2 Unix SMB/CIFS implementation.
3 Samba internal messaging functions
4 Copyright (C) 2007 by Volker Lendecke
5 Copyright (C) 2007 by Andrew Tridgell
7 This program is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 3 of the License, or
10 (at your option) any later version.
12 This program is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with this program. If not, see <http://www.gnu.org/licenses/>.
24 #include "ctdbd_conn.h"
25 #include "system/select.h"
26 #include "lib/sys_rw_data.h"
27 #include "lib/util/iov_buf.h"
31 /* paths to these include files come from --with-ctdb= in configure */
34 #include "ctdb_private.h"
36 struct ctdbd_srvid_cb {
38 int (*cb)(uint32_t src_vnn, uint32_t dst_vnn,
40 const uint8_t *msg, size_t msglen,
45 struct ctdbd_connection {
46 const char *sockname; /* Needed in ctdbd_traverse */
47 struct messaging_context *msg_ctx;
51 struct ctdbd_srvid_cb *callbacks;
53 struct tevent_fd *fde;
57 static uint32_t ctdbd_next_reqid(struct ctdbd_connection *conn)
60 if (conn->reqid == 0) {
66 static NTSTATUS ctdbd_control(struct ctdbd_connection *conn,
67 uint32_t vnn, uint32_t opcode,
68 uint64_t srvid, uint32_t flags, TDB_DATA data,
69 TALLOC_CTX *mem_ctx, TDB_DATA *outdata,
71 static int ctdbd_control_unix(struct ctdbd_connection *conn,
72 uint32_t vnn, uint32_t opcode,
73 uint64_t srvid, uint32_t flags,
75 TALLOC_CTX *mem_ctx, TDB_DATA *outdata,
79 * exit on fatal communications errors with the ctdbd daemon
81 static void cluster_fatal(const char *why)
83 DEBUG(0,("cluster fatal event: %s - exiting immediately\n", why));
84 /* we don't use smb_panic() as we don't want to delay to write
85 a core file. We need to release this process id immediately
86 so that someone else can take over without getting sharing
94 static void ctdb_packet_dump(struct ctdb_req_header *hdr)
96 if (DEBUGLEVEL < 11) {
99 DEBUGADD(11, ("len=%d, magic=%x, vers=%d, gen=%d, op=%d, reqid=%d\n",
100 (int)hdr->length, (int)hdr->ctdb_magic,
101 (int)hdr->ctdb_version, (int)hdr->generation,
102 (int)hdr->operation, (int)hdr->reqid));
106 * Register a srvid with ctdbd
108 NTSTATUS register_with_ctdbd(struct ctdbd_connection *conn, uint64_t srvid,
109 int (*cb)(uint32_t src_vnn, uint32_t dst_vnn,
111 const uint8_t *msg, size_t msglen,
117 size_t num_callbacks;
118 struct ctdbd_srvid_cb *tmp;
120 ret = ctdbd_control_unix(conn, CTDB_CURRENT_NODE,
121 CTDB_CONTROL_REGISTER_SRVID, srvid, 0,
122 tdb_null, NULL, NULL, &cstatus);
124 return map_nt_error_from_unix(ret);
127 num_callbacks = talloc_array_length(conn->callbacks);
129 tmp = talloc_realloc(conn, conn->callbacks, struct ctdbd_srvid_cb,
132 return NT_STATUS_NO_MEMORY;
134 conn->callbacks = tmp;
136 conn->callbacks[num_callbacks] = (struct ctdbd_srvid_cb) {
137 .srvid = srvid, .cb = cb, .private_data = private_data
143 static int ctdbd_msg_call_back(struct ctdbd_connection *conn,
144 struct ctdb_req_message *msg)
147 size_t i, num_callbacks;
149 msg_len = msg->hdr.length;
150 if (msg_len < offsetof(struct ctdb_req_message, data)) {
151 DEBUG(10, ("%s: len %u too small\n", __func__,
155 msg_len -= offsetof(struct ctdb_req_message, data);
157 if (msg_len < msg->datalen) {
158 DEBUG(10, ("%s: msg_len=%u < msg->datalen=%u\n", __func__,
159 (unsigned)msg_len, (unsigned)msg->datalen));
163 num_callbacks = talloc_array_length(conn->callbacks);
165 for (i=0; i<num_callbacks; i++) {
166 struct ctdbd_srvid_cb *cb = &conn->callbacks[i];
168 if ((cb->srvid == msg->srvid) && (cb->cb != NULL)) {
171 ret = cb->cb(msg->hdr.srcnode, msg->hdr.destnode,
172 msg->srvid, msg->data, msg->datalen,
183 * get our vnn from the cluster
185 static NTSTATUS get_cluster_vnn(struct ctdbd_connection *conn, uint32_t *vnn)
189 ret = ctdbd_control_unix(conn,
190 CTDB_CURRENT_NODE, CTDB_CONTROL_GET_PNN, 0, 0,
191 tdb_null, NULL, NULL, &cstatus);
193 DEBUG(1, ("ctdbd_control failed: %s\n", strerror(ret)));
194 return map_nt_error_from_unix(ret);
196 *vnn = (uint32_t)cstatus;
201 * Are we active (i.e. not banned or stopped?)
