FreeBSD Helgrind: turn off check for locks held on exit for FreeBSD 14.2

[valgrind.git] / coregrind / m_gdbserver / remote-utils.c

/* Remote utility routines for the remote server for GDB.
   Copyright (C) 1986, 1989, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
   2001, 2002, 2003, 2004, 2005, 2006, 2011
   Free Software Foundation, Inc.

   This file is part of GDB.
   It has been modified to integrate it in valgrind

   This program is free software; you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
   the Free Software Foundation; either version 2 of the License, or
   (at your option) any later version.

   This program is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   GNU General Public License for more details.

   You should have received a copy of the GNU General Public License
   along with this program; if not, write to the Free Software
   Foundation, Inc., 51 Franklin Street, Fifth Floor,
   Boston, MA 02110-1301, USA.  */

#include "pub_core_basics.h"
#include "pub_core_vki.h"
#include "pub_core_vkiscnums.h"
#include "pub_core_libcsignal.h"
#include "pub_core_options.h"
#include "pub_core_aspacemgr.h"
#include "pub_core_syswrap.h"

#include "server.h"

#  if defined(VGO_linux)
#include <sys/prctl.h>
#  endif

/* Calls sr_perror with msg.
   Outputs more information about Valgrind state if verbosity > 0
   or debuglog_getlevel > 0. */
static
void sr_extended_perror (SysRes sr, const HChar *msg)
{
   sr_perror (sr, "%s", msg);
   if (VG_(clo_verbosity) > 0 || VG_(debugLog_getLevel)() >= 1) {
      Int i;
      vki_sigset_t cursigset;
      VG_(show_sched_status) (True,  // host_stacktrace
                              True,  // stack_usage
                              True); // exited_threads
      VG_(sigprocmask) (0,           // dummy how.
                        NULL,        // do not change the sigmask
                        &cursigset); //
      VG_(dmsg)("current sigmask value { ");
      for (i = 1; i <= _VKI_NSIG; i++) {
         if (VG_(sigismember)(&cursigset, i))
            VG_(dmsg)("%d ", i);
      }
      VG_(dmsg)("}\n");
   }
}

/* Calls VG_(poll) with given arguments. If VG_(poll) fails due to EINTR,
   restarts the syscall.
   Normally, VG_(poll) gdbsrv syscalls are not supposed to be interrupted :
     either gdbsrv has been called by the scheduler (so all async signals
     are masked)
     or gdbsrv has been forced invoked by vgdb+ptrace, and vgdb is queuing
     the signals.

   However, on old kernels (such as on RHEL5.5 2.6.18), when vgdb+ptrace
   intercepts and queues an async signal, the poll syscall is not properly
   restarted. Instead, it returns EINTR even if no signal was effectively
   received by the ptraced process.
   See red-hat "Bug 679129 - Change in behaviour between RH5.5 and RH6
   with ptrace and syscalls bugzilla"
   e.g. "Why rhel5 differs? Because unlike in rhel6, sys_poll() returns
         -EINTR if interrupted, that is all. This old implementation does
         not support the restart-if-eintr-is-spurious."

   So in case VG_(poll) fails with EINTR, we retry. */
static SysRes VG_(poll_no_eintr) (struct vki_pollfd *fds, Int nfds, Int timeout)
{
  const HChar* msg = "VG_(poll) failed (old kernel ?) retrying ... \n";
  SysRes sr;
  do {
     sr = VG_(poll) (fds, nfds, timeout);
     if (!sr_isError(sr) || sr_Err(sr) != VKI_EINTR)
        return sr;
     sr_perror (sr, "%s", msg);
     if (VG_(debugLog_getLevel)() >= 1) {
        sr_extended_perror (sr, msg);
     }
  } while (1);
  /*NOTREACHED*/
}

Bool noack_mode;

static int readchar (int single);

void remote_utils_output_status(void);

#define INVALID_DESCRIPTOR -1
static int remote_desc = INVALID_DESCRIPTOR;

static VgdbShared *shared;
static int  last_looked_cntr = -1;
static struct vki_pollfd remote_desc_pollfdread_activity;

/* for a gdbserver embedded in valgrind, we read from a FIFO and write
   to another FIFO So, we need two descriptors */
static int write_remote_desc = INVALID_DESCRIPTOR;
static int pid_from_to_creator;
/* only this pid will remove the FIFOs: if an exec fails, we must avoid
   that the exiting child believes it has to remove the FIFOs of its parent */
static int mknod_done = 0;

static char *from_gdb = NULL;
static char *to_gdb = NULL;
static char *shared_mem = NULL;

static
int open_fifo (const char *side, const char *path, int flags)
{
  SysRes o;
  int fd;
  dlog(1, "Opening %s side %s\n", side, path);
  o = VG_(open) (path, flags, 0);
  if (sr_isError (o)) {
     sr_perror(o, "open fifo %s\n", path);
     fatal ("valgrind: fatal error: vgdb FIFO cannot be opened.\n");
  } else {
     fd = sr_Res(o);
     dlog(1, "result fd %d\n", fd);
  }
  fd = VG_(safe_fd)(fd);
  dlog(1, "result safe_fd %d\n", fd);
  if (fd == -1)
     fatal("safe_fd for vgdb FIFO failed\n");
  return fd;
}

void remote_utils_output_status(void)
{
   if (shared == NULL)
      VG_(umsg)("remote communication not initialized\n");
   else
      VG_(umsg)("shared->written_by_vgdb %d shared->seen_by_valgrind %d\n",
                shared->written_by_vgdb, shared->seen_by_valgrind);
}

