path: root/exec/hashexec.c

/* GNU Hurd standard exec server, #! script execution support.
   Copyright (C) 1995 Free Software Foundation, Inc.
   Written by Roland McGrath.

This file is part of the GNU Hurd.

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

The GNU Hurd 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 the GNU Hurd; see the file COPYING.  If not, write to
the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.  */


#include "priv.h"
#include <hurd/signal.h>
#include <unistd.h>

/* This is called to check E for a #! interpreter specification.  E has
   already been prepared (successfully) and checked (unsuccessfully).  If
   we return success, our caller just returns success for the RPC; we must
   handle all the RPC argument details ourselves.  If we return ENOEXEC, we
   should leave everything as it was.  If we return failure other than
   ENOEXEC, our caller will just fail the RPC.  */

void
check_hashbang (struct execdata *e,
		file_t file,
		task_t oldtask,
		int flags,
		char *argv, u_int argvlen, boolean_t argv_copy,
		char *envp, u_int envplen, boolean_t envp_copy,
		mach_port_t *dtable, u_int dtablesize, boolean_t dtable_copy,
		mach_port_t *portarray, u_int nports, boolean_t portarray_copy,
		int *intarray, u_int nints, boolean_t intarray_copy,
		mach_port_t *deallocnames, u_int ndeallocnames,
		mach_port_t *destroynames, u_int ndestroynames)
{
  char *ibuf = NULL;
  size_t ibufsiz = 0;
  char *p;
  char *interp, *arg;		/* Interpreter file name, and first argument */
  size_t interp_len, len;
  file_t interp_file;		/* Port open on the interpreter file.  */
  FILE *f = &e->stream;
  char *new_argv;
  size_t new_argvlen;
  mach_port_t *new_dtable = NULL;
  u_int new_dtablesize;

  file_t user_fd (int fd)
    {
      if (fd < 0 || fd >= dtablesize || dtable[fd] == MACH_PORT_NULL)
	{
	  errno = EBADF;
	  return MACH_PORT_NULL;
	}
      return dtable[fd];
    }
  file_t user_crdir, user_cwdir, lookup_cwdir;
  error_t user_port (int which, error_t (*operate) (mach_port_t))
    {
      error_t reauthenticate (file_t unauth, file_t *result)
	{
	  error_t err;
	  mach_port_t ref;
	  error_t uauth (auth_t auth)
	    {
	      return auth_user_authenticate (auth,
					     unauth,
					     ref, MACH_MSG_TYPE_MAKE_SEND,
					     result);
	    }
	  if (*result != MACH_PORT_NULL)
	    return 0;
	  ref = mach_reply_port ();
	  err = io_reauthenticate (unauth, ref, MACH_MSG_TYPE_MAKE_SEND);
	  if (!err)
	    err = user_port (INIT_PORT_AUTH, &uauth);
	  mach_port_destroy (mach_task_self (), ref);
	  return err;
	}

      mach_port_t port = ((which < nports &&
			   portarray[which] != MACH_PORT_NULL)
			  ? portarray[which] :
			  (flags & EXEC_DEFAULTS) ? std_ports[which]
			  : MACH_PORT_NULL);

      switch (which)
	{
	case INIT_PORT_CRDIR:
	  if ((flags & EXEC_SECURE) || port == std_ports[which])
	    return (reauthenticate (port, &user_crdir) ?:
		    (*operate) (user_crdir));
	  break;
	case INIT_PORT_CWDIR:
	  if (lookup_cwdir != MACH_PORT_NULL)
	    return (*operate) (lookup_cwdir);
	  if ((flags & EXEC_SECURE) || port == std_ports[which])
	    return (reauthenticate (port, &user_cwdir) ?:
		    (*operate) (user_cwdir));
	  break;
	}
      return (*operate) (port);
    }
  /* Look up NAME on behalf of the client.  */
  inline error_t lookup (const char *name, int flags, mach_port_t *result)
    {
      return hurd_file_name_lookup (&user_port, &user_fd,
				    name, flags, 0, result);
    }

  rewind (f);

