that's the wins replication daemon !

there are still some work to do on it but it's already functionnal.

        J.F.
(This used to be commit 2506c98d19263bd5f367a488c2238dcdfec46ee9)

1 files changed, 935 insertions, 0 deletions

diff --git a/source3/wrepld/process.c b/source3/wrepld/process.c
new file mode 100644
index 0000000000..149ff3cd35
--- /dev/null
+++ b/source3/wrepld/process.c
@@ -0,0 +1,935 @@
+/* 
+   Unix SMB/Netbios implementation.
+   Version 1.9.
+   process incoming packets - main loop
+   Copyright (C) Jean François Micouleau      1998-2002.
+   
+   This program 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 of the License, or
+   (at your option) any later version.
+   
+   This program 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 "includes.h"
+#include "wins_repl.h"
+
+extern fd_set *listen_set;
+extern int listen_number;
+extern int *sock_array;
+
+WINS_OWNER global_wins_table[64][64];
+int partner_count;
+
+TALLOC_CTX *mem_ctx;
+
+#define WINS_LIST "wins.tdb"
+#define INFO_VERSION	"INFO/version"
+#define INFO_COUNT	"INFO/num_entries"
+#define INFO_ID_HIGH	"INFO/id_high"
+#define INFO_ID_LOW	"INFO/id_low"
+#define ENTRY_PREFIX 	"ENTRY/"
+
+
+/*******************************************************************
+fill the header of a reply.
+********************************************************************/
+static void fill_header(GENERIC_PACKET *g, int opcode, int ctx, int mess)
+{
+	if (g==NULL)
+		return;
+
+	g->header.opcode=opcode;
+	g->header.assoc_ctx=ctx;
+	g->header.mess_type=mess;
+}
+
+/*******************************************************************
+dump the global table, that's a debug code.
+********************************************************************/
+static void dump_global_table(void)
+{
+	int i,j;
+	
+	for (i=0;i<partner_count;i++) {
+		DEBUG(0,("\n%d ", i));
+		for (j=0; global_wins_table[i][j].address.s_addr!=0; j++)
+			DEBUG(0,("%s:%d \t", inet_ntoa(global_wins_table[i][j].address),
+				(int)global_wins_table[i][j].max_version));
+	}
+	DEBUG(0,("\n"));
+}
+
+/*******************************************************************
+start association
+********************************************************************/
+static void start_assoc_process(GENERIC_PACKET *q, GENERIC_PACKET *r)
+{
+	/*
+	 * add this request to our current wins partners list
+	 * this list is used to know with who we are in contact
+	 *
+	 */
+	r->sa_rp.assoc_ctx=time(NULL);
+	fill_header(r, OPCODE_NON_NBT, q->sa_rq.assoc_ctx, MESSAGE_TYPE_START_ASSOC_REPLY);
+
+	/* reply we are a NT4 server */
+	
+	/* w2K is min=2, maj=5 */
+	
+	r->sa_rp.min_ver=1;
+	r->sa_rp.maj_ver=1;
+
+	add_partner(r->sa_rp.assoc_ctx, q->sa_rq.assoc_ctx, False, False);
+}
+
+/*******************************************************************
+start association reply
+********************************************************************/
+static void start_assoc_reply(GENERIC_PACKET *q, GENERIC_PACKET *r)
+{
+	int i;
+
+	/* check if we have already registered this client */
+	if (!check_partner(q->header.assoc_ctx)) {
+		DEBUG(0,("start_assoc_reply: unknown client\n"));
+		stop_packet(q, r, STOP_REASON_USER_REASON);
+		return;
+	}
+
+	if (!update_server_partner(q->header.assoc_ctx, q->sa_rp.assoc_ctx)) {
+		DEBUG(0,("start_assoc_reply: can't update server ctx\n"));
+		stop_packet(q, r, STOP_REASON_USER_REASON);
+		return;
+	}
+
+	/* if pull, request map table */	
+	if (check_pull_partner(q->header.assoc_ctx)) {
+		fill_header(r, OPCODE_NON_NBT, get_server_assoc(q->header.assoc_ctx), MESSAGE_TYPE_REPLICATE);
+
+		r->rep.msg_type=MESSAGE_REP_ADD_VERSION_REQUEST;
+		DEBUG(0,("start_assoc_reply: requesting map table\n"));
+
+		return;
+	}
+
+	/* if push, send our table */
+	if (check_push_partner(q->header.assoc_ctx)) {
+		fill_header(r, OPCODE_NON_NBT, get_server_assoc(q->header.assoc_ctx), MESSAGE_TYPE_REPLICATE);
+		r->rep.msg_type=MESSAGE_REP_UPDATE_NOTIFY_REQUEST;
+		r->rep.un_rq.partner_count=partner_count;
+		
+		r->rep.un_rq.wins_owner=(WINS_OWNER *)talloc(mem_ctx, partner_count*sizeof(WINS_OWNER));
+		if (r->rep.un_rq.wins_owner==NULL) {
+			DEBUG(0,("start_assoc_reply: can't alloc memory\n"));
+			stop_packet(q, r, STOP_REASON_USER_REASON);
+			return;
+		}
+
+		for (i=0; i<partner_count; i++)
+			r->rep.un_rq.wins_owner[i]=global_wins_table[0][i];
+		
+		DEBUG(0,("start_assoc_reply: sending update table\n"));
+		return;
+	}
+	
+	/* neither push/pull, stop */
+	/* we should not come here */
+	DEBUG(0,("we have a partner which is neither push nor pull !\n"));
+	stop_packet(q, r, STOP_REASON_USER_REASON);
+}
+
+/****************************************************************************
+initialise and fill the in-memory partner table.