/* Returns 0 if vgdb and connection state looks good,
   otherwise returns an int value telling which check failed. */
static
int vgdb_state_looks_bad(const char* where)
{
   if (VG_(kill)(shared->vgdb_pid, 0) != 0)
      return 1; // vgdb process does not exist anymore.

   if (remote_desc_activity(where) == 2)
      return 2; // check for error on remote desc shows a problem

   if (remote_desc == INVALID_DESCRIPTOR)
      return 3; // after check, remote_desc not ok anymore
       
   return 0; // all is ok.
}

void VG_(set_ptracer)(void)
{
#ifdef PR_SET_PTRACER
   SysRes o;
   const char *ptrace_scope_setting_file = "/proc/sys/kernel/yama/ptrace_scope";
   int fd;
   char ptrace_scope;
   int ret;

   o = VG_(open) (ptrace_scope_setting_file, VKI_O_RDONLY, 0);
   if (sr_isError(o)) {
      if (VG_(debugLog_getLevel)() >= 1) {
         sr_perror(o, "error VG_(open) %s\n", ptrace_scope_setting_file);
      }
      /* can't read setting. Assuming ptrace can be called by vgdb. */
      return;
   }
   fd = sr_Res(o);
   if (VG_(read) (fd, &ptrace_scope, 1) == 1) {
      dlog(1, "ptrace_scope %c\n", ptrace_scope);
      if (ptrace_scope != '0') {
         /* insufficient default ptrace_scope.
            Indicate to the kernel that we accept to be ptraced. */
#ifdef PR_SET_PTRACER_ANY
         ret = VG_(prctl) (PR_SET_PTRACER, PR_SET_PTRACER_ANY, 0, 0, 0);
         dlog(1, "set_ptracer to PR_SET_PTRACER_ANY result %d\n", ret);
#else
         ret = VG_(prctl) (PR_SET_PTRACER, 1, 0, 0, 0);
         dlog(1, "set_ptracer to 1 result %d\n", ret);
#endif
         if (ret)
            VG_(umsg)("error calling PR_SET_PTRACER, vgdb might block\n");
      }
   } else {
      dlog(0, "Could not read the ptrace_scope setting from %s\n",
           ptrace_scope_setting_file);
   }

   VG_(close) (fd);
#endif
}

/* returns 1 if one or more poll "errors" is set.
   Errors are: VKI_POLLERR or VKI_POLLHUP or VKI_POLLNAL */
static
int poll_cond (short revents)
{
   return (revents & (VKI_POLLERR | VKI_POLLHUP | VKI_POLLNVAL));
}

/* Ensures we have a valid write file descriptor.
   Returns 1 if we have a valid write file descriptor,
   0 if the write fd is not valid/cannot be opened. */
static
int ensure_write_remote_desc(void)
{
   struct vki_pollfd write_remote_desc_ok;
   SysRes ret;
   if (write_remote_desc != INVALID_DESCRIPTOR) {
      write_remote_desc_ok.fd = write_remote_desc;
      write_remote_desc_ok.events = VKI_POLLOUT;
      write_remote_desc_ok.revents = 0;
      ret = VG_(poll_no_eintr)(&write_remote_desc_ok, 1, 0);
      if (sr_isError(ret) 
          || (sr_Res(ret) > 0 && poll_cond(write_remote_desc_ok.revents))) {
        if (sr_isError(ret)) {
          sr_extended_perror(ret, "ensure_write_remote_desc: poll error\n");
        } else {
          dlog(0, "POLLcond %d closing write_remote_desc %d\n", 
               write_remote_desc_ok.revents, write_remote_desc);
        }
        VG_(close) (write_remote_desc);
        write_remote_desc = INVALID_DESCRIPTOR;
      }
   }
   if (write_remote_desc == INVALID_DESCRIPTOR) {
      /* open_fifo write will block if the receiving vgdb
         process is dead.  So, let's check for vgdb state to
         be reasonably sure someone is reading on the other
         side of the fifo. */
      if (!vgdb_state_looks_bad("bad?@ensure_write_remote_desc")) {
         write_remote_desc = open_fifo ("write", to_gdb, VKI_O_WRONLY);
      }
   }

   return (write_remote_desc != INVALID_DESCRIPTOR);
}

#if defined(VGO_darwin)
#define VKI_S_IFIFO 0010000
#endif
static
void safe_mknod (char *nod)
{
   SysRes m;
   m = VG_(mknod) (nod, VKI_S_IFIFO|0600, 0);
   if (sr_isError (m)) {
      if (sr_Err (m) == VKI_EEXIST) {
         if (VG_(clo_verbosity) > 1) {
            VG_(umsg)("%s already created\n", nod);
         }
      } else {
         sr_perror(m, "mknod %s\n", nod);
         VG_(umsg) ("valgrind: fatal error: vgdb FIFOs cannot be created.\n");
         VG_(exit)(1);
      }
   }
}

