diff mbox

posix: Fix open file action for posix_spawn on Linux

Message ID c37852d4-a5b4-6fe4-67d4-0616ba62b278@redhat.com
State New
Headers show

Commit Message

Florian Weimer Sept. 19, 2016, 3:25 p.m. UTC
On 09/19/2016 04:39 PM, Adhemerval Zanella wrote:
>
>
> On 19/09/2016 09:53, Florian Weimer wrote:
>> On 09/19/2016 02:41 PM, Adhemerval Zanella wrote:
>>
>>> +/* Return if file descriptor is opened.  */
>>> +static inline int
>>> +fd_is_valid (int fd)
>>> +{
>>> +  return fcntl_not_cancel_2 (fd, F_GETFD) != -1 || errno != EBADF;
>>> +}
>>> +
>>>  /* Function used in the clone call to setup the signals mask, posix_spawn
>>>     attributes, and file actions.  It run on its own stack (provided by the
>>>     posix_spawn call).  */
>>> @@ -219,6 +226,15 @@ __spawni_child (void *arguments)
>>>
>>>          case spawn_do_open:
>>>            {
>>> +        /* POSIX states that if fildes was already an open file descriptor,
>>> +           it shall be closed before the new file is opened.  This avoid
>>> +           pontential issues when posix_spawn plus addopen action is called
>>> +           with the process already at maximum number of file descriptor
>>> +           opened and also for multiple actions on single-open special
>>> +           paths (like /dev/watchdog).  */
>>> +        if (fd_is_valid (action->action.open_action.fd))
>>> +          close_not_cancel (action->action.open_action.fd);
>>
>> It's not clear to me why you can't just close the file descriptor unconditionally.  It does not seem to matter whether you perform fcntl or close on an invalid file descriptor.
>
> Indeed it seems superfluous to check for file descriptor validity before
> calling close, I will remove it.  With this changes, do you have any
> pending issues for the patch?

I'd rather suggest something like this, avoiding the close in the common 
case when the open call succeeds.

The sysdeps/posix implementation would need a similar change.

However, the new test case consistently fails for me because of the 
pipe2 call in the Linux implementation.

Thanks,
Florian

Comments

Adhemerval Zanella Netto Sept. 19, 2016, 5:14 p.m. UTC | #1
On 19/09/2016 12:25, Florian Weimer wrote:
> On 09/19/2016 04:39 PM, Adhemerval Zanella wrote:
>>
>>
>> On 19/09/2016 09:53, Florian Weimer wrote:
>>> On 09/19/2016 02:41 PM, Adhemerval Zanella wrote:
>>>
>>>> +/* Return if file descriptor is opened.  */
>>>> +static inline int
>>>> +fd_is_valid (int fd)
>>>> +{
>>>> +  return fcntl_not_cancel_2 (fd, F_GETFD) != -1 || errno != EBADF;
>>>> +}
>>>> +
>>>>  /* Function used in the clone call to setup the signals mask, posix_spawn
>>>>     attributes, and file actions.  It run on its own stack (provided by the
>>>>     posix_spawn call).  */
>>>> @@ -219,6 +226,15 @@ __spawni_child (void *arguments)
>>>>
>>>>          case spawn_do_open:
>>>>            {
>>>> +        /* POSIX states that if fildes was already an open file descriptor,
>>>> +           it shall be closed before the new file is opened.  This avoid
>>>> +           pontential issues when posix_spawn plus addopen action is called
>>>> +           with the process already at maximum number of file descriptor
>>>> +           opened and also for multiple actions on single-open special
>>>> +           paths (like /dev/watchdog).  */
>>>> +        if (fd_is_valid (action->action.open_action.fd))
>>>> +          close_not_cancel (action->action.open_action.fd);
>>>
>>> It's not clear to me why you can't just close the file descriptor unconditionally.  It does not seem to matter whether you perform fcntl or close on an invalid file descriptor.
>>
>> Indeed it seems superfluous to check for file descriptor validity before
>> calling close, I will remove it.  With this changes, do you have any
>> pending issues for the patch?
> 
> I'd rather suggest something like this, avoiding the close in the common case when the open call succeeds.