203 static bool ctdbd_working(struct ctdbd_connection *conn, uint32_t vnn)
207 struct ctdb_node_map *m;
208 uint32_t failure_flags;
212 ret = ctdbd_control_unix(conn, CTDB_CURRENT_NODE,
213 CTDB_CONTROL_GET_NODEMAP, 0, 0,
214 tdb_null, talloc_tos(), &outdata, &cstatus);
216 DEBUG(1, ("ctdbd_control failed: %s\n", strerror(ret)));
219 if ((cstatus != 0) || (outdata.dptr == NULL)) {
220 DEBUG(2, ("Received invalid ctdb data\n"));
224 m = (struct ctdb_node_map *)outdata.dptr;
226 for (i=0; i<m->num; i++) {
227 if (vnn == m->nodes[i].pnn) {
233 DEBUG(2, ("Did not find ourselves (node %d) in nodemap\n",
238 failure_flags = NODE_FLAGS_BANNED | NODE_FLAGS_DISCONNECTED
239 | NODE_FLAGS_PERMANENTLY_DISABLED | NODE_FLAGS_STOPPED;
241 if ((m->nodes[i].flags & failure_flags) != 0) {
242 DEBUG(2, ("Node has status %x, not active\n",
243 (int)m->nodes[i].flags));
249 TALLOC_FREE(outdata.dptr);
253 uint32_t ctdbd_vnn(const struct ctdbd_connection *conn)
255 return conn->our_vnn;
259 * Get us a ctdb connection
262 static int ctdbd_connect(const char *sockname, int *pfd)
264 struct sockaddr_un addr = { 0, };
269 fd = socket(AF_UNIX, SOCK_STREAM, 0);
272 DEBUG(3, ("Could not create socket: %s\n", strerror(err)));
276 addr.sun_family = AF_UNIX;
278 namelen = strlcpy(addr.sun_path, sockname, sizeof(addr.sun_path));
279 if (namelen >= sizeof(addr.sun_path)) {
280 DEBUG(3, ("%s: Socket name too long: %s\n", __func__,
286 salen = sizeof(struct sockaddr_un);
288 if (connect(fd, (struct sockaddr *)(void *)&addr, salen) == -1) {
290 DEBUG(1, ("connect(%s) failed: %s\n", sockname,
300 static int ctdb_read_packet(int fd, int timeout, TALLOC_CTX *mem_ctx,
301 struct ctdb_req_header **result)
303 struct ctdb_req_header *req;
309 ret = poll_one_fd(fd, POLLIN, timeout, &revents);
321 nread = read_data(fd, &msglen, sizeof(msglen));
329 if (msglen < sizeof(struct ctdb_req_header)) {
333 req = talloc_size(mem_ctx, msglen);
337 talloc_set_name_const(req, "struct ctdb_req_header");
339 req->length = msglen;
341 nread = read_data(fd, ((char *)req) + sizeof(msglen),
342 msglen - sizeof(msglen));
357 * Read a full ctdbd request. If we have a messaging context, defer incoming
358 * messages that might come in between.
