Linux: don't resume new LWPs until we've pulled all events out of the kernel

[binutils-gdb.git] / gdb / gdbserver / linux-low.c

/* Low level interface to ptrace, for the remote server for GDB.
   Copyright (C) 1995-2015 Free Software Foundation, Inc.

   This file is part of GDB.

   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 3 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, see <http://www.gnu.org/licenses/>.  */

#include "server.h"
#include "linux-low.h"
#include "nat/linux-osdata.h"
#include "agent.h"

#include "nat/linux-nat.h"
#include "nat/linux-waitpid.h"
#include "gdb_wait.h"
#include <sys/ptrace.h>
#include "nat/linux-ptrace.h"
#include "nat/linux-procfs.h"
#include "nat/linux-personality.h"
#include <signal.h>
#include <sys/ioctl.h>
#include <fcntl.h>
#include <unistd.h>
#include <sys/syscall.h>
#include <sched.h>
#include <ctype.h>
#include <pwd.h>
#include <sys/types.h>
#include <dirent.h>
#include <sys/stat.h>
#include <sys/vfs.h>
#include <sys/uio.h>
#include "filestuff.h"
#include "tracepoint.h"
#include "hostio.h"
#ifndef ELFMAG0
/* Don't include <linux/elf.h> here.  If it got included by gdb_proc_service.h
   then ELFMAG0 will have been defined.  If it didn't get included by
   gdb_proc_service.h then including it will likely introduce a duplicate
   definition of elf_fpregset_t.  */
#include <elf.h>
#endif

#ifndef SPUFS_MAGIC
#define SPUFS_MAGIC 0x23c9b64e
#endif

#ifdef HAVE_PERSONALITY
# include <sys/personality.h>
# if !HAVE_DECL_ADDR_NO_RANDOMIZE
#  define ADDR_NO_RANDOMIZE 0x0040000
# endif
#endif

#ifndef O_LARGEFILE
#define O_LARGEFILE 0
#endif

#ifndef W_STOPCODE
#define W_STOPCODE(sig) ((sig) << 8 | 0x7f)
#endif

/* This is the kernel's hard limit.  Not to be confused with
   SIGRTMIN.  */
#ifndef __SIGRTMIN
#define __SIGRTMIN 32
#endif

/* Some targets did not define these ptrace constants from the start,
   so gdbserver defines them locally here.  In the future, these may
   be removed after they are added to asm/ptrace.h.  */
#if !(defined(PT_TEXT_ADDR) \
      || defined(PT_DATA_ADDR) \
      || defined(PT_TEXT_END_ADDR))
#if defined(__mcoldfire__)
/* These are still undefined in 3.10 kernels.  */
#define PT_TEXT_ADDR 49*4
#define PT_DATA_ADDR 50*4
#define PT_TEXT_END_ADDR  51*4
/* BFIN already defines these since at least 2.6.32 kernels.  */
#elif defined(BFIN)
#define PT_TEXT_ADDR 220
#define PT_TEXT_END_ADDR 224
#define PT_DATA_ADDR 228
/* These are still undefined in 3.10 kernels.  */
#elif defined(__TMS320C6X__)
#define PT_TEXT_ADDR     (0x10000*4)
#define PT_DATA_ADDR     (0x10004*4)
#define PT_TEXT_END_ADDR (0x10008*4)
#endif
#endif

#ifdef HAVE_LINUX_BTRACE
# include "nat/linux-btrace.h"
#endif

#ifndef HAVE_ELF32_AUXV_T
/* Copied from glibc's elf.h.  */
typedef struct
{
  uint32_t a_type;              /* Entry type */
  union
    {
      uint32_t a_val;           /* Integer value */
      /* We use to have pointer elements added here.  We cannot do that,
         though, since it does not work when using 32-bit definitions
         on 64-bit platforms and vice versa.  */
    } a_un;
} Elf32_auxv_t;
#endif

#ifndef HAVE_ELF64_AUXV_T
/* Copied from glibc's elf.h.  */
typedef struct
{
  uint64_t a_type;              /* Entry type */
  union
    {
      uint64_t a_val;           /* Integer value */
      /* We use to have pointer elements added here.  We cannot do that,
         though, since it does not work when using 32-bit definitions
         on 64-bit platforms and vice versa.  */
    } a_un;
} Elf64_auxv_t;
#endif

/* A list of all unknown processes which receive stop signals.  Some
   other process will presumably claim each of these as forked
   children momentarily.  */

struct simple_pid_list
{
  /* The process ID.  */
  int pid;

  /* The status as reported by waitpid.  */
  int status;

  /* Next in chain.  */
  struct simple_pid_list *next;
};
struct simple_pid_list *stopped_pids;

/* Trivial list manipulation functions to keep track of a list of new
   stopped processes.  */

static void
add_to_pid_list (struct simple_pid_list **listp, int pid, int status)
{
  struct simple_pid_list *new_pid = xmalloc (sizeof (struct simple_pid_list));

  new_pid->pid = pid;
  new_pid->status = status;
  new_pid->next = *listp;
  *listp = new_pid;
}

static int
pull_pid_from_list (struct simple_pid_list **listp, int pid, int *statusp)
{
  struct simple_pid_list **p;

  for (p = listp; *p != NULL; p = &(*p)->next)
    if ((*p)->pid == pid)
      {
        struct simple_pid_list *next = (*p)->next;

        *statusp = (*p)->status;
        xfree (*p);
        *p = next;
        return 1;
      }
  return 0;
}

enum stopping_threads_kind
  {
    /* Not stopping threads presently.  */
    NOT_STOPPING_THREADS,

    /* Stopping threads.  */
    STOPPING_THREADS,

    /* Stopping and suspending threads.  */
    STOPPING_AND_SUSPENDING_THREADS
  };

/* This is set while stop_all_lwps is in effect.  */
enum stopping_threads_kind stopping_threads = NOT_STOPPING_THREADS;

/* FIXME make into a target method?  */
int using_threads = 1;

/* True if we're presently stabilizing threads (moving them out of
   jump pads).  */
static int stabilizing_threads;

static void linux_resume_one_lwp (struct lwp_info *lwp,
                                  int step, int signal, siginfo_t *info);
static void linux_resume (struct thread_resume *resume_info, size_t n);
static void stop_all_lwps (int suspend, struct lwp_info *except);
static void unstop_all_lwps (int unsuspend, struct lwp_info *except);
static int linux_wait_for_event_filtered (ptid_t wait_ptid, ptid_t filter_ptid,
                                          int *wstat, int options);
static int linux_wait_for_event (ptid_t ptid, int *wstat, int options);
static struct lwp_info *add_lwp (ptid_t ptid);
static int linux_stopped_by_watchpoint (void);
static void mark_lwp_dead (struct lwp_info *lwp, int wstat);
static void proceed_all_lwps (void);
static int finish_step_over (struct lwp_info *lwp);
static int kill_lwp (unsigned long lwpid, int signo);

/* When the event-loop is doing a step-over, this points at the thread
   being stepped.  */
ptid_t step_over_bkpt;

/* True if the low target can hardware single-step.  Such targets
   don't need a BREAKPOINT_REINSERT_ADDR callback.  */

static int
can_hardware_single_step (void)
{
  return (the_low_target.breakpoint_reinsert_addr == NULL);
}

/* True if the low target supports memory breakpoints.  If so, we'll
   have a GET_PC implementation.  */

static int
supports_breakpoints (void)
{
  return (the_low_target.get_pc != NULL);
}

/* Returns true if this target can support fast tracepoints.  This
   does not mean that the in-process agent has been loaded in the
   inferior.  */

static int
supports_fast_tracepoints (void)
{
  return the_low_target.install_fast_tracepoint_jump_pad != NULL;
}

/* True if LWP is stopped in its stepping range.  */

static int
lwp_in_step_range (struct lwp_info *lwp)
{
  CORE_ADDR pc = lwp->stop_pc;

  return (pc >= lwp->step_range_start && pc < lwp->step_range_end);
}

struct pending_signals
{
  int signal;
  siginfo_t info;
  struct pending_signals *prev;
};

/* The read/write ends of the pipe registered as waitable file in the
   event loop.  */
static int linux_event_pipe[2] = { -1, -1 };

/* True if we're currently in async mode.  */
#define target_is_async_p() (linux_event_pipe[0] != -1)

static void send_sigstop (struct lwp_info *lwp);
static void wait_for_sigstop (void);

/* Return non-zero if HEADER is a 64-bit ELF file.  */

static int
elf_64_header_p (const Elf64_Ehdr *header, unsigned int *machine)
{
  if (header->e_ident[EI_MAG0] == ELFMAG0
      && header->e_ident[EI_MAG1] == ELFMAG1
      && header->e_ident[EI_MAG2] == ELFMAG2
      && header->e_ident[EI_MAG3] == ELFMAG3)
    {
      *machine = header->e_machine;
      return header->e_ident[EI_CLASS] == ELFCLASS64;

    }
  *machine = EM_NONE;
  return -1;
}

/* Return non-zero if FILE is a 64-bit ELF file,
   zero if the file is not a 64-bit ELF file,
   and -1 if the file is not accessible or doesn't exist.  */

static int
elf_64_file_p (const char *file, unsigned int *machine)
{
  Elf64_Ehdr header;
  int fd;

  fd = open (file, O_RDONLY);
  if (fd < 0)
    return -1;

  if (read (fd, &header, sizeof (header)) != sizeof (header))
    {
      close (fd);
      return 0;
    }
  close (fd);

  return elf_64_header_p (&header, machine);
}

/* Accepts an integer PID; Returns true if the executable PID is
   running is a 64-bit ELF file..  */

int
linux_pid_exe_is_elf_64_file (int pid, unsigned int *machine)
{
  char file[PATH_MAX];

  sprintf (file, "/proc/%d/exe", pid);
  return elf_64_file_p (file, machine);
}

static void
delete_lwp (struct lwp_info *lwp)
{
  struct thread_info *thr = get_lwp_thread (lwp);

  if (debug_threads)
    debug_printf ("deleting %ld\n", lwpid_of (thr));

  remove_thread (thr);
  free (lwp->arch_private);
  free (lwp);
}

/* Add a process to the common process list, and set its private
   data.  */

static struct process_info *
linux_add_process (int pid, int attached)
{
  struct process_info *proc;

  proc = add_process (pid, attached);
  proc->private = xcalloc (1, sizeof (*proc->private));

  /* Set the arch when the first LWP stops.  */
  proc->private->new_inferior = 1;

  if (the_low_target.new_process != NULL)
    proc->private->arch_private = the_low_target.new_process ();

  return proc;
}

static CORE_ADDR get_pc (struct lwp_info *lwp);

/* Handle a GNU/Linux extended wait response.  If we see a clone
   event, we need to add the new LWP to our list (and not report the
   trap to higher layers).  */

static void
handle_extended_wait (struct lwp_info *event_child, int wstat)
{
  int event = linux_ptrace_get_extended_event (wstat);
  struct thread_info *event_thr = get_lwp_thread (event_child);
  struct lwp_info *new_lwp;

  if (event == PTRACE_EVENT_CLONE)
    {
      ptid_t ptid;
      unsigned long new_pid;
      int ret, status;

      ptrace (PTRACE_GETEVENTMSG, lwpid_of (event_thr), (PTRACE_TYPE_ARG3) 0,
              &new_pid);

      /* If we haven't already seen the new PID stop, wait for it now.  */
      if (!pull_pid_from_list (&stopped_pids, new_pid, &status))
        {
          /* The new child has a pending SIGSTOP.  We can't affect it until it
             hits the SIGSTOP, but we're already attached.  */

          ret = my_waitpid (new_pid, &status, __WALL);

          if (ret == -1)
            perror_with_name ("waiting for new child");
          else if (ret != new_pid)
            warning ("wait returned unexpected PID %d", ret);
          else if (!WIFSTOPPED (status))
            warning ("wait returned unexpected status 0x%x", status);
        }

      if (debug_threads)
        debug_printf ("HEW: Got clone event "
                      "from LWP %ld, new child is LWP %ld\n",
                      lwpid_of (event_thr), new_pid);

      ptid = ptid_build (pid_of (event_thr), new_pid, 0);
      new_lwp = add_lwp (ptid);

      /* Either we're going to immediately resume the new thread
         or leave it stopped.  linux_resume_one_lwp is a nop if it
         thinks the thread is currently running, so set this first
         before calling linux_resume_one_lwp.  */
      new_lwp->stopped = 1;

     /* If we're suspending all threads, leave this one suspended
        too.  */
      if (stopping_threads == STOPPING_AND_SUSPENDING_THREADS)
        new_lwp->suspended = 1;

      /* Normally we will get the pending SIGSTOP.  But in some cases
         we might get another signal delivered to the group first.
         If we do get another signal, be sure not to lose it.  */
      if (WSTOPSIG (status) != SIGSTOP)
        {
          new_lwp->stop_expected = 1;
          new_lwp->status_pending_p = 1;
          new_lwp->status_pending = status;
        }
    }
}

/* Return the PC as read from the regcache of LWP, without any
   adjustment.  */

static CORE_ADDR
get_pc (struct lwp_info *lwp)
{
  struct thread_info *saved_thread;
  struct regcache *regcache;
  CORE_ADDR pc;

  if (the_low_target.get_pc == NULL)
    return 0;

  saved_thread = current_thread;
  current_thread = get_lwp_thread (lwp);

  regcache = get_thread_regcache (current_thread, 1);
  pc = (*the_low_target.get_pc) (regcache);

  if (debug_threads)
    debug_printf ("pc is 0x%lx\n", (long) pc);

  current_thread = saved_thread;
  return pc;
}

/* This function should only be called if LWP got a SIGTRAP.
   The SIGTRAP could mean several things.

   On i386, where decr_pc_after_break is non-zero:

   If we were single-stepping this process using PTRACE_SINGLESTEP, we
   will get only the one SIGTRAP.  The value of $eip will be the next
   instruction.  If the instruction we stepped over was a breakpoint,
   we need to decrement the PC.

   If we continue the process using PTRACE_CONT, we will get a
   SIGTRAP when we hit a breakpoint.  The value of $eip will be
   the instruction after the breakpoint (i.e. needs to be
   decremented).  If we report the SIGTRAP to GDB, we must also
   report the undecremented PC.  If the breakpoint is removed, we
   must resume at the decremented PC.

   On a non-decr_pc_after_break machine with hardware or kernel
   single-step:

   If we either single-step a breakpoint instruction, or continue and
   hit a breakpoint instruction, our PC will point at the breakpoint
   instruction.  */

static int
check_stopped_by_breakpoint (struct lwp_info *lwp)
{
  CORE_ADDR pc;
  CORE_ADDR sw_breakpoint_pc;
  struct thread_info *saved_thread;

  if (the_low_target.get_pc == NULL)
    return 0;

  pc = get_pc (lwp);
  sw_breakpoint_pc = pc - the_low_target.decr_pc_after_break;

  /* breakpoint_at reads from the current thread.  */
  saved_thread = current_thread;
  current_thread = get_lwp_thread (lwp);

  /* We may have just stepped a breakpoint instruction.  E.g., in
     non-stop mode, GDB first tells the thread A to step a range, and
     then the user inserts a breakpoint inside the range.  In that
     case, we need to report the breakpoint PC.  But, when we're
     trying to step past one of our own breakpoints, that happens to
     have been placed on top of a permanent breakpoint instruction, we
     shouldn't adjust the PC, otherwise the program would keep
     trapping the permanent breakpoint forever.  */
  if ((!lwp->stepping
       || (!ptid_equal (ptid_of (current_thread), step_over_bkpt)
           && lwp->stop_pc == sw_breakpoint_pc))
      && (*the_low_target.breakpoint_at) (sw_breakpoint_pc))
    {
      if (debug_threads)
        {
          struct thread_info *thr = get_lwp_thread (lwp);

          debug_printf ("CSBB: %s stopped by software breakpoint\n",
                        target_pid_to_str (ptid_of (thr)));
        }

      /* Back up the PC if necessary.  */
      if (pc != sw_breakpoint_pc)
        {
          struct regcache *regcache
            = get_thread_regcache (current_thread, 1);
          (*the_low_target.set_pc) (regcache, sw_breakpoint_pc);
        }

      lwp->stop_pc = sw_breakpoint_pc;
      lwp->stop_reason = LWP_STOPPED_BY_SW_BREAKPOINT;
      current_thread = saved_thread;
      return 1;
    }

  if (hardware_breakpoint_inserted_here (pc))
    {
      if (debug_threads)
        {
          struct thread_info *thr = get_lwp_thread (lwp);

          debug_printf ("CSBB: %s stopped by hardware breakpoint\n",
                        target_pid_to_str (ptid_of (thr)));
        }

      lwp->stop_pc = pc;
      lwp->stop_reason = LWP_STOPPED_BY_HW_BREAKPOINT;
      current_thread = saved_thread;
      return 1;
    }

  current_thread = saved_thread;
  return 0;
}

static struct lwp_info *
add_lwp (ptid_t ptid)
{
  struct lwp_info *lwp;

  lwp = (struct lwp_info *) xmalloc (sizeof (*lwp));
  memset (lwp, 0, sizeof (*lwp));

  if (the_low_target.new_thread != NULL)
    lwp->arch_private = the_low_target.new_thread ();

  lwp->thread = add_thread (ptid, lwp);

  return lwp;
}

/* Start an inferior process and returns its pid.
   ALLARGS is a vector of program-name and args. */

static int
linux_create_inferior (char *program, char **allargs)
{
  struct lwp_info *new_lwp;
  int pid;
  ptid_t ptid;
  struct cleanup *restore_personality
    = maybe_disable_address_space_randomization (disable_randomization);

#if defined(__UCLIBC__) && defined(HAS_NOMMU)
  pid = vfork ();
#else
  pid = fork ();
#endif
  if (pid < 0)
    perror_with_name ("fork");

  if (pid == 0)
    {
      close_most_fds ();
      ptrace (PTRACE_TRACEME, 0, (PTRACE_TYPE_ARG3) 0, (PTRACE_TYPE_ARG4) 0);

#ifndef __ANDROID__ /* Bionic doesn't use SIGRTMIN the way glibc does.  */
      signal (__SIGRTMIN + 1, SIG_DFL);
#endif

      setpgid (0, 0);

      /* If gdbserver is connected to gdb via stdio, redirect the inferior's
         stdout to stderr so that inferior i/o doesn't corrupt the connection.
         Also, redirect stdin to /dev/null.  */
      if (remote_connection_is_stdio ())
        {
          close (0);
          open ("/dev/null", O_RDONLY);
          dup2 (2, 1);
          if (write (2, "stdin/stdout redirected\n",
                     sizeof ("stdin/stdout redirected\n") - 1) < 0)
            {
              /* Errors ignored.  */;
            }
        }

      execv (program, allargs);
      if (errno == ENOENT)
        execvp (program, allargs);

      fprintf (stderr, "Cannot exec %s: %s.\n", program,
               strerror (errno));
      fflush (stderr);
      _exit (0177);
    }

  do_cleanups (restore_personality);

  linux_add_process (pid, 0);

  ptid = ptid_build (pid, pid, 0);
  new_lwp = add_lwp (ptid);
  new_lwp->must_set_ptrace_flags = 1;

  return pid;
}

/* Attach to an inferior process.  Returns 0 on success, ERRNO on
   error.  */

int
linux_attach_lwp (ptid_t ptid)
{
  struct lwp_info *new_lwp;
  int lwpid = ptid_get_lwp (ptid);

  if (ptrace (PTRACE_ATTACH, lwpid, (PTRACE_TYPE_ARG3) 0, (PTRACE_TYPE_ARG4) 0)
      != 0)
    return errno;

  new_lwp = add_lwp (ptid);

  /* We need to wait for SIGSTOP before being able to make the next
     ptrace call on this LWP.  */
  new_lwp->must_set_ptrace_flags = 1;

  if (linux_proc_pid_is_stopped (lwpid))
    {
      if (debug_threads)
        debug_printf ("Attached to a stopped process\n");

      /* The process is definitely stopped.  It is in a job control
         stop, unless the kernel predates the TASK_STOPPED /
         TASK_TRACED distinction, in which case it might be in a
         ptrace stop.  Make sure it is in a ptrace stop; from there we
         can kill it, signal it, et cetera.

         First make sure there is a pending SIGSTOP.  Since we are
         already attached, the process can not transition from stopped
         to running without a PTRACE_CONT; so we know this signal will
         go into the queue.  The SIGSTOP generated by PTRACE_ATTACH is
         probably already in the queue (unless this kernel is old
         enough to use TASK_STOPPED for ptrace stops); but since
         SIGSTOP is not an RT signal, it can only be queued once.  */
      kill_lwp (lwpid, SIGSTOP);

      /* Finally, resume the stopped process.  This will deliver the
         SIGSTOP (or a higher priority signal, just like normal
         PTRACE_ATTACH), which we'll catch later on.  */
      ptrace (PTRACE_CONT, lwpid, (PTRACE_TYPE_ARG3) 0, (PTRACE_TYPE_ARG4) 0);
    }

  /* The next time we wait for this LWP we'll see a SIGSTOP as PTRACE_ATTACH
     brings it to a halt.

     There are several cases to consider here:

     1) gdbserver has already attached to the process and is being notified
        of a new thread that is being created.
        In this case we should ignore that SIGSTOP and resume the
        process.  This is handled below by setting stop_expected = 1,
        and the fact that add_thread sets last_resume_kind ==
        resume_continue.

     2) This is the first thread (the process thread), and we're attaching
        to it via attach_inferior.
        In this case we want the process thread to stop.
        This is handled by having linux_attach set last_resume_kind ==
        resume_stop after we return.

        If the pid we are attaching to is also the tgid, we attach to and
        stop all the existing threads.  Otherwise, we attach to pid and
        ignore any other threads in the same group as this pid.

     3) GDB is connecting to gdbserver and is requesting an enumeration of all
        existing threads.
        In this case we want the thread to stop.
        FIXME: This case is currently not properly handled.
        We should wait for the SIGSTOP but don't.  Things work apparently
        because enough time passes between when we ptrace (ATTACH) and when
        gdb makes the next ptrace call on the thread.