/* If remote_desc is not opened, open it.
   Setup remote_desc_pollfdread_activity. */
static void setup_remote_desc_for_reading (void)
{
   int save_fcntl_flags;

   if (remote_desc == INVALID_DESCRIPTOR) {
      /* we open the read side FIFO in non blocking mode
         We then set the fd in blocking mode.
         Opening in non-blocking read mode always succeeds while opening
         in non-blocking write mode succeeds only if the fifo is already
         opened in read mode. So, we wait till we have read the first
         character from the read side before opening the write side. */
      remote_desc = open_fifo ("read", from_gdb, VKI_O_RDONLY|VKI_O_NONBLOCK);
      save_fcntl_flags = VG_(fcntl) (remote_desc, VKI_F_GETFL, 0);
      VG_(fcntl) (remote_desc, VKI_F_SETFL, save_fcntl_flags & ~VKI_O_NONBLOCK);
   }
   remote_desc_pollfdread_activity.fd = remote_desc;
   remote_desc_pollfdread_activity.events = VKI_POLLIN;
   remote_desc_pollfdread_activity.revents = 0;
}

/* Open a connection to a remote debugger.
   NAME is the filename used for communication.  
   For Valgrind, name is the prefix for the two read and write FIFOs
   The two FIFOs names will be build by appending 
   -from-vgdb-to-pid-by-user-on-host and -to-vgdb-from-pid-by-user-on-host
   with pid being the pidnr of the valgrind process These two FIFOs
   will be created if not existing yet. They will be removed when
   the gdbserver connection is closed or the process exits */

void remote_open (const HChar *name)
{
   const HChar *user, *host;
   int len;
   VgdbShared vgdbinit;
   const int pid = VG_(getpid)();
   Addr addr_shared;
   SysRes o;
   int shared_mem_fd = INVALID_DESCRIPTOR;

   VG_(memset) (&vgdbinit, 0, sizeof (VgdbShared));
   vgdbinit = (VgdbShared) 
      {0, 0, (Addr) VG_(invoke_gdbserver),
       (Addr) VG_(threads), VG_N_THREADS, sizeof(ThreadState), 
       offsetof(ThreadState, status),
       offsetof(ThreadState, os_state) + offsetof(ThreadOSstate, lwpid),
       0
#if VEX_HOST_WORDSIZE == 8
         , 0
#endif
   };

   user = VG_(getenv)("LOGNAME");
   if (user == NULL) user = VG_(getenv)("USER");
   if (user == NULL) user = "???";
   if (VG_(strchr)(user, '/')) user = "???";

   host = VG_(getenv)("HOST");
   if (host == NULL) host = VG_(getenv)("HOSTNAME");
   if (host == NULL) host = "???";
   if (VG_(strchr)(host, '/')) host = "???";

   len = strlen(name) + strlen(user) + strlen(host) + 40;

   if (from_gdb != NULL) 
      free (from_gdb);
   from_gdb = malloc (len);
   if (to_gdb != NULL)
      free (to_gdb);
   to_gdb = malloc (len);
   if (shared_mem != NULL)
      free (shared_mem);
   shared_mem = malloc (len);
   /* below 3 lines must match the equivalent in vgdb.c */
   VG_(sprintf) (from_gdb,   "%s-from-vgdb-to-%d-by-%s-on-%s",    name,
                 pid, user, host);
   VG_(sprintf) (to_gdb,     "%s-to-vgdb-from-%d-by-%s-on-%s",    name,
                 pid, user, host);
   VG_(sprintf) (shared_mem, "%s-shared-mem-vgdb-%d-by-%s-on-%s", name,
                 pid, user, host);
   if (VG_(clo_verbosity) > 1) {
      VG_(umsg)("embedded gdbserver: reading from %s\n", from_gdb);
      VG_(umsg)("embedded gdbserver: writing to   %s\n", to_gdb);
      VG_(umsg)("embedded gdbserver: shared mem   %s\n", shared_mem);
      VG_(umsg)("\n");
      VG_(umsg)("TO CONTROL THIS PROCESS USING vgdb (which you probably\n"
                "don't want to do, unless you know exactly what you're doing,\n"
                "or are doing some strange experiment):\n"
                "  %s/../../bin/vgdb%s%s --pid=%d ...command...\n",
                VG_(libdir),
                (VG_(arg_vgdb_prefix) ? " " : ""),
                (VG_(arg_vgdb_prefix) ? VG_(arg_vgdb_prefix) : ""),
                pid);
   }
   if (VG_(clo_verbosity) > 1 
       || ((VG_(clo_vgdb_error) < 999999999
            || VG_(clo_vgdb_stop_at) != 0)
           && !(VG_(clo_launched_with_multi)))) {
      VG_(umsg)("\n");
      VG_(umsg)(
         "TO DEBUG THIS PROCESS USING GDB: start GDB like this\n"
         "  /path/to/gdb %s\n"
         "and then give GDB the following command\n"
         "  target remote | %s/../../bin/vgdb%s%s --pid=%d\n",
         VG_(args_the_exename),
         VG_(libdir),
         (VG_(arg_vgdb_prefix) ? " " : ""),
         (VG_(arg_vgdb_prefix) ? VG_(arg_vgdb_prefix) : ""),
         pid
      );
      VG_(umsg)("--pid is optional if only one valgrind process is running\n");
      VG_(umsg)("\n");
   }

   if (!mknod_done) {
      mknod_done++;
      VG_(set_ptracer)();
      /*
       * Unlink just in case a previous process with the same PID had been
       * killed and hence Valgrind hasn't had the chance yet to remove these.
       */
      VG_(unlink)(from_gdb);
      VG_(unlink)(to_gdb);
      VG_(unlink)(shared_mem);

      o = VG_(open) (shared_mem, VKI_O_CREAT|VKI_O_RDWR, 0600);
      if (sr_isError (o)) {
         sr_perror(o, "cannot create shared_mem file %s\n", shared_mem);
         fatal("Cannot recover from previous error. Good-bye.");
      } else {
         shared_mem_fd = sr_Res(o);
      }
      
