firebird-mirror/src/jrd/event.cpp

/*
 *	PROGRAM:	JRD Access Method
 *	MODULE:		event.cpp
 *	DESCRIPTION:	Event Manager
 *
 * The contents of this file are subject to the Interbase Public
 * License Version 1.0 (the "License"); you may not use this file
 * except in compliance with the License. You may obtain a copy
 * of the License at http://www.Inprise.com/IPL.html
 *
 * Software distributed under the License is distributed on an
 * "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, either express
 * or implied. See the License for the specific language governing
 * rights and limitations under the License.
 *
 * The Original Code was created by Inprise Corporation
 * and its predecessors. Portions created by Inprise Corporation are
 * Copyright (C) Inprise Corporation.
 *
 * All Rights Reserved.
 * Contributor(s): ______________________________________.
 *
 * 23-Feb-2002 Dmitry Yemanov - Events wildcarding
 *
 * 2002.10.29 Sean Leyne - Removed obsolete "Netware" port
 *
 */

#include "firebird.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "../jrd/common.h"
#include "gen/iberror.h"
#include "../jrd/ThreadStart.h"
#include "../jrd/event.h"
#include "../jrd/gdsassert.h"
#include "../jrd/event_proto.h"
#include "../jrd/gds_proto.h"
#include "../jrd/isc_proto.h"
#include "../jrd/isc_s_proto.h"
#include "../jrd/thread_proto.h"
#include "../jrd/err_proto.h"
#include "../jrd/os/isc_i_proto.h"
#include "../common/classes/init.h"
#include "../common/config/config.h"
#include "../common/utils_proto.h"

#ifdef HAVE_SYS_TYPES_H
#include <sys/types.h>
#endif

#ifdef HAVE_UNISTD_H
#include <unistd.h>
#endif

#ifdef UNIX
#include <signal.h>
#endif

#ifdef WIN_NT
#include <process.h>
#include <windows.h>

#define MUTEX		event_mutex
#endif

#ifndef MUTEX
#define MUTEX		EVENT_header->evh_mutex
#endif

#define SRQ_BASE                  ((UCHAR *) EVENT_header)

const int MAX_EVENT_BUFFER = 65500;


static EVH acquire();
static FRB alloc_global(UCHAR type, ULONG length, bool recurse);
static SLONG create_process();
static void delete_event(EVNT);
static void delete_process(SLONG);
static void delete_request(EVT_REQ);
static void delete_session(SLONG);
static void deliver();
static void deliver_request(EVT_REQ);
static void exit_handler(void *);
static EVNT find_event(USHORT, const TEXT*, EVNT);
static void free_global(FRB);
static RINT historical_interest(SES, SLONG);
static void init_shmem(void*, SH_MEM, bool);
static void insert_tail(srq *, srq *);
static EVNT make_event(USHORT, const TEXT*, SLONG);
static void mutex_bugcheck(const TEXT*, int);
static void post_process(PRB);
static void probe_processes();
static void punt(const TEXT*);
static void release();
static void remove_que(srq *);
static bool request_completed(EVT_REQ);
static THREAD_ENTRY_DECLARE watcher_thread(THREAD_ENTRY_PARAM);

namespace
{
	EVH EVENT_header = NULL;
	SLONG EVENT_process_offset = 0;
#ifdef SOLARIS_MT
	PRB EVENT_process = NULL;
#endif
	SH_MEM_T EVENT_data;

#if defined(WIN_NT)
	struct mtx event_mutex;
#endif

	class EventStartup
	{
	public:
		static void init()
		{
			TEXT event_file[MAXPATHLEN];
			gds__prefix_lock(event_file, EVENT_FILE);

			ISC_STATUS_ARRAY local_status;
			if (!(EVENT_header = (EVH) ISC_map_file(local_status, event_file, init_shmem, 0,
													Config::getEventMemSize(), &EVENT_data)))
			{
				Firebird::status_exception::raise(local_status);
			}

			gds__register_cleanup(exit_handler, NULL);
		}

		static void cleanup()
		{
			if (EVENT_process_offset)
			{
				if (EVENT_header->evh_current_process != EVENT_process_offset)
					acquire();
				delete_process(EVENT_process_offset);
				release();
			}

			ISC_STATUS_ARRAY local_status;
#ifdef SOLARIS_MT
			ISC_unmap_object(local_status, &EVENT_data, (UCHAR**) &EVENT_process, sizeof(prb));
#endif
			ISC_unmap_file(local_status, &EVENT_data);

			EVENT_header = NULL;
		}
	};

	Firebird::InitMutex<EventStartup> eventStartup;
}


void EVENT_cancel(SLONG request_id)
{
/**************************************
 *
 *	E V E N T _ c a n c e l
 *
 **************************************
 *
 * Functional description
 *	Cancel an outstanding event.
 *
 **************************************/
	if (!EVENT_header)
		return;

	acquire();