This approach has the issue that to 1. potentially trigger the dup + close
case more often because the new fd won't be equal to the expected one,
while closing the file descriptor first might avoid this issue; and 2.
to always trigger a failure in the behaviour described in the patch
(EMFILE).

> 
> The sysdeps/posix implementation would need a similar change.

Yes, that's why I state this patch only fixes the Linux part.
 
> However, the new test case consistently fails for me because of the pipe2 call in the Linux implementation.

Sorry, I should have stated that it requires the pipe removal
patch [1] on posix_spawn.

[1] https://sourceware.org/ml/libc-alpha/2016-09/msg00006.html
Florian Weimer Sept. 20, 2016, 7:57 p.m. UTC | #2
* Adhemerval Zanella:

>>> Indeed it seems superfluous to check for file descriptor validity before
>>> calling close, I will remove it.  With this changes, do you have any
>>> pending issues for the patch?
>> 
>> I'd rather suggest something like this, avoiding the close in the
>> common case when the open call succeeds.
>
> This approach has the issue that to 1. potentially trigger the dup + close
> case more often because the new fd won't be equal to the expected one,
> while closing the file descriptor first might avoid this issue; and 2.
> to always trigger a failure in the behaviour described in the patch
> (EMFILE).

1. seems a valid argument to me.  It may pollute the strace output in
some cases, but it's true that it's a net savings in the most common
cases. Suggestion withdrawn.

2. I do not understand.

>> However, the new test case consistently fails for me because of the
>> pipe2 call in the Linux implementation.
>
> Sorry, I should have stated that it requires the pipe removal
> patch [1] on posix_spawn.
>
> [1] https://sourceware.org/ml/libc-alpha/2016-09/msg00006.html

Okay, we'll have to get that patch in first obviously.
Adhemerval Zanella Netto Sept. 20, 2016, 8:04 p.m. UTC | #3
On 20/09/2016 16:57, Florian Weimer wrote:
> * Adhemerval Zanella:
> 
>>>> Indeed it seems superfluous to check for file descriptor validity before
>>>> calling close, I will remove it.  With this changes, do you have any
>>>> pending issues for the patch?
>>>
>>> I'd rather suggest something like this, avoiding the close in the
>>> common case when the open call succeeds.
>>
>> This approach has the issue that to 1. potentially trigger the dup + close
>> case more often because the new fd won't be equal to the expected one,
>> while closing the file descriptor first might avoid this issue; and 2.
>> to always trigger a failure in the behaviour described in the patch
>> (EMFILE).
> 
> 1. seems a valid argument to me.  It may pollute the strace output in
> some cases, but it's true that it's a net savings in the most common
> cases. Suggestion withdrawn.
> 
> 2. I do not understand.
> 

Issuing open and check its result in a program that reached its maximum
file descriptor count (EMFILE situation such as the testcase aims to 
reproduce) will always trigger the exception case where it will require 
to close and retry the open (so instead of just close plus open, it will
be open failure, close, open).


>>> However, the new test case consistently fails for me because of the
>>> pipe2 call in the Linux implementation.
>>
>> Sorry, I should have stated that it requires the pipe removal
>> patch [1] on posix_spawn.
>>
>> [1] https://sourceware.org/ml/libc-alpha/2016-09/msg00006.html
> 
> Okay, we'll have to get that patch in first obviously.
> 

I think we just need to align which would be best error option for
unknown issues that do not set the errno (because the error code
relies on errno value).  Original patch suggested returning a
bogus value (EHOSTDOWN), where I would prefer just -1 to avoid
implement specific behaviour.
Florian Weimer Sept. 20, 2016, 8:16 p.m. UTC | #4
* Adhemerval Zanella:

> On 20/09/2016 16:57, Florian Weimer wrote:
>> * Adhemerval Zanella:
>> 
>>>>> Indeed it seems superfluous to check for file descriptor validity before
>>>>> calling close, I will remove it.  With this changes, do you have any
>>>>> pending issues for the patch?
>>>>
>>>> I'd rather suggest something like this, avoiding the close in the
>>>> common case when the open call succeeds.
>>>
>>> This approach has the issue that to 1. potentially trigger the dup + close
>>> case more often because the new fd won't be equal to the expected one,
>>> while closing the file descriptor first might avoid this issue; and 2.
>>> to always trigger a failure in the behaviour described in the patch
>>> (EMFILE).
>> 
>> 1. seems a valid argument to me.  It may pollute the strace output in
>> some cases, but it's true that it's a net savings in the most common
>> cases. Suggestion withdrawn.
>> 
>> 2. I do not understand.
>> 
>
> Issuing open and check its result in a program that reached its maximum
> file descriptor count (EMFILE situation such as the testcase aims to 
> reproduce) will always trigger the exception case where it will require 
> to close and retry the open (so instead of just close plus open, it will
> be open failure, close, open).

Yes, but this case will be exceedingly rare.

>>>> However, the new test case consistently fails for me because of the
>>>> pipe2 call in the Linux implementation.
>>>
>>> Sorry, I should have stated that it requires the pipe removal
>>> patch [1] on posix_spawn.
>>>
>>> [1] https://sourceware.org/ml/libc-alpha/2016-09/msg00006.html
>> 
>> Okay, we'll have to get that patch in first obviously.
>> 
>
> I think we just need to align which would be best error option for
> unknown issues that do not set the errno (because the error code
> relies on errno value).  Original patch suggested returning a
> bogus value (EHOSTDOWN), where I would prefer just -1 to avoid
> implement specific behaviour.

But we have to set a different errno value eventually, for the
application to read.  Let's continue that discussion on the other
thread.
diff mbox

Patch

/* POSIX spawn interface.  Linux version.
   Copyright (C) 2016 Free Software Foundation, Inc.
   This file is part of the GNU C Library.

   The GNU C Library is free software; you can redistribute it and/or
   modify it under the terms of the GNU Lesser General Public
   License as published by the Free Software Foundation; either
   version 2.1 of the License, or (at your option) any later version.

   The GNU C Library 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
   Lesser General Public License for more details.

   You should have received a copy of the GNU Lesser General Public
   License along with the GNU C Library; if not, see
   <http://www.gnu.org/licenses/>.  */

#include <spawn.h>
#include <fcntl.h>
#include <paths.h>
#include <string.h>
#include <sys/resource.h>
#include <sys/wait.h>
#include <sys/param.h>
#include <sys/mman.h>
#include <not-cancel.h>
#include <local-setxid.h>
#include <shlib-compat.h>
#include <nptl/pthreadP.h>
#include <dl-sysdep.h>
#include <ldsodefs.h>
#include <libc-internal.h>
#include "spawn_int.h"

/* The Linux implementation of posix_spawn{p} uses the clone syscall directly
   with CLONE_VM and CLONE_VFORK flags and an allocated stack.  The new stack
   and start function solves most the vfork limitation (possible parent
   clobber due stack spilling). The remaining issue are:

   1. That no signal handlers must run in child context, to avoid corrupting
      parent's state.
   2. The parent must ensure child's stack freeing.
   3. Child must synchronize with parent to enforce 2. and to possible
      return execv issues.

   The first issue is solved by blocking all signals in child, even the
   NPTL-internal ones (SIGCANCEL and SIGSETXID).  The second and third issue
   is done by a stack allocation in parent and a synchronization with using
   a pipe or waitpid (in case or error).  The pipe has the advantage of
   allowing the child the communicate an exec error.  */


/* The Unix standard contains a long explanation of the way to signal
   an error after the fork() was successful.  Since no new wait status
   was wanted there is no way to signal an error using one of the
   available methods.  The committee chose to signal an error by a
   normal program exit with the exit code 127.  */
#define SPAWN_ERROR	127

/* We need to block both SIGCANCEL and SIGSETXID.  */
#define SIGALL_SET \
  ((__sigset_t) { .__val = {[0 ...  _SIGSET_NWORDS-1 ] =  -1 } })