+****************************************************************************/
+int init_wins_partner_table(void)
+{
+	int i=1,j=0,k;
+	char **partner = lp_list_make(lp_wins_partners());
+
+	DEBUG(0, ("init_wins_partner_table: partners: %s\n", lp_wins_partners()));
+
+	global_wins_table[0][0].address=*iface_n_ip(0);
+	global_wins_table[0][0].max_version=0;
+	global_wins_table[0][0].min_version=0;
+	global_wins_table[0][0].type=0;
+
+	while (partner[j]!=NULL) {
+		DEBUG(0,("init_wins_partner_table, adding partner: %s\n", partner[j]));
+		
+		global_wins_table[0][i].address=*interpret_addr2(partner[j]);
+		global_wins_table[0][i].max_version=0;
+		global_wins_table[0][i].min_version=0;
+		global_wins_table[0][i].type=0;
+		global_wins_table[0][i].last_pull=0;
+		global_wins_table[0][i].last_push=0;
+
+		i++;
+		j++;
+	}
+	
+	for (k=1; k<i;k++)
+		for (j=0; j<i; j++)
+			global_wins_table[k][j]=global_wins_table[0][j];
+	
+	lp_list_free (&partner);
+	
+	return i;
+}
+
+/****************************************************************************
+read the last ID from the wins tdb file.
+****************************************************************************/
+static void get_our_last_id(WINS_OWNER *wins_owner)
+{
+	TDB_CONTEXT *tdb;
+
+	tdb = tdb_open_log(lock_path(WINS_LIST), 0, TDB_DEFAULT, O_RDONLY, 0600);
+	if (!tdb) {
+		DEBUG(2,("get_our_last_id: Can't open wins database file %s. Error was %s\n", WINS_LIST, strerror(errno) ));
+		return;
+	}
+	
+	wins_owner->max_version=((SMB_BIG_UINT)tdb_fetch_int(tdb, INFO_ID_HIGH))<<32 | 
+				 (SMB_BIG_UINT)tdb_fetch_int(tdb, INFO_ID_LOW);
+
+	tdb_close(tdb);
+}
+
+/****************************************************************************
+send the list of wins server we know.
+****************************************************************************/
+static void send_version_number_map_table(GENERIC_PACKET *q, GENERIC_PACKET *r)
+{
+	int i;
+	int s_ctx=get_server_assoc(q->header.assoc_ctx);
+
+	if (s_ctx==0) {
+		DEBUG(0, ("send_entry_reply: request for a partner not in our table\n"));
+		stop_packet(q, r, STOP_REASON_USER_REASON);
+		return;
+	}
+
+	/*
+	 * return an array of wins servers, we are partner with.
+	 * each entry contains the IP address and the version info
+	 * version: ID of the last entry we've got
+	 */
+
+	/* the first wins server must be self */
+
+	/*
+	 * get our last ID from the wins database
+	 * it can have been updated since last read
+	 * as nmbd got registration/release.