      if (VG_(write)(shared_mem_fd, &vgdbinit, sizeof(VgdbShared)) 
          != sizeof(VgdbShared)) {
         fatal("error writing %d bytes to shared mem %s\n",
               (int) sizeof(VgdbShared), shared_mem);
      }
      {
         SysRes res = VG_(am_shared_mmap_file_float_valgrind)
            (sizeof(VgdbShared), VKI_PROT_READ|VKI_PROT_WRITE, 
             shared_mem_fd, (Off64T)0);
         if (sr_isError(res)) {
            sr_perror(res, "error VG_(am_shared_mmap_file_float_valgrind) %s\n",
                      shared_mem);
            fatal("Cannot recover from previous error. Good-bye.");
         }  
         addr_shared = sr_Res (res);
      }
      shared = (VgdbShared*) addr_shared;
      VG_(close) (shared_mem_fd);

      safe_mknod(to_gdb);
      safe_mknod(from_gdb);
      /* from_gdb is the last resource created: vgdb searches such FIFOs
         to detect the presence of a valgrind process.
         So, we better create this resource when all the rest needed by
         vgdb is ready : the other FIFO and the shared memory. */

      pid_from_to_creator = pid;
   }
   
   setup_remote_desc_for_reading ();
}

/* sync_gdb_connection wait a time long enough to let the connection
   be properly closed if needed when closing the connection (in case
   of detach or error), if we reopen it too quickly, it seems there
   are some events queued in the kernel concerning the "old"
   connection/remote_desc which are discovered with poll or select on
   the "new" connection/remote_desc.  We bypass this by waiting some
   time to let a proper cleanup to be donex */
void sync_gdb_connection(void)
{
   SysRes ret;
   ret = VG_(poll_no_eintr)(0, 0, 100);
   if (sr_isError(ret))
      sr_extended_perror(ret, "sync_gdb_connection: poll error\n");
}

static
const char * ppFinishReason (FinishReason reason)
{
   switch (reason) {
   case orderly_finish:    return "orderly_finish";
   case reset_after_error: return "reset_after_error";
   case reset_after_fork:  return "reset_after_fork";
   default: vg_assert (0);
   }
}

void remote_finish (FinishReason reason)
{
   dlog(1, "remote_finish (reason %s) %d %d\n", 
        ppFinishReason(reason), remote_desc, write_remote_desc);
   reset_valgrind_sink(ppFinishReason(reason));
   if (write_remote_desc != INVALID_DESCRIPTOR)
      VG_(close) (write_remote_desc);
   write_remote_desc = INVALID_DESCRIPTOR;
   
   if (remote_desc != INVALID_DESCRIPTOR) {
      /* Fully close the connection, either due to orderly_finish or
         to reset_after_fork or reset_after_error.  For
         reset_after_error, the FIFO will be re-opened soon.  This
         leaves a small window during which a race condition can
         happen between vgdb and a forking process: Just after fork,
         both the parent and the child have the FIFO open.  The child
         will close it asap (as part of the 'after fork cleanup').  If
         2 vgdbs are launched very quickly just after the fork, the
         parent will close its FIFO when the 1st vgdb exits.  Then if
         the 2nd vgdb is started before the parent has the time to
         re-open the FIFO, the 2nd vgdb will be able to open the FIFO
         (as it is still opened by the child).  The 2nd vgdb can then
         have a 'write' error when the child closes the FIFO.  After
         the 1st vgdb closes its FIFO write side, the parent gets EOF
         on its reading FIFO till it is closed and re-opened.  Opening
         a 2nd time the FIFO before closing the 'previous fd' solves
         this race condition, but causes other (not understood)
         problems due to too early re-invocation of gdbsrv.  Rather
         than to handle this race condition in gdbsrv side, we put a
         'retry' loop in vgdb for the initial write on the write
         FIFO. */
      remote_desc_pollfdread_activity.fd = INVALID_DESCRIPTOR;
      remote_desc_pollfdread_activity.events = 0;
      remote_desc_pollfdread_activity.revents = 0;
      VG_(close) (remote_desc);
      remote_desc = INVALID_DESCRIPTOR;
   }
   noack_mode = False;
   
   /* ensure the child will create its own FIFOs */
   if (reason == reset_after_fork)
      mknod_done = 0;
   
   if (reason == reset_after_error)
      sync_gdb_connection();
}

/* orderly close, cleans up everything */
void remote_close (void)
{
   const int pid = VG_(getpid)();
   remote_finish(orderly_finish);
   dlog(1, "%d (creator %d) maybe unlinking \n    %s\n    %s\n    %s\n", 
        pid, pid_from_to_creator,
        from_gdb ? from_gdb : "NULL",
        to_gdb ? to_gdb : "NULL",
        shared_mem ? shared_mem : "NULL");

   // PJF this is not ideal
   // if the guest enters capability mode then the unlink calls will fail
   // this may well also apply to Linux and seccomp
   // I don't have any thoughts on how to fix it, other than forking early on
   // having the child run the guest and the parent wait()ing and then
   // the parent doing the cleanup