#ifdef __ia64__
# define CLONE(__fn, __stack, __stacksize, __flags, __args) \
  __clone2 (__fn, __stack, __stacksize, __flags, __args, 0, 0, 0)
#else
# define CLONE(__fn, __stack, __stacksize, __flags, __args) \
  __clone (__fn, __stack, __flags, __args)
#endif

#if _STACK_GROWS_DOWN
# define STACK(__stack, __stack_size) (__stack + __stack_size)
#elif _STACK_GROWS_UP
# define STACK(__stack, __stack_size) (__stack)
#endif


struct posix_spawn_args
{
  int pipe[2];
  sigset_t oldmask;
  const char *file;
  int (*exec) (const char *, char *const *, char *const *);
  const posix_spawn_file_actions_t *fa;
  const posix_spawnattr_t *restrict attr;
  char *const *argv;
  ptrdiff_t argc;
  char *const *envp;
  int xflags;
};

/* Older version requires that shell script without shebang definition
   to be called explicitly using /bin/sh (_PATH_BSHELL).  */
static void
maybe_script_execute (struct posix_spawn_args *args)
{
  if (SHLIB_COMPAT (libc, GLIBC_2_2, GLIBC_2_15)
      && (args->xflags & SPAWN_XFLAGS_TRY_SHELL) && errno == ENOEXEC)
    {
      char *const *argv = args->argv;
      ptrdiff_t argc = args->argc;

      /* Construct an argument list for the shell.  */
      char *new_argv[argc + 1];
      new_argv[0] = (char *) _PATH_BSHELL;
      new_argv[1] = (char *) args->file;
      if (argc > 1)
	memcpy (new_argv + 2, argv + 1, argc * sizeof(char *));
      else
	new_argv[2] = NULL;

      /* Execute the shell.  */
      args->exec (new_argv[0], new_argv, args->envp);
    }
}

/* Open the file requested in the action.  */
static int
action_file_open (const struct __spawn_action *action)
{
  return open_not_cancel (action->action.open_action.path,
			  action->action.open_action.oflag
			  | O_LARGEFILE,
			  action->action.open_action.mode);
}

/* Function used in the clone call to setup the signals mask, posix_spawn
   attributes, and file actions.  It run on its own stack (provided by the
   posix_spawn call).  */
static int
__spawni_child (void *arguments)
{
  struct posix_spawn_args *args = arguments;
  const posix_spawnattr_t *restrict attr = args->attr;
  const posix_spawn_file_actions_t *file_actions = args->fa;
  int p = args->pipe[1];
  int ret;

  close_not_cancel (args->pipe[0]);

  /* The child must ensure that no signal handler are enabled because it shared
     memory with parent, so the signal disposition must be either SIG_DFL or
     SIG_IGN.  It does by iterating over all signals and although it could
     possibly be more optimized (by tracking which signal potentially have a
     signal handler), it might requires system specific solutions (since the
     sigset_t data type can be very different on different architectures).  */
  struct sigaction sa;
  memset (&sa, '\0', sizeof (sa));

  sigset_t hset;
  __sigprocmask (SIG_BLOCK, 0, &hset);
  for (int sig = 1; sig < _NSIG; ++sig)
    {
      if ((attr->__flags & POSIX_SPAWN_SETSIGDEF)
	  && sigismember (&attr->__sd, sig))
	{
	  sa.sa_handler = SIG_DFL;
	}
      else if (sigismember (&hset, sig))
	{
	  if (__nptl_is_internal_signal (sig))
	    sa.sa_handler = SIG_IGN;
	  else
	    {
	      __libc_sigaction (sig, 0, &sa);
	      if (sa.sa_handler == SIG_IGN)
		continue;
	      sa.sa_handler = SIG_DFL;
	    }
	}
      else
	continue;