+	 */ 
+	get_our_last_id(&global_wins_table[0][0]);
+
+	r->rep.avmt_rep.wins_owner=(WINS_OWNER *)talloc(mem_ctx, partner_count*sizeof(WINS_OWNER));
+	if (r->rep.avmt_rep.wins_owner==NULL) {
+		stop_packet(q, r, STOP_REASON_USER_REASON);
+		return;
+	}
+	
+	DEBUG(0,("send_version_number_map_table: partner_count: %d\n", partner_count));
+
+	for (i=0; i<partner_count; i++) {
+		DEBUG(0,("send_version_number_map_table, partner: %d -> %s, \n", i, inet_ntoa(global_wins_table[0][i].address)));
+		r->rep.avmt_rep.wins_owner[i]=global_wins_table[0][i];
+	}
+	
+	r->rep.msg_type=1;
+	r->rep.avmt_rep.partner_count=partner_count;
+	r->rep.avmt_rep.initiating_wins_server.s_addr=0; /* blatant lie, NT4/w2K do the same ! */
+	fill_header(r, OPCODE_NON_NBT, s_ctx, MESSAGE_TYPE_REPLICATE);
+}
+
+/****************************************************************************
+for a given partner, ask it to send entries we don't have.
+****************************************************************************/
+static BOOL check_partners_and_send_entries(GENERIC_PACKET *q, GENERIC_PACKET *r, int partner)
+{
+	int server;
+	int other;
+	SMB_BIG_UINT temp;
+	SMB_BIG_UINT current;
+
+
+	/*
+	 * we check if our partner has more records than us.
+	 * we need to check more than our direct partners as
+	 * we can have this case:
+	 * us: A, partners: B,C, indirect partner: D
+	 * A<->B, A<->C, B<->D, C<->D
+	 *
+	 * So if we're talking to B, we need to check if between
+	 * B and C, which one have more records about D.
+	 * and also check if we don't already have the records.
+	 */
+
+
+	 /* check all servers even indirect */
+	 for (server=1; global_wins_table[0][server].address.s_addr!=0; server++) {
+		current = global_wins_table[partner][server].max_version;
+		
+		temp=0;
+		
+		for (other=1; other<partner_count; other++) {
+			/* skip the partner itself */
+			if (other==partner)
+				continue;
+
+			if (global_wins_table[other][server].max_version > temp)
+				temp=global_wins_table[other][server].max_version;
+		}
+		
+		if (current >= temp && current > global_wins_table[0][server].max_version) {
+			/* 
+			 * it has more records than every body else and more than us,
+			 * ask it the difference between what we have and what it has
+			 */
+			fill_header(r, OPCODE_NON_NBT, get_server_assoc(q->header.assoc_ctx), MESSAGE_TYPE_REPLICATE);
+
+			r->rep.msg_type=MESSAGE_REP_SEND_ENTRIES_REQUEST;
+			r->rep.se_rq.wins_owner.address=global_wins_table[partner][server].address;
+			
+			r->rep.se_rq.wins_owner.max_version=global_wins_table[partner][server].max_version;
+			r->rep.se_rq.wins_owner.min_version=global_wins_table[0][server].max_version;
+			r->rep.se_rq.wins_owner.type=0;
+			
+			write_server_assoc_table(q->header.assoc_ctx, global_wins_table[0][partner].address, global_wins_table[partner][server].address);
+			
+			/*
+			 * and we update our version for this server
+			 * as we can't use the IDs returned in the send_entries function
+			 * the max ID can be larger than the largest ID returned
+			 */
+			
+			global_wins_table[0][server].max_version=global_wins_table[partner][server].max_version;
+
+			return True;
+		}
+	}
+	return False;
+}
+	
+/****************************************************************************
+receive the list of wins server we know.