	PRB process = (PRB) SRQ_ABS_PTR(EVENT_process_offset);

	srq* que2;
	SRQ_LOOP(process->prb_sessions, que2) {
		SES session = (SES) ((UCHAR *) que2 - OFFSET(SES, ses_sessions));
		srq* event_srq;
		SRQ_LOOP(session->ses_requests, event_srq) {
			EVT_REQ request = (EVT_REQ) ((UCHAR *) event_srq - OFFSET(EVT_REQ, req_requests));
			if (request->req_request_id == request_id) {
				delete_request(request);
				release();
				return;
			}
		}
	}

	release();
}


SLONG EVENT_create_session()
{
/**************************************
 *
 *	E V E N T _ c r e a t e _ s e s s i o n
 *
 **************************************
 *
 * Functional description
 *	Create session.
 *
 **************************************/

	eventStartup.init();

	if (!EVENT_process_offset)
	{
		create_process();
	}

	acquire();
	SES session = (SES) alloc_global(type_ses, (SLONG) sizeof(ses), false);
	PRB process = (PRB) SRQ_ABS_PTR(EVENT_process_offset);
	session->ses_flags = 0;

	insert_tail(&process->prb_sessions, &session->ses_sessions);
	SRQ_INIT(session->ses_requests);
	const SLONG id = SRQ_REL_PTR(session);
	release();

	return id;
}


void EVENT_delete_session(SLONG session_id)
{
/**************************************
 *
 *	E V E N T _ d e l e t e _ s e s s i o n
 *
 **************************************
 *
 * Functional description
 *	Delete a session.
 *
 **************************************/

	if (!EVENT_header)
		return;

	acquire();
	delete_session(session_id);
	release();
}


void EVENT_deliver()
{
/**************************************
 *
 *	E V E N T _ d e l i v e r
 *
 **************************************
 *
 * Functional description
 *	Post an event (step 2).
 *
 *  This code was primarily located in
 *  EVENT_post (see above). This
 *  routine is called by DFW_perform_post_commit_work
 *  once all pending events are prepared
 *  for delivery with EVENT_post.
 *
 **************************************/

	if (!EVENT_header)
    	return;

	acquire();

	/* Deliver requests for posted events */

	bool flag = true;

	while (flag) {
		flag = false;
		srq* event_srq;
		SRQ_LOOP (EVENT_header->evh_processes, event_srq)	{
			PRB process = (PRB) ((UCHAR*) event_srq - OFFSET (PRB, prb_processes));
			if (process->prb_flags & PRB_wakeup) {
				post_process(process);
				flag = true;
				break;
			}
		}
	}

	release();
}


void EVENT_post(USHORT major_length,
			    const TEXT * major_code,
			    USHORT minor_length,
			    const TEXT * minor_code,
			    USHORT count)
{
/**************************************
 *
 *	E V E N T _ p o s t
 *
 **************************************
 *
 * Functional description
 *	Post an event.
 *
 **************************************/

	eventStartup.init();

	acquire();

	EVNT event;
	EVNT parent = find_event(major_length, major_code, 0);
	if (parent && (event = find_event(minor_length, minor_code, parent)))
	{
		event->evnt_count += count;
		srq* event_srq;
		SRQ_LOOP(event->evnt_interests, event_srq) {
			RINT interest = (RINT) ((UCHAR *) event_srq - OFFSET(RINT, rint_interests));
			if (interest->rint_request) {
				EVT_REQ request = (EVT_REQ) SRQ_ABS_PTR(interest->rint_request);

				if (interest->rint_count <= event->evnt_count) {
					PRB process = (PRB) SRQ_ABS_PTR(request->req_process);
					process->prb_flags |= PRB_wakeup;
				}
			}
		}
	}

	release();
}


SLONG EVENT_que(SLONG session_id,
				USHORT string_length,
				const TEXT* string,
				USHORT events_length,
				const UCHAR* events,
				FPTR_EVENT_CALLBACK ast_routine, void* ast_arg)
{
/**************************************
 *
 *	E V E N T _ q u e
 *
 **************************************
 *
 * Functional description
 *
 **************************************/
// Allocate request block

	acquire();
	EVT_REQ request = (EVT_REQ) alloc_global(type_reqb, sizeof(evt_req), false);
	SES session = (SES) SRQ_ABS_PTR(session_id);
	insert_tail(&session->ses_requests, &request->req_requests);
	request->req_session = session_id;
	request->req_process = EVENT_process_offset;
	request->req_ast = ast_routine;
	request->req_ast_arg = ast_arg;
	const SLONG id = ++EVENT_header->evh_request_id;
	request->req_request_id = id;

	const SLONG request_offset = SRQ_REL_PTR(request);

/* Find parent block */
	EVNT parent = find_event(string_length, string, 0);
	if (!parent) {
		parent = make_event(string_length, string, 0);
		request = (EVT_REQ) SRQ_ABS_PTR(request_offset);
		session = (SES) SRQ_ABS_PTR(session_id);
	}

	const SLONG parent_offset = SRQ_REL_PTR(parent);