   Bool unlinkPossible = True;
#if defined(VGO_freebsd)
   unlinkPossible = (VG_(get_capability_mode)() == False);
#endif

   if (unlinkPossible == True) {
      if (pid == pid_from_to_creator && from_gdb && VG_(unlink) (from_gdb) == -1)
         warning ("could not unlink %s\n", from_gdb);
      if (pid == pid_from_to_creator && to_gdb && VG_(unlink) (to_gdb) == -1)
         warning ("could not unlink %s\n", to_gdb);
      if (pid == pid_from_to_creator && shared_mem && VG_(unlink) (shared_mem) == -1)
         warning ("could not unlink %s\n", shared_mem);
   } else {
       VG_(debugLog)(1, "remote close",
                        "cannot unlink gdb pipes\n");
   }
   free (from_gdb);
   from_gdb = NULL;
   free (to_gdb);
   to_gdb = NULL;
   free (shared_mem);
   shared_mem = NULL;
}

Bool remote_connected(void)
{
   return write_remote_desc != INVALID_DESCRIPTOR;
}

/* cleanup after an error detected by poll_cond */
static
void error_poll_cond(void)
{
   /* if we will close the connection, we assume either that
      all characters have been seen or that they will be dropped. */
   shared->seen_by_valgrind = shared->written_by_vgdb;
   remote_finish(reset_after_error);
}

/* remote_desc_activity might be used at high frequency if the user
   gives a small value to --vgdb-poll. So, the function avoids
   doing repetitively system calls by rather looking at the
   counter values maintained in shared memory by vgdb. */
int remote_desc_activity(const char *msg)
{
   int retval;
   SysRes ret;
   const int looking_at = shared->written_by_vgdb;
   if (shared->seen_by_valgrind == looking_at)
      return 0;
   if (remote_desc == INVALID_DESCRIPTOR)
      return 0;

   /* poll the remote desc */
   remote_desc_pollfdread_activity.revents = 0;
   ret = VG_(poll_no_eintr) (&remote_desc_pollfdread_activity, 1, 0);
   if (sr_isError(ret)
       || (sr_Res(ret) && poll_cond(remote_desc_pollfdread_activity.revents))) {
     if (sr_isError(ret)) {
       sr_extended_perror(ret, "remote_desc_activity: poll error\n");
     } else {
       dlog(0, "POLLcond %d remote_desc_pollfdread %d\n", 
            remote_desc_pollfdread_activity.revents, remote_desc);
       error_poll_cond();
     }
     retval = 2;
   } else {
     retval = sr_Res(ret);
   }
   dlog(1,
        "remote_desc_activity %s %d last_looked_cntr %d looking_at %d"
        " shared->written_by_vgdb %d shared->seen_by_valgrind %d"
        " retval %d\n", 
        msg, remote_desc, last_looked_cntr, looking_at, 
        shared->written_by_vgdb, shared->seen_by_valgrind,
        retval);
   /* if no error from poll, indicate we have "seen" up to looking_at */
   if (retval == 1)
      last_looked_cntr = looking_at;
   return retval;
}

/* Convert hex digit A to a number.  */

static
int fromhex (int a)
{
   if (a >= '0' && a <= '9')
      return a - '0';
   else if (a >= 'a' && a <= 'f')
      return a - 'a' + 10;
   else
     error ("Reply contains invalid hex digit 0x%x\n", (unsigned)a);
   return 0;
}

int unhexify (char *bin, const char *hex, int count)
{
   int i;
   
   for (i = 0; i < count; i++) {
      if (hex[0] == 0 || hex[1] == 0) {
         /* Hex string is short, or of uneven length.
            Return the count that has been converted so far. */
         return i;
      }
      *bin++ = fromhex (hex[0]) * 16 + fromhex (hex[1]);
      hex += 2;
   }
   return i;
}

void decode_address (CORE_ADDR *addrp, const char *start, int len)
{
   CORE_ADDR addr;
   char ch;
   int i;
   
   addr = 0;
   for (i = 0; i < len; i++) {
      ch = start[i];
      addr = addr << 4;
      addr = addr | (fromhex (ch) & 0x0f);
   }
   *addrp = addr;
}

/* builds an image of bin according to byte order of the architecture 
   Useful for register and int image */
char* heximage (char *buf, const char *bin, int count)
{
#if (VKI_LITTLE_ENDIAN)
   char rev[count]; 
   /* note: no need for trailing \0, length is known with count */
   int i;
   for (i = 0; i < count; i++)
      rev[i] = bin[count - i - 1];
   hexify (buf, rev, count);
#else
   hexify (buf, bin, count);
#endif
   return buf;
}

void* C2v(CORE_ADDR addr)
{
   return (void*) addr;
}


/* Convert BUFFER, binary data at least LEN bytes long, into escaped
   binary data in OUT_BUF.  Set *OUT_LEN to the length of the data
   encoded in OUT_BUF, and return the number of bytes in OUT_BUF
   (which may be more than *OUT_LEN due to escape characters).  The
   total number of bytes in the output buffer will be at most
   OUT_MAXLEN.  */

int
remote_escape_output (const gdb_byte *buffer, int len,
                      gdb_byte *out_buf, int *out_len,
                      int out_maxlen)
{
   int input_index, output_index;
   
   output_index = 0;
   for (input_index = 0; input_index < len; input_index++) {
      gdb_byte b = buffer[input_index];

      if (b == '$' || b == '#' || b == '}' || b == '*') {
         /* These must be escaped.  */
         if (output_index + 2 > out_maxlen)
            break;
         out_buf[output_index++] = '}';
         out_buf[output_index++] = b ^ 0x20;
      } else {
         if (output_index + 1 > out_maxlen)
            break;
         out_buf[output_index++] = b;
      }
   }

   *out_len = input_index;
   return output_index;
}

/* Convert BUFFER, escaped data LEN bytes long, into binary data
   in OUT_BUF.  Return the number of bytes written to OUT_BUF.
   Raise an error if the total number of bytes exceeds OUT_MAXLEN.

