binutils-gdb/gdb/nat/fork-inferior.c

/* Fork a Unix child process, and set up to debug it, for GDB and GDBserver.

   Copyright (C) 1990-2021 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 "gdbsupport/common-defs.h"
#include "fork-inferior.h"
#include "target/waitstatus.h"
#include "gdbsupport/filestuff.h"
#include "target/target.h"
#include "gdbsupport/common-inferior.h"
#include "gdbsupport/common-gdbthread.h"
#include "gdbsupport/pathstuff.h"
#include "gdbsupport/signals-state-save-restore.h"
#include "gdbsupport/gdb_tilde_expand.h"
#include "gdbsupport/scoped_ignore_sigttou.h"
#include "gdbsupport/managed-tty.h"
#include <vector>
#include <sys/types.h>
#include <sys/wait.h>

extern char **environ;

/* Build the argument vector for execv(3).  */

class execv_argv
{
public:
  /* EXEC_FILE is the file to run.  ALLARGS is a string containing the
     arguments to the program.  If starting with a shell, SHELL_FILE
     is the shell to run.  Otherwise, SHELL_FILE is NULL.  */
  execv_argv (const char *exec_file, const std::string &allargs,
	      const char *shell_file);

  /* Return a pointer to the built argv, in the type expected by
     execv.  The result is (only) valid for as long as this execv_argv
     object is live.  We return a "char **" because that's the type
     that the execv functions expect.  Note that it is guaranteed that
     the execv functions do not modify the argv[] array nor the
     strings to which the array point.  */
  char **argv ()
  {
    return const_cast<char **> (&m_argv[0]);
  }

private:
  DISABLE_COPY_AND_ASSIGN (execv_argv);

  /* Helper methods for constructing the argument vector.  */

  /* Used when building an argv for a straight execv call, without
     going via the shell.  */
  void init_for_no_shell (const char *exec_file,
			  const std::string &allargs);

  /* Used when building an argv for execing a shell that execs the
     child program.  */
  void init_for_shell (const char *exec_file,
		       const std::string &allargs,
		       const char *shell_file);

  /* The argument vector built.  Holds non-owning pointers.  Elements
     either point to the strings passed to the execv_argv ctor, or
     inside M_STORAGE.  */
  std::vector<const char *> m_argv;

  /* Storage.  In the no-shell case, this contains a copy of the
     arguments passed to the ctor, split by '\0'.  In the shell case,
     this contains the quoted shell command.  I.e., SHELL_COMMAND in
     {"$SHELL" "-c", SHELL_COMMAND, NULL}.  */
  std::string m_storage;
};

#if GDB_MANAGED_TERMINALS

/* SIGHUP handler for the session leader processes.  GDB sends this
   explicitly, though it'll also be called if GDB crashes and the
   terminal is abruptly closed.  */

static void
session_leader_hup (int sig)
{
  scoped_ignore_sigttou ignore_sigttou;

  /* We put the inferior (a child of the session leader) in the
     foreground, so by default, on detach, if we did nothing else, the
     inferior would get a SIGHUP when the terminal is closed by GDB.
     That SIGHUP would likely kill the inferior.  To avoid it, we put
     the session leader in the foreground before the terminal is
     closed.  Only processes in the foreground process group get the
     automatic SIGHUP, so the detached process doesn't get it.  */
  int res = tcsetpgrp (0, getpid ());
  if (res == -1)
    trace_start_error (_("tcsetpgrp failed in session leader\n"));

  _exit (0);
}

#endif

/* Create argument vector for straight call to execvp.  Breaks up
   ALLARGS into an argument vector suitable for passing to execvp and
   stores it in M_ARGV.  E.g., on "run a b c d" this routine would get
   as input the string "a b c d", and as output it would fill in
   M_ARGV with the four arguments "a", "b", "c", "d".  Each argument
   in M_ARGV points to a substring of a copy of ALLARGS stored in
   M_STORAGE.  */

void
execv_argv::init_for_no_shell (const char *exec_file,
			       const std::string &allargs)
{

  /* Save/work with a copy stored in our storage.  The pointers pushed
     to M_ARGV point directly into M_STORAGE, which is modified in
     place with the necessary NULL terminators.  This avoids N heap
     allocations and string dups when 1 is sufficient.  */
  std::string &args_copy = m_storage = allargs;

  m_argv.push_back (exec_file);

  for (size_t cur_pos = 0; cur_pos < args_copy.size ();)
    {
      /* Skip whitespace-like chars.  */
      std::size_t pos = args_copy.find_first_not_of (" \t\n", cur_pos);

      if (pos != std::string::npos)
	cur_pos = pos;