+****************************************************************************/
+static void receive_version_number_map_table(GENERIC_PACKET *q, GENERIC_PACKET *r)
+{
+	fstring peer;
+	struct in_addr addr;
+	int i,j,k,l;
+	int s_ctx=get_server_assoc(q->header.assoc_ctx);
+
+	if (s_ctx==0) {
+		DEBUG(0, ("receive_version_number_map_table: request for a partner not in our table\n"));
+		stop_packet(q, r, STOP_REASON_USER_REASON);
+		return;
+	}
+
+	fstrcpy(peer,get_socket_addr(q->fd));
+	addr=*interpret_addr2(peer);
+
+	get_our_last_id(&global_wins_table[0][0]);
+	
+	DEBUG(0,("receive_version_number_map_table: received a map of %d server from: %s\n", 
+	          q->rep.avmt_rep.partner_count ,inet_ntoa(q->rep.avmt_rep.initiating_wins_server)));
+	DEBUG(0,("real peer is: %s\n", peer));
+
+	for (i=0; global_wins_table[0][i].address.s_addr!=addr.s_addr && i<partner_count;i++)
+		;
+
+	if (i==partner_count) {
+		DEBUG(0,("receive_version_number_map_table: unknown partner: %s\n", peer));
+		stop_packet(q, r, STOP_REASON_USER_REASON);
+		return;
+	}
+
+	for (j=0; j<q->rep.avmt_rep.partner_count;j++) {
+		/*
+		 * search if we already have this entry or if it's a new one
+		 * it can be a new one in case of propagation
+		 */
+		for (k=0; global_wins_table[0][k].address.s_addr!=0 && 
+			  global_wins_table[0][k].address.s_addr!=q->rep.avmt_rep.wins_owner[j].address.s_addr; k++);
+
+		global_wins_table[i][k].address.s_addr=q->rep.avmt_rep.wins_owner[j].address.s_addr;
+		global_wins_table[i][k].max_version=q->rep.avmt_rep.wins_owner[j].max_version;
+		global_wins_table[i][k].min_version=q->rep.avmt_rep.wins_owner[j].min_version;
+		global_wins_table[i][k].type=q->rep.avmt_rep.wins_owner[j].type;
+		
+		/*
+		 * in case it's a new one, rewrite the address for all the partner
+		 * to reserve the slot.
+		 */
+
+		for(l=0; l<partner_count; l++)
+			global_wins_table[l][k].address.s_addr=q->rep.avmt_rep.wins_owner[j].address.s_addr;
+	}
+
+	dump_global_table();
+
+	/*
+	 * if this server have newer records than what we have
+	 * for several wins servers, we need to ask it.
+	 * Alas a send entry request is only on one server.
+	 * So in the send entry reply, we'll ask for the next server is required.
+	 */
+
+	if (check_partners_and_send_entries(q, r, i))
+		return;
+
+	/* it doesn't have more entries than us */
+	stop_packet(q, r, STOP_REASON_USER_REASON);
+}
+
+/****************************************************************************
+add an entry to the wins list we'll send.
+****************************************************************************/
+static BOOL add_record_to_winsname(WINS_NAME **wins_name, int *max_names, char *name, int type, int wins_flags, int id, struct in_addr *ip_list, int num_ips)
+{
+	WINS_NAME *temp_list;
+	int i;
+	int current=*max_names;
+
+	temp_list=talloc_realloc(mem_ctx, *wins_name, (current+1)*sizeof(WINS_NAME));
+	if (temp_list==NULL)
+		return False;
+
+	temp_list[current].name_len=0x11;
+	
+	safe_strcpy(temp_list[current].name, name, 15);
+
+	temp_list[current].type=type;
+	temp_list[current].empty=0;
+
+	temp_list[current].name_flag=wins_flags;
+
+	if ( (wins_flags&0x03) == 1 || (wins_flags&0x03)==2)
+		temp_list[current].group_flag=0x01000000;
+	else
+		temp_list[current].group_flag=0x00000000;
+	
+	temp_list[current].id=id;
+	
+	temp_list[current].owner.s_addr=ip_list[0].s_addr;
+
+	if (temp_list[current].name_flag & 2) {
+		temp_list[current].num_ip=num_ips;
+		temp_list[current].others=(struct in_addr *)talloc(mem_ctx, sizeof(struct in_addr)*num_ips);
+		if (temp_list[current].others==NULL)
+			return False;
+	
+		for (i=0; i<num_ips; i++)
+			temp_list[current].others[i].s_addr=ip_list[i].s_addr;
+
+	} else 
+		temp_list[current].num_ip=1;
+
+	temp_list[current].foo=0xffffffff;
+
+	*wins_name=temp_list;
+	
+	return True;
+}
+
+/****************************************************************************
+send the list of name we have.