/* Process event block */

	SRQ_PTR* ptr = &request->req_interests;
	SLONG ptr_offset = SRQ_REL_PTR(ptr);
	const UCHAR* p = events + 1;
	const UCHAR* const end = events + events_length;
	bool flag = false;

	while (p < end) {

		const USHORT count = *p++;

		/* The data in the event block may have trailing blanks.  Strip them off. */

        const UCHAR* find_end = p + count;
		while (--find_end >= p && *find_end == ' ')
			; // nothing to do.
		const USHORT len = find_end - p + 1;
		EVNT event = find_event(len, reinterpret_cast<const char*>(p), parent);
		if (!event) {
			event = make_event(len, reinterpret_cast<const char*>(p), parent_offset);
			parent = (EVNT) SRQ_ABS_PTR(parent_offset);
			session = (SES) SRQ_ABS_PTR(session_id);
			request = (EVT_REQ) SRQ_ABS_PTR(request_offset);
			ptr = (SRQ_PTR *) SRQ_ABS_PTR(ptr_offset);
		}
		p += count;
		const SLONG event_offset = SRQ_REL_PTR(event);
		RINT interest, prior;
		if (interest = historical_interest(session, event_offset)) {
			for (SRQ_PTR* ptr2 = &session->ses_interests;
				 *ptr2 && (prior = (RINT) SRQ_ABS_PTR(*ptr2));
				 ptr2 = &prior->rint_next)
			{
				if (prior == interest) {
					*ptr2 = interest->rint_next;
					interest->rint_next = 0;
					break;
				}
			}
		}
		else {
			interest = (RINT) alloc_global(type_rint, (SLONG) sizeof(req_int), false);
			event = (EVNT) SRQ_ABS_PTR(event_offset);
			insert_tail(&event->evnt_interests, &interest->rint_interests);
			interest->rint_event = event_offset;

			parent = (EVNT) SRQ_ABS_PTR(parent_offset);
			request = (EVT_REQ) SRQ_ABS_PTR(request_offset);
			ptr = (SRQ_PTR *) SRQ_ABS_PTR(ptr_offset);
			session = (SES) SRQ_ABS_PTR(session_id);
		}
		*ptr = SRQ_REL_PTR(interest);
		ptr = &interest->rint_next;
		ptr_offset = SRQ_REL_PTR(ptr);
		interest->rint_request = request_offset;
		interest->rint_count = gds__vax_integer(p, 4);
		p += 4;
		if (interest->rint_count <= event->evnt_count)
			flag = true;
	}

	if (flag)
		post_process((PRB) SRQ_ABS_PTR(EVENT_process_offset));

	release();

	return id;
}


static EVH acquire()
{
/**************************************
 *
 *	a c q u i r e
 *
 **************************************
 *
 * Functional description
 *	Acquire exclusive access to shared global region.
 *
 **************************************/

	int mutex_state;
	if (mutex_state = ISC_mutex_lock(&MUTEX))
		mutex_bugcheck("mutex lock", mutex_state);
	EVENT_header->evh_current_process = EVENT_process_offset;

	if (EVENT_header->evh_length > EVENT_data.sh_mem_length_mapped)
	{
#if (!(defined SUPERSERVER) && (defined HAVE_MMAP))
		const SLONG length = EVENT_header->evh_length;
#endif

#ifdef WIN_NT
		/* Before remapping the memory, wakeup the watcher thread.
		 * Then remap the shared memory and allow the watcher thread
		 * to remap.
		 */
		/* Need to make the following code generic to all SUPERSERVER
		 * platforms. Postponed for now. B.Sriram, 10-Jul-1997
		 */

		PRB process = (PRB) SRQ_ABS_PTR(EVENT_process_offset);
		process->prb_flags |= PRB_remap;
		//event_t* event = &process->prb_event;

		post_process(process);

		while (true) {
			release();
			Sleep(3);
			acquire();

			process = (PRB) SRQ_ABS_PTR(EVENT_process_offset);
			if (!(process->prb_flags & PRB_remap))
				break;
		}
#endif /* WIN_NT */

	EVH header = NULL;
#if (!(defined SUPERSERVER) && (defined HAVE_MMAP))
		ISC_STATUS_ARRAY local_status;
		header = (evh*) ISC_remap_file(local_status, &EVENT_data, length, false);
#endif
		if (!header) {
			release();
			gds__log("acquire: Event table remap failed");
			exit(FINI_ERROR);
		}
		EVENT_header = header;

#ifdef WIN_NT
		process = (PRB) SRQ_ABS_PTR(EVENT_process_offset);
		process->prb_flags &= ~PRB_remap_over;
#endif /* WIN_NT */
	}

	return EVENT_header;
}


static FRB alloc_global(UCHAR type, ULONG length, bool recurse)
{
/**************************************
 *
 *	a l l o c _ g l o b a l
 *
 **************************************
 *
 * Functional description
 *	Allocate a block in shared global region.
 *
 **************************************/
	SRQ_PTR *ptr;
	FRB free;
	SLONG best_tail = MAX_SLONG;

	length = ROUNDUP(length, sizeof(IPTR));
	SRQ_PTR* best = NULL;

	for (ptr = &EVENT_header->evh_free; (free = (FRB) SRQ_ABS_PTR(*ptr)) && *ptr;
		ptr = &free->frb_next)
	{
		const SLONG tail = free->frb_header.hdr_length - length;
		if (tail >= 0 && (!best || tail < best_tail)) {
			best = ptr;
			best_tail = tail;
		}
	}

	if (!best && !recurse) {
		const SLONG old_length = EVENT_data.sh_mem_length_mapped;
		const SLONG ev_length = old_length + EVENT_EXTEND_SIZE;