     On the other hand, if we are currently trying to stop all threads, we
     should treat the new thread as if we had sent it a SIGSTOP.  This works
     because we are guaranteed that the add_lwp call above added us to the
     end of the list, and so the new thread has not yet reached
     wait_for_sigstop (but will).  */
  new_lwp->stop_expected = 1;

  return 0;
}

/* Callback for linux_proc_attach_tgid_threads.  Attach to PTID if not
   already attached.  Returns true if a new LWP is found, false
   otherwise.  */

static int
attach_proc_task_lwp_callback (ptid_t ptid)
{
  /* Is this a new thread?  */
  if (find_thread_ptid (ptid) == NULL)
    {
      int lwpid = ptid_get_lwp (ptid);
      int err;

      if (debug_threads)
        debug_printf ("Found new lwp %d\n", lwpid);

      err = linux_attach_lwp (ptid);

      /* Be quiet if we simply raced with the thread exiting.  EPERM
         is returned if the thread's task still exists, and is marked
         as exited or zombie, as well as other conditions, so in that
         case, confirm the status in /proc/PID/status.  */
      if (err == ESRCH
          || (err == EPERM && linux_proc_pid_is_gone (lwpid)))
        {
          if (debug_threads)
            {
              debug_printf ("Cannot attach to lwp %d: "
                            "thread is gone (%d: %s)\n",
                            lwpid, err, strerror (err));
            }
        }
      else if (err != 0)
        {
          warning (_("Cannot attach to lwp %d: %s"),
                   lwpid,
                   linux_ptrace_attach_fail_reason_string (ptid, err));
        }

      return 1;
    }
  return 0;
}

/* Attach to PID.  If PID is the tgid, attach to it and all
   of its threads.  */

static int
linux_attach (unsigned long pid)
{
  ptid_t ptid = ptid_build (pid, pid, 0);
  int err;

  /* Attach to PID.  We will check for other threads
     soon.  */
  err = linux_attach_lwp (ptid);
  if (err != 0)
    error ("Cannot attach to process %ld: %s",
           pid, linux_ptrace_attach_fail_reason_string (ptid, err));

  linux_add_process (pid, 1);

  if (!non_stop)
    {
      struct thread_info *thread;

     /* Don't ignore the initial SIGSTOP if we just attached to this
        process.  It will be collected by wait shortly.  */
      thread = find_thread_ptid (ptid_build (pid, pid, 0));
      thread->last_resume_kind = resume_stop;
    }

  /* We must attach to every LWP.  If /proc is mounted, use that to
     find them now.  On the one hand, the inferior may be using raw
     clone instead of using pthreads.  On the other hand, even if it
     is using pthreads, GDB may not be connected yet (thread_db needs
     to do symbol lookups, through qSymbol).  Also, thread_db walks
     structures in the inferior's address space to find the list of
     threads/LWPs, and those structures may well be corrupted.  Note
     that once thread_db is loaded, we'll still use it to list threads
     and associate pthread info with each LWP.  */
  linux_proc_attach_tgid_threads (pid, attach_proc_task_lwp_callback);
  return 0;
}

struct counter
{
  int pid;
  int count;
};

static int
second_thread_of_pid_p (struct inferior_list_entry *entry, void *args)
{
  struct counter *counter = args;

  if (ptid_get_pid (entry->id) == counter->pid)
    {
      if (++counter->count > 1)
        return 1;
    }

  return 0;
}

static int
last_thread_of_process_p (int pid)
{
  struct counter counter = { pid , 0 };

  return (find_inferior (&all_threads,
                         second_thread_of_pid_p, &counter) == NULL);
}

/* Kill LWP.  */

static void
linux_kill_one_lwp (struct lwp_info *lwp)
{
  struct thread_info *thr = get_lwp_thread (lwp);
  int pid = lwpid_of (thr);

  /* PTRACE_KILL is unreliable.  After stepping into a signal handler,
     there is no signal context, and ptrace(PTRACE_KILL) (or
     ptrace(PTRACE_CONT, SIGKILL), pretty much the same) acts like
     ptrace(CONT, pid, 0,0) and just resumes the tracee.  A better
     alternative is to kill with SIGKILL.  We only need one SIGKILL
     per process, not one for each thread.  But since we still support
     linuxthreads, and we also support debugging programs using raw
     clone without CLONE_THREAD, we send one for each thread.  For
     years, we used PTRACE_KILL only, so we're being a bit paranoid
     about some old kernels where PTRACE_KILL might work better
     (dubious if there are any such, but that's why it's paranoia), so
     we try SIGKILL first, PTRACE_KILL second, and so we're fine
     everywhere.  */

  errno = 0;
  kill_lwp (pid, SIGKILL);
  if (debug_threads)
    {
      int save_errno = errno;

      debug_printf ("LKL:  kill_lwp (SIGKILL) %s, 0, 0 (%s)\n",
                    target_pid_to_str (ptid_of (thr)),
                    save_errno ? strerror (save_errno) : "OK");
    }

  errno = 0;
  ptrace (PTRACE_KILL, pid, (PTRACE_TYPE_ARG3) 0, (PTRACE_TYPE_ARG4) 0);
  if (debug_threads)
    {
      int save_errno = errno;

      debug_printf ("LKL:  PTRACE_KILL %s, 0, 0 (%s)\n",
                    target_pid_to_str (ptid_of (thr)),
                    save_errno ? strerror (save_errno) : "OK");
    }
}

/* Kill LWP and wait for it to die.  */

static void
kill_wait_lwp (struct lwp_info *lwp)
{
  struct thread_info *thr = get_lwp_thread (lwp);
  int pid = ptid_get_pid (ptid_of (thr));
  int lwpid = ptid_get_lwp (ptid_of (thr));
  int wstat;
  int res;

  if (debug_threads)
    debug_printf ("kwl: killing lwp %d, for pid: %d\n", lwpid, pid);

  do
    {
      linux_kill_one_lwp (lwp);

      /* Make sure it died.  Notes:

         - The loop is most likely unnecessary.

         - We don't use linux_wait_for_event as that could delete lwps
           while we're iterating over them.  We're not interested in
           any pending status at this point, only in making sure all
           wait status on the kernel side are collected until the
           process is reaped.

         - We don't use __WALL here as the __WALL emulation relies on
           SIGCHLD, and killing a stopped process doesn't generate
           one, nor an exit status.
      */
      res = my_waitpid (lwpid, &wstat, 0);
      if (res == -1 && errno == ECHILD)
        res = my_waitpid (lwpid, &wstat, __WCLONE);
    } while (res > 0 && WIFSTOPPED (wstat));

  gdb_assert (res > 0);
}

/* Callback for `find_inferior'.  Kills an lwp of a given process,
   except the leader.  */

static int
kill_one_lwp_callback (struct inferior_list_entry *entry, void *args)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lwp = get_thread_lwp (thread);
  int pid = * (int *) args;

  if (ptid_get_pid (entry->id) != pid)
    return 0;

  /* We avoid killing the first thread here, because of a Linux kernel (at
     least 2.6.0-test7 through 2.6.8-rc4) bug; if we kill the parent before
     the children get a chance to be reaped, it will remain a zombie
     forever.  */

  if (lwpid_of (thread) == pid)
    {
      if (debug_threads)
        debug_printf ("lkop: is last of process %s\n",
                      target_pid_to_str (entry->id));
      return 0;
    }

  kill_wait_lwp (lwp);
  return 0;
}

static int
linux_kill (int pid)
{
  struct process_info *process;
  struct lwp_info *lwp;

  process = find_process_pid (pid);
  if (process == NULL)
    return -1;

  /* If we're killing a running inferior, make sure it is stopped
     first, as PTRACE_KILL will not work otherwise.  */
  stop_all_lwps (0, NULL);

  find_inferior (&all_threads, kill_one_lwp_callback , &pid);

  /* See the comment in linux_kill_one_lwp.  We did not kill the first
     thread in the list, so do so now.  */
  lwp = find_lwp_pid (pid_to_ptid (pid));

  if (lwp == NULL)
    {
      if (debug_threads)
        debug_printf ("lk_1: cannot find lwp for pid: %d\n",
                      pid);
    }
  else
    kill_wait_lwp (lwp);

  the_target->mourn (process);

  /* Since we presently can only stop all lwps of all processes, we
     need to unstop lwps of other processes.  */
  unstop_all_lwps (0, NULL);
  return 0;
}

/* Get pending signal of THREAD, for detaching purposes.  This is the
   signal the thread last stopped for, which we need to deliver to the
   thread when detaching, otherwise, it'd be suppressed/lost.  */

static int
get_detach_signal (struct thread_info *thread)
{
  enum gdb_signal signo = GDB_SIGNAL_0;
  int status;
  struct lwp_info *lp = get_thread_lwp (thread);

  if (lp->status_pending_p)
    status = lp->status_pending;
  else
    {
      /* If the thread had been suspended by gdbserver, and it stopped
         cleanly, then it'll have stopped with SIGSTOP.  But we don't
         want to deliver that SIGSTOP.  */
      if (thread->last_status.kind != TARGET_WAITKIND_STOPPED
          || thread->last_status.value.sig == GDB_SIGNAL_0)
        return 0;

      /* Otherwise, we may need to deliver the signal we
         intercepted.  */
      status = lp->last_status;
    }

  if (!WIFSTOPPED (status))
    {
      if (debug_threads)
        debug_printf ("GPS: lwp %s hasn't stopped: no pending signal\n",
                      target_pid_to_str (ptid_of (thread)));
      return 0;
    }

  /* Extended wait statuses aren't real SIGTRAPs.  */
  if (WSTOPSIG (status) == SIGTRAP && linux_is_extended_waitstatus (status))
    {
      if (debug_threads)
        debug_printf ("GPS: lwp %s had stopped with extended "
                      "status: no pending signal\n",
                      target_pid_to_str (ptid_of (thread)));
      return 0;
    }

  signo = gdb_signal_from_host (WSTOPSIG (status));

  if (program_signals_p && !program_signals[signo])
    {
      if (debug_threads)
        debug_printf ("GPS: lwp %s had signal %s, but it is in nopass state\n",
                      target_pid_to_str (ptid_of (thread)),
                      gdb_signal_to_string (signo));
      return 0;
    }
  else if (!program_signals_p
           /* If we have no way to know which signals GDB does not
              want to have passed to the program, assume
              SIGTRAP/SIGINT, which is GDB's default.  */
           && (signo == GDB_SIGNAL_TRAP || signo == GDB_SIGNAL_INT))
    {
      if (debug_threads)
        debug_printf ("GPS: lwp %s had signal %s, "
                      "but we don't know if we should pass it. "
                      "Default to not.\n",
                      target_pid_to_str (ptid_of (thread)),
                      gdb_signal_to_string (signo));
      return 0;
    }
  else
    {
      if (debug_threads)
        debug_printf ("GPS: lwp %s has pending signal %s: delivering it.\n",
                      target_pid_to_str (ptid_of (thread)),
                      gdb_signal_to_string (signo));

      return WSTOPSIG (status);
    }
}

static int
linux_detach_one_lwp (struct inferior_list_entry *entry, void *args)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lwp = get_thread_lwp (thread);
  int pid = * (int *) args;
  int sig;

  if (ptid_get_pid (entry->id) != pid)
    return 0;

  /* If there is a pending SIGSTOP, get rid of it.  */
  if (lwp->stop_expected)
    {
      if (debug_threads)
        debug_printf ("Sending SIGCONT to %s\n",
                      target_pid_to_str (ptid_of (thread)));

      kill_lwp (lwpid_of (thread), SIGCONT);
      lwp->stop_expected = 0;
    }

  /* Flush any pending changes to the process's registers.  */
  regcache_invalidate_thread (thread);

  /* Pass on any pending signal for this thread.  */
  sig = get_detach_signal (thread);

  /* Finally, let it resume.  */
  if (the_low_target.prepare_to_resume != NULL)
    the_low_target.prepare_to_resume (lwp);
  if (ptrace (PTRACE_DETACH, lwpid_of (thread), (PTRACE_TYPE_ARG3) 0,
              (PTRACE_TYPE_ARG4) (long) sig) < 0)
    error (_("Can't detach %s: %s"),
           target_pid_to_str (ptid_of (thread)),
           strerror (errno));

  delete_lwp (lwp);
  return 0;
}

static int
linux_detach (int pid)
{
  struct process_info *process;

  process = find_process_pid (pid);
  if (process == NULL)
    return -1;

  /* Stop all threads before detaching.  First, ptrace requires that
     the thread is stopped to sucessfully detach.  Second, thread_db
     may need to uninstall thread event breakpoints from memory, which
     only works with a stopped process anyway.  */
  stop_all_lwps (0, NULL);

#ifdef USE_THREAD_DB
  thread_db_detach (process);
#endif

  /* Stabilize threads (move out of jump pads).  */
  stabilize_threads ();

  find_inferior (&all_threads, linux_detach_one_lwp, &pid);

  the_target->mourn (process);

  /* Since we presently can only stop all lwps of all processes, we
     need to unstop lwps of other processes.  */
  unstop_all_lwps (0, NULL);
  return 0;
}

/* Remove all LWPs that belong to process PROC from the lwp list.  */

static int
delete_lwp_callback (struct inferior_list_entry *entry, void *proc)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lwp = get_thread_lwp (thread);
  struct process_info *process = proc;

  if (pid_of (thread) == pid_of (process))
    delete_lwp (lwp);

  return 0;
}

static void
linux_mourn (struct process_info *process)
{
  struct process_info_private *priv;

#ifdef USE_THREAD_DB
  thread_db_mourn (process);
#endif

  find_inferior (&all_threads, delete_lwp_callback, process);

  /* Freeing all private data.  */
  priv = process->private;
  free (priv->arch_private);
  free (priv);
  process->private = NULL;

  remove_process (process);
}

static void
linux_join (int pid)
{
  int status, ret;

  do {
    ret = my_waitpid (pid, &status, 0);
    if (WIFEXITED (status) || WIFSIGNALED (status))
      break;
  } while (ret != -1 || errno != ECHILD);
}

/* Return nonzero if the given thread is still alive.  */
static int
linux_thread_alive (ptid_t ptid)
{
  struct lwp_info *lwp = find_lwp_pid (ptid);

  /* We assume we always know if a thread exits.  If a whole process
     exited but we still haven't been able to report it to GDB, we'll
     hold on to the last lwp of the dead process.  */
  if (lwp != NULL)
    return !lwp->dead;
  else
    return 0;
}

/* Return 1 if this lwp still has an interesting status pending.  If
   not (e.g., it had stopped for a breakpoint that is gone), return
   false.  */

static int
thread_still_has_status_pending_p (struct thread_info *thread)
{
  struct lwp_info *lp = get_thread_lwp (thread);

  if (!lp->status_pending_p)
    return 0;

  /* If we got a `vCont;t', but we haven't reported a stop yet, do
     report any status pending the LWP may have.  */
  if (thread->last_resume_kind == resume_stop
      && thread->last_status.kind != TARGET_WAITKIND_IGNORE)
    return 0;

  if (thread->last_resume_kind != resume_stop
      && (lp->stop_reason == LWP_STOPPED_BY_SW_BREAKPOINT
          || lp->stop_reason == LWP_STOPPED_BY_HW_BREAKPOINT))
    {
      struct thread_info *saved_thread;
      CORE_ADDR pc;
      int discard = 0;

      gdb_assert (lp->last_status != 0);

      pc = get_pc (lp);

      saved_thread = current_thread;
      current_thread = thread;

      if (pc != lp->stop_pc)
        {
          if (debug_threads)
            debug_printf ("PC of %ld changed\n",
                          lwpid_of (thread));
          discard = 1;
        }
      else if (lp->stop_reason == LWP_STOPPED_BY_SW_BREAKPOINT
               && !(*the_low_target.breakpoint_at) (pc))
        {
          if (debug_threads)
            debug_printf ("previous SW breakpoint of %ld gone\n",
                          lwpid_of (thread));
          discard = 1;
        }
      else if (lp->stop_reason == LWP_STOPPED_BY_HW_BREAKPOINT
               && !hardware_breakpoint_inserted_here (pc))
        {
          if (debug_threads)
            debug_printf ("previous HW breakpoint of %ld gone\n",
                          lwpid_of (thread));
          discard = 1;
        }

      current_thread = saved_thread;

      if (discard)
        {
          if (debug_threads)
            debug_printf ("discarding pending breakpoint status\n");
          lp->status_pending_p = 0;
          return 0;
        }
    }

  return 1;
}

/* Return 1 if this lwp has an interesting status pending.  */
static int
status_pending_p_callback (struct inferior_list_entry *entry, void *arg)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lp = get_thread_lwp (thread);
  ptid_t ptid = * (ptid_t *) arg;

  /* Check if we're only interested in events from a specific process
     or its lwps.  */
  if (!ptid_equal (minus_one_ptid, ptid)
      && ptid_get_pid (ptid) != ptid_get_pid (thread->entry.id))
    return 0;

  if (lp->status_pending_p
      && !thread_still_has_status_pending_p (thread))
    {
      linux_resume_one_lwp (lp, lp->stepping, GDB_SIGNAL_0, NULL);
      return 0;
    }

  return lp->status_pending_p;
}

static int
same_lwp (struct inferior_list_entry *entry, void *data)
{
  ptid_t ptid = *(ptid_t *) data;
  int lwp;

  if (ptid_get_lwp (ptid) != 0)
    lwp = ptid_get_lwp (ptid);
  else
    lwp = ptid_get_pid (ptid);

  if (ptid_get_lwp (entry->id) == lwp)
    return 1;

  return 0;
}

struct lwp_info *
find_lwp_pid (ptid_t ptid)
{
  struct inferior_list_entry *thread
    = find_inferior (&all_threads, same_lwp, &ptid);

  if (thread == NULL)
    return NULL;

  return get_thread_lwp ((struct thread_info *) thread);
}

/* Return the number of known LWPs in the tgid given by PID.  */

static int
num_lwps (int pid)
{
  struct inferior_list_entry *inf, *tmp;
  int count = 0;

  ALL_INFERIORS (&all_threads, inf, tmp)
    {
      if (ptid_get_pid (inf->id) == pid)
        count++;
    }

  return count;
}

/* Detect zombie thread group leaders, and "exit" them.  We can't reap
   their exits until all other threads in the group have exited.  */

static void
check_zombie_leaders (void)
{
  struct process_info *proc, *tmp;

  ALL_PROCESSES (proc, tmp)
    {
      pid_t leader_pid = pid_of (proc);
      struct lwp_info *leader_lp;

      leader_lp = find_lwp_pid (pid_to_ptid (leader_pid));

      if (debug_threads)
        debug_printf ("leader_pid=%d, leader_lp!=NULL=%d, "
                      "num_lwps=%d, zombie=%d\n",
                      leader_pid, leader_lp!= NULL, num_lwps (leader_pid),
                      linux_proc_pid_is_zombie (leader_pid));

      if (leader_lp != NULL
          /* Check if there are other threads in the group, as we may
             have raced with the inferior simply exiting.  */
          && !last_thread_of_process_p (leader_pid)
          && linux_proc_pid_is_zombie (leader_pid))
        {
          /* A leader zombie can mean one of two things:

             - It exited, and there's an exit status pending
             available, or only the leader exited (not the whole
             program).  In the latter case, we can't waitpid the
             leader's exit status until all other threads are gone.

             - There are 3 or more threads in the group, and a thread
             other than the leader exec'd.  On an exec, the Linux
             kernel destroys all other threads (except the execing
             one) in the thread group, and resets the execing thread's
             tid to the tgid.  No exit notification is sent for the
             execing thread -- from the ptracer's perspective, it
             appears as though the execing thread just vanishes.
             Until we reap all other threads except the leader and the
             execing thread, the leader will be zombie, and the
             execing thread will be in `D (disc sleep)'.  As soon as
             all other threads are reaped, the execing thread changes
             it's tid to the tgid, and the previous (zombie) leader
             vanishes, giving place to the "new" leader.  We could try
             distinguishing the exit and exec cases, by waiting once
             more, and seeing if something comes out, but it doesn't
             sound useful.  The previous leader _does_ go away, and
             we'll re-add the new one once we see the exec event
             (which is just the same as what would happen if the
             previous leader did exit voluntarily before some other
             thread execs).  */

          if (debug_threads)
            fprintf (stderr,
                     "CZL: Thread group leader %d zombie "
                     "(it exited, or another thread execd).\n",
                     leader_pid);

          delete_lwp (leader_lp);
        }
    }
}

/* Callback for `find_inferior'.  Returns the first LWP that is not
   stopped.  ARG is a PTID filter.  */

static int
not_stopped_callback (struct inferior_list_entry *entry, void *arg)
{
  struct thread_info *thr = (struct thread_info *) entry;
  struct lwp_info *lwp;
  ptid_t filter = *(ptid_t *) arg;

  if (!ptid_match (ptid_of (thr), filter))
    return 0;

  lwp = get_thread_lwp (thr);
  if (!lwp->stopped)
    return 1;

  return 0;
}

/* This function should only be called if the LWP got a SIGTRAP.