+****************************************************************************/
+static void send_entry_request(GENERIC_PACKET *q, GENERIC_PACKET *r)
+{
+	int max_names=0;
+	int i;
+	time_t time_now = time(NULL);
+	WINS_OWNER *wins_owner;
+	TDB_CONTEXT *tdb;
+	TDB_DATA kbuf, dbuf, newkey;
+	int s_ctx=get_server_assoc(q->header.assoc_ctx);
+	int num_interfaces = iface_count();
+
+	if (s_ctx==0) {
+		DEBUG(0, ("send_entry_request: request for a partner not in our table\n"));
+		stop_packet(q, r, STOP_REASON_USER_REASON);
+		return;
+	}
+
+
+	wins_owner=&q->rep.se_rq.wins_owner;
+	r->rep.se_rp.wins_name=NULL;
+
+	DEBUG(0,("send_entry_request: we have been asked to send the list of wins records\n"));
+	DEBUGADD(0,("owned by: %s and between min: %d and max: %d\n", inet_ntoa(wins_owner->address),
+		    (int)wins_owner->min_version, (int)wins_owner->max_version));
+
+	/*
+	 * if we are asked to send records owned by us
+	 * we overwrite the wins ip with 0.0.0.0
+	 * to make it easy in case of multihomed
+	 */
+
+	for (i=0; i<num_interfaces; i++)
+		if (ip_equal(wins_owner->address, *iface_n_ip(i))) {
+			wins_owner->address=*interpret_addr2("0.0.0.0");
+			break;
+		}
+
+
+	tdb = tdb_open_log(lock_path(WINS_LIST), 0, TDB_DEFAULT, O_RDONLY, 0600);
+	if (!tdb) {
+		DEBUG(2,("send_entry_request: Can't open wins database file %s. Error was %s\n", WINS_LIST, strerror(errno) ));
+		return;
+	}
+
+	for (kbuf = tdb_firstkey(tdb); 
+	     kbuf.dptr; 
+	     newkey = tdb_nextkey(tdb, kbuf), safe_free(kbuf.dptr), kbuf=newkey) {
+		pstring name_type, name, ip_str;
+		char *p;
+		int type = 0;
+		int nb_flags;
+		int ttl;
+		unsigned int num_ips;
+		int low, high;
+		SMB_BIG_UINT version;
+		struct in_addr wins_ip;
+		struct in_addr *ip_list;
+		int wins_flags;
+		int len;
+
+		if (strncmp(kbuf.dptr, ENTRY_PREFIX, strlen(ENTRY_PREFIX)) != 0)
+			continue;
+		
+		
+		dbuf = tdb_fetch(tdb, kbuf);
+		if (!dbuf.dptr)
+			continue;
+
+		fstrcpy(name_type, kbuf.dptr+strlen(ENTRY_PREFIX));
+		pstrcpy(name, name_type);
+
+		if((p = strchr(name,'#')) != NULL) {
+			*p = 0;
+			sscanf(p+1,"%x",&type);
+		}
+
+		len = tdb_unpack(dbuf.dptr, dbuf.dsize, "dddfddd",
+				&nb_flags,
+				&high,
+				&low,
+				ip_str,
+				&ttl, 
+				&num_ips,
+				&wins_flags);
+
+		wins_ip=*interpret_addr2(ip_str);
+
+ 		/* Allocate the space for the ip_list. */
+		if((ip_list = (struct in_addr *)talloc(mem_ctx,  num_ips * sizeof(struct in_addr))) == NULL) {
+			DEBUG(0,("initialise_wins: talloc fail !\n"));
+			return;
+		}
+
+		for (i = 0; i < num_ips; i++) {
+			len += tdb_unpack(dbuf.dptr+len, dbuf.dsize-len, "f", ip_str);
+			ip_list[i] = *interpret_addr2(ip_str);
+		}
+
+		/* add all entries that have 60 seconds or more to live */
+		if ((ttl - 60) > time_now || ttl == PERMANENT_TTL) {
+			if(ttl != PERMANENT_TTL)
+				ttl -= time_now;
+    
+			DEBUG( 4, ("send_entry_request: add name: %s#%02x ttl = %d first IP %s flags = %2x\n",
+			    name, type, ttl, inet_ntoa(ip_list[0]), nb_flags));
+
+			/* add the record to the list to send */
+			version=((SMB_BIG_UINT)high)<<32 | low;
+			
+			if (wins_owner->min_version<=version && wins_owner->max_version>=version &&
+			    wins_owner->address.s_addr==wins_ip.s_addr) {
+				if(!add_record_to_winsname(&r->rep.se_rp.wins_name, &max_names, name, type, wins_flags, version, ip_list, num_ips))
+					return;
+				max_names++;
+			}
+
+		} else {
+			DEBUG(4, ("send_entry_request: not adding name (ttl problem) %s#%02x ttl = %d first IP %s flags = %2x\n",
+				  name, type, ttl, inet_ntoa(ip_list[0]), nb_flags));
+		}
+	}
+    
+	tdb_close(tdb);
+
+	DEBUG(0,("send_entry_request, sending %d records\n", max_names));
+	fill_header(r, OPCODE_NON_NBT, s_ctx, MESSAGE_TYPE_REPLICATE);
+	r->rep.msg_type=MESSAGE_REP_SEND_ENTRIES_REPLY; /* reply */
+	r->rep.se_rp.max_names=max_names;
+}
+
+
+/****************************************************************************
+.