#ifdef WIN_NT

		/* Before remapping the memory, wakeup the watcher thread.
		 * Then remap the shared memory and allow the watcher thread
		 * to remap.
		 */

		PRB process = (PRB) SRQ_ABS_PTR(EVENT_process_offset);
		process->prb_flags |= PRB_remap;
		//event_t* event = &process->prb_event;
		post_process(process);

		while (true) {
			release();
			Sleep(3);
			acquire();

			process = (PRB) SRQ_ABS_PTR(EVENT_process_offset);
			if (!(process->prb_flags & PRB_remap))
				break;
		}
#endif /* WIN_NT */

		EVH header = 0;
#if !((defined SUPERSERVER) && (defined HAVE_MMAP))
		ISC_STATUS_ARRAY local_status;
		header = reinterpret_cast<EVH>(ISC_remap_file(local_status, &EVENT_data, ev_length, true));
#endif
		if (header) {
			free = (FRB) ((UCHAR *) header + old_length);
/**
	free->frb_header.hdr_length = EVENT_EXTEND_SIZE - sizeof (struct evh);
**/
			free->frb_header.hdr_length = EVENT_data.sh_mem_length_mapped - old_length;
			free->frb_header.hdr_type = type_frb;
			free->frb_next = 0;

			EVENT_header = header;
			EVENT_header->evh_length = EVENT_data.sh_mem_length_mapped;

			free_global(free);

#ifdef WIN_NT
			process = (PRB) SRQ_ABS_PTR(EVENT_process_offset);
			process->prb_flags &= ~PRB_remap_over;
#endif /* WIN_NT */

			return alloc_global(type, length, true);
		}
	}

	if (!best) {
		release();
		gds__log("alloc_global: Event table space exhausted");
		exit(FINI_ERROR);
	}

	free = (FRB) SRQ_ABS_PTR(*best);

	if (best_tail < (SLONG) sizeof(frb))
		*best = free->frb_next;
	else {
		free->frb_header.hdr_length -= length;
		free = (FRB) ((UCHAR *) free + free->frb_header.hdr_length);
		free->frb_header.hdr_length = length;
	}

	memset((UCHAR*) free + sizeof(event_hdr), 0, free->frb_header.hdr_length - sizeof(event_hdr));
	free->frb_header.hdr_type = type;

	return free;
}


static SLONG create_process()
{
/**************************************
 *
 *	c r e a t e _ p r o c e s s
 *
 **************************************
 *
 * Functional description
 *	Create process block unless it already exists.
 *
 **************************************/
	if (EVENT_process_offset)
		return EVENT_process_offset;

	acquire();
	PRB process = (PRB) alloc_global(type_prb, (SLONG) sizeof(prb), false);
	insert_tail(&EVENT_header->evh_processes, &process->prb_processes);
	SRQ_INIT(process->prb_sessions);
	EVENT_process_offset = SRQ_REL_PTR(process);

	ISC_event_init(&process->prb_event);

#ifdef SOLARIS_MT
	ISC_STATUS_ARRAY local_status;
	EVENT_process = (PRB) ISC_map_object(local_status, &EVENT_data, EVENT_process_offset, sizeof(prb));
#endif

	process->prb_process_id = getpid();

	probe_processes();
	release();

	if (gds__thread_start(watcher_thread, NULL, THREAD_medium, 0, 0))
		ERR_bugcheck_msg("cannot start thread");

	return EVENT_process_offset;
}


static void delete_event(EVNT event)
{
/**************************************
 *
 *	d e l e t e _ e v e n t
 *
 **************************************
 *
 * Functional description
 *	Delete an unused and unloved event.
 *
 **************************************/
	remove_que(&event->evnt_events);

	if (event->evnt_parent) {
		EVNT parent = (EVNT) SRQ_ABS_PTR(event->evnt_parent);
		if (!--parent->evnt_count)
			delete_event(parent);
	}

	free_global((FRB) event);
}


static void delete_process(SLONG process_offset)
{
/**************************************
 *
 *	d e l e t e _ p r o c e s s
 *
 **************************************
 *
 * Functional description
 *	Delete a process block including friends and relations.
 *
 **************************************/
	PRB process = (PRB) SRQ_ABS_PTR(process_offset);

/* Delete any open sessions */

	while (!SRQ_EMPTY(process->prb_sessions)) {
		SES session = (SES) ((UCHAR *) SRQ_NEXT(process->prb_sessions) - OFFSET(SES, ses_sessions));
		delete_session(SRQ_REL_PTR(session));
	}