   This function reverses remote_escape_output.  It allows more
   escaped characters than that function does, in particular because
   '*' must be escaped to avoid the run-length encoding processing
   in reading packets.  */

static
int remote_unescape_input (const gdb_byte *buffer, int len,
                       gdb_byte *out_buf, int out_maxlen)
{
   int input_index, output_index;
   int escaped;
   
   output_index = 0;
   escaped = 0;
   for (input_index = 0; input_index < len; input_index++) {
      gdb_byte b = buffer[input_index];

      if (output_index + 1 > out_maxlen)
         error ("Received too much data (len %d) from the target.\n", len);

      if (escaped) {
         out_buf[output_index++] = b ^ 0x20;
         escaped = 0;
      } else if (b == '}') {
         escaped = 1;
      } else {
         out_buf[output_index++] = b;
      }
   }

   if (escaped)
      error ("Unmatched escape character in target response.\n");

   return output_index;
}

/* Look for a sequence of characters which can be run-length encoded.
   If there are any, update *CSUM and *P.  Otherwise, output the
   single character.  Return the number of characters consumed.  */

static
int try_rle (char *buf, int remaining, unsigned char *csum, char **p)
{
   int n;

   /* Always output the character.  */
   *csum += buf[0];
   *(*p)++ = buf[0];

   /* Don't go past '~'.  */
   if (remaining > 97)
      remaining = 97;
   
   for (n = 1; n < remaining; n++)
      if (buf[n] != buf[0])
         break;
   
   /* N is the index of the first character not the same as buf[0].
      buf[0] is counted twice, so by decrementing N, we get the number
      of characters the RLE sequence will replace.  */
   n--;
   
   if (n < 3)
      return 1;
   
   /* Skip the frame characters.  The manual says to skip '+' and '-'
      also, but there's no reason to.  Unfortunately these two unusable
      characters double the encoded length of a four byte zero
      value.  */
   while (n + 29 == '$' || n + 29 == '#')
      n--;
   
   *csum += '*';
   *(*p)++ = '*';
   *csum += n + 29;
   *(*p)++ = n + 29;
  
   return n + 1;
}

/* Send a packet to the remote machine, with error checking.
   The data of the packet is in BUF, and the length of the
   packet is in CNT.  Returns >= 0 on success, -1 otherwise.  */

int putpkt_binary (char *buf, int cnt)
{
   int i;
   unsigned char csum = 0;
   char *buf2;
   char *p;
   int cc;

   buf2 = malloc (PBUFSIZ+POVERHSIZ);
   // should malloc PBUFSIZ, but bypass GDB bug (see gdbserver_init in server.c)
   vg_assert (5 == POVERHSIZ);
   vg_assert (cnt <= PBUFSIZ); // be tolerant for GDB bug.

   /* Copy the packet into buffer BUF2, encapsulating it
      and giving it a checksum.  */

   p = buf2;
   *p++ = '$';

   for (i = 0; i < cnt;)
      i += try_rle (buf + i, cnt - i, &csum, &p);

   *p++ = '#';
   *p++ = tohex ((csum >> 4) & 0xf);
   *p++ = tohex (csum & 0xf);

   *p = '\0';

   /* we might have to write a pkt when out FIFO not yet/anymore opened */
   if (!ensure_write_remote_desc()) {
      warning ("putpkt(write) error: no write_remote_desc\n");
      free (buf2);
      return -1;
   }

   /* Send it once (noack_mode)
      or send it over and over until we get a positive ack.  */

   do {
      if (VG_(write) (write_remote_desc, buf2, p - buf2) != p - buf2) {
         warning ("putpkt(write) error\n");
         free (buf2);
         return -1;
      }

      if (VG_(debugLog_getLevel)() >= 3) {
         char *tracebuf = malloc(4 * (p - buf2) + 1); // worst case
         char *tr = tracebuf;
                                                    
         for (UInt npr = 0; npr < p - buf2; npr++) {
            UChar uc = (unsigned char)buf2[npr];
            if (uc > 31 && uc < 127) {
               *tr++ = uc;
            } else {
               *tr++ = '\\';
               VG_(sprintf)(tr, "%03o", uc);
               tr += 3;
            }
         }
         *tr++ = 0;
         dlog(3, "putpkt (\"%s\"); (%slen %d) %s\n", tracebuf,
              strlen(tracebuf) == p - buf2 ? "binary " : "", 
              (int)(p - buf2),
              noack_mode ? "[no ack]" : "[looking for ack]");
         free (tracebuf);
      }

      if (noack_mode)
         break;

      cc = readchar (1);
      if (cc > 0)
         dlog(3, "[received '%c' (0x%x)]\n", cc, (unsigned)cc);

      if (cc <= 0) {
         if (cc == 0)
            dlog(1, "putpkt(read): Got EOF\n");
         else
            warning ("putpkt(read) error\n");

         free (buf2);
         return -1;
      }

      /* Check for an input interrupt while we're here.  */
      if (cc == '\003')
         dlog(1, "Received 0x03 character (SIGINT)\n");
   }
   while (cc != '+');

   free (buf2);
   return 1;                    /* Success! */
}

/* Send a packet to the remote machine, with error checking.  The data
   of the packet is in BUF, and the packet should be a NUL-terminated
   string.  Returns >= 0 on success, -1 otherwise.  */