   Handle any tracepoint steps or hits.  Return true if a tracepoint
   event was handled, 0 otherwise.  */

static int
handle_tracepoints (struct lwp_info *lwp)
{
  struct thread_info *tinfo = get_lwp_thread (lwp);
  int tpoint_related_event = 0;

  gdb_assert (lwp->suspended == 0);

  /* If this tracepoint hit causes a tracing stop, we'll immediately
     uninsert tracepoints.  To do this, we temporarily pause all
     threads, unpatch away, and then unpause threads.  We need to make
     sure the unpausing doesn't resume LWP too.  */
  lwp->suspended++;

  /* And we need to be sure that any all-threads-stopping doesn't try
     to move threads out of the jump pads, as it could deadlock the
     inferior (LWP could be in the jump pad, maybe even holding the
     lock.)  */

  /* Do any necessary step collect actions.  */
  tpoint_related_event |= tracepoint_finished_step (tinfo, lwp->stop_pc);

  tpoint_related_event |= handle_tracepoint_bkpts (tinfo, lwp->stop_pc);

  /* See if we just hit a tracepoint and do its main collect
     actions.  */
  tpoint_related_event |= tracepoint_was_hit (tinfo, lwp->stop_pc);

  lwp->suspended--;

  gdb_assert (lwp->suspended == 0);
  gdb_assert (!stabilizing_threads || lwp->collecting_fast_tracepoint);

  if (tpoint_related_event)
    {
      if (debug_threads)
        debug_printf ("got a tracepoint event\n");
      return 1;
    }

  return 0;
}

/* Convenience wrapper.  Returns true if LWP is presently collecting a
   fast tracepoint.  */

static int
linux_fast_tracepoint_collecting (struct lwp_info *lwp,
                                  struct fast_tpoint_collect_status *status)
{
  CORE_ADDR thread_area;
  struct thread_info *thread = get_lwp_thread (lwp);

  if (the_low_target.get_thread_area == NULL)
    return 0;

  /* Get the thread area address.  This is used to recognize which
     thread is which when tracing with the in-process agent library.
     We don't read anything from the address, and treat it as opaque;
     it's the address itself that we assume is unique per-thread.  */
  if ((*the_low_target.get_thread_area) (lwpid_of (thread), &thread_area) == -1)
    return 0;

  return fast_tracepoint_collecting (thread_area, lwp->stop_pc, status);
}

/* The reason we resume in the caller, is because we want to be able
   to pass lwp->status_pending as WSTAT, and we need to clear
   status_pending_p before resuming, otherwise, linux_resume_one_lwp
   refuses to resume.  */

static int
maybe_move_out_of_jump_pad (struct lwp_info *lwp, int *wstat)
{
  struct thread_info *saved_thread;

  saved_thread = current_thread;
  current_thread = get_lwp_thread (lwp);

  if ((wstat == NULL
       || (WIFSTOPPED (*wstat) && WSTOPSIG (*wstat) != SIGTRAP))
      && supports_fast_tracepoints ()
      && agent_loaded_p ())
    {
      struct fast_tpoint_collect_status status;
      int r;

      if (debug_threads)
        debug_printf ("Checking whether LWP %ld needs to move out of the "
                      "jump pad.\n",
                      lwpid_of (current_thread));

      r = linux_fast_tracepoint_collecting (lwp, &status);

      if (wstat == NULL
          || (WSTOPSIG (*wstat) != SIGILL
              && WSTOPSIG (*wstat) != SIGFPE
              && WSTOPSIG (*wstat) != SIGSEGV
              && WSTOPSIG (*wstat) != SIGBUS))
        {
          lwp->collecting_fast_tracepoint = r;

          if (r != 0)
            {
              if (r == 1 && lwp->exit_jump_pad_bkpt == NULL)
                {
                  /* Haven't executed the original instruction yet.
                     Set breakpoint there, and wait till it's hit,
                     then single-step until exiting the jump pad.  */
                  lwp->exit_jump_pad_bkpt
                    = set_breakpoint_at (status.adjusted_insn_addr, NULL);
                }

              if (debug_threads)
                debug_printf ("Checking whether LWP %ld needs to move out of "
                              "the jump pad...it does\n",
                              lwpid_of (current_thread));
              current_thread = saved_thread;

              return 1;
            }
        }
      else
        {
          /* If we get a synchronous signal while collecting, *and*
             while executing the (relocated) original instruction,
             reset the PC to point at the tpoint address, before
             reporting to GDB.  Otherwise, it's an IPA lib bug: just
             report the signal to GDB, and pray for the best.  */

          lwp->collecting_fast_tracepoint = 0;

          if (r != 0
              && (status.adjusted_insn_addr <= lwp->stop_pc
                  && lwp->stop_pc < status.adjusted_insn_addr_end))
            {
              siginfo_t info;
              struct regcache *regcache;

              /* The si_addr on a few signals references the address
                 of the faulting instruction.  Adjust that as
                 well.  */
              if ((WSTOPSIG (*wstat) == SIGILL
                   || WSTOPSIG (*wstat) == SIGFPE
                   || WSTOPSIG (*wstat) == SIGBUS
                   || WSTOPSIG (*wstat) == SIGSEGV)
                  && ptrace (PTRACE_GETSIGINFO, lwpid_of (current_thread),
                             (PTRACE_TYPE_ARG3) 0, &info) == 0
                  /* Final check just to make sure we don't clobber
                     the siginfo of non-kernel-sent signals.  */
                  && (uintptr_t) info.si_addr == lwp->stop_pc)
                {
                  info.si_addr = (void *) (uintptr_t) status.tpoint_addr;
                  ptrace (PTRACE_SETSIGINFO, lwpid_of (current_thread),
                          (PTRACE_TYPE_ARG3) 0, &info);
                }

              regcache = get_thread_regcache (current_thread, 1);
              (*the_low_target.set_pc) (regcache, status.tpoint_addr);
              lwp->stop_pc = status.tpoint_addr;

              /* Cancel any fast tracepoint lock this thread was
                 holding.  */
              force_unlock_trace_buffer ();
            }

          if (lwp->exit_jump_pad_bkpt != NULL)
            {
              if (debug_threads)
                debug_printf ("Cancelling fast exit-jump-pad: removing bkpt. "
                              "stopping all threads momentarily.\n");

              stop_all_lwps (1, lwp);

              delete_breakpoint (lwp->exit_jump_pad_bkpt);
              lwp->exit_jump_pad_bkpt = NULL;

              unstop_all_lwps (1, lwp);

              gdb_assert (lwp->suspended >= 0);
            }
        }
    }

  if (debug_threads)
    debug_printf ("Checking whether LWP %ld needs to move out of the "
                  "jump pad...no\n",
                  lwpid_of (current_thread));

  current_thread = saved_thread;
  return 0;
}

/* Enqueue one signal in the "signals to report later when out of the
   jump pad" list.  */

static void
enqueue_one_deferred_signal (struct lwp_info *lwp, int *wstat)
{
  struct pending_signals *p_sig;
  struct thread_info *thread = get_lwp_thread (lwp);

  if (debug_threads)
    debug_printf ("Deferring signal %d for LWP %ld.\n",
                  WSTOPSIG (*wstat), lwpid_of (thread));

  if (debug_threads)
    {
      struct pending_signals *sig;

      for (sig = lwp->pending_signals_to_report;
           sig != NULL;
           sig = sig->prev)
        debug_printf ("   Already queued %d\n",
                      sig->signal);

      debug_printf ("   (no more currently queued signals)\n");
    }

  /* Don't enqueue non-RT signals if they are already in the deferred
     queue.  (SIGSTOP being the easiest signal to see ending up here
     twice)  */
  if (WSTOPSIG (*wstat) < __SIGRTMIN)
    {
      struct pending_signals *sig;

      for (sig = lwp->pending_signals_to_report;
           sig != NULL;
           sig = sig->prev)
        {
          if (sig->signal == WSTOPSIG (*wstat))
            {
              if (debug_threads)
                debug_printf ("Not requeuing already queued non-RT signal %d"
                              " for LWP %ld\n",
                              sig->signal,
                              lwpid_of (thread));
              return;
            }
        }
    }

  p_sig = xmalloc (sizeof (*p_sig));
  p_sig->prev = lwp->pending_signals_to_report;
  p_sig->signal = WSTOPSIG (*wstat);
  memset (&p_sig->info, 0, sizeof (siginfo_t));
  ptrace (PTRACE_GETSIGINFO, lwpid_of (thread), (PTRACE_TYPE_ARG3) 0,
          &p_sig->info);

  lwp->pending_signals_to_report = p_sig;
}

/* Dequeue one signal from the "signals to report later when out of
   the jump pad" list.  */

static int
dequeue_one_deferred_signal (struct lwp_info *lwp, int *wstat)
{
  struct thread_info *thread = get_lwp_thread (lwp);

  if (lwp->pending_signals_to_report != NULL)
    {
      struct pending_signals **p_sig;

      p_sig = &lwp->pending_signals_to_report;
      while ((*p_sig)->prev != NULL)
        p_sig = &(*p_sig)->prev;

      *wstat = W_STOPCODE ((*p_sig)->signal);
      if ((*p_sig)->info.si_signo != 0)
        ptrace (PTRACE_SETSIGINFO, lwpid_of (thread), (PTRACE_TYPE_ARG3) 0,
                &(*p_sig)->info);
      free (*p_sig);
      *p_sig = NULL;

      if (debug_threads)
        debug_printf ("Reporting deferred signal %d for LWP %ld.\n",
                      WSTOPSIG (*wstat), lwpid_of (thread));

      if (debug_threads)
        {
          struct pending_signals *sig;

          for (sig = lwp->pending_signals_to_report;
               sig != NULL;
               sig = sig->prev)
            debug_printf ("   Still queued %d\n",
                          sig->signal);

          debug_printf ("   (no more queued signals)\n");
        }

      return 1;
    }

  return 0;
}

/* Return true if the event in LP may be caused by breakpoint.  */

static int
wstatus_maybe_breakpoint (int wstatus)
{
  return (WIFSTOPPED (wstatus)
          && (WSTOPSIG (wstatus) == SIGTRAP
              /* SIGILL and SIGSEGV are also treated as traps in case a
                 breakpoint is inserted at the current PC.  */
              || WSTOPSIG (wstatus) == SIGILL
              || WSTOPSIG (wstatus) == SIGSEGV));
}

/* Fetch the possibly triggered data watchpoint info and store it in
   CHILD.

   On some archs, like x86, that use debug registers to set
   watchpoints, it's possible that the way to know which watched
   address trapped, is to check the register that is used to select
   which address to watch.  Problem is, between setting the watchpoint
   and reading back which data address trapped, the user may change
   the set of watchpoints, and, as a consequence, GDB changes the
   debug registers in the inferior.  To avoid reading back a stale
   stopped-data-address when that happens, we cache in LP the fact
   that a watchpoint trapped, and the corresponding data address, as
   soon as we see CHILD stop with a SIGTRAP.  If GDB changes the debug
   registers meanwhile, we have the cached data we can rely on.  */

static int
check_stopped_by_watchpoint (struct lwp_info *child)
{
  if (the_low_target.stopped_by_watchpoint != NULL)
    {
      struct thread_info *saved_thread;

      saved_thread = current_thread;
      current_thread = get_lwp_thread (child);

      if (the_low_target.stopped_by_watchpoint ())
        {
          child->stop_reason = LWP_STOPPED_BY_WATCHPOINT;

          if (the_low_target.stopped_data_address != NULL)
            child->stopped_data_address
              = the_low_target.stopped_data_address ();
          else
            child->stopped_data_address = 0;
        }

      current_thread = saved_thread;
    }

  return child->stop_reason == LWP_STOPPED_BY_WATCHPOINT;
}

/* Do low-level handling of the event, and check if we should go on
   and pass it to caller code.  Return the affected lwp if we are, or
   NULL otherwise.  */

static struct lwp_info *
linux_low_filter_event (int lwpid, int wstat)
{
  struct lwp_info *child;
  struct thread_info *thread;
  int have_stop_pc = 0;

  child = find_lwp_pid (pid_to_ptid (lwpid));

  /* If we didn't find a process, one of two things presumably happened:
     - A process we started and then detached from has exited.  Ignore it.
     - A process we are controlling has forked and the new child's stop
     was reported to us by the kernel.  Save its PID.  */
  if (child == NULL && WIFSTOPPED (wstat))
    {
      add_to_pid_list (&stopped_pids, lwpid, wstat);
      return NULL;
    }
  else if (child == NULL)
    return NULL;

  thread = get_lwp_thread (child);

  child->stopped = 1;

  child->last_status = wstat;

  /* Check if the thread has exited.  */
  if ((WIFEXITED (wstat) || WIFSIGNALED (wstat)))
    {
      if (debug_threads)
        debug_printf ("LLFE: %d exited.\n", lwpid);
      if (num_lwps (pid_of (thread)) > 1)
        {

          /* If there is at least one more LWP, then the exit signal was
             not the end of the debugged application and should be
             ignored.  */
          delete_lwp (child);
          return NULL;
        }
      else
        {
          /* This was the last lwp in the process.  Since events are
             serialized to GDB core, and we can't report this one
             right now, but GDB core and the other target layers will
             want to be notified about the exit code/signal, leave the
             status pending for the next time we're able to report
             it.  */
          mark_lwp_dead (child, wstat);
          return child;
        }
    }

  gdb_assert (WIFSTOPPED (wstat));

  if (WIFSTOPPED (wstat))
    {
      struct process_info *proc;

      /* Architecture-specific setup after inferior is running.  This
         needs to happen after we have attached to the inferior and it
         is stopped for the first time, but before we access any
         inferior registers.  */
      proc = find_process_pid (pid_of (thread));
      if (proc->private->new_inferior)
        {
          struct thread_info *saved_thread;

          saved_thread = current_thread;
          current_thread = thread;

          the_low_target.arch_setup ();

          current_thread = saved_thread;

          proc->private->new_inferior = 0;
        }
    }

  if (WIFSTOPPED (wstat) && child->must_set_ptrace_flags)
    {
      struct process_info *proc = find_process_pid (pid_of (thread));

      linux_enable_event_reporting (lwpid, proc->attached);
      child->must_set_ptrace_flags = 0;
    }

  /* Be careful to not overwrite stop_pc until
     check_stopped_by_breakpoint is called.  */
  if (WIFSTOPPED (wstat) && WSTOPSIG (wstat) == SIGTRAP
      && linux_is_extended_waitstatus (wstat))
    {
      child->stop_pc = get_pc (child);
      handle_extended_wait (child, wstat);
      return NULL;
    }

  if (WIFSTOPPED (wstat) && WSTOPSIG (wstat) == SIGTRAP
      && check_stopped_by_watchpoint (child))
    ;
  else if (WIFSTOPPED (wstat) && wstatus_maybe_breakpoint (wstat))
    {
      if (check_stopped_by_breakpoint (child))
        have_stop_pc = 1;
    }

  if (!have_stop_pc)
    child->stop_pc = get_pc (child);

  if (WIFSTOPPED (wstat) && WSTOPSIG (wstat) == SIGSTOP
      && child->stop_expected)
    {
      if (debug_threads)
        debug_printf ("Expected stop.\n");
      child->stop_expected = 0;

      if (thread->last_resume_kind == resume_stop)
        {
          /* We want to report the stop to the core.  Treat the
             SIGSTOP as a normal event.  */
        }
      else if (stopping_threads != NOT_STOPPING_THREADS)
        {
          /* Stopping threads.  We don't want this SIGSTOP to end up
             pending.  */
          return NULL;
        }
      else
        {
          /* Filter out the event.  */
          linux_resume_one_lwp (child, child->stepping, 0, NULL);
          return NULL;
        }
    }

  child->status_pending_p = 1;
  child->status_pending = wstat;
  return child;
}

/* Resume LWPs that are currently stopped without any pending status
   to report, but are resumed from the core's perspective.  */

static void
resume_stopped_resumed_lwps (struct inferior_list_entry *entry)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lp = get_thread_lwp (thread);

  if (lp->stopped
      && !lp->status_pending_p
      && thread->last_resume_kind != resume_stop
      && thread->last_status.kind == TARGET_WAITKIND_IGNORE)
    {
      int step = thread->last_resume_kind == resume_step;

      if (debug_threads)
        debug_printf ("RSRL: resuming stopped-resumed LWP %s at %s: step=%d\n",
                      target_pid_to_str (ptid_of (thread)),
                      paddress (lp->stop_pc),
                      step);

      linux_resume_one_lwp (lp, step, GDB_SIGNAL_0, NULL);
    }
}

/* Wait for an event from child(ren) WAIT_PTID, and return any that
   match FILTER_PTID (leaving others pending).  The PTIDs can be:
   minus_one_ptid, to specify any child; a pid PTID, specifying all
   lwps of a thread group; or a PTID representing a single lwp.  Store
   the stop status through the status pointer WSTAT.  OPTIONS is
   passed to the waitpid call.  Return 0 if no event was found and
   OPTIONS contains WNOHANG.  Return -1 if no unwaited-for children
   was found.  Return the PID of the stopped child otherwise.  */

static int
linux_wait_for_event_filtered (ptid_t wait_ptid, ptid_t filter_ptid,
                               int *wstatp, int options)
{
  struct thread_info *event_thread;
  struct lwp_info *event_child, *requested_child;
  sigset_t block_mask, prev_mask;

 retry:
  /* N.B. event_thread points to the thread_info struct that contains
     event_child.  Keep them in sync.  */
  event_thread = NULL;
  event_child = NULL;
  requested_child = NULL;

  /* Check for a lwp with a pending status.  */

  if (ptid_equal (filter_ptid, minus_one_ptid) || ptid_is_pid (filter_ptid))
    {
      event_thread = (struct thread_info *)
        find_inferior (&all_threads, status_pending_p_callback, &filter_ptid);
      if (event_thread != NULL)
        event_child = get_thread_lwp (event_thread);
      if (debug_threads && event_thread)
        debug_printf ("Got a pending child %ld\n", lwpid_of (event_thread));
    }
  else if (!ptid_equal (filter_ptid, null_ptid))
    {
      requested_child = find_lwp_pid (filter_ptid);

      if (stopping_threads == NOT_STOPPING_THREADS
          && requested_child->status_pending_p
          && requested_child->collecting_fast_tracepoint)
        {
          enqueue_one_deferred_signal (requested_child,
                                       &requested_child->status_pending);
          requested_child->status_pending_p = 0;
          requested_child->status_pending = 0;
          linux_resume_one_lwp (requested_child, 0, 0, NULL);
        }

      if (requested_child->suspended
          && requested_child->status_pending_p)
        {
          internal_error (__FILE__, __LINE__,
                          "requesting an event out of a"
                          " suspended child?");
        }

      if (requested_child->status_pending_p)
        {
          event_child = requested_child;
          event_thread = get_lwp_thread (event_child);
        }
    }

  if (event_child != NULL)
    {
      if (debug_threads)
        debug_printf ("Got an event from pending child %ld (%04x)\n",
                      lwpid_of (event_thread), event_child->status_pending);
      *wstatp = event_child->status_pending;
      event_child->status_pending_p = 0;
      event_child->status_pending = 0;
      current_thread = event_thread;
      return lwpid_of (event_thread);
    }

  /* But if we don't find a pending event, we'll have to wait.

     We only enter this loop if no process has a pending wait status.
     Thus any action taken in response to a wait status inside this
     loop is responding as soon as we detect the status, not after any
     pending events.  */

  /* Make sure SIGCHLD is blocked until the sigsuspend below.  Block
     all signals while here.  */
  sigfillset (&block_mask);
  sigprocmask (SIG_BLOCK, &block_mask, &prev_mask);

  /* Always pull all events out of the kernel.  We'll randomly select
     an event LWP out of all that have events, to prevent
     starvation.  */
  while (event_child == NULL)
    {
      pid_t ret = 0;

      /* Always use -1 and WNOHANG, due to couple of a kernel/ptrace
         quirks:

         - If the thread group leader exits while other threads in the
           thread group still exist, waitpid(TGID, ...) hangs.  That
           waitpid won't return an exit status until the other threads
           in the group are reaped.

         - When a non-leader thread execs, that thread just vanishes
           without reporting an exit (so we'd hang if we waited for it
           explicitly in that case).  The exec event is reported to
           the TGID pid (although we don't currently enable exec
           events).  */
      errno = 0;
      ret = my_waitpid (-1, wstatp, options | WNOHANG);

      if (debug_threads)
        debug_printf ("LWFE: waitpid(-1, ...) returned %d, %s\n",
                      ret, errno ? strerror (errno) : "ERRNO-OK");

      if (ret > 0)
        {
          if (debug_threads)
            {
              debug_printf ("LLW: waitpid %ld received %s\n",
                            (long) ret, status_to_str (*wstatp));
            }

          /* Filter all events.  IOW, leave all events pending.  We'll
             randomly select an event LWP out of all that have events
             below.  */
          linux_low_filter_event (ret, *wstatp);
          /* Retry until nothing comes out of waitpid.  A single
             SIGCHLD can indicate more than one child stopped.  */
          continue;
        }

      /* Now that we've pulled all events out of the kernel, resume
         LWPs that don't have an interesting event to report.  */
      if (stopping_threads == NOT_STOPPING_THREADS)
        for_each_inferior (&all_threads, resume_stopped_resumed_lwps);

      /* ... and find an LWP with a status to report to the core, if
         any.  */
      event_thread = (struct thread_info *)
        find_inferior (&all_threads, status_pending_p_callback, &filter_ptid);
      if (event_thread != NULL)
        {
          event_child = get_thread_lwp (event_thread);
          *wstatp = event_child->status_pending;
          event_child->status_pending_p = 0;
          event_child->status_pending = 0;
          break;
        }

      /* Check for zombie thread group leaders.  Those can't be reaped
         until all other threads in the thread group are.  */
      check_zombie_leaders ();

      /* If there are no resumed children left in the set of LWPs we
         want to wait for, bail.  We can't just block in
         waitpid/sigsuspend, because lwps might have been left stopped
         in trace-stop state, and we'd be stuck forever waiting for
         their status to change (which would only happen if we resumed
         them).  Even if WNOHANG is set, this return code is preferred
         over 0 (below), as it is more detailed.  */
      if ((find_inferior (&all_threads,
                          not_stopped_callback,
                          &wait_ptid) == NULL))
        {
          if (debug_threads)
            debug_printf ("LLW: exit (no unwaited-for LWP)\n");
          sigprocmask (SIG_SETMASK, &prev_mask, NULL);
          return -1;
        }

      /* No interesting event to report to the caller.  */
      if ((options & WNOHANG))
        {
          if (debug_threads)
            debug_printf ("WNOHANG set, no event found\n");

          sigprocmask (SIG_SETMASK, &prev_mask, NULL);
          return 0;
        }

      /* Block until we get an event reported with SIGCHLD.  */
      if (debug_threads)
        debug_printf ("sigsuspend'ing\n");

      sigsuspend (&prev_mask);
      sigprocmask (SIG_SETMASK, &prev_mask, NULL);
      goto retry;
    }

  sigprocmask (SIG_SETMASK, &prev_mask, NULL);

  current_thread = event_thread;

  /* Check for thread exit.  */
  if (! WIFSTOPPED (*wstatp))
    {
      gdb_assert (last_thread_of_process_p (pid_of (event_thread)));

      if (debug_threads)
        debug_printf ("LWP %d is the last lwp of process.  "
                      "Process %ld exiting.\n",
                      pid_of (event_thread), lwpid_of (event_thread));
      return lwpid_of (event_thread);
    }

  return lwpid_of (event_thread);
}

/* Wait for an event from child(ren) PTID.  PTIDs can be:
   minus_one_ptid, to specify any child; a pid PTID, specifying all
   lwps of a thread group; or a PTID representing a single lwp.  Store
   the stop status through the status pointer WSTAT.  OPTIONS is
   passed to the waitpid call.  Return 0 if no event was found and
   OPTIONS contains WNOHANG.  Return -1 if no unwaited-for children
   was found.  Return the PID of the stopped child otherwise.  */

static int
linux_wait_for_event (ptid_t ptid, int *wstatp, int options)
{
  return linux_wait_for_event_filtered (ptid, ptid, wstatp, options);
}

/* Count the LWP's that have had events.  */

static int
count_events_callback (struct inferior_list_entry *entry, void *data)
{
  struct thread_info *thread = (struct thread_info *) entry;
  int *count = data;

  gdb_assert (count != NULL);

  /* Count only resumed LWPs that have an event pending. */
  if (thread->last_status.kind == TARGET_WAITKIND_IGNORE
      && thread->last_resume_kind != resume_stop
      && thread->status_pending_p)
    (*count)++;

  return 0;
}

/* Select the LWP (if any) that is currently being single-stepped.  */

static int
select_singlestep_lwp_callback (struct inferior_list_entry *entry, void *data)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lp = get_thread_lwp (thread);

  if (thread->last_status.kind == TARGET_WAITKIND_IGNORE
      && thread->last_resume_kind == resume_step
      && lp->status_pending_p)
    return 1;
  else
    return 0;
}

/* Select the Nth LWP that has had a SIGTRAP event that should be
   reported to GDB.  */

static int
select_event_lwp_callback (struct inferior_list_entry *entry, void *data)
{
  struct thread_info *thread = (struct thread_info *) entry;
  int *selector = data;

  gdb_assert (selector != NULL);

  /* Select only resumed LWPs that have an event pending. */
  if (thread->last_resume_kind != resume_stop
      && thread->last_status.kind == TARGET_WAITKIND_IGNORE
      && thread->status_pending_p)
    if ((*selector)-- == 0)
      return 1;

  return 0;
}

/* Select one LWP out of those that have events pending.  */

static void
select_event_lwp (struct lwp_info **orig_lp)
{
  int num_events = 0;
  int random_selector;
  struct thread_info *event_thread = NULL;

  /* In all-stop, give preference to the LWP that is being
     single-stepped.  There will be at most one, and it's the LWP that
     the core is most interested in.  If we didn't do this, then we'd
     have to handle pending step SIGTRAPs somehow in case the core
     later continues the previously-stepped thread, otherwise we'd
     report the pending SIGTRAP, and the core, not having stepped the
     thread, wouldn't understand what the trap was for, and therefore
     would report it to the user as a random signal.  */
  if (!non_stop)
    {
      event_thread
        = (struct thread_info *) find_inferior (&all_threads,
                                                select_singlestep_lwp_callback,
                                                NULL);
      if (event_thread != NULL)
        {
          if (debug_threads)
            debug_printf ("SEL: Select single-step %s\n",
                          target_pid_to_str (ptid_of (event_thread)));
        }
    }
  if (event_thread == NULL)
    {
      /* No single-stepping LWP.  Select one at random, out of those
         which have had SIGTRAP events.  */

      /* First see how many SIGTRAP events we have.  */
      find_inferior (&all_threads, count_events_callback, &num_events);

      /* Now randomly pick a LWP out of those that have had a SIGTRAP.  */
      random_selector = (int)
        ((num_events * (double) rand ()) / (RAND_MAX + 1.0));

      if (debug_threads && num_events > 1)
        debug_printf ("SEL: Found %d SIGTRAP events, selecting #%d\n",
                      num_events, random_selector);

      event_thread
        = (struct thread_info *) find_inferior (&all_threads,
                                                select_event_lwp_callback,
                                                &random_selector);
    }

  if (event_thread != NULL)
    {
      struct lwp_info *event_lp = get_thread_lwp (event_thread);

      /* Switch the event LWP.  */
      *orig_lp = event_lp;
    }
}

/* Decrement the suspend count of an LWP.  */

static int
unsuspend_one_lwp (struct inferior_list_entry *entry, void *except)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lwp = get_thread_lwp (thread);

  /* Ignore EXCEPT.  */
  if (lwp == except)
    return 0;

  lwp->suspended--;

  gdb_assert (lwp->suspended >= 0);
  return 0;
}

/* Decrement the suspend count of all LWPs, except EXCEPT, if non
   NULL.  */

static void
unsuspend_all_lwps (struct lwp_info *except)
{
  find_inferior (&all_threads, unsuspend_one_lwp, except);
}

static void move_out_of_jump_pad_callback (struct inferior_list_entry *entry);
static int stuck_in_jump_pad_callback (struct inferior_list_entry *entry,
                                       void *data);
static int lwp_running (struct inferior_list_entry *entry, void *data);
static ptid_t linux_wait_1 (ptid_t ptid,
                            struct target_waitstatus *ourstatus,
                            int target_options);

/* Stabilize threads (move out of jump pads).

