/* npth.c - a lightweight implementation of pth over pthread.

 * Copyright (C) 2011 g10 Code GmbH

 *

 * This file is part of nPth.

 *

 * nPth 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.

 *

 * nPth 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 this program; if not, see <https://www.gnu.org/licenses/>.

 */

#ifdef HAVE_CONFIG_H

#include <config.h>

#endif

#include <stdlib.h>

#include <stdio.h>

#include <string.h>

#include <assert.h>

#include <errno.h>

#include <pthread.h>

#include <fcntl.h>

#include <sys/stat.h>

#ifdef HAVE_LIB_DISPATCH

# include <dispatch/dispatch.h>

typedef dispatch_semaphore_t sem_t;

/* This glue code is for macOS which does not have full implementation

   of POSIX semaphore.  On macOS, using semaphore in Grand Central

   Dispatch library is better than using the partial implementation of

   POSIX semaphore where sem_init doesn't work well.

 */

static int

sem_init (sem_t *sem, int is_shared, unsigned int value)

{

  (void)is_shared;

  if ((*sem = dispatch_semaphore_create (value)) == NULL)

    return -1;

  else

    return 0;

}

static int

sem_post (sem_t *sem)

{

  dispatch_semaphore_signal (*sem);

  return 0;

}

static int

sem_wait (sem_t *sem)

{

  dispatch_semaphore_wait (*sem, DISPATCH_TIME_FOREVER);

  return 0;

}

#else

# include <semaphore.h>

#endif

#ifdef HAVE_UNISTD_H

# include <unistd.h>

#endif

#ifndef HAVE_PSELECT

# include <signal.h>

#endif

#ifdef HAVE_POLL_H

#include <poll.h>

#endif

#include "npth.h"

/* The global lock that excludes all threads but one.  This is a

   semaphore, because these can be safely used in a library even if

   the application or other libraries call fork(), including from a

   signal handler.  sem_post is async-signal-safe.  (The reason a

   semaphore is safe and a mutex is not safe is that a mutex has an

   owner, while a semaphore does not.)  We init sceptre to a static

   buffer for use by sem_init; in case sem_open is used instead

   SCEPTRE will changed to the value returned by sem_open.

   GOT_SCEPTRE is a flag used for debugging to tell wether we hold

   SCEPTRE.  */

static sem_t sceptre_buffer;

static sem_t *sceptre = &sceptre_buffer;

static int got_sceptre;

/* Configure defines HAVE_FORK_UNSAFE_SEMAPHORE if child process can't

   access non-shared unnamed semaphore which is created by its parent.

   We use unnamed semaphore (if available) for the global lock.  The

   specific semaphore is only valid for those threads in a process,

   and it is no use by other processes.  Thus, PSHARED argument for

   sem_init is naturally 0.

   However, there are daemon-like applications which use fork after

   npth's initialization by npth_init.  In this case, a child process

   uses the semaphore which was created by its parent process, while

   parent does nothing with the semaphore.  In some system (e.g. AIX),

   access by child process to non-shared unnamed semaphore is

   prohibited.  For such a system, HAVE_FORK_UNSAFE_SEMAPHORE should

   be defined, so that unnamed semaphore will be created with the

   option PSHARED=1.  The purpose of the setting of PSHARED=1 is only

   for allowing the access of the lock by child process.  For NPTH, it

   does not mean any other interactions between processes.

 */

#ifdef HAVE_FORK_UNSAFE_SEMAPHORE

#define NPTH_SEMAPHORE_PSHARED 1

#else

#define NPTH_SEMAPHORE_PSHARED 0

#endif

/* The main thread is the active thread at the time pth_init was

   called.  As of now it is only useful for debugging.  The volatile

   make sure the compiler does not eliminate this set but not used

   variable.  */

static volatile pthread_t main_thread;

/* This flag is set as soon as npth_init has been called or if any

 * thread has been created.  It will never be cleared again.  The only

 * purpose is to make npth_protect and npth_unprotect more robust in

 * that they can be shortcut when npth_init has not yet been called.