      /* Find the position of the next separator.  */
      std::size_t next_sep = args_copy.find_first_of (" \t\n", cur_pos);

      if (next_sep == std::string::npos)
	{
	  /* No separator found, which means this is the last
	     argument.  */
	  next_sep = args_copy.size ();
	}
      else
	{
	  /* Replace the separator with a terminator.  */
	  args_copy[next_sep++] = '\0';
	}

      m_argv.push_back (&args_copy[cur_pos]);

      cur_pos = next_sep;
    }

  /* NULL-terminate the vector.  */
  m_argv.push_back (NULL);
}

/* When executing a command under the given shell, return true if the
   '!' character should be escaped when embedded in a quoted
   command-line argument.  */

static bool
escape_bang_in_quoted_argument (const char *shell_file)
{
  size_t shell_file_len = strlen (shell_file);

  /* Bang should be escaped only in C Shells.  For now, simply check
     that the shell name ends with 'csh', which covers at least csh
     and tcsh.  This should be good enough for now.  */

  if (shell_file_len < 3)
    return false;

  if (shell_file[shell_file_len - 3] == 'c'
      && shell_file[shell_file_len - 2] == 's'
      && shell_file[shell_file_len - 1] == 'h')
    return true;

  return false;
}

/* See declaration.  */

execv_argv::execv_argv (const char *exec_file,
			const std::string &allargs,
			const char *shell_file)
{
  if (shell_file == NULL)
    init_for_no_shell (exec_file, allargs);
  else
    init_for_shell (exec_file, allargs, shell_file);
}

/* See declaration.  */

void
execv_argv::init_for_shell (const char *exec_file,
			    const std::string &allargs,
			    const char *shell_file)
{
  const char *exec_wrapper = get_exec_wrapper ();

  /* We're going to call a shell.  */
  bool escape_bang = escape_bang_in_quoted_argument (shell_file);

  /* We need to build a new shell command string, and make argv point
     to it.  So build it in the storage.  */
  std::string &shell_command = m_storage;

  shell_command = "exec ";

  /* Add any exec wrapper.  That may be a program name with arguments,
     so the user must handle quoting.  */
  if (exec_wrapper != NULL)
    {
      shell_command += exec_wrapper;
      shell_command += ' ';
    }

  /* Now add exec_file, quoting as necessary.  */

  /* Quoting in this style is said to work with all shells.  But csh
     on IRIX 4.0.1 can't deal with it.  So we only quote it if we need
     to.  */
  bool need_to_quote;
  const char *p = exec_file;
  while (1)
    {
      switch (*p)
	{
	case '\'':
	case '!':
	case '"':
	case '(':
	case ')':
	case '$':
	case '&':
	case ';':
	case '<':
	case '>':
	case ' ':
	case '\n':
	case '\t':
	  need_to_quote = true;
	  goto end_scan;

	case '\0':
	  need_to_quote = false;
	  goto end_scan;

	default:
	  break;
	}
      ++p;
    }
 end_scan:
  if (need_to_quote)
    {
      shell_command += '\'';
      for (p = exec_file; *p != '\0'; ++p)
	{
	  if (*p == '\'')
	    shell_command += "'\\''";
	  else if (*p == '!' && escape_bang)
	    shell_command += "\\!";
	  else
	    shell_command += *p;
	}
      shell_command += '\'';
    }
  else
    shell_command += exec_file;

  shell_command += ' ' + allargs;

  /* If we decided above to start up with a shell, we exec the shell.
     "-c" says to interpret the next arg as a shell command to
     execute, and this command is "exec <target-program> <args>".  */
  m_argv.reserve (4);
  m_argv.push_back (shell_file);
  m_argv.push_back ("-c");
  m_argv.push_back (shell_command.c_str ());
  m_argv.push_back (NULL);
}

/* See nat/fork-inferior.h.  */

pid_t
fork_inferior (const char *exec_file_arg, const std::string &allargs,
	       char **env, void (*traceme_fun) (),
	       gdb::function_view<void (int)> init_trace_fun,
	       void (*pre_trace_fun) (),
	       const char *shell_file_arg,
	       void (*exec_fun)(const char *file, char * const *argv,
				char * const *env),
	       pid_t (*handle_session_leader_fork) (pid_t sl_pid))
{
  pid_t pid;
  /* Set debug_fork then attach to the child while it sleeps, to debug.  */
  int debug_fork = 0;
  const char *shell_file;
  const char *exec_file;
  char **save_our_env;
  int i;
  int save_errno;
  const char *inferior_cwd;
  std::string expanded_inferior_cwd;