/* Untangle and release process block */

	remove_que(&process->prb_processes);
	free_global((FRB) process);

	if (EVENT_process_offset == process_offset) {
		/* Terminate the event watcher thread */
		/* When we come through the exit handler, the event semaphore might
		   have already been released by another exit handler.  So we cannot
		   use that semaphore to post the event.  Besides, the watcher thread
		   would be terminated anyway because the whole NLM is being unloaded. */
		// CVC: NLM??? is this Novell Netware specific code???

		process->prb_flags |= PRB_exiting;
		bool timeout = false;
		while (process->prb_flags & PRB_exiting && !timeout) {
			ISC_event_post(&process->prb_event);
			const SLONG value = ISC_event_clear(&process->prb_event);
			release();
#ifdef SOLARIS_MT
			event_t* event = &EVENT_process->prb_event;
#else
			event_t* event = &process->prb_event;
#endif
			timeout = ISC_event_wait(event, value, 5 * 1000000) == FB_FAILURE;
			acquire();
		}
		EVENT_process_offset = 0;
	}
}


static void delete_request(EVT_REQ request)
{
/**************************************
 *
 *	d e l e t e _ r e q u e s t
 *
 **************************************
 *
 * Functional description
 *	Release an unwanted and unloved request.
 *
 **************************************/
	SES session = (SES) SRQ_ABS_PTR(request->req_session);

	while (request->req_interests) {
		RINT interest = (RINT) SRQ_ABS_PTR(request->req_interests);

		request->req_interests = interest->rint_next;
		if (historical_interest(session, interest->rint_event)) {
			remove_que(&interest->rint_interests);
			free_global((FRB) interest);
		}
		else {
			interest->rint_next = session->ses_interests;
			session->ses_interests = SRQ_REL_PTR(interest);
			interest->rint_request = (SRQ_PTR)0;
		}
	}

	remove_que(&request->req_requests);
	free_global((FRB) request);
}


static void delete_session(SLONG session_id)
{
/**************************************
 *
 *	d e l e t e _ s e s s i o n
 *
 **************************************
 *
 * Functional description
 *	Delete a session.
 *
 **************************************/
	SES session = (SES) SRQ_ABS_PTR(session_id);

	// if session currently delivered events, delay its deletion until deliver ends
	if (session->ses_flags & SES_delivering)
	{
		session->ses_flags |= SES_purge;

		// give a chance for delivering thread to detect SES_purge flag we just set
		release();
		THREAD_SLEEP(100);
		acquire();

		return;
	}

/* Delete all requests */

	while (!SRQ_EMPTY(session->ses_requests)) {
		srq requests = session->ses_requests;
		EVT_REQ request = (EVT_REQ) ((UCHAR *) SRQ_NEXT(requests) - OFFSET(EVT_REQ, req_requests));
		delete_request(request);
	}

/* Delete any historical interests */

	while (session->ses_interests) {
		RINT interest = (RINT) SRQ_ABS_PTR(session->ses_interests);
		EVNT event = (EVNT) SRQ_ABS_PTR(interest->rint_event);
		session->ses_interests = interest->rint_next;
		remove_que(&interest->rint_interests);
		free_global((FRB) interest);
		if (SRQ_EMPTY(event->evnt_interests))
			delete_event(event);
	}

	remove_que(&session->ses_sessions);
	free_global((FRB) session);
}


static void deliver()
{
/**************************************
 *
 *	d e l i v e r
 *
 **************************************
 *
 * Functional description
 *	We've been poked -- deliver any satisfying requests.
 *
 **************************************/
	acquire();
	PRB process = (PRB) SRQ_ABS_PTR(EVENT_process_offset);
	process->prb_flags &= ~PRB_pending;

	srq* que2 = SRQ_NEXT(process->prb_sessions);
	while (que2 != &process->prb_sessions)
	{
		SES session = (SES) ((UCHAR *) que2 - OFFSET(SES, ses_sessions));
		session->ses_flags |= SES_delivering;
		const SLONG session_offset = SRQ_REL_PTR(session);
		const SLONG que2_offset = SRQ_REL_PTR(que2);
		for (bool flag = true; flag;) {
			flag = false;
			srq* event_srq;
			SRQ_LOOP(session->ses_requests, event_srq) {
				EVT_REQ request = (EVT_REQ) ((UCHAR *) event_srq - OFFSET(EVT_REQ, req_requests));
				if (request_completed(request)) {
					deliver_request(request);
					process = (PRB) SRQ_ABS_PTR(EVENT_process_offset);
					session = (SES) SRQ_ABS_PTR(session_offset);
					que2 = (srq *) SRQ_ABS_PTR(que2_offset);
					flag = !(session->ses_flags & SES_purge);
					break;
				}
			}
		}
		session->ses_flags &= ~SES_delivering;
		if (session->ses_flags & SES_purge)
		{
			que2 = SRQ_NEXT((*que2));
			delete_session(SRQ_REL_PTR(session));
			break;
		}
		else
			que2 = SRQ_NEXT((*que2));
	}

	release();
}


static void deliver_request(EVT_REQ request)
{
/**************************************
 *
 *	d e l i v e r _ r e q u e s t
 *
 **************************************
 *
 * Functional description
 *	Request has been satisfied, send updated event block to user, then
 *	Clean up request.
 *
 **************************************/
	UCHAR buffer[512];