  /* Check for our ``magic number''--"#!".  */

  errno = 0;
  if (getc (f) != '#' || getc (f) != '!')
    {
      /* No `#!' here.  If there was a read error (not including EOF),
	 return that error indication.  Otherwise return ENOEXEC to
	 say it's not a file we know how to execute.  */
      e->error = ferror (f) ? errno : ENOEXEC;
      return;
    }

  /* Read the rest of the first line of the file.  */

  interp_len = getline (&ibuf, &ibufsiz, f);
  if (ferror (f))
    {
      e->error = errno ?: EIO;
      return;
    }
  if (ibuf[interp_len - 1] == '\n')
    ibuf[--interp_len] = '\0';

  /* Find the name of the interpreter.  */
  p = ibuf + strspn (ibuf, " \t");
  interp = strsep (&p, " \t");
  /* Skip remaining blanks, and the rest of the line is the argument.  */
  p += strspn (p, " \t");
  arg = p;
  len = interp_len - (arg - ibuf);
  if (len == 0)
    arg = NULL;
  else
    ++len;			/* Include the terminating null.  */

  user_crdir = user_cwdir = lookup_cwdir = MACH_PORT_NULL;

  rwlock_reader_lock (&std_lock);

  /* Open a port on the interpreter file.  */
  e->error = lookup (interp, O_EXEC, &interp_file);

  if (! e->error)
    {
      /* This code is in a local function here for convenience.  Some things in
	 this function need to be protected against faults while accessing ARGV
	 and ENVP; below, we register to preempt signals on these fault areas
	 before calling prepare_args, and unregister afterwards.  When such a
	 fault is detected, the handler does `longjmp (args_faulted, 1)'.  */

      jmp_buf args_faulted;

      inline void prepare_args (void)
	{

	  volatile char *file_name = NULL;
	  size_t namelen;

	  if (! (flags & EXEC_SECURE))
	    {
	      /* Try to figure out the file's name.  We guess that if ARGV[0]
		 contains a slash, it might be the name of the file; and that
		 if it contains no slash, looking for files named by ARGV[0] in
		 the `PATH' environment variable might find it.  */

	      volatile error_t error;
	      volatile char *name;
	      file_t name_file;
	      struct stat st;
	      int file_fstype;
	      fsid_t file_fsid;
	      ino_t file_fileno;

	      error = io_stat (file, &st); /* XXX insecure */
	      if (error)
		goto out;
	      file_fstype = st.st_fstype;
	      file_fsid = st.st_fsid;
	      file_fileno = st.st_ino;

	      if (memchr (argv, '\0', argvlen) == NULL)
		{
		  name = alloca (argvlen + 1);
		  bcopy (argv, name, argvlen);
		  name[argvlen] = '\0';
		}
	      else
		name = argv;

	      if (strchr (name, '/') != NULL)
		error = lookup (name, 0, &name_file);
	      else if (! setjmp (args_faulted))
		{
		  /* Search PATH for it.  If we fault accessing ENVP, setjmp
		     will return again, nonzero this time, and we give up.  */

		  const char envar[] = "\0PATH=";
		  char *path, *p;
		  if (envplen >= sizeof (envar) &&
		      !memcmp (&envar[1], envp, sizeof (envar) - 2))
		    p = envp - 1;
		  else
		    p = memmem (envp, envplen, envar, sizeof (envar) - 1);
		  if (p != NULL)
		    {
		      size_t len;
		      p += sizeof (envar) - 1;
		      len = strlen (p) + 1;
		      path = alloca (len);
		      bcopy (p, path, len);
		    }
		  else
		    {
		      const size_t len = confstr (_CS_PATH, NULL, 0);
		      path = alloca (len);
		      confstr (_CS_PATH, path, len);
		    }