 * This is important for libraries which want to support nPth by using

 * those two functions but may have be initialized before pPth. */

static int initialized_or_any_threads;

/* Systems that don't have pthread_mutex_timedlock get a busy wait

   implementation that probes the lock every BUSY_WAIT_INTERVAL

   milliseconds.  */

#define BUSY_WAIT_INTERVAL 200

typedef int (*trylock_func_t) (void *);

#ifndef HAVE_PTHREAD_MUTEX_TIMEDLOCK

static int

busy_wait_for (trylock_func_t trylock, void *lock,

         const struct timespec *abstime)

{

  int err;

  /* This is not great, but better than nothing.  Only works for locks

     which are mostly uncontested.  Provides absolutely no fairness at

     all.  Creates many wake-ups.  */

  while (1)

    {

      struct timespec ts;

      err = npth_clock_gettime (&ts);

      if (err < 0)

  {

    /* Just for safety make sure we return some error.  */

    err = errno ? errno : EINVAL;

    break;

  }

      if (npth_timercmp (abstime, &ts, <))

  {

    err = ETIMEDOUT;

    break;

  }

      err = (*trylock) (lock);

      if (err != EBUSY)

  break;

      /* Try again after waiting a bit.  We could calculate the

   maximum wait time from ts and abstime, but we don't

   bother, as our granularity is pretty fine.  */

      usleep (BUSY_WAIT_INTERVAL * 1000);

    }

  return err;

}

#endif

static void

enter_npth (void)

{

  int res;

  got_sceptre = 0;

  res = sem_post (sceptre);

  assert (res == 0);

}

static void

leave_npth (void)

{

  int res;

  int save_errno = errno;

  do {

    res = sem_wait (sceptre);

  } while (res < 0 && errno == EINTR);

  assert (!res);

  got_sceptre = 1;

  errno = save_errno;

}

#define ENTER() enter_npth ()

#define LEAVE() leave_npth ()

int

npth_init (void)

{

  int res;

  main_thread = pthread_self();

  /* Track that we have been initialized.  */

  initialized_or_any_threads |= 1;

  /* Better reset ERRNO so that we know that it has been set by

     sem_init.  */

  errno = 0;

  /* The semaphore is binary.  */

  res = sem_init (sceptre, NPTH_SEMAPHORE_PSHARED, 1);

  /* There are some versions of operating systems which have sem_init

     symbol defined but the call actually returns ENOSYS at runtime.

     We know this problem for older versions of AIX (<= 4.3.3) and

     macOS.  For macOS, we use semaphore in Grand Central Dispatch

     library, so ENOSYS doesn't happen.  We only support AIX >= 5.2,

     where sem_init is supported.

   */

  if (res < 0)

    {

      /* POSIX.1-2001 defines the semaphore interface but does not

         specify the return value for success.  Thus we better

         bail out on error only on a POSIX.1-2008 system.  */

#if _POSIX_C_SOURCE >= 200809L

      return errno;

#endif

    }

  LEAVE();

  return 0;

}

int

npth_getname_np (npth_t target_thread, char *buf, size_t buflen)

{

#ifdef HAVE_PTHREAD_GETNAME_NP

  return pthread_getname_np (target_thread, buf, buflen);

#else

  (void)target_thread;

  (void)buf;

  (void)buflen;

  return ENOSYS;

#endif

}

int

npth_setname_np (npth_t target_thread, const char *name)

{

#ifdef HAVE_PTHREAD_SETNAME_NP

#ifdef __NetBSD__

  return pthread_setname_np (target_thread, "%s", (void*) name);

#else

#ifdef __APPLE__

  if (target_thread == npth_self ())

    return pthread_setname_np (name);

  else

    return ENOTSUP;

#else

  return pthread_setname_np (target_thread, name);

#endif

#endif

#else

  (void)target_thread;

  (void)name;

  return ENOSYS;

#endif

}

struct startup_s

{

  void *(*start_routine) (void *);

  void *arg;

};

static void *

thread_start (void *startup_arg)