int putpkt (char *buf)
{
   return putpkt_binary (buf, strlen (buf));
}

void monitor_output (char *s)
{
   if (remote_connected()) {
      const int len = strlen(s);
      char *buf = malloc(1 + 2*len + 1);
      
      buf[0] = 'O';
      hexify(buf+1, s, len);
      if (putpkt (buf) < 0) {
         /* We probably have lost the connection with vgdb. */
         reset_valgrind_sink("Error writing monitor output");
         /* write again after reset */
         VG_(printf) ("%s", s);
      }
      
      free (buf);
   } else {
      print_to_initial_valgrind_sink (s);
   }
}

/* Returns next char from remote GDB.  -1 if error.  */
/* if single, only one character maximum can be read with
   read system call. Otherwise, when reading an ack character
   we might pile up the next gdb command in the static buf.
   The read loop is then blocked in poll till gdb times out. */
static
int readchar (int single)
{
   static unsigned char buf[PBUFSIZ];
   static int bufcnt = 0;
   static unsigned char *bufp;
   SysRes ret;
  
   if (bufcnt-- > 0)
      return *bufp++;

   if (remote_desc == INVALID_DESCRIPTOR)
      return -1;

   /* No characters available in buf =>
      wait for some characters to arrive */
   remote_desc_pollfdread_activity.revents = 0;
   ret = VG_(poll_no_eintr)(&remote_desc_pollfdread_activity, 1, -1);
   if (sr_isError(ret) || sr_Res(ret) != 1) {
     if (sr_isError(ret)) {
        sr_extended_perror(ret, "readchar: poll error\n");
     } else {
        dlog(0, "readchar: poll got %d, expecting 1\n", (int)sr_Res(ret));
     }
     return -1;
   }
   if (single)
      bufcnt = VG_(read) (remote_desc, buf, 1);
   else
      bufcnt = VG_(read) (remote_desc, buf, sizeof (buf));

   if (bufcnt <= 0) {
      if (bufcnt == 0)
         dlog (1, "readchar: Got EOF\n");
      else
         warning ("readchar read error\n");
      
      return -1;
   }

   shared->seen_by_valgrind += bufcnt;
   
   /* If we have received a character and we do not yet have a
      connection, we better open our "write" fifo to let vgdb open its
      read fifo side */
   if (write_remote_desc == INVALID_DESCRIPTOR 
       && !ensure_write_remote_desc()) {
      dlog(1, "reachar: write_remote_desc could not be created");
   }

   bufp = buf;
   bufcnt--;

   if (poll_cond(remote_desc_pollfdread_activity.revents)) {
      dlog(1, "readchar: POLLcond got %d\n",
           remote_desc_pollfdread_activity.revents);
      error_poll_cond();
   }

   return *bufp++;
}


/* Read a packet from the remote machine, with error checking,
   and store it in BUF.  Returns length of packet, or negative if error. */

int getpkt (char *buf)
{
   char *bp;
   unsigned char csum, c1, c2;
   int c;
  
   while (1) {
      csum = 0;
      
      while (1) {
         c = readchar (0);
         if (c == '$')
            break;
         dlog(3, "[getpkt: discarding char '%c']\n", c);
         if (c < 0)
            return -1;
      }

      bp = buf;
      while (1) {
         c = readchar (0);
         if (c < 0)
            return -1;
         if (c == '#')
            break;
         *bp++ = c;
         csum += c;
      }
      *bp = 0;

      c1 = fromhex (readchar (0));
      c2 = fromhex (readchar (0));

      if (csum == (c1 << 4) + c2)
         break;

      dlog (0, "Bad checksum, sentsum=0x%x, csum=0x%x, buf=%s\n",
            (unsigned)(c1 << 4) + c2, (unsigned)csum, buf);
      if (!ensure_write_remote_desc()) {
         dlog(1, "getpkt(write nack) no write_remote_desc");
      }
      VG_(write) (write_remote_desc, "-", 1);
   }

   if (noack_mode)
      dlog(3, "getpkt (\"%s\");  [no ack] \n", buf);
   else
      dlog(3, "getpkt (\"%s\");  [sending ack] \n", buf);

   if (!noack_mode) {
      if (!ensure_write_remote_desc()) {
         dlog(1, "getpkt(write ack) no write_remote_desc");
      }
      VG_(write) (write_remote_desc, "+", 1);
      dlog(3, "[sent ack]\n");
   }

   return bp - buf;
}

void write_ok (char *buf)
{
   buf[0] = 'O';
   buf[1] = 'K';
   buf[2] = '\0';
}

void write_enn (char *buf)
{
   /* Some day, we should define the meanings of the error codes... */
   buf[0] = 'E';
   buf[1] = '0';
   buf[2] = '1';
   buf[3] = '\0';
}

void convert_int_to_ascii (const unsigned char *from, char *to, int n)
{
   int nib;
   int ch;
   while (n--) {
      ch = *from++;
      nib = ((ch & 0xf0) >> 4) & 0x0f;
      *to++ = tohex (nib);
      nib = ch & 0x0f;
      *to++ = tohex (nib);
   }
   *to++ = 0;
}


void convert_ascii_to_int (const char *from, unsigned char *to, int n)
{
   int nib1, nib2;
   while (n--) {
      nib1 = fromhex (*from++);
      nib2 = fromhex (*from++);
      *to++ = (((nib1 & 0x0f) << 4) & 0xf0) | (nib2 & 0x0f);
   }
}

static
char * outreg (int regno, char *buf)
{
   if ((regno >> 12) != 0)
      *buf++ = tohex ((regno >> 12) & 0xf);
   if ((regno >> 8) != 0)
      *buf++ = tohex ((regno >> 8) & 0xf);
   *buf++ = tohex ((regno >> 4) & 0xf);
   *buf++ = tohex (regno & 0xf);
   *buf++ = ':';
   collect_register_as_string (regno, buf);
   buf += 2 * register_size (regno);
   *buf++ = ';';

   return buf;
}

void prepare_resume_reply (char *buf, char status, unsigned char sig)
{
   int nib;
   