		  p = path;
		  do
		    {
		      path = strchr (p, ':');
		      if (path)
			*path++ = '\0';
		      if (*p == '\0')
			lookup_cwdir = MACH_PORT_NULL;
		      else if (lookup (p, O_EXEC, &lookup_cwdir))
			continue;
		      error = lookup (name, O_EXEC, &name_file);
		      if (*p != '\0')
			{
			  mach_port_deallocate (mach_task_self (),
						lookup_cwdir);
			  lookup_cwdir = MACH_PORT_NULL;
			}
		      if (!error)
			{
			  if (*p != '\0')
			    {
			      size_t dirlen = strlen (p);
			      size_t namelen = strlen (name);
			      char *new = alloca (dirlen + 1 + namelen + 1);
			      memcpy (new, p, dirlen);
			      new[dirlen] = '/';
			      memcpy (&new[dirlen + 1], name, namelen + 1);
			      name = new;
			    }
			  break;
			}
		    } while ((p = path) != NULL);
		}
	      else
		name_file = MACH_PORT_NULL;

	      if (!error && name_file != MACH_PORT_NULL)
		{
		  if (!io_stat (name_file, &st) && /* XXX insecure */
		      st.st_fstype == file_fstype &&
		      st.st_fsid == file_fsid &&
		      st.st_ino == file_fileno)
		    file_name = name;
		  mach_port_deallocate (mach_task_self (), name_file);
		}
	    }

	  if (file_name == NULL)
	    {
	      /* We can't easily find the file.
		 Put it in a file descriptor and pass /dev/fd/N.  */
	      int fd;
	    out:

	      for (fd = 0; fd < dtablesize; ++fd)
		if (dtable[fd] == MACH_PORT_NULL)
		  break;
	      if (fd == dtablesize)
		{
		  /* Extend the descriptor table.  */
		  new_dtable = alloca ((dtablesize + 1) * sizeof (file_t));
		  memcpy (new_dtable, dtable, dtablesize * sizeof (file_t));
		  new_dtablesize = dtablesize + 1;
		  new_dtable[fd] = file;
		}
	      else
		dtable[fd] = file;
	      mach_port_mod_refs (mach_task_self (), file,
				  MACH_PORT_RIGHT_SEND, +1);

	      file_name = alloca (100);
	      sprintf (file_name, "/dev/fd/%d", fd);
	    }

	  /* Prepare the arguments to pass to the interpreter from the original
	     arguments and the name of the script file.  The args will look
	     like `ARGV[0] {ARG} FILE_NAME ARGV[1..n]' (ARG might have been
	     omitted). */

	  namelen = strlen (file_name) + 1;

	  if (! setjmp (args_faulted))
	    {
	      /* XXX leaks below if fault */
	      char *other_args;
	      new_argvlen = argvlen + len + namelen;
	      e->error = vm_allocate (mach_task_self (),
				      (vm_address_t *) &new_argv,
				      new_argvlen, 1);
	      if (e->error)
		return;
	      other_args = memccpy (new_argv, argv, '\0', argvlen);
	      p = &new_argv[other_args ? other_args - new_argv : argvlen];
	      if (arg)
		{
		  memcpy (p, arg, len);
		  p += len;
		}
	      memcpy (p, file_name, namelen);
	      p += namelen;
	      if (other_args)
		memcpy (p, other_args - new_argv + argv,
			argvlen - (other_args - new_argv));
	    }
	  else
	    {
	      /* We got a fault reading ARGV.  So don't use it.  */
	      static const char loser[]
		= "**fault in exec server reading argv[0]**";
	      new_argvlen = sizeof (loser) + len + namelen;
	      new_argv = alloca (argvlen);
	      memcpy (new_argv, loser, sizeof (loser));
	      memcpy (new_argv + sizeof (loser), arg, len);
	      memcpy (new_argv + sizeof (loser) + len, file_name, namelen);
	    }
	}