361 static int ctdb_read_req(struct ctdbd_connection *conn, uint32_t reqid,
362 TALLOC_CTX *mem_ctx, struct ctdb_req_header **result)
364 struct ctdb_req_header *hdr;
369 ret = ctdb_read_packet(conn->fd, conn->timeout, mem_ctx, &hdr);
371 DEBUG(0, ("ctdb_read_packet failed: %s\n", strerror(ret)));
372 cluster_fatal("ctdbd died\n");
375 DEBUG(11, ("Received ctdb packet\n"));
376 ctdb_packet_dump(hdr);
378 if (hdr->operation == CTDB_REQ_MESSAGE) {
379 struct ctdb_req_message *msg = (struct ctdb_req_message *)hdr;
381 if (conn->msg_ctx == NULL) {
382 DEBUG(1, ("Got a message without having a msg ctx, "
383 "dropping msg %llu\n",
384 (long long unsigned)msg->srvid));
389 ret = ctdbd_msg_call_back(conn, msg);
399 if ((reqid != 0) && (hdr->reqid != reqid)) {
400 /* we got the wrong reply */
401 DEBUG(0,("Discarding mismatched ctdb reqid %u should have "
402 "been %u\n", hdr->reqid, reqid));
407 *result = talloc_move(mem_ctx, &hdr);
412 static int ctdbd_connection_destructor(struct ctdbd_connection *c)
422 * Get us a ctdbd connection
425 static NTSTATUS ctdbd_init_connection(TALLOC_CTX *mem_ctx,
426 const char *sockname, int timeout,
427 struct ctdbd_connection **pconn)
429 struct ctdbd_connection *conn;
433 if (!(conn = talloc_zero(mem_ctx, struct ctdbd_connection))) {
434 DEBUG(0, ("talloc failed\n"));
435 return NT_STATUS_NO_MEMORY;
438 conn->sockname = talloc_strdup(conn, sockname);
439 if (conn->sockname == NULL) {
440 DBG_ERR("%s: talloc failed\n", __func__);
441 status = NT_STATUS_NO_MEMORY;
445 conn->timeout = timeout;
447 if (conn->timeout == 0) {
451 ret = ctdbd_connect(conn->sockname, &conn->fd);
453 status = map_nt_error_from_unix(ret);
454 DEBUG(1, ("ctdbd_connect failed: %s\n", strerror(ret)));
457 talloc_set_destructor(conn, ctdbd_connection_destructor);
459 status = get_cluster_vnn(conn, &conn->our_vnn);
461 if (!NT_STATUS_IS_OK(status)) {
462 DEBUG(10, ("get_cluster_vnn failed: %s\n", nt_errstr(status)));
466 if (!ctdbd_working(conn, conn->our_vnn)) {
467 DEBUG(2, ("Node is not working, can not connect\n"));
468 status = NT_STATUS_INTERNAL_DB_ERROR;
472 generate_random_buffer((unsigned char *)&conn->rand_srvid,
473 sizeof(conn->rand_srvid));
475 status = register_with_ctdbd(conn, conn->rand_srvid, NULL, NULL);
477 if (!NT_STATUS_IS_OK(status)) {
478 DEBUG(5, ("Could not register random srvid: %s\n",
492 * Get us a ctdbd connection and register us as a process
495 NTSTATUS ctdbd_messaging_connection(TALLOC_CTX *mem_ctx,
496 const char *sockname, int timeout,
497 struct ctdbd_connection **pconn)
499 struct ctdbd_connection *conn;
502 status = ctdbd_init_connection(mem_ctx, sockname, timeout, &conn);
504 if (!NT_STATUS_IS_OK(status)) {
508 status = register_with_ctdbd(conn, MSG_SRVID_SAMBA, NULL, NULL);
509 if (!NT_STATUS_IS_OK(status)) {
521 struct messaging_context *ctdb_conn_msg_ctx(struct ctdbd_connection *conn)
523 return conn->msg_ctx;
526 int ctdbd_conn_get_fd(struct ctdbd_connection *conn)
532 * Packet handler to receive and handle a ctdb message
534 static int ctdb_handle_message(struct ctdbd_connection *conn,
535 struct ctdb_req_header *hdr)
537 struct ctdb_req_message *msg;
539 if (hdr->operation != CTDB_REQ_MESSAGE) {
540 DEBUG(0, ("Received async msg of type %u, discarding\n",