   If a thread is midway collecting a fast tracepoint, we need to
   finish the collection and move it out of the jump pad before
   reporting the signal.

   This avoids recursion while collecting (when a signal arrives
   midway, and the signal handler itself collects), which would trash
   the trace buffer.  In case the user set a breakpoint in a signal
   handler, this avoids the backtrace showing the jump pad, etc..
   Most importantly, there are certain things we can't do safely if
   threads are stopped in a jump pad (or in its callee's).  For
   example:

     - starting a new trace run.  A thread still collecting the
   previous run, could trash the trace buffer when resumed.  The trace
   buffer control structures would have been reset but the thread had
   no way to tell.  The thread could even midway memcpy'ing to the
   buffer, which would mean that when resumed, it would clobber the
   trace buffer that had been set for a new run.

     - we can't rewrite/reuse the jump pads for new tracepoints
   safely.  Say you do tstart while a thread is stopped midway while
   collecting.  When the thread is later resumed, it finishes the
   collection, and returns to the jump pad, to execute the original
   instruction that was under the tracepoint jump at the time the
   older run had been started.  If the jump pad had been rewritten
   since for something else in the new run, the thread would now
   execute the wrong / random instructions.  */

static void
linux_stabilize_threads (void)
{
  struct thread_info *saved_thread;
  struct thread_info *thread_stuck;

  thread_stuck
    = (struct thread_info *) find_inferior (&all_threads,
                                            stuck_in_jump_pad_callback,
                                            NULL);
  if (thread_stuck != NULL)
    {
      if (debug_threads)
        debug_printf ("can't stabilize, LWP %ld is stuck in jump pad\n",
                      lwpid_of (thread_stuck));
      return;
    }

  saved_thread = current_thread;

  stabilizing_threads = 1;

  /* Kick 'em all.  */
  for_each_inferior (&all_threads, move_out_of_jump_pad_callback);

  /* Loop until all are stopped out of the jump pads.  */
  while (find_inferior (&all_threads, lwp_running, NULL) != NULL)
    {
      struct target_waitstatus ourstatus;
      struct lwp_info *lwp;
      int wstat;

      /* Note that we go through the full wait even loop.  While
         moving threads out of jump pad, we need to be able to step
         over internal breakpoints and such.  */
      linux_wait_1 (minus_one_ptid, &ourstatus, 0);

      if (ourstatus.kind == TARGET_WAITKIND_STOPPED)
        {
          lwp = get_thread_lwp (current_thread);

          /* Lock it.  */
          lwp->suspended++;

          if (ourstatus.value.sig != GDB_SIGNAL_0
              || current_thread->last_resume_kind == resume_stop)
            {
              wstat = W_STOPCODE (gdb_signal_to_host (ourstatus.value.sig));
              enqueue_one_deferred_signal (lwp, &wstat);
            }
        }
    }

  find_inferior (&all_threads, unsuspend_one_lwp, NULL);

  stabilizing_threads = 0;

  current_thread = saved_thread;

  if (debug_threads)
    {
      thread_stuck
        = (struct thread_info *) find_inferior (&all_threads,
                                                stuck_in_jump_pad_callback,
                                                NULL);
      if (thread_stuck != NULL)
        debug_printf ("couldn't stabilize, LWP %ld got stuck in jump pad\n",
                      lwpid_of (thread_stuck));
    }
}

static void async_file_mark (void);

/* Convenience function that is called when the kernel reports an
   event that is not passed out to GDB.  */

static ptid_t
ignore_event (struct target_waitstatus *ourstatus)
{
  /* If we got an event, there may still be others, as a single
     SIGCHLD can indicate more than one child stopped.  This forces
     another target_wait call.  */
  async_file_mark ();

  ourstatus->kind = TARGET_WAITKIND_IGNORE;
  return null_ptid;
}

/* Wait for process, returns status.  */

static ptid_t
linux_wait_1 (ptid_t ptid,
              struct target_waitstatus *ourstatus, int target_options)
{
  int w;
  struct lwp_info *event_child;
  int options;
  int pid;
  int step_over_finished;
  int bp_explains_trap;
  int maybe_internal_trap;
  int report_to_gdb;
  int trace_event;
  int in_step_range;

  if (debug_threads)
    {
      debug_enter ();
      debug_printf ("linux_wait_1: [%s]\n", target_pid_to_str (ptid));
    }

  /* Translate generic target options into linux options.  */
  options = __WALL;
  if (target_options & TARGET_WNOHANG)
    options |= WNOHANG;

  bp_explains_trap = 0;
  trace_event = 0;
  in_step_range = 0;
  ourstatus->kind = TARGET_WAITKIND_IGNORE;

  if (ptid_equal (step_over_bkpt, null_ptid))
    pid = linux_wait_for_event (ptid, &w, options);
  else
    {
      if (debug_threads)
        debug_printf ("step_over_bkpt set [%s], doing a blocking wait\n",
                      target_pid_to_str (step_over_bkpt));
      pid = linux_wait_for_event (step_over_bkpt, &w, options & ~WNOHANG);
    }

  if (pid == 0)
    {
      gdb_assert (target_options & TARGET_WNOHANG);

      if (debug_threads)
        {
          debug_printf ("linux_wait_1 ret = null_ptid, "
                        "TARGET_WAITKIND_IGNORE\n");
          debug_exit ();
        }

      ourstatus->kind = TARGET_WAITKIND_IGNORE;
      return null_ptid;
    }
  else if (pid == -1)
    {
      if (debug_threads)
        {
          debug_printf ("linux_wait_1 ret = null_ptid, "
                        "TARGET_WAITKIND_NO_RESUMED\n");
          debug_exit ();
        }

      ourstatus->kind = TARGET_WAITKIND_NO_RESUMED;
      return null_ptid;
    }

  event_child = get_thread_lwp (current_thread);

  /* linux_wait_for_event only returns an exit status for the last
     child of a process.  Report it.  */
  if (WIFEXITED (w) || WIFSIGNALED (w))
    {
      if (WIFEXITED (w))
        {
          ourstatus->kind = TARGET_WAITKIND_EXITED;
          ourstatus->value.integer = WEXITSTATUS (w);

          if (debug_threads)
            {
              debug_printf ("linux_wait_1 ret = %s, exited with "
                            "retcode %d\n",
                            target_pid_to_str (ptid_of (current_thread)),
                            WEXITSTATUS (w));
              debug_exit ();
            }
        }
      else
        {
          ourstatus->kind = TARGET_WAITKIND_SIGNALLED;
          ourstatus->value.sig = gdb_signal_from_host (WTERMSIG (w));

          if (debug_threads)
            {
              debug_printf ("linux_wait_1 ret = %s, terminated with "
                            "signal %d\n",
                            target_pid_to_str (ptid_of (current_thread)),
                            WTERMSIG (w));
              debug_exit ();
            }
        }

      return ptid_of (current_thread);
    }

  /* If this event was not handled before, and is not a SIGTRAP, we
     report it.  SIGILL and SIGSEGV are also treated as traps in case
     a breakpoint is inserted at the current PC.  If this target does
     not support internal breakpoints at all, we also report the
     SIGTRAP without further processing; it's of no concern to us.  */
  maybe_internal_trap
    = (supports_breakpoints ()
       && (WSTOPSIG (w) == SIGTRAP
           || ((WSTOPSIG (w) == SIGILL
                || WSTOPSIG (w) == SIGSEGV)
               && (*the_low_target.breakpoint_at) (event_child->stop_pc))));

  if (maybe_internal_trap)
    {
      /* Handle anything that requires bookkeeping before deciding to
         report the event or continue waiting.  */

      /* First check if we can explain the SIGTRAP with an internal
         breakpoint, or if we should possibly report the event to GDB.
         Do this before anything that may remove or insert a
         breakpoint.  */
      bp_explains_trap = breakpoint_inserted_here (event_child->stop_pc);

      /* We have a SIGTRAP, possibly a step-over dance has just
         finished.  If so, tweak the state machine accordingly,
         reinsert breakpoints and delete any reinsert (software
         single-step) breakpoints.  */
      step_over_finished = finish_step_over (event_child);

      /* Now invoke the callbacks of any internal breakpoints there.  */
      check_breakpoints (event_child->stop_pc);

      /* Handle tracepoint data collecting.  This may overflow the
         trace buffer, and cause a tracing stop, removing
         breakpoints.  */
      trace_event = handle_tracepoints (event_child);

      if (bp_explains_trap)
        {
          /* If we stepped or ran into an internal breakpoint, we've
             already handled it.  So next time we resume (from this
             PC), we should step over it.  */
          if (debug_threads)
            debug_printf ("Hit a gdbserver breakpoint.\n");

          if (breakpoint_here (event_child->stop_pc))
            event_child->need_step_over = 1;
        }
    }
  else
    {
      /* We have some other signal, possibly a step-over dance was in
         progress, and it should be cancelled too.  */
      step_over_finished = finish_step_over (event_child);
    }

  /* We have all the data we need.  Either report the event to GDB, or
     resume threads and keep waiting for more.  */

  /* If we're collecting a fast tracepoint, finish the collection and
     move out of the jump pad before delivering a signal.  See
     linux_stabilize_threads.  */

  if (WIFSTOPPED (w)
      && WSTOPSIG (w) != SIGTRAP
      && supports_fast_tracepoints ()
      && agent_loaded_p ())
    {
      if (debug_threads)
        debug_printf ("Got signal %d for LWP %ld.  Check if we need "
                      "to defer or adjust it.\n",
                      WSTOPSIG (w), lwpid_of (current_thread));

      /* Allow debugging the jump pad itself.  */
      if (current_thread->last_resume_kind != resume_step
          && maybe_move_out_of_jump_pad (event_child, &w))
        {
          enqueue_one_deferred_signal (event_child, &w);

          if (debug_threads)
            debug_printf ("Signal %d for LWP %ld deferred (in jump pad)\n",
                          WSTOPSIG (w), lwpid_of (current_thread));

          linux_resume_one_lwp (event_child, 0, 0, NULL);

          return ignore_event (ourstatus);
        }
    }

  if (event_child->collecting_fast_tracepoint)
    {
      if (debug_threads)
        debug_printf ("LWP %ld was trying to move out of the jump pad (%d). "
                      "Check if we're already there.\n",
                      lwpid_of (current_thread),
                      event_child->collecting_fast_tracepoint);

      trace_event = 1;

      event_child->collecting_fast_tracepoint
        = linux_fast_tracepoint_collecting (event_child, NULL);

      if (event_child->collecting_fast_tracepoint != 1)
        {
          /* No longer need this breakpoint.  */
          if (event_child->exit_jump_pad_bkpt != NULL)
            {
              if (debug_threads)
                debug_printf ("No longer need exit-jump-pad bkpt; removing it."
                              "stopping all threads momentarily.\n");

              /* Other running threads could hit this breakpoint.
                 We don't handle moribund locations like GDB does,
                 instead we always pause all threads when removing
                 breakpoints, so that any step-over or
                 decr_pc_after_break adjustment is always taken
                 care of while the breakpoint is still
                 inserted.  */
              stop_all_lwps (1, event_child);

              delete_breakpoint (event_child->exit_jump_pad_bkpt);
              event_child->exit_jump_pad_bkpt = NULL;

              unstop_all_lwps (1, event_child);

              gdb_assert (event_child->suspended >= 0);
            }
        }

      if (event_child->collecting_fast_tracepoint == 0)
        {
          if (debug_threads)
            debug_printf ("fast tracepoint finished "
                          "collecting successfully.\n");

          /* We may have a deferred signal to report.  */
          if (dequeue_one_deferred_signal (event_child, &w))
            {
              if (debug_threads)
                debug_printf ("dequeued one signal.\n");
            }
          else
            {
              if (debug_threads)
                debug_printf ("no deferred signals.\n");

              if (stabilizing_threads)
                {
                  ourstatus->kind = TARGET_WAITKIND_STOPPED;
                  ourstatus->value.sig = GDB_SIGNAL_0;

                  if (debug_threads)
                    {
                      debug_printf ("linux_wait_1 ret = %s, stopped "
                                    "while stabilizing threads\n",
                                    target_pid_to_str (ptid_of (current_thread)));
                      debug_exit ();
                    }

                  return ptid_of (current_thread);
                }
            }
        }
    }

  /* Check whether GDB would be interested in this event.  */

  /* If GDB is not interested in this signal, don't stop other
     threads, and don't report it to GDB.  Just resume the inferior
     right away.  We do this for threading-related signals as well as
     any that GDB specifically requested we ignore.  But never ignore
     SIGSTOP if we sent it ourselves, and do not ignore signals when
     stepping - they may require special handling to skip the signal
     handler.  */
  /* FIXME drow/2002-06-09: Get signal numbers from the inferior's
     thread library?  */
  if (WIFSTOPPED (w)
      && current_thread->last_resume_kind != resume_step
      && (
#if defined (USE_THREAD_DB) && !defined (__ANDROID__)
          (current_process ()->private->thread_db != NULL
           && (WSTOPSIG (w) == __SIGRTMIN
               || WSTOPSIG (w) == __SIGRTMIN + 1))
          ||
#endif
          (pass_signals[gdb_signal_from_host (WSTOPSIG (w))]
           && !(WSTOPSIG (w) == SIGSTOP
                && current_thread->last_resume_kind == resume_stop))))
    {
      siginfo_t info, *info_p;

      if (debug_threads)
        debug_printf ("Ignored signal %d for LWP %ld.\n",
                      WSTOPSIG (w), lwpid_of (current_thread));

      if (ptrace (PTRACE_GETSIGINFO, lwpid_of (current_thread),
                  (PTRACE_TYPE_ARG3) 0, &info) == 0)
        info_p = &info;
      else
        info_p = NULL;
      linux_resume_one_lwp (event_child, event_child->stepping,
                            WSTOPSIG (w), info_p);
      return ignore_event (ourstatus);
    }

  /* Note that all addresses are always "out of the step range" when
     there's no range to begin with.  */
  in_step_range = lwp_in_step_range (event_child);

  /* If GDB wanted this thread to single step, and the thread is out
     of the step range, we always want to report the SIGTRAP, and let
     GDB handle it.  Watchpoints should always be reported.  So should
     signals we can't explain.  A SIGTRAP we can't explain could be a
     GDB breakpoint --- we may or not support Z0 breakpoints.  If we
     do, we're be able to handle GDB breakpoints on top of internal
     breakpoints, by handling the internal breakpoint and still
     reporting the event to GDB.  If we don't, we're out of luck, GDB
     won't see the breakpoint hit.  */
  report_to_gdb = (!maybe_internal_trap
                   || (current_thread->last_resume_kind == resume_step
                       && !in_step_range)
                   || event_child->stop_reason == LWP_STOPPED_BY_WATCHPOINT
                   || (!step_over_finished && !in_step_range
                       && !bp_explains_trap && !trace_event)
                   || (gdb_breakpoint_here (event_child->stop_pc)
                       && gdb_condition_true_at_breakpoint (event_child->stop_pc)
                       && gdb_no_commands_at_breakpoint (event_child->stop_pc)));

  run_breakpoint_commands (event_child->stop_pc);

  /* We found no reason GDB would want us to stop.  We either hit one
     of our own breakpoints, or finished an internal step GDB
     shouldn't know about.  */
  if (!report_to_gdb)
    {
      if (debug_threads)
        {
          if (bp_explains_trap)
            debug_printf ("Hit a gdbserver breakpoint.\n");
          if (step_over_finished)
            debug_printf ("Step-over finished.\n");
          if (trace_event)
            debug_printf ("Tracepoint event.\n");
          if (lwp_in_step_range (event_child))
            debug_printf ("Range stepping pc 0x%s [0x%s, 0x%s).\n",
                          paddress (event_child->stop_pc),
                          paddress (event_child->step_range_start),
                          paddress (event_child->step_range_end));
        }

      /* We're not reporting this breakpoint to GDB, so apply the
         decr_pc_after_break adjustment to the inferior's regcache
         ourselves.  */

      if (the_low_target.set_pc != NULL)
        {
          struct regcache *regcache
            = get_thread_regcache (current_thread, 1);
          (*the_low_target.set_pc) (regcache, event_child->stop_pc);
        }

      /* We may have finished stepping over a breakpoint.  If so,
         we've stopped and suspended all LWPs momentarily except the
         stepping one.  This is where we resume them all again.  We're
         going to keep waiting, so use proceed, which handles stepping
         over the next breakpoint.  */
      if (debug_threads)
        debug_printf ("proceeding all threads.\n");

      if (step_over_finished)
        unsuspend_all_lwps (event_child);

      proceed_all_lwps ();
      return ignore_event (ourstatus);
    }

  if (debug_threads)
    {
      if (current_thread->last_resume_kind == resume_step)
        {
          if (event_child->step_range_start == event_child->step_range_end)
            debug_printf ("GDB wanted to single-step, reporting event.\n");
          else if (!lwp_in_step_range (event_child))
            debug_printf ("Out of step range, reporting event.\n");
        }
      if (event_child->stop_reason == LWP_STOPPED_BY_WATCHPOINT)
        debug_printf ("Stopped by watchpoint.\n");
      else if (gdb_breakpoint_here (event_child->stop_pc))
        debug_printf ("Stopped by GDB breakpoint.\n");
      if (debug_threads)
        debug_printf ("Hit a non-gdbserver trap event.\n");
    }

  /* Alright, we're going to report a stop.  */

  if (!stabilizing_threads)
    {
      /* In all-stop, stop all threads.  */
      if (!non_stop)
        stop_all_lwps (0, NULL);

      /* If we're not waiting for a specific LWP, choose an event LWP
         from among those that have had events.  Giving equal priority
         to all LWPs that have had events helps prevent
         starvation.  */
      if (ptid_equal (ptid, minus_one_ptid))
        {
          event_child->status_pending_p = 1;
          event_child->status_pending = w;

          select_event_lwp (&event_child);

          /* current_thread and event_child must stay in sync.  */
          current_thread = get_lwp_thread (event_child);

          event_child->status_pending_p = 0;
          w = event_child->status_pending;
        }

      if (step_over_finished)
        {
          if (!non_stop)
            {
              /* If we were doing a step-over, all other threads but
                 the stepping one had been paused in start_step_over,
                 with their suspend counts incremented.  We don't want
                 to do a full unstop/unpause, because we're in
                 all-stop mode (so we want threads stopped), but we
                 still need to unsuspend the other threads, to
                 decrement their `suspended' count back.  */
              unsuspend_all_lwps (event_child);
            }
          else
            {
              /* If we just finished a step-over, then all threads had
                 been momentarily paused.  In all-stop, that's fine,
                 we want threads stopped by now anyway.  In non-stop,
                 we need to re-resume threads that GDB wanted to be
                 running.  */
              unstop_all_lwps (1, event_child);
            }
        }

      /* Stabilize threads (move out of jump pads).  */
      if (!non_stop)
        stabilize_threads ();
    }
  else
    {
      /* If we just finished a step-over, then all threads had been
         momentarily paused.  In all-stop, that's fine, we want
         threads stopped by now anyway.  In non-stop, we need to
         re-resume threads that GDB wanted to be running.  */
      if (step_over_finished)
        unstop_all_lwps (1, event_child);
    }

  ourstatus->kind = TARGET_WAITKIND_STOPPED;

  /* Now that we've selected our final event LWP, un-adjust its PC if
     it was a software breakpoint.  */
  if (event_child->stop_reason == LWP_STOPPED_BY_SW_BREAKPOINT)
    {
      int decr_pc = the_low_target.decr_pc_after_break;

      if (decr_pc != 0)
        {
          struct regcache *regcache
            = get_thread_regcache (current_thread, 1);
          (*the_low_target.set_pc) (regcache, event_child->stop_pc + decr_pc);
        }
    }

  if (current_thread->last_resume_kind == resume_stop
      && WSTOPSIG (w) == SIGSTOP)
    {
      /* A thread that has been requested to stop by GDB with vCont;t,
         and it stopped cleanly, so report as SIG0.  The use of
         SIGSTOP is an implementation detail.  */
      ourstatus->value.sig = GDB_SIGNAL_0;
    }
  else if (current_thread->last_resume_kind == resume_stop
           && WSTOPSIG (w) != SIGSTOP)
    {
      /* A thread that has been requested to stop by GDB with vCont;t,
         but, it stopped for other reasons.  */
      ourstatus->value.sig = gdb_signal_from_host (WSTOPSIG (w));
    }
  else
    {
      ourstatus->value.sig = gdb_signal_from_host (WSTOPSIG (w));
    }

  gdb_assert (ptid_equal (step_over_bkpt, null_ptid));

  if (debug_threads)
    {
      debug_printf ("linux_wait_1 ret = %s, %d, %d\n",
                    target_pid_to_str (ptid_of (current_thread)),
                    ourstatus->kind, ourstatus->value.sig);
      debug_exit ();
    }

  return ptid_of (current_thread);
}

/* Get rid of any pending event in the pipe.  */
static void
async_file_flush (void)
{
  int ret;
  char buf;

  do
    ret = read (linux_event_pipe[0], &buf, 1);
  while (ret >= 0 || (ret == -1 && errno == EINTR));
}

/* Put something in the pipe, so the event loop wakes up.  */
static void
async_file_mark (void)
{
  int ret;

  async_file_flush ();

  do
    ret = write (linux_event_pipe[1], "+", 1);
  while (ret == 0 || (ret == -1 && errno == EINTR));