      /* Preempt SIGSEGV signals for the address ranges of ARGV and ENVP.
	 When such a signal arrives, `preempter' is called to decide what
	 handler to run; it always returns `handler', which is then invoked
	 as a normal signal handler.  Our handler always simply longjmps to
	 ARGS_FAULTED.  */

      void handler (int sig)
	{
	  longjmp (args_faulted, 1);
	}
      const thread_t mythread = hurd_thread_self ();
      sighandler_t preempter (thread_t thread,
			      int signo, long int sigcode, int sigerror)
	{
	  return thread == mythread ? handler : SIG_DFL;
	}

      struct hurd_signal_preempt argv_preempter, envp_preempter;

      /* Register the preemptions.  */
      hurd_preempt_signals (&argv_preempter, SIGSEGV,
			    (long int) argv, (long int) argv + argvlen - 1,
			    preempter);
      hurd_preempt_signals (&envp_preempter, SIGSEGV,
			    (long int) envp, (long int) envp + envplen - 1,
			    preempter);

      /* Do the work.  Everywhere we might get a page fault inside ARGV or
	 ENVP, is inside the zero-return case of an `if' on setjmp
	 (ARGS_FAULTED).  */
      prepare_args ();

      /* Unregister the preemptions.  */
      hurd_unpreempt_signals (&argv_preempter, SIGSEGV);
      hurd_unpreempt_signals (&envp_preempter, SIGSEGV);
    }

  /* We are now done reading the script file.  */
  finish (e, 0);
  free (ibuf);

  rwlock_reader_unlock (&std_lock);

  if (user_crdir != MACH_PORT_NULL)
    mach_port_deallocate (mach_task_self (), user_crdir);
  if (user_cwdir != MACH_PORT_NULL)
    mach_port_deallocate (mach_task_self (), user_cwdir);

  if (e->error)
    /* We cannot open the interpreter file to execute it.  Lose!  */
    return;

  /* Execute the interpreter program.  */
  e->error = file_exec (interp_file,
			oldtask, flags,
			new_argv, new_argvlen, envp, envplen,
			new_dtable ?: dtable, MACH_MSG_TYPE_COPY_SEND,
			new_dtable ? new_dtablesize : dtablesize,
			portarray, MACH_MSG_TYPE_COPY_SEND, nports,
			intarray, nints,
			deallocnames, ndeallocnames,
			destroynames, ndestroynames);
  mach_port_deallocate (mach_task_self (), interp_file);

  if (! e->error)
    {
      /* The exec of the interpreter succeeded!  Deallocate the resources
	 we passed to that exec.  We don't need to save them in a bootinfo
	 structure; the exec of the interpreter takes care of that.  */
      u_int i;
      mach_port_deallocate (mach_task_self (), file);
      task_resume (oldtask);	/* Our caller suspended it.  */
      mach_port_deallocate (mach_task_self (), oldtask);
      if (! argv_copy)
	vm_deallocate (mach_task_self (), (vm_address_t) argv, argvlen);
      if (! envp_copy)
	vm_deallocate (mach_task_self (), (vm_address_t) envp, envplen);
      for (i = 0; i < dtablesize; ++i)
	mach_port_deallocate (mach_task_self (), dtable[i]);
      if (! dtable_copy)
	vm_deallocate (mach_task_self (), (vm_address_t) dtable,
		       dtablesize * sizeof *dtable);
      for (i = 0; i < nports; ++i)
	mach_port_deallocate (mach_task_self (), portarray[i]);
      if (! portarray_copy)
	vm_deallocate (mach_task_self (), (vm_address_t) portarray,
		       nports * sizeof *portarray);
      if (! intarray_copy)
	vm_deallocate (mach_task_self (), (vm_address_t) intarray,
		       nints * sizeof *intarray);
    }
}