/*
   Copyright (C) 1995, 1996, 1997 Free Software Foundation, Inc.
   Written by Michael I. Bushnell.

   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 this program; if not, write to the Free Software
   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */

#include "ports.h"
#include <spin-lock.h>
#include <assert.h>
#include <cthreads.h>
#include <mach/message.h>
#include <mach/thread_info.h>
#include <mach/thread_switch.h>

/* A very high priority is assigned to reduce thread storms in core servers. */
#define THREAD_PRI 2

static void
adjust_priority (void)
{
  mach_port_t host_priv, self, pset, pset_priv;
  error_t err;

  self = MACH_PORT_NULL;

  err = get_privileged_ports (&host_priv, NULL);
  if (err)
    goto out;

  self = mach_thread_self ();
  err = thread_get_assignment (self, &pset);
  if (err)
    goto out;

  err = host_processor_set_priv (host_priv, pset, &pset_priv);
  if (err)
    goto out;

  err = thread_max_priority (self, pset_priv, 0);
  if (err)
    goto out;

  err = thread_priority (self, THREAD_PRI, 0);

out:
  if (self != MACH_PORT_NULL)
    mach_port_deallocate (mach_task_self (), self);

  if (err)
    {
      errno = err;
      perror ("unable to adjust libports thread priority");
    }
}

void
ports_manage_port_operations_multithread (struct port_bucket *bucket,
					  ports_demuxer_type demuxer,
					  int thread_timeout,
					  int global_timeout,
					  void (*hook)())
{
  volatile int nreqthreads;
  volatile int totalthreads;
  spin_lock_t lock = SPIN_LOCK_INITIALIZER;

  auto int thread_function (int);

  int
  internal_demuxer (mach_msg_header_t *inp,
		    mach_msg_header_t *outheadp)
    {
      int spawn = 0;
      int status;
      struct port_info *pi;
      struct rpc_info link;
      register mig_reply_header_t *outp = (mig_reply_header_t *) outheadp;
      static const mach_msg_type_t RetCodeType = {
		/* msgt_name = */		MACH_MSG_TYPE_INTEGER_32,
		/* msgt_size = */		32,
		/* msgt_number = */		1,
		/* msgt_inline = */		TRUE,
		/* msgt_longform = */		FALSE,
		/* msgt_deallocate = */		FALSE,
		/* msgt_unused = */		0
	};

      spin_lock (&lock);
      assert (nreqthreads);
      nreqthreads--;
      if (nreqthreads == 0)
	/* No thread would be listening for requests, spawn one. */
	spawn = 1;
      spin_unlock (&lock);

      if (spawn)
	{
	  spin_lock (&lock);
	  totalthreads++;
	  nreqthreads++;
	  spin_unlock (&lock);
	  cthread_detach (cthread_fork ((cthread_fn_t) thread_function, 0));
	}
      
      /* Fill in default response. */
      outp->Head.msgh_bits 
	= MACH_MSGH_BITS(MACH_MSGH_BITS_REMOTE(inp->msgh_bits), 0);
      outp->Head.msgh_size = sizeof *outp;
      outp->Head.msgh_remote_port = inp->msgh_remote_port;
      outp->Head.msgh_local_port = MACH_PORT_NULL;
      outp->Head.msgh_seqno = 0;
      outp->Head.msgh_id = inp->msgh_id + 100;
      outp->RetCodeType = RetCodeType;
      outp->RetCode = MIG_BAD_ID;

      pi = ports_lookup_port (bucket, inp->msgh_local_port, 0);
      if (pi)
	{
	  error_t err = ports_begin_rpc (pi, inp->msgh_id, &link);
	  if (err)
	    {
	      outp->RetCode = err;
	      status = 1;
	    }
	  else
	    {
	      mutex_lock (&_ports_lock);
	      if (inp->msgh_seqno < pi->cancel_threshold)
		hurd_thread_cancel (link.thread);
	      mutex_unlock (&_ports_lock);
	      status = demuxer (inp, outheadp);
	      ports_end_rpc (pi, &link);
	    }
	  ports_port_deref (pi);
	}
      else
	{
	  outp->RetCode = EOPNOTSUPP;
	  status = 1;
	}

      spin_lock (&lock);
      nreqthreads++;
      spin_unlock (&lock);

      return status;
    }

  int
  thread_function (int master)
    {
      int timeout;
      error_t err;

      /* XXX To reduce starvation, the priority of new threads is initially
	 depressed. This helps already existing threads complete their job
	 and be recycled to handle new messages. The duration of this
	 depression is made a function of the total number of threads because
	 more threads implies more contention, and the priority of threads
	 blocking on a contented spin lock is also implicitely depressed.
	 The lock isn't needed, since an approximation is sufficient. */
      timeout = (((totalthreads - 1) / 100) + 1) * 10;
      thread_switch (MACH_PORT_NULL, SWITCH_OPTION_DEPRESS, timeout);

      /* XXX Give servers a greater priority when possible. */
      adjust_priority ();

      if (hook)
	(*hook) ();

      if (master)
	timeout = global_timeout;
      else
	timeout = thread_timeout;

    startover:

      do
	err = mach_msg_server_timeout (internal_demuxer, 0, bucket->portset,
				       timeout ? MACH_RCV_TIMEOUT : 0,
				       timeout);
      while (err != MACH_RCV_TIMED_OUT);

      if (master)
	{
	  spin_lock (&lock);
	  if (totalthreads != 1)
	    {
	      spin_unlock (&lock);
	      goto startover;
	    }
	}
      else
	{
	  spin_lock (&lock);
	  if (nreqthreads == 1)
	    {
	      /* No other thread is listening for requests, continue. */
	      spin_unlock (&lock);
	      goto startover;
	    }
	  nreqthreads--;
	  totalthreads--;
	  spin_unlock (&lock);
	}
      return 0;
    }

  nreqthreads = 1;
  totalthreads = 1;
  thread_function (1);
}