{

  struct startup_s *startup = startup_arg;

  void *(*start_routine) (void *);

  void *arg;

  void *result;

  start_routine = startup->start_routine;

  arg = startup->arg;

  free (startup);

  LEAVE();

  result = (*start_routine) (arg);

  /* Note: instead of returning here, we might end up in

     npth_exit() instead.  */

  ENTER();

  return result;

}

int

npth_create (npth_t *thread, const npth_attr_t *attr,

       void *(*start_routine) (void *), void *arg)

{

  int err;

  struct startup_s *startup;

  startup = malloc (sizeof (*startup));

  if (!startup)

    return errno;

  initialized_or_any_threads |= 2;

  startup->start_routine = start_routine;

  startup->arg = arg;

  err = pthread_create (thread, attr, thread_start, startup);

  if (err)

    {

      free (startup);

      return err;

    }

  /* Memory is released in thread_start.  */

  return 0;

}

int

npth_join (npth_t thread, void **retval)

{

  int err;

#ifdef HAVE_PTHREAD_TRYJOIN_NP

  /* No need to allow competing threads to enter when we can get the

     lock immediately.  pthread_tryjoin_np is a GNU extension.  */

  err = pthread_tryjoin_np (thread, retval);

  if (err != EBUSY)

    return err;

#endif /*HAVE_PTHREAD_TRYJOIN_NP*/

  ENTER();

  err = pthread_join (thread, retval);

  LEAVE();

  return err;

}

void

npth_exit (void *retval)

{

  ENTER();

  pthread_exit (retval);

  /* Never reached.  But just in case pthread_exit does return... */

  LEAVE();

}

int

npth_mutex_lock (npth_mutex_t *mutex)

{

  int err;

  /* No need to allow competing threads to enter when we can get the

     lock immediately.  */

  err = pthread_mutex_trylock (mutex);

  if (err != EBUSY)

    return err;

  ENTER();

  err = pthread_mutex_lock (mutex);

  LEAVE();

  return err;

}

int

npth_mutex_timedlock (npth_mutex_t *mutex, const struct timespec *abstime)

{

  int err;

  /* No need to allow competing threads to enter when we can get the

     lock immediately.  */

  err = pthread_mutex_trylock (mutex);

  if (err != EBUSY)

    return err;

  ENTER();

#if HAVE_PTHREAD_MUTEX_TIMEDLOCK

  err = pthread_mutex_timedlock (mutex, abstime);

#else

  err = busy_wait_for ((trylock_func_t) pthread_mutex_trylock, mutex, abstime);

#endif

  LEAVE();

  return err;

}

#ifndef _NPTH_NO_RWLOCK

int

npth_rwlock_rdlock (npth_rwlock_t *rwlock)

{

  int err;

#ifdef HAVE_PTHREAD_RWLOCK_TRYRDLOCK

  /* No need to allow competing threads to enter when we can get the

     lock immediately.  */

  err = pthread_rwlock_tryrdlock (rwlock);

  if (err != EBUSY)

    return err;

#endif

  ENTER();

  err = pthread_rwlock_rdlock (rwlock);

  LEAVE();

  return err;

}

int

npth_rwlock_timedrdlock (npth_rwlock_t *rwlock, const struct timespec *abstime)

{

  int err;

#ifdef HAVE_PTHREAD_RWLOCK_TRYRDLOCK

  /* No need to allow competing threads to enter when we can get the

     lock immediately.  */

  err = pthread_rwlock_tryrdlock (rwlock);

  if (err != EBUSY)

    return err;

#endif

  ENTER();

#if HAVE_PTHREAD_RWLOCK_TIMEDRDLOCK

  err = pthread_rwlock_timedrdlock (rwlock, abstime);

#else

  err = busy_wait_for ((trylock_func_t) pthread_rwlock_tryrdlock, rwlock,

           abstime);

#endif

  LEAVE();

  return err;

}

int

npth_rwlock_wrlock (npth_rwlock_t *rwlock)

{

  int err;

#ifdef HAVE_PTHREAD_RWLOCK_TRYWRLOCK

  /* No need to allow competing threads to enter when we can get the

     lock immediately.  */

  err = pthread_rwlock_trywrlock (rwlock);

  if (err != EBUSY)

    return err;