  /* Ignore EAGAIN.  If the pipe is full, the event loop will already
     be awakened anyway.  */
}

static ptid_t
linux_wait (ptid_t ptid,
            struct target_waitstatus *ourstatus, int target_options)
{
  ptid_t event_ptid;

  /* Flush the async file first.  */
  if (target_is_async_p ())
    async_file_flush ();

  do
    {
      event_ptid = linux_wait_1 (ptid, ourstatus, target_options);
    }
  while ((target_options & TARGET_WNOHANG) == 0
         && ptid_equal (event_ptid, null_ptid)
         && ourstatus->kind == TARGET_WAITKIND_IGNORE);

  /* If at least one stop was reported, there may be more.  A single
     SIGCHLD can signal more than one child stop.  */
  if (target_is_async_p ()
      && (target_options & TARGET_WNOHANG) != 0
      && !ptid_equal (event_ptid, null_ptid))
    async_file_mark ();

  return event_ptid;
}

/* Send a signal to an LWP.  */

static int
kill_lwp (unsigned long lwpid, int signo)
{
  /* Use tkill, if possible, in case we are using nptl threads.  If tkill
     fails, then we are not using nptl threads and we should be using kill.  */

#ifdef __NR_tkill
  {
    static int tkill_failed;

    if (!tkill_failed)
      {
        int ret;

        errno = 0;
        ret = syscall (__NR_tkill, lwpid, signo);
        if (errno != ENOSYS)
          return ret;
        tkill_failed = 1;
      }
  }
#endif

  return kill (lwpid, signo);
}

void
linux_stop_lwp (struct lwp_info *lwp)
{
  send_sigstop (lwp);
}

static void
send_sigstop (struct lwp_info *lwp)
{
  int pid;

  pid = lwpid_of (get_lwp_thread (lwp));

  /* If we already have a pending stop signal for this process, don't
     send another.  */
  if (lwp->stop_expected)
    {
      if (debug_threads)
        debug_printf ("Have pending sigstop for lwp %d\n", pid);

      return;
    }

  if (debug_threads)
    debug_printf ("Sending sigstop to lwp %d\n", pid);

  lwp->stop_expected = 1;
  kill_lwp (pid, SIGSTOP);
}

static int
send_sigstop_callback (struct inferior_list_entry *entry, void *except)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lwp = get_thread_lwp (thread);

  /* Ignore EXCEPT.  */
  if (lwp == except)
    return 0;

  if (lwp->stopped)
    return 0;

  send_sigstop (lwp);
  return 0;
}

/* Increment the suspend count of an LWP, and stop it, if not stopped
   yet.  */
static int
suspend_and_send_sigstop_callback (struct inferior_list_entry *entry,
                                   void *except)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lwp = get_thread_lwp (thread);

  /* Ignore EXCEPT.  */
  if (lwp == except)
    return 0;

  lwp->suspended++;

  return send_sigstop_callback (entry, except);
}

static void
mark_lwp_dead (struct lwp_info *lwp, int wstat)
{
  /* It's dead, really.  */
  lwp->dead = 1;

  /* Store the exit status for later.  */
  lwp->status_pending_p = 1;
  lwp->status_pending = wstat;

  /* Prevent trying to stop it.  */
  lwp->stopped = 1;

  /* No further stops are expected from a dead lwp.  */
  lwp->stop_expected = 0;
}

/* Wait for all children to stop for the SIGSTOPs we just queued.  */

static void
wait_for_sigstop (void)
{
  struct thread_info *saved_thread;
  ptid_t saved_tid;
  int wstat;
  int ret;

  saved_thread = current_thread;
  if (saved_thread != NULL)
    saved_tid = saved_thread->entry.id;
  else
    saved_tid = null_ptid; /* avoid bogus unused warning */

  if (debug_threads)
    debug_printf ("wait_for_sigstop: pulling events\n");

  /* Passing NULL_PTID as filter indicates we want all events to be
     left pending.  Eventually this returns when there are no
     unwaited-for children left.  */
  ret = linux_wait_for_event_filtered (minus_one_ptid, null_ptid,
                                       &wstat, __WALL);
  gdb_assert (ret == -1);

  if (saved_thread == NULL || linux_thread_alive (saved_tid))
    current_thread = saved_thread;
  else
    {
      if (debug_threads)
        debug_printf ("Previously current thread died.\n");

      if (non_stop)
        {
          /* We can't change the current inferior behind GDB's back,
             otherwise, a subsequent command may apply to the wrong
             process.  */
          current_thread = NULL;
        }
      else
        {
          /* Set a valid thread as current.  */
          set_desired_thread (0);
        }
    }
}

/* Returns true if LWP ENTRY is stopped in a jump pad, and we can't
   move it out, because we need to report the stop event to GDB.  For
   example, if the user puts a breakpoint in the jump pad, it's
   because she wants to debug it.  */

static int
stuck_in_jump_pad_callback (struct inferior_list_entry *entry, void *data)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lwp = get_thread_lwp (thread);

  gdb_assert (lwp->suspended == 0);
  gdb_assert (lwp->stopped);

  /* Allow debugging the jump pad, gdb_collect, etc..  */
  return (supports_fast_tracepoints ()
          && agent_loaded_p ()
          && (gdb_breakpoint_here (lwp->stop_pc)
              || lwp->stop_reason == LWP_STOPPED_BY_WATCHPOINT
              || thread->last_resume_kind == resume_step)
          && linux_fast_tracepoint_collecting (lwp, NULL));
}

static void
move_out_of_jump_pad_callback (struct inferior_list_entry *entry)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lwp = get_thread_lwp (thread);
  int *wstat;

  gdb_assert (lwp->suspended == 0);
  gdb_assert (lwp->stopped);

  wstat = lwp->status_pending_p ? &lwp->status_pending : NULL;

  /* Allow debugging the jump pad, gdb_collect, etc.  */
  if (!gdb_breakpoint_here (lwp->stop_pc)
      && lwp->stop_reason != LWP_STOPPED_BY_WATCHPOINT
      && thread->last_resume_kind != resume_step
      && maybe_move_out_of_jump_pad (lwp, wstat))
    {
      if (debug_threads)
        debug_printf ("LWP %ld needs stabilizing (in jump pad)\n",
                      lwpid_of (thread));

      if (wstat)
        {
          lwp->status_pending_p = 0;
          enqueue_one_deferred_signal (lwp, wstat);

          if (debug_threads)
            debug_printf ("Signal %d for LWP %ld deferred "
                          "(in jump pad)\n",
                          WSTOPSIG (*wstat), lwpid_of (thread));
        }

      linux_resume_one_lwp (lwp, 0, 0, NULL);
    }
  else
    lwp->suspended++;
}

static int
lwp_running (struct inferior_list_entry *entry, void *data)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lwp = get_thread_lwp (thread);

  if (lwp->dead)
    return 0;
  if (lwp->stopped)
    return 0;
  return 1;
}

/* Stop all lwps that aren't stopped yet, except EXCEPT, if not NULL.
   If SUSPEND, then also increase the suspend count of every LWP,
   except EXCEPT.  */

static void
stop_all_lwps (int suspend, struct lwp_info *except)
{
  /* Should not be called recursively.  */
  gdb_assert (stopping_threads == NOT_STOPPING_THREADS);

  if (debug_threads)
    {
      debug_enter ();
      debug_printf ("stop_all_lwps (%s, except=%s)\n",
                    suspend ? "stop-and-suspend" : "stop",
                    except != NULL
                    ? target_pid_to_str (ptid_of (get_lwp_thread (except)))
                    : "none");
    }

  stopping_threads = (suspend
                      ? STOPPING_AND_SUSPENDING_THREADS
                      : STOPPING_THREADS);

  if (suspend)
    find_inferior (&all_threads, suspend_and_send_sigstop_callback, except);
  else
    find_inferior (&all_threads, send_sigstop_callback, except);
  wait_for_sigstop ();
  stopping_threads = NOT_STOPPING_THREADS;

  if (debug_threads)
    {
      debug_printf ("stop_all_lwps done, setting stopping_threads "
                    "back to !stopping\n");
      debug_exit ();
    }
}

/* Resume execution of the inferior process.
   If STEP is nonzero, single-step it.
   If SIGNAL is nonzero, give it that signal.  */

static void
linux_resume_one_lwp (struct lwp_info *lwp,
                      int step, int signal, siginfo_t *info)
{
  struct thread_info *thread = get_lwp_thread (lwp);
  struct thread_info *saved_thread;
  int fast_tp_collecting;

  if (lwp->stopped == 0)
    return;

  fast_tp_collecting = lwp->collecting_fast_tracepoint;

  gdb_assert (!stabilizing_threads || fast_tp_collecting);

  /* Cancel actions that rely on GDB not changing the PC (e.g., the
     user used the "jump" command, or "set $pc = foo").  */
  if (lwp->stop_pc != get_pc (lwp))
    {
      /* Collecting 'while-stepping' actions doesn't make sense
         anymore.  */
      release_while_stepping_state_list (thread);
    }

  /* If we have pending signals or status, and a new signal, enqueue the
     signal.  Also enqueue the signal if we are waiting to reinsert a
     breakpoint; it will be picked up again below.  */
  if (signal != 0
      && (lwp->status_pending_p
          || lwp->pending_signals != NULL
          || lwp->bp_reinsert != 0
          || fast_tp_collecting))
    {
      struct pending_signals *p_sig;
      p_sig = xmalloc (sizeof (*p_sig));
      p_sig->prev = lwp->pending_signals;
      p_sig->signal = signal;
      if (info == NULL)
        memset (&p_sig->info, 0, sizeof (siginfo_t));
      else
        memcpy (&p_sig->info, info, sizeof (siginfo_t));
      lwp->pending_signals = p_sig;
    }

  if (lwp->status_pending_p)
    {
      if (debug_threads)
        debug_printf ("Not resuming lwp %ld (%s, signal %d, stop %s);"
                      " has pending status\n",
                      lwpid_of (thread), step ? "step" : "continue", signal,
                      lwp->stop_expected ? "expected" : "not expected");
      return;
    }

  saved_thread = current_thread;
  current_thread = thread;

  if (debug_threads)
    debug_printf ("Resuming lwp %ld (%s, signal %d, stop %s)\n",
                  lwpid_of (thread), step ? "step" : "continue", signal,
                  lwp->stop_expected ? "expected" : "not expected");

  /* This bit needs some thinking about.  If we get a signal that
     we must report while a single-step reinsert is still pending,
     we often end up resuming the thread.  It might be better to
     (ew) allow a stack of pending events; then we could be sure that
     the reinsert happened right away and not lose any signals.

     Making this stack would also shrink the window in which breakpoints are
     uninserted (see comment in linux_wait_for_lwp) but not enough for
     complete correctness, so it won't solve that problem.  It may be
     worthwhile just to solve this one, however.  */
  if (lwp->bp_reinsert != 0)
    {
      if (debug_threads)
        debug_printf ("  pending reinsert at 0x%s\n",
                      paddress (lwp->bp_reinsert));

      if (can_hardware_single_step ())
        {
          if (fast_tp_collecting == 0)
            {
              if (step == 0)
                fprintf (stderr, "BAD - reinserting but not stepping.\n");
              if (lwp->suspended)
                fprintf (stderr, "BAD - reinserting and suspended(%d).\n",
                         lwp->suspended);
            }

          step = 1;
        }

      /* Postpone any pending signal.  It was enqueued above.  */
      signal = 0;
    }

  if (fast_tp_collecting == 1)
    {
      if (debug_threads)
        debug_printf ("lwp %ld wants to get out of fast tracepoint jump pad"
                      " (exit-jump-pad-bkpt)\n",
                      lwpid_of (thread));

      /* Postpone any pending signal.  It was enqueued above.  */
      signal = 0;
    }
  else if (fast_tp_collecting == 2)
    {
      if (debug_threads)
        debug_printf ("lwp %ld wants to get out of fast tracepoint jump pad"
                      " single-stepping\n",
                      lwpid_of (thread));

      if (can_hardware_single_step ())
        step = 1;
      else
        {
          internal_error (__FILE__, __LINE__,
                          "moving out of jump pad single-stepping"
                          " not implemented on this target");
        }

      /* Postpone any pending signal.  It was enqueued above.  */
      signal = 0;
    }

  /* If we have while-stepping actions in this thread set it stepping.
     If we have a signal to deliver, it may or may not be set to
     SIG_IGN, we don't know.  Assume so, and allow collecting
     while-stepping into a signal handler.  A possible smart thing to
     do would be to set an internal breakpoint at the signal return
     address, continue, and carry on catching this while-stepping
     action only when that breakpoint is hit.  A future
     enhancement.  */
  if (thread->while_stepping != NULL
      && can_hardware_single_step ())
    {
      if (debug_threads)
        debug_printf ("lwp %ld has a while-stepping action -> forcing step.\n",
                      lwpid_of (thread));
      step = 1;
    }

  if (the_low_target.get_pc != NULL)
    {
      struct regcache *regcache = get_thread_regcache (current_thread, 1);

      lwp->stop_pc = (*the_low_target.get_pc) (regcache);

      if (debug_threads)
        {
          debug_printf ("  %s from pc 0x%lx\n", step ? "step" : "continue",
                        (long) lwp->stop_pc);
        }
    }

  /* If we have pending signals, consume one unless we are trying to
     reinsert a breakpoint or we're trying to finish a fast tracepoint
     collect.  */
  if (lwp->pending_signals != NULL
      && lwp->bp_reinsert == 0
      && fast_tp_collecting == 0)
    {
      struct pending_signals **p_sig;

      p_sig = &lwp->pending_signals;
      while ((*p_sig)->prev != NULL)
        p_sig = &(*p_sig)->prev;

      signal = (*p_sig)->signal;
      if ((*p_sig)->info.si_signo != 0)
        ptrace (PTRACE_SETSIGINFO, lwpid_of (thread), (PTRACE_TYPE_ARG3) 0,
                &(*p_sig)->info);

      free (*p_sig);
      *p_sig = NULL;
    }

  if (the_low_target.prepare_to_resume != NULL)
    the_low_target.prepare_to_resume (lwp);

  regcache_invalidate_thread (thread);
  errno = 0;
  lwp->stopped = 0;
  lwp->stop_reason = LWP_STOPPED_BY_NO_REASON;
  lwp->stepping = step;
  ptrace (step ? PTRACE_SINGLESTEP : PTRACE_CONT, lwpid_of (thread),
          (PTRACE_TYPE_ARG3) 0,
          /* Coerce to a uintptr_t first to avoid potential gcc warning
             of coercing an 8 byte integer to a 4 byte pointer.  */
          (PTRACE_TYPE_ARG4) (uintptr_t) signal);

  current_thread = saved_thread;
  if (errno)
    {
      /* ESRCH from ptrace either means that the thread was already
         running (an error) or that it is gone (a race condition).  If
         it's gone, we will get a notification the next time we wait,
         so we can ignore the error.  We could differentiate these
         two, but it's tricky without waiting; the thread still exists
         as a zombie, so sending it signal 0 would succeed.  So just
         ignore ESRCH.  */
      if (errno == ESRCH)
        return;

      perror_with_name ("ptrace");
    }
}

struct thread_resume_array
{
  struct thread_resume *resume;
  size_t n;
};

/* This function is called once per thread via find_inferior.
   ARG is a pointer to a thread_resume_array struct.
   We look up the thread specified by ENTRY in ARG, and mark the thread
   with a pointer to the appropriate resume request.

   This algorithm is O(threads * resume elements), but resume elements
   is small (and will remain small at least until GDB supports thread
   suspension).  */

static int
linux_set_resume_request (struct inferior_list_entry *entry, void *arg)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lwp = get_thread_lwp (thread);
  int ndx;
  struct thread_resume_array *r;

  r = arg;

  for (ndx = 0; ndx < r->n; ndx++)
    {
      ptid_t ptid = r->resume[ndx].thread;
      if (ptid_equal (ptid, minus_one_ptid)
          || ptid_equal (ptid, entry->id)
          /* Handle both 'pPID' and 'pPID.-1' as meaning 'all threads
             of PID'.  */
          || (ptid_get_pid (ptid) == pid_of (thread)
              && (ptid_is_pid (ptid)
                  || ptid_get_lwp (ptid) == -1)))
        {
          if (r->resume[ndx].kind == resume_stop
              && thread->last_resume_kind == resume_stop)
            {
              if (debug_threads)
                debug_printf ("already %s LWP %ld at GDB's request\n",
                              (thread->last_status.kind
                               == TARGET_WAITKIND_STOPPED)
                              ? "stopped"
                              : "stopping",
                              lwpid_of (thread));

              continue;
            }

          lwp->resume = &r->resume[ndx];
          thread->last_resume_kind = lwp->resume->kind;

          lwp->step_range_start = lwp->resume->step_range_start;
          lwp->step_range_end = lwp->resume->step_range_end;

          /* If we had a deferred signal to report, dequeue one now.
             This can happen if LWP gets more than one signal while
             trying to get out of a jump pad.  */
          if (lwp->stopped
              && !lwp->status_pending_p
              && dequeue_one_deferred_signal (lwp, &lwp->status_pending))
            {
              lwp->status_pending_p = 1;

              if (debug_threads)
                debug_printf ("Dequeueing deferred signal %d for LWP %ld, "
                              "leaving status pending.\n",
                              WSTOPSIG (lwp->status_pending),
                              lwpid_of (thread));
            }

          return 0;
        }
    }

  /* No resume action for this thread.  */
  lwp->resume = NULL;

  return 0;
}

/* find_inferior callback for linux_resume.
   Set *FLAG_P if this lwp has an interesting status pending.  */

static int
resume_status_pending_p (struct inferior_list_entry *entry, void *flag_p)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lwp = get_thread_lwp (thread);

  /* LWPs which will not be resumed are not interesting, because
     we might not wait for them next time through linux_wait.  */
  if (lwp->resume == NULL)
    return 0;

  if (thread_still_has_status_pending_p (thread))
    * (int *) flag_p = 1;

  return 0;
}

/* Return 1 if this lwp that GDB wants running is stopped at an
   internal breakpoint that we need to step over.  It assumes that any
   required STOP_PC adjustment has already been propagated to the
   inferior's regcache.  */

static int
need_step_over_p (struct inferior_list_entry *entry, void *dummy)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lwp = get_thread_lwp (thread);
  struct thread_info *saved_thread;
  CORE_ADDR pc;

  /* LWPs which will not be resumed are not interesting, because we
     might not wait for them next time through linux_wait.  */

  if (!lwp->stopped)
    {
      if (debug_threads)
        debug_printf ("Need step over [LWP %ld]? Ignoring, not stopped\n",
                      lwpid_of (thread));
      return 0;
    }

  if (thread->last_resume_kind == resume_stop)
    {
      if (debug_threads)
        debug_printf ("Need step over [LWP %ld]? Ignoring, should remain"
                      " stopped\n",
                      lwpid_of (thread));
      return 0;
    }

  gdb_assert (lwp->suspended >= 0);

  if (lwp->suspended)
    {
      if (debug_threads)
        debug_printf ("Need step over [LWP %ld]? Ignoring, suspended\n",
                      lwpid_of (thread));
      return 0;
    }

  if (!lwp->need_step_over)
    {
      if (debug_threads)
        debug_printf ("Need step over [LWP %ld]? No\n", lwpid_of (thread));
    }

  if (lwp->status_pending_p)
    {
      if (debug_threads)
        debug_printf ("Need step over [LWP %ld]? Ignoring, has pending"
                      " status.\n",
                      lwpid_of (thread));
      return 0;
    }

  /* Note: PC, not STOP_PC.  Either GDB has adjusted the PC already,
     or we have.  */
  pc = get_pc (lwp);

  /* If the PC has changed since we stopped, then don't do anything,
     and let the breakpoint/tracepoint be hit.  This happens if, for
     instance, GDB handled the decr_pc_after_break subtraction itself,
     GDB is OOL stepping this thread, or the user has issued a "jump"
     command, or poked thread's registers herself.  */
  if (pc != lwp->stop_pc)
    {
      if (debug_threads)
        debug_printf ("Need step over [LWP %ld]? Cancelling, PC was changed. "
                      "Old stop_pc was 0x%s, PC is now 0x%s\n",
                      lwpid_of (thread),
                      paddress (lwp->stop_pc), paddress (pc));

      lwp->need_step_over = 0;
      return 0;
    }

  saved_thread = current_thread;
  current_thread = thread;

  /* We can only step over breakpoints we know about.  */
  if (breakpoint_here (pc) || fast_tracepoint_jump_here (pc))
    {
      /* Don't step over a breakpoint that GDB expects to hit
         though.  If the condition is being evaluated on the target's side
         and it evaluate to false, step over this breakpoint as well.  */
      if (gdb_breakpoint_here (pc)
          && gdb_condition_true_at_breakpoint (pc)
          && gdb_no_commands_at_breakpoint (pc))
        {
          if (debug_threads)
            debug_printf ("Need step over [LWP %ld]? yes, but found"
                          " GDB breakpoint at 0x%s; skipping step over\n",
                          lwpid_of (thread), paddress (pc));

          current_thread = saved_thread;
          return 0;
        }
      else
        {
          if (debug_threads)
            debug_printf ("Need step over [LWP %ld]? yes, "
                          "found breakpoint at 0x%s\n",
                          lwpid_of (thread), paddress (pc));

          /* We've found an lwp that needs stepping over --- return 1 so
             that find_inferior stops looking.  */
          current_thread = saved_thread;

          /* If the step over is cancelled, this is set again.  */
          lwp->need_step_over = 0;
          return 1;
        }
    }

  current_thread = saved_thread;

  if (debug_threads)
    debug_printf ("Need step over [LWP %ld]? No, no breakpoint found"
                  " at 0x%s\n",
                  lwpid_of (thread), paddress (pc));

  return 0;
}

/* Start a step-over operation on LWP.  When LWP stopped at a
   breakpoint, to make progress, we need to remove the breakpoint out
   of the way.  If we let other threads run while we do that, they may
   pass by the breakpoint location and miss hitting it.  To avoid
   that, a step-over momentarily stops all threads while LWP is
   single-stepped while the breakpoint is temporarily uninserted from
   the inferior.  When the single-step finishes, we reinsert the
   breakpoint, and let all threads that are supposed to be running,
   run again.