  /* If no exec file handed to us, get it from the exec-file command
     -- with a good, common error message if none is specified.  */
  if (exec_file_arg == NULL)
    exec_file = get_exec_file (1);
  else
    exec_file = exec_file_arg;

  /* 'startup_with_shell' is declared in inferior.h and bound to the
     "set startup-with-shell" option.  If 0, we'll just do a
     fork/exec, no shell, so don't bother figuring out what shell.  */
  if (startup_with_shell)
    {
      shell_file = shell_file_arg;

      /* Figure out what shell to start up the user program under.  */
      if (shell_file == NULL)
	shell_file = get_shell ();

      gdb_assert (shell_file != NULL);
    }
  else
    shell_file = NULL;

  /* Build the argument vector.  */
  execv_argv child_argv (exec_file, allargs, shell_file);

  /* Retain a copy of our environment variables, since the child will
     replace the value of environ and if we're vforked, we have to
     restore it.  */
  save_our_env = environ;

  /* Perform any necessary actions regarding to TTY before the
     fork/vfork call.  */
  prefork_hook ();

  /* It is generally good practice to flush any possible pending stdio
     output prior to doing a fork, to avoid the possibility of both
     the parent and child flushing the same data after the fork.  */
  gdb_flush_out_err ();

  /* Check if the user wants to set a different working directory for
     the inferior.  */
  inferior_cwd = get_inferior_cwd ();

  if (inferior_cwd != NULL)
    {
      /* Expand before forking because between fork and exec, the child
	 process may only execute async-signal-safe operations.  */
      expanded_inferior_cwd = gdb_tilde_expand (inferior_cwd);
      inferior_cwd = expanded_inferior_cwd.c_str ();
    }

  /* If there's any initialization of the target layers that must
     happen to prepare to handle the child we're about fork, do it
     now...  */
  if (pre_trace_fun != NULL)
    (*pre_trace_fun) ();

  /* Create the child process.  Since the child process is going to
     exec(3) shortly afterwards, try to reduce the overhead by
     calling vfork(2).  However, if PRE_TRACE_FUN is non-null, it's
     likely that this optimization won't work since there's too much
     work to do between the vfork(2) and the exec(3).  This is known
     to be the case on ttrace(2)-based HP-UX, where some handshaking
     between parent and child needs to happen between fork(2) and
     exec(2).  However, since the parent is suspended in the vforked
     state, this doesn't work.  Also note that the vfork(2) call might
     actually be a call to fork(2) due to the fact that autoconf will
     ``#define vfork fork'' on certain platforms.  */
#if !(defined(__UCLIBC__) && defined(HAS_NOMMU))
  if (pre_trace_fun || debug_fork || child_has_managed_tty_hook ())
    pid = fork ();
  else
#endif
    pid = vfork ();

  if (pid < 0)
    perror_with_name (("vfork"));

  if (pid == 0)
    {
      /* Close all file descriptors except those that gdb inherited
	 (usually 0/1/2), so they don't leak to the inferior.  Note
	 that this closes the file descriptors of all secondary
	 UIs.  */
      close_most_fds ();

      /* Change to the requested working directory if the user
	 requested it.  */
      if (inferior_cwd != NULL)
	{
	  if (chdir (inferior_cwd) < 0)
	    trace_start_error_with_name (inferior_cwd);
	}

      if (debug_fork)
	sleep (debug_fork);

      /* Execute any necessary post-fork actions before we exec.  */
      postfork_child_hook ();

      /* Changing the signal handlers for the inferior after
	 a vfork can also change them for the superior, so we don't mess
	 with signals here.  See comments in
	 initialize_signals for how we get the right signal handlers
	 for the inferior.  */

      /* "Trace me, Dr. Memory!"  */
      (*traceme_fun) ();

      /* The call above set this process (the "child") as debuggable
	by the original gdb process (the "parent").  Since processes
	(unlike people) can have only one parent, if you are debugging
	gdb itself (and your debugger is thus _already_ the
	controller/parent for this child), code from here on out is
	undebuggable.  Indeed, you probably got an error message
	saying "not parent".  Sorry; you'll have to use print
	statements!  */

      restore_original_signals_state ();