+****************************************************************************/
+static void update_notify_request(GENERIC_PACKET *q, GENERIC_PACKET *r)
+{
+	int i,j,k,l;
+	UPDATE_NOTIFY_REQUEST *u;
+	int s_ctx=get_server_assoc(q->header.assoc_ctx);
+	
+	if (s_ctx==0) {
+		DEBUG(0, ("send_entry_reply: request for a partner not in our table\n"));
+		stop_packet(q, r, STOP_REASON_USER_REASON);
+		return;
+	}
+
+	u=&q->rep.un_rq;
+
+	/* check if we already have the range of records */
+
+	DEBUG(0,("update_notify_request: wins server: %s offered this list of %d records:\n",
+		inet_ntoa(u->initiating_wins_server), u->partner_count));
+
+	get_our_last_id(&global_wins_table[0][0]);
+
+	for (i=0; i<partner_count; i++) {
+		if (global_wins_table[0][i].address.s_addr==u->initiating_wins_server.s_addr) {
+			DEBUG(0,("update_notify_request: found initiator at index %d\n", i));
+			break;
+		}
+	}
+
+	/*
+	 * some explanation is required, before someone say it's crap.
+	 *
+	 * let's take an example, we have 2 wins partners, we already now
+	 * that our max id is 10, partner 1 ID is 20 and partner 2 ID is 30
+	 * the array looks like:
+	 *
+	 * 	0	1	2
+	 * 0	10	20	30  
+	 * 1
+	 * 2
+	 *
+	 * we receive an update from partner 2 saying he has: 1:15, 2:40, 3:50
+	 * we must enlarge the array to add partner 3, it will look like:
+	 *
+	 * 	0	1	2	3
+	 * 0	10	20	30
+	 * 1
+	 * 2		15	40	50
+	 *
+	 * now we know, we should pull from partner 2, the records 30->40 of 2 and 0->50 of 3.
+	 * once the pull will be over, our table will look like:
+	 *
+	 * 	0	1	2	3
+	 * 0	10	20	40	50
+	 * 1
+	 * 2		15	40	50
+	 *
+	 *
+	 */
+
+	for (j=0; j<u->partner_count;j++) {
+		/*
+		 * search if we already have this entry or if it's a new one
+		 * it can be a new one in case of propagation
+		 */
+
+		for (k=0; global_wins_table[0][k].address.s_addr!=0 && 
+			  global_wins_table[0][k].address.s_addr!=u->wins_owner[j].address.s_addr; k++);
+
+		global_wins_table[i][k].address.s_addr=u->wins_owner[j].address.s_addr;
+		global_wins_table[i][k].max_version=u->wins_owner[j].max_version;
+		global_wins_table[i][k].min_version=u->wins_owner[j].min_version;
+		global_wins_table[i][k].type=u->wins_owner[j].type;
+		
+		/*
+		 * in case it's a new one, rewrite the address for all the partner
+		 * to reserve the slot.
+		 */
+
+		for(l=0; l<partner_count; l++)
+			global_wins_table[l][k].address.s_addr=u->wins_owner[j].address.s_addr;
+	}
+
+	dump_global_table();
+
+	stop_packet(q, r, STOP_REASON_USER_REASON);
+}
+
+/****************************************************************************
+.