	FPTR_EVENT_CALLBACK ast = request->req_ast;
	void* arg = request->req_ast_arg;
	UCHAR* event_buffer = buffer;
	UCHAR* p = buffer;
	const UCHAR* end = buffer + sizeof(buffer);
	*p++ = EPB_version1;

/* Loop thru interest block picking up event name, counts, and unlinking
   stuff */

	RINT interest;
	for (SRQ_PTR next = request->req_interests;
		 next && (interest = (RINT) SRQ_ABS_PTR(next));
		 next = interest->rint_next)
	{
		interest = (RINT) SRQ_ABS_PTR(next);
		EVNT event = (EVNT) SRQ_ABS_PTR(interest->rint_event);
		if (end < p + event->evnt_length + 5)
		{
			/* Running out of space - allocate some more and copy it over */
			fb_assert(event_buffer == buffer);	/* we're in this block only once */
			// CVC: We don't check at run-time in the release version, it's
			// likely to produce a buffer overrun if we reach MAX_EVENT_BUFFER,
			// since new_buffer will be again and again this number. FIXED.
			UCHAR* new_buffer = 0;
			if (event_buffer == buffer)
				new_buffer = (UCHAR*) gds__alloc((SLONG) MAX_EVENT_BUFFER);
			/* FREE: at procedure exit */
			if (!new_buffer)
			{	/* NOMEM: */
				gds__log("failed to post all events");
				break;			/* exit loop and send what we have */
			}
			event_buffer = new_buffer;
			memcpy(event_buffer, buffer, p - buffer);
			p = event_buffer + (p - buffer);
			end = event_buffer + MAX_EVENT_BUFFER;
		}
		*p++ = event->evnt_length;
		memcpy(p, event->evnt_name, event->evnt_length);
		p += event->evnt_length;
		const SLONG count = event->evnt_count + 1;
		*p++ = (UCHAR) (count);
		*p++ = (UCHAR) (count >> 8);
		*p++ = (UCHAR) (count >> 16);
		*p++ = (UCHAR) (count >> 24);
	}

	delete_request(request);
	release();
	(*ast)(arg, p - event_buffer, event_buffer);
	if (event_buffer != buffer)
	{
		gds__free(event_buffer);
	}
	acquire();
}


static void exit_handler(void* arg)
{
/**************************************
 *
 *	e x i t _ h a n d l e r
 *
 **************************************
 *
 * Functional description
 *	Cleanup on exit.
 *
 **************************************/

	eventStartup.cleanup();
}


static EVNT find_event(USHORT length, const TEXT* string, EVNT parent)
{
/**************************************
 *
 *	f i n d _ e v e n t
 *
 **************************************
 *
 * Functional description
 *	Lookup an event.
 *
 **************************************/
	SRQ_PTR parent_offset = (parent) ? SRQ_REL_PTR(parent) : 0;

	srq* event_srq;
	SRQ_LOOP(EVENT_header->evh_events, event_srq) {
		EVNT event = (EVNT) ((UCHAR *) event_srq - OFFSET(EVNT, evnt_events));
		if (event->evnt_parent == parent_offset && event->evnt_length == length &&
			!memcmp(string, event->evnt_name, length))
		{
			return event;
		}
	}

	return NULL;
}


static void free_global(FRB block)
{
/**************************************
 *
 *	f r e e _ g l o b a l
 *
 **************************************
 *
 * Functional description
 *	Free a previous allocated block.
 *
 **************************************/
	SRQ_PTR *ptr;
	FRB free;

	FRB prior = NULL;
	SRQ_PTR offset = SRQ_REL_PTR(block);
	block->frb_header.hdr_type = type_frb;

	for (ptr = &EVENT_header->evh_free; (free = (FRB) SRQ_ABS_PTR(*ptr)) && *ptr;
		 prior = free, ptr = &free->frb_next)
	{
		if ((SCHAR *) block < (SCHAR *) free)
			break;
	}

	if (offset <= 0 || offset > EVENT_header->evh_length ||
		(prior && (UCHAR*) block < (UCHAR*) prior + prior->frb_header.hdr_length))
	{
		punt("free_global: bad block");
		return;
	}

/* Start by linking block into chain */

	block->frb_next = *ptr;
	*ptr = offset;

/* Try to merge free block with next block */

	if (free && (SCHAR *) block + block->frb_header.hdr_length == (SCHAR *) free)
	{
		block->frb_header.hdr_length += free->frb_header.hdr_length;
		block->frb_next = free->frb_next;
	}