      __libc_sigaction (sig, &sa, 0);
    }

#ifdef _POSIX_PRIORITY_SCHEDULING
  /* Set the scheduling algorithm and parameters.  */
  if ((attr->__flags & (POSIX_SPAWN_SETSCHEDPARAM | POSIX_SPAWN_SETSCHEDULER))
      == POSIX_SPAWN_SETSCHEDPARAM)
    {
      if ((ret = __sched_setparam (0, &attr->__sp)) == -1)
	goto fail;
    }
  else if ((attr->__flags & POSIX_SPAWN_SETSCHEDULER) != 0)
    {
      if ((ret = __sched_setscheduler (0, attr->__policy, &attr->__sp)) == -1)
	goto fail;
    }
#endif

  /* Set the process group ID.  */
  if ((attr->__flags & POSIX_SPAWN_SETPGROUP) != 0
      && (ret = __setpgid (0, attr->__pgrp)) != 0)
    goto fail;

  /* Set the effective user and group IDs.  */
  if ((attr->__flags & POSIX_SPAWN_RESETIDS) != 0
      && ((ret = local_seteuid (__getuid ())) != 0
	  || (ret = local_setegid (__getgid ())) != 0))
    goto fail;

  /* Execute the file actions.  */
  if (file_actions != 0)
    {
      int cnt;
      struct rlimit64 fdlimit;
      bool have_fdlimit = false;

      for (cnt = 0; cnt < file_actions->__used; ++cnt)
	{
	  struct __spawn_action *action = &file_actions->__actions[cnt];

	  /* Dup the pipe fd onto an unoccupied one to avoid any file
	     operation to clobber it.  */
	  if ((action->action.close_action.fd == p)
	      || (action->action.open_action.fd == p)
	      || (action->action.dup2_action.fd == p))
	    {
	      if ((ret = __dup (p)) < 0)
		goto fail;
	      p = ret;
	    }

	  switch (action->tag)
	    {
	    case spawn_do_close:
	      if ((ret =
		   close_not_cancel (action->action.close_action.fd)) != 0)
		{
		  if (!have_fdlimit)
		    {
		      __getrlimit64 (RLIMIT_NOFILE, &fdlimit);
		      have_fdlimit = true;
		    }

		  /* Signal errors only for file descriptors out of range.  */
		  if (action->action.close_action.fd < 0
		      || action->action.close_action.fd >= fdlimit.rlim_cur)
		    goto fail;
		}
	      break;

	    case spawn_do_open:
	      {
		ret = action_file_open (action);
		if (ret == -1)
		  {
		    /* Attempt to close the target descriptor.  The
		       open call may succeed after that if all the
		       file descriptors were exhausted.  */
		    (void) close_not_cancel (action->action.open_action.fd);
		    ret = action_file_open (action);
		  }
		if (ret == -1)
		  goto fail;

		int new_fd = ret;

		/* Make sure the desired file descriptor is used.  */
		if (ret != action->action.open_action.fd)
		  {
		    if ((ret = __dup2 (new_fd, action->action.open_action.fd))
			!= action->action.open_action.fd)
		      goto fail;

		    if ((ret = close_not_cancel (new_fd)) != 0)
		      goto fail;
		  }
	      }
	      break;

	    case spawn_do_dup2:
	      if ((ret = __dup2 (action->action.dup2_action.fd,
				 action->action.dup2_action.newfd))
		  != action->action.dup2_action.newfd)
		goto fail;
	      break;
	    }
	}
    }

  /* Set the initial signal mask of the child if POSIX_SPAWN_SETSIGMASK
     is set, otherwise restore the previous one.  */
  __sigprocmask (SIG_SETMASK, (attr->__flags & POSIX_SPAWN_SETSIGMASK)
		 ? &attr->__ss : &args->oldmask, 0);

  args->exec (args->file, args->argv, args->envp);

  /* This is compatibility function required to enable posix_spawn run
     script without shebang definition for older posix_spawn versions
     (2.15).  */
  maybe_script_execute (args);

  ret = -errno;

fail:
  /* Since sizeof errno < PIPE_BUF, the write is atomic. */
  ret = -ret;
  if (ret)
    while (write_not_cancel (p, &ret, sizeof ret) < 0)
      continue;
  _exit (SPAWN_ERROR);
}