   *buf++ = status;

   nib = ((sig & 0xf0) >> 4);
   *buf++ = tohex (nib);
   nib = sig & 0x0f;
   *buf++ = tohex (nib);

   if (status == 'T') {
      const char **regp = gdbserver_expedite_regs;
      
      if (valgrind_stopped_by_watchpoint()) {
         CORE_ADDR addr;
         int i;

         strncpy (buf, "watch:", 6);
         buf += 6;

         addr = valgrind_stopped_data_address ();

         /* Convert each byte of the address into two hexadecimal chars.
            Note that we take sizeof (void *) instead of sizeof (addr);
            this is to avoid sending a 64-bit address to a 32-bit GDB.  */
         for (i = sizeof (void *) * 2; i > 0; i--) {
            *buf++ = tohex ((addr >> (i - 1) * 4) & 0xf);
         }
         *buf++ = ';';
      }

      if (valgrind_stopped_by_syscall () >= 0) {
         VG_(sprintf) (buf, "%s:%x;",
                       valgrind_stopped_before_syscall ()
                       ? "syscall_entry" : "syscall_return",
                       (UInt)valgrind_stopped_by_syscall ());
         buf += strlen (buf);
      }

      while (*regp) {
         buf = outreg (find_regno (*regp), buf);
         regp ++;
      }

      {
         unsigned int gdb_id_from_wait;
         
         /* FIXME right place to set this? */
         thread_from_wait = 
            ((struct inferior_list_entry *)current_inferior)->id;
         gdb_id_from_wait = thread_to_gdb_id (current_inferior);
         
         dlog(1, "Writing resume reply for %lu\n", thread_from_wait);
         /* This if (1) ought to be unnecessary.  But remote_wait in GDB
            will claim this event belongs to inferior_ptid if we do not
            specify a thread, and there's no way for gdbserver to know
            what inferior_ptid is.  */
         if (1 || old_thread_from_wait != thread_from_wait) {
            general_thread = thread_from_wait;
            VG_(sprintf) (buf, "thread:%x;", gdb_id_from_wait);
            buf += strlen (buf);
            old_thread_from_wait = thread_from_wait;
         }
      }
   }
   /* For W and X, we're done.  */
   *buf++ = 0;
}

void decode_m_packet (char *from, CORE_ADDR *mem_addr_ptr, unsigned int *len_ptr)
{
   int i = 0, j = 0;
   char ch;
   *mem_addr_ptr = *len_ptr = 0;

   while ((ch = from[i++]) != ',') {
      *mem_addr_ptr = *mem_addr_ptr << 4;
      *mem_addr_ptr |= fromhex (ch) & 0x0f;
   }

   for (j = 0; j < 4; j++) {
      if ((ch = from[i++]) == 0)
         break;
      *len_ptr = *len_ptr << 4;
      *len_ptr |= fromhex (ch) & 0x0f;
   }
}

void decode_M_packet (char *from, CORE_ADDR *mem_addr_ptr, unsigned int *len_ptr,
                 unsigned char *to)
{
   int i = 0;
   char ch;
   *mem_addr_ptr = *len_ptr = 0;

   while ((ch = from[i++]) != ',') {
      *mem_addr_ptr = *mem_addr_ptr << 4;
      *mem_addr_ptr |= fromhex (ch) & 0x0f;
   }

   while ((ch = from[i++]) != ':') {
      *len_ptr = *len_ptr << 4;
      *len_ptr |= fromhex (ch) & 0x0f;
   }

   convert_ascii_to_int (&from[i++], to, *len_ptr);
}

int decode_X_packet (char *from, int packet_len, CORE_ADDR *mem_addr_ptr,
                 unsigned int *len_ptr, unsigned char *to)
{
   int i = 0;
   char ch;
   *mem_addr_ptr = *len_ptr = 0;
   
   while ((ch = from[i++]) != ',') {
      *mem_addr_ptr = *mem_addr_ptr << 4;
      *mem_addr_ptr |= fromhex (ch) & 0x0f;
   }

   while ((ch = from[i++]) != ':') {
      *len_ptr = *len_ptr << 4;
      *len_ptr |= fromhex (ch) & 0x0f;
   }

   if (remote_unescape_input ((const gdb_byte *) &from[i], packet_len - i,
                              to, *len_ptr) != *len_ptr)
      return -1;

   return 0;
}


/* Return the default path prefix for the named pipes (FIFOs) used by vgdb/gdb
   to communicate with valgrind */
HChar *
VG_(vgdb_prefix_default)(void)
{
   static HChar *prefix;
   
   if (prefix == NULL) {
     const HChar *tmpdir = VG_(tmpdir)();
     prefix = malloc(strlen(tmpdir) + strlen("/vgdb-pipe") + 1);
     strcpy(prefix, tmpdir);
     strcat(prefix, "/vgdb-pipe");
   }
   return prefix;
}