   On targets that don't support hardware single-step, we don't
   currently support full software single-stepping.  Instead, we only
   support stepping over the thread event breakpoint, by asking the
   low target where to place a reinsert breakpoint.  Since this
   routine assumes the breakpoint being stepped over is a thread event
   breakpoint, it usually assumes the return address of the current
   function is a good enough place to set the reinsert breakpoint.  */

static int
start_step_over (struct lwp_info *lwp)
{
  struct thread_info *thread = get_lwp_thread (lwp);
  struct thread_info *saved_thread;
  CORE_ADDR pc;
  int step;

  if (debug_threads)
    debug_printf ("Starting step-over on LWP %ld.  Stopping all threads\n",
                  lwpid_of (thread));

  stop_all_lwps (1, lwp);
  gdb_assert (lwp->suspended == 0);

  if (debug_threads)
    debug_printf ("Done stopping all threads for step-over.\n");

  /* Note, we should always reach here with an already adjusted PC,
     either by GDB (if we're resuming due to GDB's request), or by our
     caller, if we just finished handling an internal breakpoint GDB
     shouldn't care about.  */
  pc = get_pc (lwp);

  saved_thread = current_thread;
  current_thread = thread;

  lwp->bp_reinsert = pc;
  uninsert_breakpoints_at (pc);
  uninsert_fast_tracepoint_jumps_at (pc);

  if (can_hardware_single_step ())
    {
      step = 1;
    }
  else
    {
      CORE_ADDR raddr = (*the_low_target.breakpoint_reinsert_addr) ();
      set_reinsert_breakpoint (raddr);
      step = 0;
    }

  current_thread = saved_thread;

  linux_resume_one_lwp (lwp, step, 0, NULL);

  /* Require next event from this LWP.  */
  step_over_bkpt = thread->entry.id;
  return 1;
}

/* Finish a step-over.  Reinsert the breakpoint we had uninserted in
   start_step_over, if still there, and delete any reinsert
   breakpoints we've set, on non hardware single-step targets.  */

static int
finish_step_over (struct lwp_info *lwp)
{
  if (lwp->bp_reinsert != 0)
    {
      if (debug_threads)
        debug_printf ("Finished step over.\n");

      /* Reinsert any breakpoint at LWP->BP_REINSERT.  Note that there
         may be no breakpoint to reinsert there by now.  */
      reinsert_breakpoints_at (lwp->bp_reinsert);
      reinsert_fast_tracepoint_jumps_at (lwp->bp_reinsert);

      lwp->bp_reinsert = 0;

      /* Delete any software-single-step reinsert breakpoints.  No
         longer needed.  We don't have to worry about other threads
         hitting this trap, and later not being able to explain it,
         because we were stepping over a breakpoint, and we hold all
         threads but LWP stopped while doing that.  */
      if (!can_hardware_single_step ())
        delete_reinsert_breakpoints ();

      step_over_bkpt = null_ptid;
      return 1;
    }
  else
    return 0;
}

/* This function is called once per thread.  We check the thread's resume
   request, which will tell us whether to resume, step, or leave the thread
   stopped; and what signal, if any, it should be sent.

   For threads which we aren't explicitly told otherwise, we preserve
   the stepping flag; this is used for stepping over gdbserver-placed
   breakpoints.

   If pending_flags was set in any thread, we queue any needed
   signals, since we won't actually resume.  We already have a pending
   event to report, so we don't need to preserve any step requests;
   they should be re-issued if necessary.  */

static int
linux_resume_one_thread (struct inferior_list_entry *entry, void *arg)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lwp = get_thread_lwp (thread);
  int step;
  int leave_all_stopped = * (int *) arg;
  int leave_pending;

  if (lwp->resume == NULL)
    return 0;

  if (lwp->resume->kind == resume_stop)
    {
      if (debug_threads)
        debug_printf ("resume_stop request for LWP %ld\n", lwpid_of (thread));

      if (!lwp->stopped)
        {
          if (debug_threads)
            debug_printf ("stopping LWP %ld\n", lwpid_of (thread));

          /* Stop the thread, and wait for the event asynchronously,
             through the event loop.  */
          send_sigstop (lwp);
        }
      else
        {
          if (debug_threads)
            debug_printf ("already stopped LWP %ld\n",
                          lwpid_of (thread));

          /* The LWP may have been stopped in an internal event that
             was not meant to be notified back to GDB (e.g., gdbserver
             breakpoint), so we should be reporting a stop event in
             this case too.  */

          /* If the thread already has a pending SIGSTOP, this is a
             no-op.  Otherwise, something later will presumably resume
             the thread and this will cause it to cancel any pending
             operation, due to last_resume_kind == resume_stop.  If
             the thread already has a pending status to report, we
             will still report it the next time we wait - see
             status_pending_p_callback.  */

          /* If we already have a pending signal to report, then
             there's no need to queue a SIGSTOP, as this means we're
             midway through moving the LWP out of the jumppad, and we
             will report the pending signal as soon as that is
             finished.  */
          if (lwp->pending_signals_to_report == NULL)
            send_sigstop (lwp);
        }

      /* For stop requests, we're done.  */
      lwp->resume = NULL;
      thread->last_status.kind = TARGET_WAITKIND_IGNORE;
      return 0;
    }

  /* If this thread which is about to be resumed has a pending status,
     then don't resume any threads - we can just report the pending
     status.  Make sure to queue any signals that would otherwise be
     sent.  In all-stop mode, we do this decision based on if *any*
     thread has a pending status.  If there's a thread that needs the
     step-over-breakpoint dance, then don't resume any other thread
     but that particular one.  */
  leave_pending = (lwp->status_pending_p || leave_all_stopped);

  if (!leave_pending)
    {
      if (debug_threads)
        debug_printf ("resuming LWP %ld\n", lwpid_of (thread));

      step = (lwp->resume->kind == resume_step);
      linux_resume_one_lwp (lwp, step, lwp->resume->sig, NULL);
    }
  else
    {
      if (debug_threads)
        debug_printf ("leaving LWP %ld stopped\n", lwpid_of (thread));

      /* If we have a new signal, enqueue the signal.  */
      if (lwp->resume->sig != 0)
        {
          struct pending_signals *p_sig;
          p_sig = xmalloc (sizeof (*p_sig));
          p_sig->prev = lwp->pending_signals;
          p_sig->signal = lwp->resume->sig;
          memset (&p_sig->info, 0, sizeof (siginfo_t));

          /* If this is the same signal we were previously stopped by,
             make sure to queue its siginfo.  We can ignore the return
             value of ptrace; if it fails, we'll skip
             PTRACE_SETSIGINFO.  */
          if (WIFSTOPPED (lwp->last_status)
              && WSTOPSIG (lwp->last_status) == lwp->resume->sig)
            ptrace (PTRACE_GETSIGINFO, lwpid_of (thread), (PTRACE_TYPE_ARG3) 0,
                    &p_sig->info);

          lwp->pending_signals = p_sig;
        }
    }

  thread->last_status.kind = TARGET_WAITKIND_IGNORE;
  lwp->resume = NULL;
  return 0;
}

static void
linux_resume (struct thread_resume *resume_info, size_t n)
{
  struct thread_resume_array array = { resume_info, n };
  struct thread_info *need_step_over = NULL;
  int any_pending;
  int leave_all_stopped;

  if (debug_threads)
    {
      debug_enter ();
      debug_printf ("linux_resume:\n");
    }

  find_inferior (&all_threads, linux_set_resume_request, &array);

  /* If there is a thread which would otherwise be resumed, which has
     a pending status, then don't resume any threads - we can just
     report the pending status.  Make sure to queue any signals that
     would otherwise be sent.  In non-stop mode, we'll apply this
     logic to each thread individually.  We consume all pending events
     before considering to start a step-over (in all-stop).  */
  any_pending = 0;
  if (!non_stop)
    find_inferior (&all_threads, resume_status_pending_p, &any_pending);

  /* If there is a thread which would otherwise be resumed, which is
     stopped at a breakpoint that needs stepping over, then don't
     resume any threads - have it step over the breakpoint with all
     other threads stopped, then resume all threads again.  Make sure
     to queue any signals that would otherwise be delivered or
     queued.  */
  if (!any_pending && supports_breakpoints ())
    need_step_over
      = (struct thread_info *) find_inferior (&all_threads,
                                              need_step_over_p, NULL);

  leave_all_stopped = (need_step_over != NULL || any_pending);

  if (debug_threads)
    {
      if (need_step_over != NULL)
        debug_printf ("Not resuming all, need step over\n");
      else if (any_pending)
        debug_printf ("Not resuming, all-stop and found "
                      "an LWP with pending status\n");
      else
        debug_printf ("Resuming, no pending status or step over needed\n");
    }

  /* Even if we're leaving threads stopped, queue all signals we'd
     otherwise deliver.  */
  find_inferior (&all_threads, linux_resume_one_thread, &leave_all_stopped);

  if (need_step_over)
    start_step_over (get_thread_lwp (need_step_over));

  if (debug_threads)
    {
      debug_printf ("linux_resume done\n");
      debug_exit ();
    }
}

/* This function is called once per thread.  We check the thread's
   last resume request, which will tell us whether to resume, step, or
   leave the thread stopped.  Any signal the client requested to be
   delivered has already been enqueued at this point.

   If any thread that GDB wants running is stopped at an internal
   breakpoint that needs stepping over, we start a step-over operation
   on that particular thread, and leave all others stopped.  */

static int
proceed_one_lwp (struct inferior_list_entry *entry, void *except)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lwp = get_thread_lwp (thread);
  int step;

  if (lwp == except)
    return 0;

  if (debug_threads)
    debug_printf ("proceed_one_lwp: lwp %ld\n", lwpid_of (thread));

  if (!lwp->stopped)
    {
      if (debug_threads)
        debug_printf ("   LWP %ld already running\n", lwpid_of (thread));
      return 0;
    }

  if (thread->last_resume_kind == resume_stop
      && thread->last_status.kind != TARGET_WAITKIND_IGNORE)
    {
      if (debug_threads)
        debug_printf ("   client wants LWP to remain %ld stopped\n",
                      lwpid_of (thread));
      return 0;
    }

  if (lwp->status_pending_p)
    {
      if (debug_threads)
        debug_printf ("   LWP %ld has pending status, leaving stopped\n",
                      lwpid_of (thread));
      return 0;
    }

  gdb_assert (lwp->suspended >= 0);

  if (lwp->suspended)
    {
      if (debug_threads)
        debug_printf ("   LWP %ld is suspended\n", lwpid_of (thread));
      return 0;
    }

  if (thread->last_resume_kind == resume_stop
      && lwp->pending_signals_to_report == NULL
      && lwp->collecting_fast_tracepoint == 0)
    {
      /* We haven't reported this LWP as stopped yet (otherwise, the
         last_status.kind check above would catch it, and we wouldn't
         reach here.  This LWP may have been momentarily paused by a
         stop_all_lwps call while handling for example, another LWP's
         step-over.  In that case, the pending expected SIGSTOP signal
         that was queued at vCont;t handling time will have already
         been consumed by wait_for_sigstop, and so we need to requeue
         another one here.  Note that if the LWP already has a SIGSTOP
         pending, this is a no-op.  */

      if (debug_threads)
        debug_printf ("Client wants LWP %ld to stop. "
                      "Making sure it has a SIGSTOP pending\n",
                      lwpid_of (thread));

      send_sigstop (lwp);
    }

  step = thread->last_resume_kind == resume_step;
  linux_resume_one_lwp (lwp, step, 0, NULL);
  return 0;
}

static int
unsuspend_and_proceed_one_lwp (struct inferior_list_entry *entry, void *except)
{
  struct thread_info *thread = (struct thread_info *) entry;
  struct lwp_info *lwp = get_thread_lwp (thread);

  if (lwp == except)
    return 0;

  lwp->suspended--;
  gdb_assert (lwp->suspended >= 0);

  return proceed_one_lwp (entry, except);
}

/* When we finish a step-over, set threads running again.  If there's
   another thread that may need a step-over, now's the time to start
   it.  Eventually, we'll move all threads past their breakpoints.  */

static void
proceed_all_lwps (void)
{
  struct thread_info *need_step_over;

  /* If there is a thread which would otherwise be resumed, which is
     stopped at a breakpoint that needs stepping over, then don't
     resume any threads - have it step over the breakpoint with all
     other threads stopped, then resume all threads again.  */

  if (supports_breakpoints ())
    {
      need_step_over
        = (struct thread_info *) find_inferior (&all_threads,
                                                need_step_over_p, NULL);

      if (need_step_over != NULL)
        {
          if (debug_threads)
            debug_printf ("proceed_all_lwps: found "
                          "thread %ld needing a step-over\n",
                          lwpid_of (need_step_over));

          start_step_over (get_thread_lwp (need_step_over));
          return;
        }
    }

  if (debug_threads)
    debug_printf ("Proceeding, no step-over needed\n");

  find_inferior (&all_threads, proceed_one_lwp, NULL);
}

/* Stopped LWPs that the client wanted to be running, that don't have
   pending statuses, are set to run again, except for EXCEPT, if not
   NULL.  This undoes a stop_all_lwps call.  */

static void
unstop_all_lwps (int unsuspend, struct lwp_info *except)
{
  if (debug_threads)
    {
      debug_enter ();
      if (except)
        debug_printf ("unstopping all lwps, except=(LWP %ld)\n",
                      lwpid_of (get_lwp_thread (except)));
      else
        debug_printf ("unstopping all lwps\n");
    }

  if (unsuspend)
    find_inferior (&all_threads, unsuspend_and_proceed_one_lwp, except);
  else
    find_inferior (&all_threads, proceed_one_lwp, except);

  if (debug_threads)
    {
      debug_printf ("unstop_all_lwps done\n");
      debug_exit ();
    }
}


#ifdef HAVE_LINUX_REGSETS

#define use_linux_regsets 1

/* Returns true if REGSET has been disabled.  */

static int
regset_disabled (struct regsets_info *info, struct regset_info *regset)
{
  return (info->disabled_regsets != NULL
          && info->disabled_regsets[regset - info->regsets]);
}

/* Disable REGSET.  */

static void
disable_regset (struct regsets_info *info, struct regset_info *regset)
{
  int dr_offset;

  dr_offset = regset - info->regsets;
  if (info->disabled_regsets == NULL)
    info->disabled_regsets = xcalloc (1, info->num_regsets);
  info->disabled_regsets[dr_offset] = 1;
}

static int
regsets_fetch_inferior_registers (struct regsets_info *regsets_info,
                                  struct regcache *regcache)
{
  struct regset_info *regset;
  int saw_general_regs = 0;
  int pid;
  struct iovec iov;

  pid = lwpid_of (current_thread);
  for (regset = regsets_info->regsets; regset->size >= 0; regset++)
    {
      void *buf, *data;
      int nt_type, res;

      if (regset->size == 0 || regset_disabled (regsets_info, regset))
        continue;

      buf = xmalloc (regset->size);

      nt_type = regset->nt_type;
      if (nt_type)
        {
          iov.iov_base = buf;
          iov.iov_len = regset->size;
          data = (void *) &iov;
        }
      else
        data = buf;

#ifndef __sparc__
      res = ptrace (regset->get_request, pid,
                    (PTRACE_TYPE_ARG3) (long) nt_type, data);
#else
      res = ptrace (regset->get_request, pid, data, nt_type);
#endif
      if (res < 0)
        {
          if (errno == EIO)
            {
              /* If we get EIO on a regset, do not try it again for
                 this process mode.  */
              disable_regset (regsets_info, regset);
            }
          else if (errno == ENODATA)
            {
              /* ENODATA may be returned if the regset is currently
                 not "active".  This can happen in normal operation,
                 so suppress the warning in this case.  */
            }
          else
            {
              char s[256];
              sprintf (s, "ptrace(regsets_fetch_inferior_registers) PID=%d",
                       pid);
              perror (s);
            }
        }
      else
        {
          if (regset->type == GENERAL_REGS)
            saw_general_regs = 1;
          regset->store_function (regcache, buf);
        }
      free (buf);
    }
  if (saw_general_regs)
    return 0;
  else
    return 1;
}

static int
regsets_store_inferior_registers (struct regsets_info *regsets_info,
                                  struct regcache *regcache)
{
  struct regset_info *regset;
  int saw_general_regs = 0;
  int pid;
  struct iovec iov;

  pid = lwpid_of (current_thread);
  for (regset = regsets_info->regsets; regset->size >= 0; regset++)
    {
      void *buf, *data;
      int nt_type, res;

      if (regset->size == 0 || regset_disabled (regsets_info, regset)
          || regset->fill_function == NULL)
        continue;

      buf = xmalloc (regset->size);

      /* First fill the buffer with the current register set contents,
         in case there are any items in the kernel's regset that are
         not in gdbserver's regcache.  */

      nt_type = regset->nt_type;
      if (nt_type)
        {
          iov.iov_base = buf;
          iov.iov_len = regset->size;
          data = (void *) &iov;
        }
      else
        data = buf;

#ifndef __sparc__
      res = ptrace (regset->get_request, pid,
                    (PTRACE_TYPE_ARG3) (long) nt_type, data);
#else
      res = ptrace (regset->get_request, pid, data, nt_type);
#endif

      if (res == 0)
        {
          /* Then overlay our cached registers on that.  */
          regset->fill_function (regcache, buf);

          /* Only now do we write the register set.  */
#ifndef __sparc__
          res = ptrace (regset->set_request, pid,
                        (PTRACE_TYPE_ARG3) (long) nt_type, data);
#else
          res = ptrace (regset->set_request, pid, data, nt_type);
#endif
        }

      if (res < 0)
        {
          if (errno == EIO)
            {
              /* If we get EIO on a regset, do not try it again for
                 this process mode.  */
              disable_regset (regsets_info, regset);
            }
          else if (errno == ESRCH)
            {
              /* At this point, ESRCH should mean the process is
                 already gone, in which case we simply ignore attempts
                 to change its registers.  See also the related
                 comment in linux_resume_one_lwp.  */
              free (buf);
              return 0;
            }
          else
            {
              perror ("Warning: ptrace(regsets_store_inferior_registers)");
            }
        }
      else if (regset->type == GENERAL_REGS)
        saw_general_regs = 1;
      free (buf);
    }
  if (saw_general_regs)
    return 0;
  else
    return 1;
}

#else /* !HAVE_LINUX_REGSETS */

#define use_linux_regsets 0
#define regsets_fetch_inferior_registers(regsets_info, regcache) 1
#define regsets_store_inferior_registers(regsets_info, regcache) 1

#endif

/* Return 1 if register REGNO is supported by one of the regset ptrace
   calls or 0 if it has to be transferred individually.  */

static int
linux_register_in_regsets (const struct regs_info *regs_info, int regno)
{
  unsigned char mask = 1 << (regno % 8);
  size_t index = regno / 8;

  return (use_linux_regsets
          && (regs_info->regset_bitmap == NULL
              || (regs_info->regset_bitmap[index] & mask) != 0));
}

#ifdef HAVE_LINUX_USRREGS

int
register_addr (const struct usrregs_info *usrregs, int regnum)
{
  int addr;

  if (regnum < 0 || regnum >= usrregs->num_regs)
    error ("Invalid register number %d.", regnum);

  addr = usrregs->regmap[regnum];

  return addr;
}

/* Fetch one register.  */
static void
fetch_register (const struct usrregs_info *usrregs,
                struct regcache *regcache, int regno)
{
  CORE_ADDR regaddr;
  int i, size;
  char *buf;
  int pid;

  if (regno >= usrregs->num_regs)
    return;
  if ((*the_low_target.cannot_fetch_register) (regno))
    return;

  regaddr = register_addr (usrregs, regno);
  if (regaddr == -1)
    return;

  size = ((register_size (regcache->tdesc, regno)
           + sizeof (PTRACE_XFER_TYPE) - 1)
          & -sizeof (PTRACE_XFER_TYPE));
  buf = alloca (size);

  pid = lwpid_of (current_thread);
  for (i = 0; i < size; i += sizeof (PTRACE_XFER_TYPE))
    {
      errno = 0;
      *(PTRACE_XFER_TYPE *) (buf + i) =
        ptrace (PTRACE_PEEKUSER, pid,
                /* Coerce to a uintptr_t first to avoid potential gcc warning
                   of coercing an 8 byte integer to a 4 byte pointer.  */
                (PTRACE_TYPE_ARG3) (uintptr_t) regaddr, (PTRACE_TYPE_ARG4) 0);
      regaddr += sizeof (PTRACE_XFER_TYPE);
      if (errno != 0)
        error ("reading register %d: %s", regno, strerror (errno));
    }

  if (the_low_target.supply_ptrace_register)
    the_low_target.supply_ptrace_register (regcache, regno, buf);
  else
    supply_register (regcache, regno, buf);
}

/* Store one register.  */
static void
store_register (const struct usrregs_info *usrregs,
                struct regcache *regcache, int regno)
{
  CORE_ADDR regaddr;
  int i, size;
  char *buf;
  int pid;

  if (regno >= usrregs->num_regs)
    return;
  if ((*the_low_target.cannot_store_register) (regno))
    return;

  regaddr = register_addr (usrregs, regno);
  if (regaddr == -1)
    return;

  size = ((register_size (regcache->tdesc, regno)
           + sizeof (PTRACE_XFER_TYPE) - 1)
          & -sizeof (PTRACE_XFER_TYPE));
  buf = alloca (size);
  memset (buf, 0, size);

  if (the_low_target.collect_ptrace_register)
    the_low_target.collect_ptrace_register (regcache, regno, buf);
  else
    collect_register (regcache, regno, buf);

  pid = lwpid_of (current_thread);
  for (i = 0; i < size; i += sizeof (PTRACE_XFER_TYPE))
    {
      errno = 0;
      ptrace (PTRACE_POKEUSER, pid,
            /* Coerce to a uintptr_t first to avoid potential gcc warning
               about coercing an 8 byte integer to a 4 byte pointer.  */
              (PTRACE_TYPE_ARG3) (uintptr_t) regaddr,
              (PTRACE_TYPE_ARG4) *(PTRACE_XFER_TYPE *) (buf + i));
      if (errno != 0)
        {
          /* At this point, ESRCH should mean the process is
             already gone, in which case we simply ignore attempts
             to change its registers.  See also the related
             comment in linux_resume_one_lwp.  */
          if (errno == ESRCH)
            return;

          if ((*the_low_target.cannot_store_register) (regno) == 0)
            error ("writing register %d: %s", regno, strerror (errno));
        }
      regaddr += sizeof (PTRACE_XFER_TYPE);
    }
}

/* Fetch all registers, or just one, from the child process.
   If REGNO is -1, do this for all registers, skipping any that are
   assumed to have been retrieved by regsets_fetch_inferior_registers,
   unless ALL is non-zero.
   Otherwise, REGNO specifies which register (so we can save time).  */
static void
usr_fetch_inferior_registers (const struct regs_info *regs_info,
                              struct regcache *regcache, int regno, int all)
{
  struct usrregs_info *usr = regs_info->usrregs;

  if (regno == -1)
    {
      for (regno = 0; regno < usr->num_regs; regno++)
        if (all || !linux_register_in_regsets (regs_info, regno))
          fetch_register (usr, regcache, regno);
    }
  else
    fetch_register (usr, regcache, regno);
}

/* Store our register values back into the inferior.
   If REGNO is -1, do this for all registers, skipping any that are
   assumed to have been saved by regsets_store_inferior_registers,
   unless ALL is non-zero.
   Otherwise, REGNO specifies which register (so we can save time).  */
static void
usr_store_inferior_registers (const struct regs_info *regs_info,
                              struct regcache *regcache, int regno, int all)
{
  struct usrregs_info *usr = regs_info->usrregs;

  if (regno == -1)
    {
      for (regno = 0; regno < usr->num_regs; regno++)
        if (all || !linux_register_in_regsets (regs_info, regno))
          store_register (usr, regcache, regno);
    }
  else
    store_register (usr, regcache, regno);
}

#else /* !HAVE_LINUX_USRREGS */

#define usr_fetch_inferior_registers(regs_info, regcache, regno, all) do {} while (0)
#define usr_store_inferior_registers(regs_info, regcache, regno, all) do {} while (0)