/* Spawn a new process executing PATH with the attributes describes in *ATTRP.
   Before running the process perform the actions described in FILE-ACTIONS. */
static int
__spawnix (pid_t * pid, const char *file,
	   const posix_spawn_file_actions_t * file_actions,
	   const posix_spawnattr_t * attrp, char *const argv[],
	   char *const envp[], int xflags,
	   int (*exec) (const char *, char *const *, char *const *))
{
  pid_t new_pid;
  struct posix_spawn_args args;
  int ec;

  if (__pipe2 (args.pipe, O_CLOEXEC))
    return errno;

  /* To avoid imposing hard limits on posix_spawn{p} the total number of
     arguments is first calculated to allocate a mmap to hold all possible
     values.  */
  ptrdiff_t argc = 0;
  /* Linux allows at most max (0x7FFFFFFF, 1/4 stack size) arguments
     to be used in a execve call.  We limit to INT_MAX minus one due the
     compatiblity code that may execute a shell script (maybe_script_execute)
     where it will construct another argument list with an additional
     argument.  */
  ptrdiff_t limit = INT_MAX - 1;
  while (argv[argc++] != NULL)
    if (argc == limit)
      {
	errno = E2BIG;
	return errno;
      }

  int prot = (PROT_READ | PROT_WRITE
	     | ((GL (dl_stack_flags) & PF_X) ? PROT_EXEC : 0));

  /* Add a slack area for child's stack.  */
  size_t argv_size = (argc * sizeof (void *)) + 512;
  size_t stack_size = ALIGN_UP (argv_size, GLRO(dl_pagesize));
  void *stack = __mmap (NULL, stack_size, prot,
			MAP_PRIVATE | MAP_ANONYMOUS | MAP_STACK, -1, 0);
  if (__glibc_unlikely (stack == MAP_FAILED))
    {
      close_not_cancel (args.pipe[0]);
      close_not_cancel (args.pipe[1]);
      return errno;
    }

  /* Disable asynchronous cancellation.  */
  int cs = LIBC_CANCEL_ASYNC ();

  args.file = file;
  args.exec = exec;
  args.fa = file_actions;
  args.attr = attrp ? attrp : &(const posix_spawnattr_t) { 0 };
  args.argv = argv;
  args.argc = argc;
  args.envp = envp;
  args.xflags = xflags;

  __sigprocmask (SIG_BLOCK, &SIGALL_SET, &args.oldmask);

  /* The clone flags used will create a new child that will run in the same
     memory space (CLONE_VM) and the execution of calling thread will be
     suspend until the child calls execve or _exit.  These condition as
     signal below either by pipe write (_exit with SPAWN_ERROR) or
     a successful execve.
     Also since the calling thread execution will be suspend, there is not
     need for CLONE_SETTLS.  Although parent and child share the same TLS
     namespace, there will be no concurrent access for TLS variables (errno
     for instance).  */
  new_pid = CLONE (__spawni_child, STACK (stack, stack_size), stack_size,
		   CLONE_VM | CLONE_VFORK | SIGCHLD, &args);

  close_not_cancel (args.pipe[1]);

  if (new_pid > 0)
    {
      if (__read (args.pipe[0], &ec, sizeof ec) != sizeof ec)
	ec = 0;
      else
	__waitpid (new_pid, NULL, 0);
    }
  else
    ec = -new_pid;

  __munmap (stack, stack_size);

  close_not_cancel (args.pipe[0]);

  if ((ec == 0) && (pid != NULL))
    *pid = new_pid;

  __sigprocmask (SIG_SETMASK, &args.oldmask, 0);

  LIBC_CANCEL_RESET (cs);

  return ec;
}

/* Spawn a new process executing PATH with the attributes describes in *ATTRP.
   Before running the process perform the actions described in FILE-ACTIONS. */
int
__spawni (pid_t * pid, const char *file,
	  const posix_spawn_file_actions_t * acts,
	  const posix_spawnattr_t * attrp, char *const argv[],
	  char *const envp[], int xflags)
{
  return __spawnix (pid, file, acts, attrp, argv, envp, xflags,
		    xflags & SPAWN_XFLAGS_USE_PATH ? __execvpe : __execve);
}