#endif

  ENTER();

  err = pthread_rwlock_wrlock (rwlock);

  LEAVE();

  return err;

}

int

npth_rwlock_timedwrlock (npth_rwlock_t *rwlock, const struct timespec *abstime)

{

  int err;

#ifdef HAVE_PTHREAD_RWLOCK_TRYWRLOCK

  /* No need to allow competing threads to enter when we can get the

     lock immediately.  */

  err = pthread_rwlock_trywrlock (rwlock);

  if (err != EBUSY)

    return err;

#endif

  ENTER();

#if HAVE_PTHREAD_RWLOCK_TIMEDWRLOCK

  err = pthread_rwlock_timedwrlock (rwlock, abstime);

#elif HAVE_PTHREAD_RWLOCK_TRYRDLOCK

  err = busy_wait_for ((trylock_func_t) pthread_rwlock_trywrlock, rwlock,

           abstime);

#else

  err = ENOSYS;

#endif

  LEAVE();

  return err;

}

#endif

int

npth_cond_wait (npth_cond_t *cond, npth_mutex_t *mutex)

{

  int err;

  ENTER();

  err = pthread_cond_wait (cond, mutex);

  LEAVE();

  return err;

}

int

npth_cond_timedwait (npth_cond_t *cond, npth_mutex_t *mutex,

         const struct timespec *abstime)

{

  int err;

  ENTER();

  err = pthread_cond_timedwait (cond, mutex, abstime);

  LEAVE();

  return err;

}

/* Standard POSIX Replacement API */

int

npth_usleep(unsigned int usec)

{

  int res;

  ENTER();

  res = usleep(usec);

  LEAVE();

  return res;

}

unsigned int

npth_sleep(unsigned int sec)

{

  unsigned res;

  ENTER();

  res = sleep(sec);

  LEAVE();

  return res;

}

int

npth_system(const char *cmd)

{

  int res;

  ENTER();

  res = system(cmd);

  LEAVE();

  return res;

}

pid_t

npth_waitpid(pid_t pid, int *status, int options)

{

  pid_t res;

  ENTER();

  res = waitpid(pid,status, options);

  LEAVE();

  return res;

}

int

npth_connect(int s, const struct sockaddr *addr, socklen_t addrlen)

{

  int res;

  ENTER();

  res = connect(s, addr, addrlen);

  LEAVE();

  return res;

}

int

npth_accept(int s, struct sockaddr *addr, socklen_t *addrlen)

{

  int res;

  ENTER();

  res = accept(s, addr, addrlen);

  LEAVE();

  return res;

}

int

npth_select(int nfd, fd_set *rfds, fd_set *wfds, fd_set *efds,

      struct timeval *timeout)

{

  int res;

  ENTER();

  res = select(nfd, rfds, wfds, efds, timeout);

  LEAVE();

  return res;

}

int

npth_pselect(int nfd, fd_set *rfds, fd_set *wfds, fd_set *efds,

       const struct timespec *timeout, const sigset_t *sigmask)

{

  int res;

  ENTER();

#ifdef HAVE_PSELECT

  res = pselect (nfd, rfds, wfds, efds, timeout, sigmask);

#else /*!HAVE_PSELECT*/

  {

    /* A better emulation of pselect would be to create a pipe, wait

       in the select for one end and have a signal handler write to

       the other end.  However, this is non-trivial to implement and

       thus we only print a compile time warning.  */

#   ifdef __GNUC__

#     warning Using a non race free pselect emulation.

#   endif

    struct timeval t, *tp;

    tp = NULL;

    if (!timeout)

      ;

    else if (timeout->tv_nsec >= 0 && timeout->tv_nsec < 1000000000)

      {

        t.tv_sec = timeout->tv_sec;

        t.tv_usec = (timeout->tv_nsec + 999) / 1000;

        tp = &t;

      }

    else

      {

        errno = EINVAL;

        res = -1;

        goto leave;

      }

    if (sigmask)