/* Next, try to merge the free block with the prior block */

	if (prior && (SCHAR *) prior + prior->frb_header.hdr_length ==	(SCHAR *) block)
	{
		prior->frb_header.hdr_length += block->frb_header.hdr_length;
		prior->frb_next = block->frb_next;
	}
}


static RINT historical_interest(SES session, SRQ_PTR event)
{
/**************************************
 *
 *	h i s t o r i c a l _ i n t e r e s t
 *
 **************************************
 *
 * Functional description
 *	Find a historical interest, if any, of an event with a session.
 *
 **************************************/
	RINT interest;

	for (SRQ_PTR ptr = session->ses_interests;
		 ptr && (interest = (RINT) SRQ_ABS_PTR(ptr)); ptr = interest->rint_next)
	{
		if (interest->rint_event == event)
			return interest;
	}

	return NULL;
}


static void init_shmem(void* arg, SH_MEM shmem_data, bool initialize)
{
/**************************************
 *
 *	i n i t _ s h m e m
 *
 **************************************
 *
 * Functional description
 *	Initialize global region header.
 *
 **************************************/
	int mutex_state;

#if defined(WIN_NT)
	char buffer[MAXPATHLEN];
	gds__prefix_lock(buffer, EVENT_FILE);
	if ( (mutex_state = ISC_mutex_init(&MUTEX, buffer)) )
		mutex_bugcheck("mutex init", mutex_state);
#endif

	if (!initialize)
		return;

	EVENT_header = (EVH) shmem_data->sh_mem_address;
	EVENT_header->evh_length = EVENT_data.sh_mem_length_mapped;
	EVENT_header->evh_version = EVENT_VERSION;
	EVENT_header->evh_request_id = 0;
	SRQ_INIT(EVENT_header->evh_processes);
	SRQ_INIT(EVENT_header->evh_events);

#if !defined(WIN_NT)
	if ( (mutex_state = ISC_mutex_init(&MUTEX)) )
		mutex_bugcheck("mutex init", mutex_state);
#endif

	FRB free = (FRB) ((UCHAR*) EVENT_header + sizeof(evh));
	free->frb_header.hdr_length = EVENT_data.sh_mem_length_mapped - sizeof(evh);
	free->frb_header.hdr_type = type_frb;
	free->frb_next = 0;

	EVENT_header->evh_free = (UCHAR *) free - (UCHAR *) EVENT_header;
}


static void insert_tail(srq * event_srq, srq * node)
{
/**************************************
 *
 *	i n s e r t _ t a i l
 *
 **************************************
 *
 * Functional description
 *	Insert a node at the tail of a event_srq.
 *
 **************************************/
	node->srq_forward = SRQ_REL_PTR(event_srq);
	node->srq_backward = event_srq->srq_backward;

	srq* prior = (srq *) SRQ_ABS_PTR(event_srq->srq_backward);
	prior->srq_forward = SRQ_REL_PTR(node);
	event_srq->srq_backward = SRQ_REL_PTR(node);
}


static EVNT make_event(USHORT length, const TEXT* string, SLONG parent_offset)
{
/**************************************
 *
 *	m a k e _ e v e n t
 *
 **************************************
 *
 * Functional description
 *	Allocate an link in an event.
 *
 **************************************/
	EVNT event = (EVNT) alloc_global(type_evnt, (SLONG) (sizeof(evnt) + length), false);
	insert_tail(&EVENT_header->evh_events, &event->evnt_events);
	SRQ_INIT(event->evnt_interests);

	if (parent_offset) {
		event->evnt_parent = parent_offset;
		EVNT parent = (EVNT) SRQ_ABS_PTR(parent_offset);
		++parent->evnt_count;
	}

	event->evnt_length = length;
	memcpy(event->evnt_name, string, length);

	return event;
}


static void mutex_bugcheck(const TEXT* string, int mutex_state)
{
/**************************************
 *
 *	m u t e x _ b u g c h e c k
 *
 **************************************
 *
 * Functional description
 *	There has been a bugcheck during a mutex operation.
 *	Post the bugcheck.
 *
 **************************************/
	TEXT msg[128];

	sprintf(msg, "EVENT: %s error, status = %d", string, mutex_state);
	gds__log(msg);

	fprintf(stderr, "%s\n", msg);
	exit(FINI_ERROR);
}


static void post_process(PRB process)
{
/**************************************
 *
 *	p o s t _ p r o c e s s
 *
 **************************************
 *
 * Functional description
 *	Wakeup process.
 *
 **************************************/

	process->prb_flags &= ~PRB_wakeup;
	process->prb_flags |= PRB_pending;
	release();
	ISC_event_post(&process->prb_event);
	acquire();
}


static void probe_processes()
{
/**************************************
 *
 *	p r o b e _ p r o c e s s e s
 *
 **************************************
 *
 * Functional description
 *	Probe a process to see if it still exists.  If it doesn't, get
 *	rid of it.
 *
 **************************************/
	srq* event_srq;