+****************************************************************************/
+static void send_entry_reply(GENERIC_PACKET *q, GENERIC_PACKET *r)
+{
+	int i,j,k;
+	struct in_addr partner, server;
+	pid_t pid;
+	int s_ctx=get_server_assoc(q->header.assoc_ctx);
+	WINS_RECORD record;
+	
+	if (s_ctx==0) {
+		DEBUG(0, ("send_entry_reply: request for a partner not in our table\n"));
+		stop_packet(q, r, STOP_REASON_USER_REASON);
+		return;
+	}
+
+	DEBUG(0,("send_entry_reply:got %d new records\n", q->rep.se_rp.max_names));
+
+	/* we got records from a wins partner but that can be from another wins server */
+	/* hopefully we track that */
+
+	/* and the only doc available from MS is wrong ! */
+
+	get_server_assoc_table(q->header.assoc_ctx, &partner, &server);
+
+	for (j=0; global_wins_table[0][j].address.s_addr!=0; j++) {
+		if (global_wins_table[0][j].address.s_addr==server.s_addr) {
+			DEBUG(0,("send_entry_reply: found server at index %d\n", j));
+			break;
+		}
+	}
+
+	pid = pidfile_pid("nmbd");
+	if (pid == 0) {
+		DEBUG(0,("send_entry_reply: Can't find pid for nmbd\n"));
+		return;
+	}
+
+	for (k=0; k<q->rep.se_rp.max_names; k++) {
+		DEBUG(0,("send_entry_reply: %s<%02x> %d\n", q->rep.se_rp.wins_name[k].name, q->rep.se_rp.wins_name[k].type,
+		         (int)q->rep.se_rp.wins_name[k].id));
+
+		safe_strcpy(record.name, q->rep.se_rp.wins_name[k].name, 16);
+		record.type=q->rep.se_rp.wins_name[k].type;
+		record.id=q->rep.se_rp.wins_name[k].id;
+		record.wins_flags=q->rep.se_rp.wins_name[k].name_flag&0x00ff;
+		record.num_ips=q->rep.se_rp.wins_name[k].num_ip;
+
+		record.wins_ip.s_addr=server.s_addr;
+
+		if (record.num_ips==1)
+			record.ip[0]=q->rep.se_rp.wins_name[k].owner;
+		else
+			for (i=0; i<record.num_ips; i++)
+				record.ip[i]=q->rep.se_rp.wins_name[k].others[i];
+
+		record.nb_flags=0;
+
+		if (record.wins_flags&WINS_NGROUP || record.wins_flags&WINS_SGROUP)
+			record.nb_flags|=NB_GROUP;
+		
+		if (record.wins_flags&WINS_ACTIVE)
+			record.nb_flags|=NB_ACTIVE;
+		
+		record.nb_flags|=record.wins_flags&WINS_HNODE;
+		
+		message_send_pid(pid, MSG_WINS_NEW_ENTRY, &record, sizeof(record), False);
+
+	}
+
+	dump_global_table();
+
+	/*
+	 * we got some entries, 
+	 * ask the partner to send us the map table again
+	 * to get the other servers entries.
+	 *
+	 * we're getting the map table 1 time more than really
+	 * required. We could remove that call, but that
+	 * would complexify the code. I prefer this trade-of. 
+	 */
+	fill_header(r, OPCODE_NON_NBT, s_ctx, MESSAGE_TYPE_REPLICATE);
+
+	r->rep.msg_type=MESSAGE_REP_ADD_VERSION_REQUEST;
+}
+
+/****************************************************************************
+decode the replication message and reply.