      {

        int save_errno;

        sigset_t savemask;

        pthread_sigmask (SIG_SETMASK, sigmask, &savemask);

        res = select (nfd, rfds, wfds, efds, tp);

        save_errno = errno;

        pthread_sigmask (SIG_SETMASK, &savemask, NULL);

        errno = save_errno;

      }

    else

      res = select (nfd, rfds, wfds, efds, tp);

  leave:

    ;

  }

#endif /*!HAVE_PSELECT*/

  LEAVE();

  return res;

}

int

npth_poll (struct pollfd *fds, unsigned long nfds, int timeout)

{

  int res;

  ENTER();

  res = poll (fds, (nfds_t)nfds, timeout);

  LEAVE();

  return res;

}

int

npth_ppoll (struct pollfd *fds, unsigned long nfds,

            const struct timespec *timeout, const sigset_t *sigmask)

{

  int res;

  ENTER();

#ifdef HAVE_PPOLL

  res = ppoll (fds, (nfds_t)nfds, timeout, sigmask);

#else /*!HAVE_PPOLL*/

  {

#   ifdef __GNUC__

#     warning Using a non race free ppoll emulation.

#   endif

    int t;

    if (!timeout)

      t = -1;

    else if (timeout->tv_nsec >= 0 && timeout->tv_nsec < 1000000000)

      t = timeout->tv_sec * 1000 + (timeout->tv_nsec + 999999) / 1000000;

    else

      {

        errno = EINVAL;

        res = -1;

        goto leave;

      }

    if (sigmask)

      {

        int save_errno;

        sigset_t savemask;

        pthread_sigmask (SIG_SETMASK, sigmask, &savemask);

        res = poll (fds, (nfds_t)nfds, t);

        save_errno = errno;

        pthread_sigmask (SIG_SETMASK, &savemask, NULL);

        errno = save_errno;

      }

    else

      res = poll (fds, (nfds_t)nfds, t);

  leave:

    ;

  }

#endif

  LEAVE();

  return res;

}

ssize_t

npth_read(int fd, void *buf, size_t nbytes)

{

  ssize_t res;

  ENTER();

  res = read(fd, buf, nbytes);

  LEAVE();

  return res;

}

ssize_t

npth_write(int fd, const void *buf, size_t nbytes)

{

  ssize_t res;

  ENTER();

  res = write(fd, buf, nbytes);

  LEAVE();

  return res;

}

int

npth_recvmsg (int fd, struct msghdr *msg, int flags)

{

  int res;

  ENTER();

  res = recvmsg (fd, msg, flags);

  LEAVE();

  return res;

}

int

npth_sendmsg (int fd, const struct msghdr *msg, int flags)

{

  int res;

  ENTER();

  res = sendmsg (fd, msg, flags);

  LEAVE();

  return res;

}

void

npth_unprotect (void)

{

  /* If we are not initialized we may not access the semaphore and

   * thus we shortcut it. Note that in this case the unprotect/protect

   * is not needed.  For failsafe reasons if an nPth thread has ever

   * been created but nPth has accidentally not initialized we do not

   * shortcut so that a stack backtrace (due to the access of the

   * uninitialized semaphore) is more expressive.  */

  if (initialized_or_any_threads)

    ENTER();

}

void

npth_protect (void)

{

  /* See npth_unprotect for commentary.  */

  if (initialized_or_any_threads)

    LEAVE();

}

int

npth_is_protected (void)

{

  return got_sceptre;

}

int

npth_clock_gettime (struct timespec *ts)

{

#if defined(CLOCK_REALTIME) && HAVE_CLOCK_GETTIME

  return clock_gettime (CLOCK_REALTIME, ts);

#elif HAVE_GETTIMEOFDAY

  {

    struct timeval tv;

    if (gettimeofday (&tv, NULL))

      return -1;

    ts->tv_sec = tv.tv_sec;

    ts->tv_nsec = tv.tv_usec * 1000;

    return 0;

  }

#else

  /* FIXME: fall back on time() with seconds resolution.  */

# error clock_gettime not available - please provide a fallback.

#endif

}