#endif


void
linux_fetch_registers (struct regcache *regcache, int regno)
{
  int use_regsets;
  int all = 0;
  const struct regs_info *regs_info = (*the_low_target.regs_info) ();

  if (regno == -1)
    {
      if (the_low_target.fetch_register != NULL
          && regs_info->usrregs != NULL)
        for (regno = 0; regno < regs_info->usrregs->num_regs; regno++)
          (*the_low_target.fetch_register) (regcache, regno);

      all = regsets_fetch_inferior_registers (regs_info->regsets_info, regcache);
      if (regs_info->usrregs != NULL)
        usr_fetch_inferior_registers (regs_info, regcache, -1, all);
    }
  else
    {
      if (the_low_target.fetch_register != NULL
          && (*the_low_target.fetch_register) (regcache, regno))
        return;

      use_regsets = linux_register_in_regsets (regs_info, regno);
      if (use_regsets)
        all = regsets_fetch_inferior_registers (regs_info->regsets_info,
                                                regcache);
      if ((!use_regsets || all) && regs_info->usrregs != NULL)
        usr_fetch_inferior_registers (regs_info, regcache, regno, 1);
    }
}

void
linux_store_registers (struct regcache *regcache, int regno)
{
  int use_regsets;
  int all = 0;
  const struct regs_info *regs_info = (*the_low_target.regs_info) ();

  if (regno == -1)
    {
      all = regsets_store_inferior_registers (regs_info->regsets_info,
                                              regcache);
      if (regs_info->usrregs != NULL)
        usr_store_inferior_registers (regs_info, regcache, regno, all);
    }
  else
    {
      use_regsets = linux_register_in_regsets (regs_info, regno);
      if (use_regsets)
        all = regsets_store_inferior_registers (regs_info->regsets_info,
                                                regcache);
      if ((!use_regsets || all) && regs_info->usrregs != NULL)
        usr_store_inferior_registers (regs_info, regcache, regno, 1);
    }
}


/* Copy LEN bytes from inferior's memory starting at MEMADDR
   to debugger memory starting at MYADDR.  */

static int
linux_read_memory (CORE_ADDR memaddr, unsigned char *myaddr, int len)
{
  int pid = lwpid_of (current_thread);
  register PTRACE_XFER_TYPE *buffer;
  register CORE_ADDR addr;
  register int count;
  char filename[64];
  register int i;
  int ret;
  int fd;

  /* Try using /proc.  Don't bother for one word.  */
  if (len >= 3 * sizeof (long))
    {
      int bytes;

      /* We could keep this file open and cache it - possibly one per
         thread.  That requires some juggling, but is even faster.  */
      sprintf (filename, "/proc/%d/mem", pid);
      fd = open (filename, O_RDONLY | O_LARGEFILE);
      if (fd == -1)
        goto no_proc;

      /* If pread64 is available, use it.  It's faster if the kernel
         supports it (only one syscall), and it's 64-bit safe even on
         32-bit platforms (for instance, SPARC debugging a SPARC64
         application).  */
#ifdef HAVE_PREAD64
      bytes = pread64 (fd, myaddr, len, memaddr);
#else
      bytes = -1;
      if (lseek (fd, memaddr, SEEK_SET) != -1)
        bytes = read (fd, myaddr, len);
#endif

      close (fd);
      if (bytes == len)
        return 0;

      /* Some data was read, we'll try to get the rest with ptrace.  */
      if (bytes > 0)
        {
          memaddr += bytes;
          myaddr += bytes;
          len -= bytes;
        }
    }

 no_proc:
  /* Round starting address down to longword boundary.  */
  addr = memaddr & -(CORE_ADDR) sizeof (PTRACE_XFER_TYPE);
  /* Round ending address up; get number of longwords that makes.  */
  count = ((((memaddr + len) - addr) + sizeof (PTRACE_XFER_TYPE) - 1)
           / sizeof (PTRACE_XFER_TYPE));
  /* Allocate buffer of that many longwords.  */
  buffer = (PTRACE_XFER_TYPE *) alloca (count * sizeof (PTRACE_XFER_TYPE));

  /* Read all the longwords */
  errno = 0;
  for (i = 0; i < count; i++, addr += sizeof (PTRACE_XFER_TYPE))
    {
      /* Coerce the 3rd arg to a uintptr_t first to avoid potential gcc warning
         about coercing an 8 byte integer to a 4 byte pointer.  */
      buffer[i] = ptrace (PTRACE_PEEKTEXT, pid,
                          (PTRACE_TYPE_ARG3) (uintptr_t) addr,
                          (PTRACE_TYPE_ARG4) 0);
      if (errno)
        break;
    }
  ret = errno;

  /* Copy appropriate bytes out of the buffer.  */
  if (i > 0)
    {
      i *= sizeof (PTRACE_XFER_TYPE);
      i -= memaddr & (sizeof (PTRACE_XFER_TYPE) - 1);
      memcpy (myaddr,
              (char *) buffer + (memaddr & (sizeof (PTRACE_XFER_TYPE) - 1)),
              i < len ? i : len);
    }

  return ret;
}

/* Copy LEN bytes of data from debugger memory at MYADDR to inferior's
   memory at MEMADDR.  On failure (cannot write to the inferior)
   returns the value of errno.  Always succeeds if LEN is zero.  */

static int
linux_write_memory (CORE_ADDR memaddr, const unsigned char *myaddr, int len)
{
  register int i;
  /* Round starting address down to longword boundary.  */
  register CORE_ADDR addr = memaddr & -(CORE_ADDR) sizeof (PTRACE_XFER_TYPE);
  /* Round ending address up; get number of longwords that makes.  */
  register int count
    = (((memaddr + len) - addr) + sizeof (PTRACE_XFER_TYPE) - 1)
    / sizeof (PTRACE_XFER_TYPE);

  /* Allocate buffer of that many longwords.  */
  register PTRACE_XFER_TYPE *buffer = (PTRACE_XFER_TYPE *)
    alloca (count * sizeof (PTRACE_XFER_TYPE));

  int pid = lwpid_of (current_thread);

  if (len == 0)
    {
      /* Zero length write always succeeds.  */
      return 0;
    }

  if (debug_threads)
    {
      /* Dump up to four bytes.  */
      unsigned int val = * (unsigned int *) myaddr;
      if (len == 1)
        val = val & 0xff;
      else if (len == 2)
        val = val & 0xffff;
      else if (len == 3)
        val = val & 0xffffff;
      debug_printf ("Writing %0*x to 0x%08lx\n", 2 * ((len < 4) ? len : 4),
                    val, (long)memaddr);
    }

  /* Fill start and end extra bytes of buffer with existing memory data.  */

  errno = 0;
  /* Coerce the 3rd arg to a uintptr_t first to avoid potential gcc warning
     about coercing an 8 byte integer to a 4 byte pointer.  */
  buffer[0] = ptrace (PTRACE_PEEKTEXT, pid,
                      (PTRACE_TYPE_ARG3) (uintptr_t) addr,
                      (PTRACE_TYPE_ARG4) 0);
  if (errno)
    return errno;

  if (count > 1)
    {
      errno = 0;
      buffer[count - 1]
        = ptrace (PTRACE_PEEKTEXT, pid,
                  /* Coerce to a uintptr_t first to avoid potential gcc warning
                     about coercing an 8 byte integer to a 4 byte pointer.  */
                  (PTRACE_TYPE_ARG3) (uintptr_t) (addr + (count - 1)
                                                  * sizeof (PTRACE_XFER_TYPE)),
                  (PTRACE_TYPE_ARG4) 0);
      if (errno)
        return errno;
    }

  /* Copy data to be written over corresponding part of buffer.  */

  memcpy ((char *) buffer + (memaddr & (sizeof (PTRACE_XFER_TYPE) - 1)),
          myaddr, len);

  /* Write the entire buffer.  */

  for (i = 0; i < count; i++, addr += sizeof (PTRACE_XFER_TYPE))
    {
      errno = 0;
      ptrace (PTRACE_POKETEXT, pid,
              /* Coerce to a uintptr_t first to avoid potential gcc warning
                 about coercing an 8 byte integer to a 4 byte pointer.  */
              (PTRACE_TYPE_ARG3) (uintptr_t) addr,
              (PTRACE_TYPE_ARG4) buffer[i]);
      if (errno)
        return errno;
    }

  return 0;
}

static void
linux_look_up_symbols (void)
{
#ifdef USE_THREAD_DB
  struct process_info *proc = current_process ();

  if (proc->private->thread_db != NULL)
    return;

  /* If the kernel supports tracing clones, then we don't need to
     use the magic thread event breakpoint to learn about
     threads.  */
  thread_db_init (!linux_supports_traceclone ());
#endif
}

static void
linux_request_interrupt (void)
{
  extern unsigned long signal_pid;

  /* Send a SIGINT to the process group.  This acts just like the user
     typed a ^C on the controlling terminal.  */
  kill (-signal_pid, SIGINT);
}

/* Copy LEN bytes from inferior's auxiliary vector starting at OFFSET
   to debugger memory starting at MYADDR.  */

static int
linux_read_auxv (CORE_ADDR offset, unsigned char *myaddr, unsigned int len)
{
  char filename[PATH_MAX];
  int fd, n;
  int pid = lwpid_of (current_thread);

  xsnprintf (filename, sizeof filename, "/proc/%d/auxv", pid);

  fd = open (filename, O_RDONLY);
  if (fd < 0)
    return -1;

  if (offset != (CORE_ADDR) 0
      && lseek (fd, (off_t) offset, SEEK_SET) != (off_t) offset)
    n = -1;
  else
    n = read (fd, myaddr, len);

  close (fd);

  return n;
}

/* These breakpoint and watchpoint related wrapper functions simply
   pass on the function call if the target has registered a
   corresponding function.  */

static int
linux_supports_z_point_type (char z_type)
{
  return (the_low_target.supports_z_point_type != NULL
          && the_low_target.supports_z_point_type (z_type));
}

static int
linux_insert_point (enum raw_bkpt_type type, CORE_ADDR addr,
                    int size, struct raw_breakpoint *bp)
{
  if (the_low_target.insert_point != NULL)
    return the_low_target.insert_point (type, addr, size, bp);
  else
    /* Unsupported (see target.h).  */
    return 1;
}

static int
linux_remove_point (enum raw_bkpt_type type, CORE_ADDR addr,
                    int size, struct raw_breakpoint *bp)
{
  if (the_low_target.remove_point != NULL)
    return the_low_target.remove_point (type, addr, size, bp);
  else
    /* Unsupported (see target.h).  */
    return 1;
}

static int
linux_stopped_by_watchpoint (void)
{
  struct lwp_info *lwp = get_thread_lwp (current_thread);

  return lwp->stop_reason == LWP_STOPPED_BY_WATCHPOINT;
}

static CORE_ADDR
linux_stopped_data_address (void)
{
  struct lwp_info *lwp = get_thread_lwp (current_thread);

  return lwp->stopped_data_address;
}

#if defined(__UCLIBC__) && defined(HAS_NOMMU)         \
    && defined(PT_TEXT_ADDR) && defined(PT_DATA_ADDR) \
    && defined(PT_TEXT_END_ADDR)

/* This is only used for targets that define PT_TEXT_ADDR,
   PT_DATA_ADDR and PT_TEXT_END_ADDR.  If those are not defined, supposedly
   the target has different ways of acquiring this information, like
   loadmaps.  */

/* Under uClinux, programs are loaded at non-zero offsets, which we need
   to tell gdb about.  */

static int
linux_read_offsets (CORE_ADDR *text_p, CORE_ADDR *data_p)
{
  unsigned long text, text_end, data;
  int pid = lwpid_of (get_thread_lwp (current_thread));

  errno = 0;

  text = ptrace (PTRACE_PEEKUSER, pid, (PTRACE_TYPE_ARG3) PT_TEXT_ADDR,
                 (PTRACE_TYPE_ARG4) 0);
  text_end = ptrace (PTRACE_PEEKUSER, pid, (PTRACE_TYPE_ARG3) PT_TEXT_END_ADDR,
                     (PTRACE_TYPE_ARG4) 0);
  data = ptrace (PTRACE_PEEKUSER, pid, (PTRACE_TYPE_ARG3) PT_DATA_ADDR,
                 (PTRACE_TYPE_ARG4) 0);

  if (errno == 0)
    {
      /* Both text and data offsets produced at compile-time (and so
         used by gdb) are relative to the beginning of the program,
         with the data segment immediately following the text segment.
         However, the actual runtime layout in memory may put the data
         somewhere else, so when we send gdb a data base-address, we
         use the real data base address and subtract the compile-time
         data base-address from it (which is just the length of the
         text segment).  BSS immediately follows data in both
         cases.  */
      *text_p = text;
      *data_p = data - (text_end - text);

      return 1;
    }
 return 0;
}
#endif

static int
linux_qxfer_osdata (const char *annex,
                    unsigned char *readbuf, unsigned const char *writebuf,
                    CORE_ADDR offset, int len)
{
  return linux_common_xfer_osdata (annex, readbuf, offset, len);
}

/* Convert a native/host siginfo object, into/from the siginfo in the
   layout of the inferiors' architecture.  */

static void
siginfo_fixup (siginfo_t *siginfo, void *inf_siginfo, int direction)
{
  int done = 0;

  if (the_low_target.siginfo_fixup != NULL)
    done = the_low_target.siginfo_fixup (siginfo, inf_siginfo, direction);

  /* If there was no callback, or the callback didn't do anything,
     then just do a straight memcpy.  */
  if (!done)
    {
      if (direction == 1)
        memcpy (siginfo, inf_siginfo, sizeof (siginfo_t));
      else
        memcpy (inf_siginfo, siginfo, sizeof (siginfo_t));
    }
}

static int
linux_xfer_siginfo (const char *annex, unsigned char *readbuf,
                    unsigned const char *writebuf, CORE_ADDR offset, int len)
{
  int pid;
  siginfo_t siginfo;
  char inf_siginfo[sizeof (siginfo_t)];

  if (current_thread == NULL)
    return -1;

  pid = lwpid_of (current_thread);

  if (debug_threads)
    debug_printf ("%s siginfo for lwp %d.\n",
                  readbuf != NULL ? "Reading" : "Writing",
                  pid);

  if (offset >= sizeof (siginfo))
    return -1;

  if (ptrace (PTRACE_GETSIGINFO, pid, (PTRACE_TYPE_ARG3) 0, &siginfo) != 0)
    return -1;

  /* When GDBSERVER is built as a 64-bit application, ptrace writes into
     SIGINFO an object with 64-bit layout.  Since debugging a 32-bit
     inferior with a 64-bit GDBSERVER should look the same as debugging it
     with a 32-bit GDBSERVER, we need to convert it.  */
  siginfo_fixup (&siginfo, inf_siginfo, 0);

  if (offset + len > sizeof (siginfo))
    len = sizeof (siginfo) - offset;

  if (readbuf != NULL)
    memcpy (readbuf, inf_siginfo + offset, len);
  else
    {
      memcpy (inf_siginfo + offset, writebuf, len);

      /* Convert back to ptrace layout before flushing it out.  */
      siginfo_fixup (&siginfo, inf_siginfo, 1);

      if (ptrace (PTRACE_SETSIGINFO, pid, (PTRACE_TYPE_ARG3) 0, &siginfo) != 0)
        return -1;
    }

  return len;
}

/* SIGCHLD handler that serves two purposes: In non-stop/async mode,
   so we notice when children change state; as the handler for the
   sigsuspend in my_waitpid.  */

static void
sigchld_handler (int signo)
{
  int old_errno = errno;

  if (debug_threads)
    {
      do
        {
          /* fprintf is not async-signal-safe, so call write
             directly.  */
          if (write (2, "sigchld_handler\n",
                     sizeof ("sigchld_handler\n") - 1) < 0)
            break; /* just ignore */
        } while (0);
    }

  if (target_is_async_p ())
    async_file_mark (); /* trigger a linux_wait */

  errno = old_errno;
}

static int
linux_supports_non_stop (void)
{
  return 1;
}

static int
linux_async (int enable)
{
  int previous = target_is_async_p ();

  if (debug_threads)
    debug_printf ("linux_async (%d), previous=%d\n",
                  enable, previous);

  if (previous != enable)
    {
      sigset_t mask;
      sigemptyset (&mask);
      sigaddset (&mask, SIGCHLD);

      sigprocmask (SIG_BLOCK, &mask, NULL);

      if (enable)
        {
          if (pipe (linux_event_pipe) == -1)
            {
              linux_event_pipe[0] = -1;
              linux_event_pipe[1] = -1;
              sigprocmask (SIG_UNBLOCK, &mask, NULL);

              warning ("creating event pipe failed.");
              return previous;
            }

          fcntl (linux_event_pipe[0], F_SETFL, O_NONBLOCK);
          fcntl (linux_event_pipe[1], F_SETFL, O_NONBLOCK);

          /* Register the event loop handler.  */
          add_file_handler (linux_event_pipe[0],
                            handle_target_event, NULL);

          /* Always trigger a linux_wait.  */
          async_file_mark ();
        }
      else
        {
          delete_file_handler (linux_event_pipe[0]);

          close (linux_event_pipe[0]);
          close (linux_event_pipe[1]);
          linux_event_pipe[0] = -1;
          linux_event_pipe[1] = -1;
        }

      sigprocmask (SIG_UNBLOCK, &mask, NULL);
    }

  return previous;
}

static int
linux_start_non_stop (int nonstop)
{
  /* Register or unregister from event-loop accordingly.  */
  linux_async (nonstop);

  if (target_is_async_p () != (nonstop != 0))
    return -1;

  return 0;
}

static int
linux_supports_multi_process (void)
{
  return 1;
}

static int
linux_supports_disable_randomization (void)
{
#ifdef HAVE_PERSONALITY
  return 1;
#else
  return 0;
#endif
}

static int
linux_supports_agent (void)
{
  return 1;
}

static int
linux_supports_range_stepping (void)
{
  if (*the_low_target.supports_range_stepping == NULL)
    return 0;

  return (*the_low_target.supports_range_stepping) ();
}

/* Enumerate spufs IDs for process PID.  */
static int
spu_enumerate_spu_ids (long pid, unsigned char *buf, CORE_ADDR offset, int len)
{
  int pos = 0;
  int written = 0;
  char path[128];
  DIR *dir;
  struct dirent *entry;

  sprintf (path, "/proc/%ld/fd", pid);
  dir = opendir (path);
  if (!dir)
    return -1;

  rewinddir (dir);
  while ((entry = readdir (dir)) != NULL)
    {
      struct stat st;
      struct statfs stfs;
      int fd;

      fd = atoi (entry->d_name);
      if (!fd)
        continue;

      sprintf (path, "/proc/%ld/fd/%d", pid, fd);
      if (stat (path, &st) != 0)
        continue;
      if (!S_ISDIR (st.st_mode))
        continue;

      if (statfs (path, &stfs) != 0)
        continue;
      if (stfs.f_type != SPUFS_MAGIC)
        continue;

      if (pos >= offset && pos + 4 <= offset + len)
        {
          *(unsigned int *)(buf + pos - offset) = fd;
          written += 4;
        }
      pos += 4;
    }

  closedir (dir);
  return written;
}

/* Implements the to_xfer_partial interface for the TARGET_OBJECT_SPU
   object type, using the /proc file system.  */
static int
linux_qxfer_spu (const char *annex, unsigned char *readbuf,
                 unsigned const char *writebuf,
                 CORE_ADDR offset, int len)
{
  long pid = lwpid_of (current_thread);
  char buf[128];
  int fd = 0;
  int ret = 0;

  if (!writebuf && !readbuf)
    return -1;

  if (!*annex)
    {
      if (!readbuf)
        return -1;
      else
        return spu_enumerate_spu_ids (pid, readbuf, offset, len);
    }

  sprintf (buf, "/proc/%ld/fd/%s", pid, annex);
  fd = open (buf, writebuf? O_WRONLY : O_RDONLY);
  if (fd <= 0)
    return -1;

  if (offset != 0
      && lseek (fd, (off_t) offset, SEEK_SET) != (off_t) offset)
    {
      close (fd);
      return 0;
    }

  if (writebuf)
    ret = write (fd, writebuf, (size_t) len);
  else
    ret = read (fd, readbuf, (size_t) len);

  close (fd);
  return ret;
}

#if defined PT_GETDSBT || defined PTRACE_GETFDPIC
struct target_loadseg
{
  /* Core address to which the segment is mapped.  */
  Elf32_Addr addr;
  /* VMA recorded in the program header.  */
  Elf32_Addr p_vaddr;
  /* Size of this segment in memory.  */
  Elf32_Word p_memsz;
};

# if defined PT_GETDSBT
struct target_loadmap
{
  /* Protocol version number, must be zero.  */
  Elf32_Word version;
  /* Pointer to the DSBT table, its size, and the DSBT index.  */
  unsigned *dsbt_table;
  unsigned dsbt_size, dsbt_index;
  /* Number of segments in this map.  */
  Elf32_Word nsegs;
  /* The actual memory map.  */
  struct target_loadseg segs[/*nsegs*/];
};
#  define LINUX_LOADMAP         PT_GETDSBT
#  define LINUX_LOADMAP_EXEC    PTRACE_GETDSBT_EXEC
#  define LINUX_LOADMAP_INTERP  PTRACE_GETDSBT_INTERP
# else
struct target_loadmap
{
  /* Protocol version number, must be zero.  */
  Elf32_Half version;
  /* Number of segments in this map.  */
  Elf32_Half nsegs;
  /* The actual memory map.  */
  struct target_loadseg segs[/*nsegs*/];
};
#  define LINUX_LOADMAP         PTRACE_GETFDPIC
#  define LINUX_LOADMAP_EXEC    PTRACE_GETFDPIC_EXEC
#  define LINUX_LOADMAP_INTERP  PTRACE_GETFDPIC_INTERP
# endif

static int
linux_read_loadmap (const char *annex, CORE_ADDR offset,
                    unsigned char *myaddr, unsigned int len)
{
  int pid = lwpid_of (current_thread);
  int addr = -1;
  struct target_loadmap *data = NULL;
  unsigned int actual_length, copy_length;

  if (strcmp (annex, "exec") == 0)
    addr = (int) LINUX_LOADMAP_EXEC;
  else if (strcmp (annex, "interp") == 0)
    addr = (int) LINUX_LOADMAP_INTERP;
  else
    return -1;

  if (ptrace (LINUX_LOADMAP, pid, addr, &data) != 0)
    return -1;

  if (data == NULL)
    return -1;

  actual_length = sizeof (struct target_loadmap)
    + sizeof (struct target_loadseg) * data->nsegs;

  if (offset < 0 || offset > actual_length)
    return -1;

  copy_length = actual_length - offset < len ? actual_length - offset : len;
  memcpy (myaddr, (char *) data + offset, copy_length);
  return copy_length;
}
#else
# define linux_read_loadmap NULL
#endif /* defined PT_GETDSBT || defined PTRACE_GETFDPIC */

static void
linux_process_qsupported (const char *query)
{
  if (the_low_target.process_qsupported != NULL)
    the_low_target.process_qsupported (query);
}

static int
linux_supports_tracepoints (void)
{
  if (*the_low_target.supports_tracepoints == NULL)
    return 0;

  return (*the_low_target.supports_tracepoints) ();
}

static CORE_ADDR
linux_read_pc (struct regcache *regcache)
{
  if (the_low_target.get_pc == NULL)
    return 0;

  return (*the_low_target.get_pc) (regcache);
}

static void
linux_write_pc (struct regcache *regcache, CORE_ADDR pc)
{
  gdb_assert (the_low_target.set_pc != NULL);