      /* Fork again so that the resulting inferior process is not the
	 session leader.  This makes it possible for the inferior to
	 exit without killing its own children.  If we instead let the
	 inferior process be the session leader, when it exits, it'd
	 cause a SIGHUP to be sent to all processes in its session
	 (i.e., it's children).  The code is disabled on no-MMU
	 machines because those can't do fork, only vfork.  In theory
	 we could make this work with vfork by making the session
	 leader process exec a helper process, probably gdb itself in
	 a special mode (e.g., something like
	   exec /proc/self/exe --session-leader-for PID
      */
#if GDB_MANAGED_TERMINALS
      if (child_has_managed_tty_hook ())
	{
	  /* Fork again, to make sure the inferior is not the new
	     session's leader.  */
	  pid_t child2 = fork ();
	  if (child2 != 0)
	    {
	      /* This is the parent / session leader process.  It just
		 stays around until GDB closes the terminal.  */

	      /* Gracefully handle SIGHUP.  */
	      signal (SIGHUP, session_leader_hup);

	      managed_tty_debug_printf
		(_("session-leader (sid=%d): waiting for child pid=%d exit\n"),
		 (int) getpid (), (int) child2);

	      /* Reap the child/inferior exit status.  */
	      int status;
	      int res = waitpid (child2, &status, 0);

	      managed_tty_debug_printf (_("session-leader (sid=%d): "
					  "wait for child pid=%d returned: "
					  "res=%d, waitstatus=0x%x\n"),
					(int) getpid (), child2, res, status);

	      if (res == -1)
		warning (_("session-leader (sid=%d): unexpected waitstatus "
			   "reaping child pid=%d: "
			   "res=-1, errno=%d (%s)"),
			 (int) getpid (), child2, errno, safe_strerror (errno));
	      else if (res != child2)
		warning (_("session-leader (sid=%d): unexpected waitstatus "
			   "reaping child pid=%d: "
			   "res=%d, status=0x%x"),
			 (int) getpid (), child2, res, status);

	      /* Don't exit yet.  While our direct child is gone,
		 there may still be grandchildren attached to our
		 session.  We'll exit when our parent (GDB) closes the
		 pty, killing us with SIGHUP.  */
	      while (1)
		pause ();
	    }
	  else
	    {
	      /* This is the child / final inferior process.  */

	      int res;

	      /* Run the inferior in its own process group, and make
		 it the session's foreground pgrp.  */

	      res = gdb_setpgid ();
	      if (res == -1)
		trace_start_error (_("setpgid failed in grandchild"));

	      scoped_ignore_sigttou ignore_sigttou;

	      res = tcsetpgrp (0, getpid ());
	      if (res == -1)
		trace_start_error (_("tcsetpgrp failed in grandchild\n"));
	    }
	}
#endif /* GDB_MANAGED_TERMINALS */

      /* There is no execlpe call, so we have to set the environment
	 for our child in the global variable.  If we've vforked, this
	 clobbers the parent, but environ is restored a few lines down
	 in the parent.  By the way, yes we do need to look down the
	 path to find $SHELL.  Rich Pixley says so, and I agree.  */
      environ = env;

      char **argv = child_argv.argv ();

      if (exec_fun != NULL)
	(*exec_fun) (argv[0], &argv[0], env);
      else
	execvp (argv[0], &argv[0]);

      /* If we get here, it's an error.  */
      save_errno = errno;
      warning ("Cannot exec %s", argv[0]);

      for (i = 1; argv[i] != NULL; i++)
	warning (" %s", argv[i]);

      warning ("Error: %s", safe_strerror (save_errno));

      _exit (0177);
    }

  /* Restore our environment in case a vforked child clob'd it.  */
  environ = save_our_env;

  if (child_has_managed_tty_hook () && handle_session_leader_fork != nullptr)
    pid = handle_session_leader_fork (pid);

  postfork_hook (pid);

  /* Now that we have a child process, make it our target, and
     initialize anything target-vector-specific that needs
     initializing.  */
  if (init_trace_fun)
    init_trace_fun (pid);

  /* We are now in the child process of interest, having exec'd the
     correct program, and are poised at the first instruction of the
     new program.  */
  return pid;
}

