference of
 476  *
 477  * This function is a helper for ptrace implementations.  It checks
 478  * permissions and then grabs a task struct for use of the actual
 479  * ptrace implementation.
 480  *
 481  * Returns the task_struct for @pid or an ERR_PTR() on failure.
 482  */
 483 struct task_struct *ptrace_get_task_struct(pid_t pid)
 484 {
 485         struct task_struct *child;
 486
 487         /*
 488          * Tracing init is not allowed.
 489          */
 490         if (pid == 1)
 491                 return ERR_PTR(-EPERM);
 492
 493         read_lock(&tasklist_lock);
 494         child = find_task_by_pid(pid);
 495         if (child)
 496                 get_task_struct(child);
 497         read_unlock(&tasklist_lock);
 498         if (!child)
 499                 return ERR_PTR(-ESRCH);
 500         return child;
 501 }
 502
 503 #ifndef __ARCH_SYS_PTRACE
 504 asmlinkage long sys_ptrace(long request, long pid, long addr, long data)
 505 {
 506         struct task_struct *child;
 507         long ret;
 508
 509         /*
 510          * This lock_kernel fixes a subtle race with suid exec
 511          */
 512         lock_kernel();
 513         if (request == PTRACE_TRACEME) {
 514                 ret = ptrace_traceme();
 515                 goto out;
 516         }
 517
 518         child = ptrace_get_task_struct(pid);
 519         if (IS_ERR(child)) {
 520                 ret = PTR_ERR(child);
 521                 goto out;
 522         }
 523
 524         if (request == PTRACE_ATTACH) {
 525                 ret = ptrace_attach(child);
 526                 goto out_put_task_struct;
 527         }
 528
 529         ret = ptrace_check_attach(child, request == PTRACE_KILL);
 530         if (ret < 0)
 531                 goto out_put_task_struct;
 532
 533         ret = arch_ptrace(child, request, addr, data);
 534         if (ret < 0)
 535                 goto out_put_task_struct;
 536
 537  out_put_task_struct:
 538         put_task_struct(child);
 539  out:
 540         unlock_kernel();
 541         return ret;
 542 }
 543 #endif /* __ARCH_SYS_PTRACE */