  (*the_low_target.set_pc) (regcache, pc);
}

static int
linux_thread_stopped (struct thread_info *thread)
{
  return get_thread_lwp (thread)->stopped;
}

/* This exposes stop-all-threads functionality to other modules.  */

static void
linux_pause_all (int freeze)
{
  stop_all_lwps (freeze, NULL);
}

/* This exposes unstop-all-threads functionality to other gdbserver
   modules.  */

static void
linux_unpause_all (int unfreeze)
{
  unstop_all_lwps (unfreeze, NULL);
}

static int
linux_prepare_to_access_memory (void)
{
  /* Neither ptrace nor /proc/PID/mem allow accessing memory through a
     running LWP.  */
  if (non_stop)
    linux_pause_all (1);
  return 0;
}

static void
linux_done_accessing_memory (void)
{
  /* Neither ptrace nor /proc/PID/mem allow accessing memory through a
     running LWP.  */
  if (non_stop)
    linux_unpause_all (1);
}

static int
linux_install_fast_tracepoint_jump_pad (CORE_ADDR tpoint, CORE_ADDR tpaddr,
                                        CORE_ADDR collector,
                                        CORE_ADDR lockaddr,
                                        ULONGEST orig_size,
                                        CORE_ADDR *jump_entry,
                                        CORE_ADDR *trampoline,
                                        ULONGEST *trampoline_size,
                                        unsigned char *jjump_pad_insn,
                                        ULONGEST *jjump_pad_insn_size,
                                        CORE_ADDR *adjusted_insn_addr,
                                        CORE_ADDR *adjusted_insn_addr_end,
                                        char *err)
{
  return (*the_low_target.install_fast_tracepoint_jump_pad)
    (tpoint, tpaddr, collector, lockaddr, orig_size,
     jump_entry, trampoline, trampoline_size,
     jjump_pad_insn, jjump_pad_insn_size,
     adjusted_insn_addr, adjusted_insn_addr_end,
     err);
}

static struct emit_ops *
linux_emit_ops (void)
{
  if (the_low_target.emit_ops != NULL)
    return (*the_low_target.emit_ops) ();
  else
    return NULL;
}

static int
linux_get_min_fast_tracepoint_insn_len (void)
{
  return (*the_low_target.get_min_fast_tracepoint_insn_len) ();
}

/* Extract &phdr and num_phdr in the inferior.  Return 0 on success.  */

static int
get_phdr_phnum_from_proc_auxv (const int pid, const int is_elf64,
                               CORE_ADDR *phdr_memaddr, int *num_phdr)
{
  char filename[PATH_MAX];
  int fd;
  const int auxv_size = is_elf64
    ? sizeof (Elf64_auxv_t) : sizeof (Elf32_auxv_t);
  char buf[sizeof (Elf64_auxv_t)];  /* The larger of the two.  */

  xsnprintf (filename, sizeof filename, "/proc/%d/auxv", pid);

  fd = open (filename, O_RDONLY);
  if (fd < 0)
    return 1;

  *phdr_memaddr = 0;
  *num_phdr = 0;
  while (read (fd, buf, auxv_size) == auxv_size
         && (*phdr_memaddr == 0 || *num_phdr == 0))
    {
      if (is_elf64)
        {
          Elf64_auxv_t *const aux = (Elf64_auxv_t *) buf;

          switch (aux->a_type)
            {
            case AT_PHDR:
              *phdr_memaddr = aux->a_un.a_val;
              break;
            case AT_PHNUM:
              *num_phdr = aux->a_un.a_val;
              break;
            }
        }
      else
        {
          Elf32_auxv_t *const aux = (Elf32_auxv_t *) buf;

          switch (aux->a_type)
            {
            case AT_PHDR:
              *phdr_memaddr = aux->a_un.a_val;
              break;
            case AT_PHNUM:
              *num_phdr = aux->a_un.a_val;
              break;
            }
        }
    }

  close (fd);

  if (*phdr_memaddr == 0 || *num_phdr == 0)
    {
      warning ("Unexpected missing AT_PHDR and/or AT_PHNUM: "
               "phdr_memaddr = %ld, phdr_num = %d",
               (long) *phdr_memaddr, *num_phdr);
      return 2;
    }

  return 0;
}

/* Return &_DYNAMIC (via PT_DYNAMIC) in the inferior, or 0 if not present.  */

static CORE_ADDR
get_dynamic (const int pid, const int is_elf64)
{
  CORE_ADDR phdr_memaddr, relocation;
  int num_phdr, i;
  unsigned char *phdr_buf;
  const int phdr_size = is_elf64 ? sizeof (Elf64_Phdr) : sizeof (Elf32_Phdr);

  if (get_phdr_phnum_from_proc_auxv (pid, is_elf64, &phdr_memaddr, &num_phdr))
    return 0;

  gdb_assert (num_phdr < 100);  /* Basic sanity check.  */
  phdr_buf = alloca (num_phdr * phdr_size);

  if (linux_read_memory (phdr_memaddr, phdr_buf, num_phdr * phdr_size))
    return 0;

  /* Compute relocation: it is expected to be 0 for "regular" executables,
     non-zero for PIE ones.  */
  relocation = -1;
  for (i = 0; relocation == -1 && i < num_phdr; i++)
    if (is_elf64)
      {
        Elf64_Phdr *const p = (Elf64_Phdr *) (phdr_buf + i * phdr_size);

        if (p->p_type == PT_PHDR)
          relocation = phdr_memaddr - p->p_vaddr;
      }
    else
      {
        Elf32_Phdr *const p = (Elf32_Phdr *) (phdr_buf + i * phdr_size);

        if (p->p_type == PT_PHDR)
          relocation = phdr_memaddr - p->p_vaddr;
      }

  if (relocation == -1)
    {
      /* PT_PHDR is optional, but necessary for PIE in general.  Fortunately
         any real world executables, including PIE executables, have always
         PT_PHDR present.  PT_PHDR is not present in some shared libraries or
         in fpc (Free Pascal 2.4) binaries but neither of those have a need for
         or present DT_DEBUG anyway (fpc binaries are statically linked).

         Therefore if there exists DT_DEBUG there is always also PT_PHDR.

         GDB could find RELOCATION also from AT_ENTRY - e_entry.  */

      return 0;
    }

  for (i = 0; i < num_phdr; i++)
    {
      if (is_elf64)
        {
          Elf64_Phdr *const p = (Elf64_Phdr *) (phdr_buf + i * phdr_size);

          if (p->p_type == PT_DYNAMIC)
            return p->p_vaddr + relocation;
        }
      else
        {
          Elf32_Phdr *const p = (Elf32_Phdr *) (phdr_buf + i * phdr_size);

          if (p->p_type == PT_DYNAMIC)
            return p->p_vaddr + relocation;
        }
    }

  return 0;
}

/* Return &_r_debug in the inferior, or -1 if not present.  Return value
   can be 0 if the inferior does not yet have the library list initialized.
   We look for DT_MIPS_RLD_MAP first.  MIPS executables use this instead of
   DT_DEBUG, although they sometimes contain an unused DT_DEBUG entry too.  */

static CORE_ADDR
get_r_debug (const int pid, const int is_elf64)
{
  CORE_ADDR dynamic_memaddr;
  const int dyn_size = is_elf64 ? sizeof (Elf64_Dyn) : sizeof (Elf32_Dyn);
  unsigned char buf[sizeof (Elf64_Dyn)];  /* The larger of the two.  */
  CORE_ADDR map = -1;

  dynamic_memaddr = get_dynamic (pid, is_elf64);
  if (dynamic_memaddr == 0)
    return map;

  while (linux_read_memory (dynamic_memaddr, buf, dyn_size) == 0)
    {
      if (is_elf64)
        {
          Elf64_Dyn *const dyn = (Elf64_Dyn *) buf;
#ifdef DT_MIPS_RLD_MAP
          union
            {
              Elf64_Xword map;
              unsigned char buf[sizeof (Elf64_Xword)];
            }
          rld_map;

          if (dyn->d_tag == DT_MIPS_RLD_MAP)
            {
              if (linux_read_memory (dyn->d_un.d_val,
                                     rld_map.buf, sizeof (rld_map.buf)) == 0)
                return rld_map.map;
              else
                break;
            }
#endif  /* DT_MIPS_RLD_MAP */

          if (dyn->d_tag == DT_DEBUG && map == -1)
            map = dyn->d_un.d_val;

          if (dyn->d_tag == DT_NULL)
            break;
        }
      else
        {
          Elf32_Dyn *const dyn = (Elf32_Dyn *) buf;
#ifdef DT_MIPS_RLD_MAP
          union
            {
              Elf32_Word map;
              unsigned char buf[sizeof (Elf32_Word)];
            }
          rld_map;

          if (dyn->d_tag == DT_MIPS_RLD_MAP)
            {
              if (linux_read_memory (dyn->d_un.d_val,
                                     rld_map.buf, sizeof (rld_map.buf)) == 0)
                return rld_map.map;
              else
                break;
            }
#endif  /* DT_MIPS_RLD_MAP */

          if (dyn->d_tag == DT_DEBUG && map == -1)
            map = dyn->d_un.d_val;

          if (dyn->d_tag == DT_NULL)
            break;
        }

      dynamic_memaddr += dyn_size;
    }

  return map;
}

/* Read one pointer from MEMADDR in the inferior.  */

static int
read_one_ptr (CORE_ADDR memaddr, CORE_ADDR *ptr, int ptr_size)
{
  int ret;

  /* Go through a union so this works on either big or little endian
     hosts, when the inferior's pointer size is smaller than the size
     of CORE_ADDR.  It is assumed the inferior's endianness is the
     same of the superior's.  */
  union
  {
    CORE_ADDR core_addr;
    unsigned int ui;
    unsigned char uc;
  } addr;

  ret = linux_read_memory (memaddr, &addr.uc, ptr_size);
  if (ret == 0)
    {
      if (ptr_size == sizeof (CORE_ADDR))
        *ptr = addr.core_addr;
      else if (ptr_size == sizeof (unsigned int))
        *ptr = addr.ui;
      else
        gdb_assert_not_reached ("unhandled pointer size");
    }
  return ret;
}

struct link_map_offsets
  {
    /* Offset and size of r_debug.r_version.  */
    int r_version_offset;

    /* Offset and size of r_debug.r_map.  */
    int r_map_offset;

    /* Offset to l_addr field in struct link_map.  */
    int l_addr_offset;

    /* Offset to l_name field in struct link_map.  */
    int l_name_offset;

    /* Offset to l_ld field in struct link_map.  */
    int l_ld_offset;

    /* Offset to l_next field in struct link_map.  */
    int l_next_offset;

    /* Offset to l_prev field in struct link_map.  */
    int l_prev_offset;
  };

/* Construct qXfer:libraries-svr4:read reply.  */

static int
linux_qxfer_libraries_svr4 (const char *annex, unsigned char *readbuf,
                            unsigned const char *writebuf,
                            CORE_ADDR offset, int len)
{
  char *document;
  unsigned document_len;
  struct process_info_private *const priv = current_process ()->private;
  char filename[PATH_MAX];
  int pid, is_elf64;

  static const struct link_map_offsets lmo_32bit_offsets =
    {
      0,     /* r_version offset. */
      4,     /* r_debug.r_map offset.  */
      0,     /* l_addr offset in link_map.  */
      4,     /* l_name offset in link_map.  */
      8,     /* l_ld offset in link_map.  */
      12,    /* l_next offset in link_map.  */
      16     /* l_prev offset in link_map.  */
    };

  static const struct link_map_offsets lmo_64bit_offsets =
    {
      0,     /* r_version offset. */
      8,     /* r_debug.r_map offset.  */
      0,     /* l_addr offset in link_map.  */
      8,     /* l_name offset in link_map.  */
      16,    /* l_ld offset in link_map.  */
      24,    /* l_next offset in link_map.  */
      32     /* l_prev offset in link_map.  */
    };
  const struct link_map_offsets *lmo;
  unsigned int machine;
  int ptr_size;
  CORE_ADDR lm_addr = 0, lm_prev = 0;
  int allocated = 1024;
  char *p;
  CORE_ADDR l_name, l_addr, l_ld, l_next, l_prev;
  int header_done = 0;

  if (writebuf != NULL)
    return -2;
  if (readbuf == NULL)
    return -1;

  pid = lwpid_of (current_thread);
  xsnprintf (filename, sizeof filename, "/proc/%d/exe", pid);
  is_elf64 = elf_64_file_p (filename, &machine);
  lmo = is_elf64 ? &lmo_64bit_offsets : &lmo_32bit_offsets;
  ptr_size = is_elf64 ? 8 : 4;

  while (annex[0] != '\0')
    {
      const char *sep;
      CORE_ADDR *addrp;
      int len;

      sep = strchr (annex, '=');
      if (sep == NULL)
        break;

      len = sep - annex;
      if (len == 5 && strncmp (annex, "start", 5) == 0)
        addrp = &lm_addr;
      else if (len == 4 && strncmp (annex, "prev", 4) == 0)
        addrp = &lm_prev;
      else
        {
          annex = strchr (sep, ';');
          if (annex == NULL)
            break;
          annex++;
          continue;
        }

      annex = decode_address_to_semicolon (addrp, sep + 1);
    }

  if (lm_addr == 0)
    {
      int r_version = 0;

      if (priv->r_debug == 0)
        priv->r_debug = get_r_debug (pid, is_elf64);

      /* We failed to find DT_DEBUG.  Such situation will not change
         for this inferior - do not retry it.  Report it to GDB as
         E01, see for the reasons at the GDB solib-svr4.c side.  */
      if (priv->r_debug == (CORE_ADDR) -1)
        return -1;

      if (priv->r_debug != 0)
        {
          if (linux_read_memory (priv->r_debug + lmo->r_version_offset,
                                 (unsigned char *) &r_version,
                                 sizeof (r_version)) != 0
              || r_version != 1)
            {
              warning ("unexpected r_debug version %d", r_version);
            }
          else if (read_one_ptr (priv->r_debug + lmo->r_map_offset,
                                 &lm_addr, ptr_size) != 0)
            {
              warning ("unable to read r_map from 0x%lx",
                       (long) priv->r_debug + lmo->r_map_offset);
            }
        }
    }

  document = xmalloc (allocated);
  strcpy (document, "<library-list-svr4 version=\"1.0\"");
  p = document + strlen (document);

  while (lm_addr
         && read_one_ptr (lm_addr + lmo->l_name_offset,
                          &l_name, ptr_size) == 0
         && read_one_ptr (lm_addr + lmo->l_addr_offset,
                          &l_addr, ptr_size) == 0
         && read_one_ptr (lm_addr + lmo->l_ld_offset,
                          &l_ld, ptr_size) == 0
         && read_one_ptr (lm_addr + lmo->l_prev_offset,
                          &l_prev, ptr_size) == 0
         && read_one_ptr (lm_addr + lmo->l_next_offset,
                          &l_next, ptr_size) == 0)
    {
      unsigned char libname[PATH_MAX];

      if (lm_prev != l_prev)
        {
          warning ("Corrupted shared library list: 0x%lx != 0x%lx",
                   (long) lm_prev, (long) l_prev);
          break;
        }

      /* Ignore the first entry even if it has valid name as the first entry
         corresponds to the main executable.  The first entry should not be
         skipped if the dynamic loader was loaded late by a static executable
         (see solib-svr4.c parameter ignore_first).  But in such case the main
         executable does not have PT_DYNAMIC present and this function already
         exited above due to failed get_r_debug.  */
      if (lm_prev == 0)
        {
          sprintf (p, " main-lm=\"0x%lx\"", (unsigned long) lm_addr);
          p = p + strlen (p);
        }
      else
        {
          /* Not checking for error because reading may stop before
             we've got PATH_MAX worth of characters.  */
          libname[0] = '\0';
          linux_read_memory (l_name, libname, sizeof (libname) - 1);
          libname[sizeof (libname) - 1] = '\0';
          if (libname[0] != '\0')
            {
              /* 6x the size for xml_escape_text below.  */
              size_t len = 6 * strlen ((char *) libname);
              char *name;

              if (!header_done)
                {
                  /* Terminate `<library-list-svr4'.  */
                  *p++ = '>';
                  header_done = 1;
                }

              while (allocated < p - document + len + 200)
                {
                  /* Expand to guarantee sufficient storage.  */
                  uintptr_t document_len = p - document;

                  document = xrealloc (document, 2 * allocated);
                  allocated *= 2;
                  p = document + document_len;
                }

              name = xml_escape_text ((char *) libname);
              p += sprintf (p, "<library name=\"%s\" lm=\"0x%lx\" "
                            "l_addr=\"0x%lx\" l_ld=\"0x%lx\"/>",
                            name, (unsigned long) lm_addr,
                            (unsigned long) l_addr, (unsigned long) l_ld);
              free (name);
            }
        }

      lm_prev = lm_addr;
      lm_addr = l_next;
    }

  if (!header_done)
    {
      /* Empty list; terminate `<library-list-svr4'.  */
      strcpy (p, "/>");
    }
  else
    strcpy (p, "</library-list-svr4>");

  document_len = strlen (document);
  if (offset < document_len)
    document_len -= offset;
  else
    document_len = 0;
  if (len > document_len)
    len = document_len;

  memcpy (readbuf, document + offset, len);
  xfree (document);

  return len;
}

#ifdef HAVE_LINUX_BTRACE

/* See to_enable_btrace target method.  */

static struct btrace_target_info *
linux_low_enable_btrace (ptid_t ptid)
{
  struct btrace_target_info *tinfo;

  tinfo = linux_enable_btrace (ptid);

  if (tinfo != NULL)
    {
      struct thread_info *thread = find_thread_ptid (ptid);
      struct regcache *regcache = get_thread_regcache (thread, 0);

      tinfo->ptr_bits = register_size (regcache->tdesc, 0) * 8;
    }

  return tinfo;
}

/* See to_disable_btrace target method.  */

static int
linux_low_disable_btrace (struct btrace_target_info *tinfo)
{
  enum btrace_error err;

  err = linux_disable_btrace (tinfo);
  return (err == BTRACE_ERR_NONE ? 0 : -1);
}

/* See to_read_btrace target method.  */

static int
linux_low_read_btrace (struct btrace_target_info *tinfo, struct buffer *buffer,
                       int type)
{
  VEC (btrace_block_s) *btrace;
  struct btrace_block *block;
  enum btrace_error err;
  int i;

  btrace = NULL;
  err = linux_read_btrace (&btrace, tinfo, type);
  if (err != BTRACE_ERR_NONE)
    {
      if (err == BTRACE_ERR_OVERFLOW)
        buffer_grow_str0 (buffer, "E.Overflow.");
      else
        buffer_grow_str0 (buffer, "E.Generic Error.");

      return -1;
    }

  buffer_grow_str (buffer, "<!DOCTYPE btrace SYSTEM \"btrace.dtd\">\n");
  buffer_grow_str (buffer, "<btrace version=\"1.0\">\n");

  for (i = 0; VEC_iterate (btrace_block_s, btrace, i, block); i++)
    buffer_xml_printf (buffer, "<block begin=\"0x%s\" end=\"0x%s\"/>\n",
                       paddress (block->begin), paddress (block->end));

  buffer_grow_str0 (buffer, "</btrace>\n");

  VEC_free (btrace_block_s, btrace);

  return 0;
}
#endif /* HAVE_LINUX_BTRACE */

static struct target_ops linux_target_ops = {
  linux_create_inferior,
  linux_attach,
  linux_kill,
  linux_detach,
  linux_mourn,
  linux_join,
  linux_thread_alive,
  linux_resume,
  linux_wait,
  linux_fetch_registers,
  linux_store_registers,
  linux_prepare_to_access_memory,
  linux_done_accessing_memory,
  linux_read_memory,
  linux_write_memory,
  linux_look_up_symbols,
  linux_request_interrupt,
  linux_read_auxv,
  linux_supports_z_point_type,
  linux_insert_point,
  linux_remove_point,
  linux_stopped_by_watchpoint,
  linux_stopped_data_address,
#if defined(__UCLIBC__) && defined(HAS_NOMMU)         \
    && defined(PT_TEXT_ADDR) && defined(PT_DATA_ADDR) \
    && defined(PT_TEXT_END_ADDR)
  linux_read_offsets,
#else
  NULL,
#endif
#ifdef USE_THREAD_DB
  thread_db_get_tls_address,
#else
  NULL,
#endif
  linux_qxfer_spu,
  hostio_last_error_from_errno,
  linux_qxfer_osdata,
  linux_xfer_siginfo,
  linux_supports_non_stop,
  linux_async,
  linux_start_non_stop,
  linux_supports_multi_process,
#ifdef USE_THREAD_DB
  thread_db_handle_monitor_command,
#else
  NULL,
#endif
  linux_common_core_of_thread,
  linux_read_loadmap,
  linux_process_qsupported,
  linux_supports_tracepoints,
  linux_read_pc,
  linux_write_pc,
  linux_thread_stopped,
  NULL,
  linux_pause_all,
  linux_unpause_all,
  linux_stabilize_threads,
  linux_install_fast_tracepoint_jump_pad,
  linux_emit_ops,
  linux_supports_disable_randomization,
  linux_get_min_fast_tracepoint_insn_len,
  linux_qxfer_libraries_svr4,
  linux_supports_agent,
#ifdef HAVE_LINUX_BTRACE
  linux_supports_btrace,
  linux_low_enable_btrace,
  linux_low_disable_btrace,
  linux_low_read_btrace,
#else
  NULL,
  NULL,
  NULL,
  NULL,
#endif
  linux_supports_range_stepping,
};

static void
linux_init_signals ()
{
  /* FIXME drow/2002-06-09: As above, we should check with LinuxThreads
     to find what the cancel signal actually is.  */
#ifndef __ANDROID__ /* Bionic doesn't use SIGRTMIN the way glibc does.  */
  signal (__SIGRTMIN+1, SIG_IGN);
#endif
}

#ifdef HAVE_LINUX_REGSETS
void
initialize_regsets_info (struct regsets_info *info)
{
  for (info->num_regsets = 0;
       info->regsets[info->num_regsets].size >= 0;
       info->num_regsets++)
    ;
}
#endif

void
initialize_low (void)
{
  struct sigaction sigchld_action;
  memset (&sigchld_action, 0, sizeof (sigchld_action));
  set_target_ops (&linux_target_ops);
  set_breakpoint_data (the_low_target.breakpoint,
                       the_low_target.breakpoint_len);
  linux_init_signals ();
  linux_ptrace_init_warnings ();

  sigchld_action.sa_handler = sigchld_handler;
  sigemptyset (&sigchld_action.sa_mask);
  sigchld_action.sa_flags = SA_RESTART;
  sigaction (SIGCHLD, &sigchld_action, NULL);

  initialize_low_arch ();
}