+****************************************************************************/
+static void replicate(GENERIC_PACKET *q, GENERIC_PACKET *r)
+{
+	switch (q->rep.msg_type) {
+		case 0:
+			/* add version number map table request */
+			send_version_number_map_table(q, r);
+			break;
+		case 1:
+			receive_version_number_map_table(q, r);
+			break;
+		case 2:
+			/* send entry request */
+			send_entry_request(q, r);
+			break;
+		case 3:
+			/* send entry reply */
+			send_entry_reply(q, r);
+			break;
+		case 4:
+			/* update notification request */
+			update_notify_request(q, r);
+			break;
+	}
+}
+
+/****************************************************************************
+do a switch on the message type, and return the response size
+****************************************************************************/
+static BOOL switch_message(GENERIC_PACKET *q, GENERIC_PACKET *r)
+{
+	switch (q->header.mess_type) {
+		case 0:
+			/* Start association type */			
+			start_assoc_process(q, r);
+			return True;
+			break;
+		case 1:
+			/* start association reply */
+			start_assoc_reply(q, r);
+			return True;
+			break;
+		case 2:
+			/* stop association message */
+			/*
+			 * remove the partner from the list and 
+			 * reply false to NOT send a packet
+			 */
+			remove_partner(q->header.assoc_ctx);
+			return False;
+			break;
+		case 3:
+			/* replication message */
+			replicate(q, r);
+			return True;
+			break;
+	}
+
+	return False;
+}
+
+
+/****************************************************************************
+  construct a reply to the incoming packet
+****************************************************************************/
+void construct_reply(struct wins_packet_struct *p)
+{
+	GENERIC_PACKET r;
+	struct BUFFER buffer;
+
+	buffer.buffer=NULL;
+	buffer.offset=0;
+	buffer.length=0;
+
+	DEBUG(0,("dump: received packet\n"));
+	dump_generic_packet(p->packet);
+
+	/* Verify if the request we got is from a listed partner */
+	if (!check_partner(p->packet->header.assoc_ctx)) {
+		fstring peer;
+		struct in_addr addr;
+		int i;
+		fstrcpy(peer,get_socket_addr(p->fd));
+		addr=*interpret_addr2(peer);
+
+		for (i=1; i<partner_count; i++)
+			if (ip_equal(addr, global_wins_table[0][i].address))
+				break;
+
+		if (i==partner_count) {
+			DEBUG(0,("construct_reply: got a request from a non peer machine: %s\n", peer));
+			stop_packet(p->packet, &r, STOP_REASON_AUTH_FAILED);
+			p->stop_packet=True;
+			encode_generic_packet(&buffer, &r);
+			if (!send_smb(p->fd, buffer.buffer))
+				exit_server("process_smb: send_smb failed.");
+			return;
+		}
+	}
+
+	if (switch_message(p->packet, &r)) {
+		encode_generic_packet(&buffer, &r);
+		DEBUG(0,("dump: sending packet\n"));
+		dump_generic_packet(&r);
+
+		if(buffer.offset > 0) {
+			if (!send_smb(p->fd, buffer.buffer))
+				exit_server("process_smb: send_smb failed.");
+		}
+	}
+
+	/* if we got a stop assoc or if we send a stop assoc, close the fd after */
+	if (p->packet->header.mess_type==MESSAGE_TYPE_STOP_ASSOC || 
+	    r.header.mess_type==MESSAGE_TYPE_STOP_ASSOC)
+		p->stop_packet=True;
+}
+
+/****************************************************************************
+  contact periodically our wins partner to do a pull replication
+****************************************************************************/
+void run_pull_replication(time_t t)
+{
+	/* we pull every 30 minutes to query about new records*/
+	int i, s;
+	struct BUFFER buffer;
+	GENERIC_PACKET p;
+
+	buffer.buffer=NULL;
+	buffer.offset=0;
+	buffer.length=0;
+
+	for (i=1; i<partner_count; i++) {
+		if (global_wins_table[0][i].last_pull < t) {
+			global_wins_table[0][i].last_pull=t+30*60; /* next in 30 minutes */
+			
+			/* contact the wins server */
+			p.header.mess_type=MESSAGE_TYPE_START_ASSOC_REQUEST;
+			p.header.opcode=OPCODE_NON_NBT;
+			p.header.assoc_ctx=0;
+			p.sa_rq.assoc_ctx=(int)t;
+			p.sa_rq.min_ver=1;
+			p.sa_rq.maj_ver=1;
+			
+			DEBUG(3,("run_pull_replication: contacting wins server %s.\n", inet_ntoa(global_wins_table[0][i].address)));
+			encode_generic_packet(&buffer, &p);
+			dump_generic_packet(&p);
+
+			/* send the packet to the server and add the descriptor to receive answers */
+			s=open_socket_out(SOCK_STREAM, &global_wins_table[0][i].address, 42, LONG_CONNECT_TIMEOUT);
+			if (s==-1) {
+				DEBUG(0,("run_pull_replication: can't contact wins server %s.\n", inet_ntoa(global_wins_table[0][i].address)));
+				return;
+			}
+			
+			if(buffer.offset > 0) {
+				if (!send_smb(s, buffer.buffer))
+					exit_server("run_pull_replication: send_smb failed.");
+			}
+			
+			add_fd_to_sock_array(s);
+			FD_SET(s, listen_set);
+
+			/* add ourself as a client */
+			add_partner((int)t, 0, True, False);
+		}
+	}
+}
+
+/****************************************************************************
+  contact periodically our wins partner to do a push replication
+****************************************************************************/
+void run_push_replication(time_t t)
+{
+	/* we push every 30 minutes or 25 new entries */
+
+}
+