	SRQ_LOOP(EVENT_header->evh_processes, event_srq) {
		PRB process = (PRB) ((UCHAR *) event_srq - OFFSET(PRB, prb_processes));
		const SLONG process_offset = SRQ_REL_PTR(process);
		if (process_offset != EVENT_process_offset &&
			!ISC_check_process_existence(process->prb_process_id))
		{
			event_srq = (srq *) SRQ_ABS_PTR(event_srq->srq_backward);
			delete_process(process_offset);
		}
	}
}


static void punt(const TEXT* string)
{
/**************************************
 *
 *	p u n t
 *
 **************************************
 *
 * Functional description
 *
 **************************************/

	printf("(EVENT) punt: global region corrupt -- %s\n", string);
}


static void release()
{
/**************************************
 *
 *	r e l e a s e
 *
 **************************************
 *
 * Functional description
 *	Release exclusive control of shared global region.
 *
 **************************************/
	int mutex_state;

#ifdef DEBUG_EVENT
	validate();
#endif

	EVENT_header->evh_current_process = 0;
	if (mutex_state = ISC_mutex_unlock(&MUTEX))
		mutex_bugcheck("mutex lock", mutex_state);
}


static void remove_que(srq* node)
{
/**************************************
 *
 *	r e m o v e _ q u e
 *
 **************************************
 *
 * Functional description
 *	Remove a node from a self-relative event_srq.
 *
 **************************************/
	srq* event_srq = (srq *) SRQ_ABS_PTR(node->srq_forward);
	event_srq->srq_backward = node->srq_backward;

	event_srq = (srq *) SRQ_ABS_PTR(node->srq_backward);
	event_srq->srq_forward = node->srq_forward;
	node->srq_forward = node->srq_backward = 0;
}


static bool request_completed(EVT_REQ request)
{
/**************************************
 *
 *	r e q u e s t _ c o m p l e t e d
 *
 **************************************
 *
 * Functional description
 *	See if request is completed.
 *
 **************************************/
	RINT interest;
	for (SRQ_PTR next = request->req_interests; next; next = interest->rint_next) {
		interest = (RINT) SRQ_ABS_PTR(next);
		EVNT event = (EVNT) SRQ_ABS_PTR(interest->rint_event);
		if (interest->rint_count <= event->evnt_count)
			return true;
	}

	return false;
}


#ifdef DEBUG_EVENT
static int validate()
{
/**************************************
 *
 *	v a l i d a t e
 *
 **************************************
 *
 * Functional description
 *	Make sure everything looks ok.
 *
 **************************************/
// Check consistency of global region (debugging only)

	SRQ_PTR next_free = 0;
	SLONG offset;

	for (offset = sizeof(evh); offset < EVENT_header->evh_length;
		offset += block->frb_header.hdr_length)
	{
		const event_hdr* block = (event_hdr*) SRQ_ABS_PTR(offset);
		if (!block->frb_header.hdr_length || !block->frb_header.hdr_type ||
			block->frb_header.hdr_type >= type_max)
		{
			punt("bad block length or type");
			break;
		}
		if (next_free)
		{
			if (offset == next_free)
				next_free = 0;
			else if (offset > next_free)
				punt("bad free chain");
		}
		if (block->frb_header.hdr_type == type_frb) {
			next_free = ((FRB) block)->frb_next;
			if (next_free >= EVENT_header->evh_length)
				punt("bad frb_next");
		}
	}

	if (offset != EVENT_header->evh_length)
		punt("bad block length");
}
#endif


static THREAD_ENTRY_DECLARE watcher_thread(THREAD_ENTRY_PARAM)
{
/**************************************
 *
 *	w a t c h e r _ t h r e a d
 *
 **************************************
 *
 * Functional description
 *	Wait for something to happen.
 *
 **************************************/
	while (EVENT_process_offset) {
		acquire();
		PRB process = (PRB) SRQ_ABS_PTR(EVENT_process_offset);
		process->prb_flags &= ~PRB_wakeup;

		if (process->prb_flags & PRB_exiting) {
			process->prb_flags &= ~PRB_exiting;
			ISC_event_post(&process->prb_event);
			release();
			break;
		}
#ifdef WIN_NT
		if (process->prb_flags & PRB_remap) {
			process->prb_flags |= PRB_remap_over;
			process->prb_flags &= ~PRB_remap;
			release();
			while (true) {
				Sleep(3);
				acquire();
				process = (PRB) SRQ_ABS_PTR(EVENT_process_offset);
				release();
				if (!(process->prb_flags & PRB_remap_over))
					break;
			}
			acquire();
			process = (PRB) SRQ_ABS_PTR(EVENT_process_offset);
		}
#endif

		const SLONG value = ISC_event_clear(&process->prb_event);
		release();
		deliver();
		acquire();
		process = (PRB) SRQ_ABS_PTR(EVENT_process_offset);
		release();
#ifdef SOLARIS_MT
		event_t* event = &EVENT_process->prb_event;
#else
		event_t* event = &process->prb_event;
#endif
		ISC_event_wait(event, value, 0);
	}
	return 0;
}