/* See nat/fork-inferior.h.  */

ptid_t
startup_inferior (process_stratum_target *proc_target, pid_t pid, int ntraps,
		  struct target_waitstatus *last_waitstatus,
		  ptid_t *last_ptid)
{
  int pending_execs = ntraps;
  int terminal_initted = 0;
  ptid_t resume_ptid;

  if (startup_with_shell)
    {
      /* One trap extra for exec'ing the shell.  */
      pending_execs++;
    }

  if (target_supports_multi_process ())
    resume_ptid = ptid_t (pid);
  else
    resume_ptid = minus_one_ptid;

  /* The process was started by the fork that created it, but it will
     have stopped one instruction after execing the shell.  Here we
     must get it up to actual execution of the real program.  */
  if (get_exec_wrapper () != NULL)
    pending_execs++;

  while (1)
    {
      enum gdb_signal resume_signal = GDB_SIGNAL_0;
      ptid_t event_ptid;

      struct target_waitstatus ws;
      memset (&ws, 0, sizeof (ws));
      event_ptid = target_wait (resume_ptid, &ws, 0);

      if (last_waitstatus != NULL)
	*last_waitstatus = ws;
      if (last_ptid != NULL)
	*last_ptid = event_ptid;

      if (ws.kind == TARGET_WAITKIND_IGNORE)
	/* The inferior didn't really stop, keep waiting.  */
	continue;

      switch (ws.kind)
	{
	  case TARGET_WAITKIND_SPURIOUS:
	  case TARGET_WAITKIND_LOADED:
	  case TARGET_WAITKIND_FORKED:
	  case TARGET_WAITKIND_VFORKED:
	  case TARGET_WAITKIND_SYSCALL_ENTRY:
	  case TARGET_WAITKIND_SYSCALL_RETURN:
	    /* Ignore gracefully during startup of the inferior.  */
	    switch_to_thread (proc_target, event_ptid);
	    break;

	  case TARGET_WAITKIND_SIGNALLED:
	    target_terminal::ours ();
	    target_mourn_inferior (event_ptid);
	    error (_("During startup program terminated with signal %s, %s."),
		   gdb_signal_to_name (ws.value.sig),
		   gdb_signal_to_string (ws.value.sig));
	    return resume_ptid;

	  case TARGET_WAITKIND_EXITED:
	    target_terminal::ours ();
	    target_mourn_inferior (event_ptid);
	    if (ws.value.integer)
	      error (_("During startup program exited with code %d."),
		     ws.value.integer);
	    else
	      error (_("During startup program exited normally."));
	    return resume_ptid;

	  case TARGET_WAITKIND_EXECD:
	    /* Handle EXEC signals as if they were SIGTRAP signals.  */
	    /* Free the exec'ed pathname, but only if this isn't the
	       waitstatus we are returning to the caller.  */
	    if (pending_execs != 1)
	      xfree (ws.value.execd_pathname);
	    resume_signal = GDB_SIGNAL_TRAP;
	    switch_to_thread (proc_target, event_ptid);
	    break;

	  case TARGET_WAITKIND_STOPPED:
	    resume_signal = ws.value.sig;
	    switch_to_thread (proc_target, event_ptid);
	    break;
	}

      if (resume_signal != GDB_SIGNAL_TRAP)
	{
	  /* Let shell child handle its own signals in its own way.  */
	  target_continue (resume_ptid, resume_signal);
	}
      else
	{
	  /* We handle SIGTRAP, however; it means child did an exec.  */
	  if (!terminal_initted)
	    {
	      /* Now that the child has exec'd we know it has already
		 set its process group.  On POSIX systems, tcsetpgrp
		 will fail with EPERM if we try it before the child's
		 setpgid.  */

	      /* Set up the "saved terminal modes" of the inferior
		 based on what modes we are starting it with.  */
	      target_terminal::init ();

	      /* Install inferior's terminal modes.  */
	      target_terminal::inferior ();

	      terminal_initted = 1;
	    }

	  if (--pending_execs == 0)
	    break;

	  /* Just make it go on.  */
	  target_continue_no_signal (resume_ptid);
	}
    }

  return resume_ptid;
}

/* See nat/fork-inferior.h.  */

void
trace_start_error (const char *fmt, ...)
{
  va_list ap;

  va_start (ap, fmt);
  warning ("Could not trace the inferior process.");
  vwarning (fmt, ap);
  va_end (ap);

  gdb_flush_out_err ();
  _exit (0177);
}

/* See nat/fork-inferior.h.  */

void
trace_start_error_with_name (const char *string)
{
  trace_start_error ("%s: %s", string, safe_strerror (errno));
}