545 msg = (struct ctdb_req_message *)hdr;
547 ctdbd_msg_call_back(conn, msg);
553 * The ctdbd socket is readable asynchronuously
556 static void ctdbd_socket_handler(struct tevent_context *event_ctx,
557 struct tevent_fd *event,
561 struct ctdbd_connection *conn = talloc_get_type_abort(
562 private_data, struct ctdbd_connection);
563 struct ctdb_req_header *hdr = NULL;
566 ret = ctdb_read_packet(conn->fd, conn->timeout, talloc_tos(), &hdr);
568 DEBUG(0, ("ctdb_read_packet failed: %s\n", strerror(ret)));
569 cluster_fatal("ctdbd died\n");
572 ret = ctdb_handle_message(conn, hdr);
577 DEBUG(10, ("could not handle incoming message: %s\n",
583 * Prepare a ctdbd connection to receive messages
586 NTSTATUS ctdbd_register_msg_ctx(struct ctdbd_connection *conn,
587 struct messaging_context *msg_ctx)
589 SMB_ASSERT(conn->msg_ctx == NULL);
590 SMB_ASSERT(conn->fde == NULL);
592 if (!(conn->fde = tevent_add_fd(messaging_tevent_context(msg_ctx),
596 ctdbd_socket_handler,
598 DEBUG(0, ("event_add_fd failed\n"));
599 return NT_STATUS_NO_MEMORY;
602 conn->msg_ctx = msg_ctx;
607 NTSTATUS ctdbd_messaging_send_iov(struct ctdbd_connection *conn,
608 uint32_t dst_vnn, uint64_t dst_srvid,
609 const struct iovec *iov, int iovlen)
611 struct ctdb_req_message r;
612 struct iovec iov2[iovlen+1];
613 size_t buflen = iov_buflen(iov, iovlen);
616 r.hdr.length = offsetof(struct ctdb_req_message, data) + buflen;
617 r.hdr.ctdb_magic = CTDB_MAGIC;
618 r.hdr.ctdb_version = CTDB_PROTOCOL;
619 r.hdr.generation = 1;
620 r.hdr.operation = CTDB_REQ_MESSAGE;
621 r.hdr.destnode = dst_vnn;
622 r.hdr.srcnode = conn->our_vnn;
627 DEBUG(10, ("ctdbd_messaging_send: Sending ctdb packet\n"));
628 ctdb_packet_dump(&r.hdr);
630 iov2[0].iov_base = &r;
631 iov2[0].iov_len = offsetof(struct ctdb_req_message, data);
632 memcpy(&iov2[1], iov, iovlen * sizeof(struct iovec));
634 nwritten = write_data_iov(conn->fd, iov2, iovlen+1);
635 if (nwritten == -1) {
636 DEBUG(3, ("write_data_iov failed: %s\n", strerror(errno)));
637 cluster_fatal("cluster dispatch daemon msg write error\n");
644 * send/recv a generic ctdb control message
646 static int ctdbd_control_unix(struct ctdbd_connection *conn,
647 uint32_t vnn, uint32_t opcode,
648 uint64_t srvid, uint32_t flags,
650 TALLOC_CTX *mem_ctx, TDB_DATA *outdata,
653 struct ctdb_req_control req;
654 struct ctdb_req_header *hdr;
655 struct ctdb_reply_control *reply = NULL;
661 req.hdr.length = offsetof(struct ctdb_req_control, data) + data.dsize;
662 req.hdr.ctdb_magic = CTDB_MAGIC;
663 req.hdr.ctdb_version = CTDB_PROTOCOL;
664 req.hdr.operation = CTDB_REQ_CONTROL;
665 req.hdr.reqid = ctdbd_next_reqid(conn);
666 req.hdr.destnode = vnn;
669 req.datalen = data.dsize;
672 DEBUG(10, ("ctdbd_control: Sending ctdb packet\n"));
673 ctdb_packet_dump(&req.hdr);
675 iov[0].iov_base = &req;
676 iov[0].iov_len = offsetof(struct ctdb_req_control, data);
677 iov[1].iov_base = data.dptr;
678 iov[1].iov_len = data.dsize;
680 nwritten = write_data_iov(conn->fd, iov, ARRAY_SIZE(iov));
681 if (nwritten == -1) {
682 DEBUG(3, ("write_data_iov failed: %s\n", strerror(errno)));
683 cluster_fatal("cluster dispatch daemon msg write error\n");
686 if (flags & CTDB_CTRL_FLAG_NOREPLY) {
693 ret = ctdb_read_req(conn, req.hdr.reqid, NULL, &hdr);
695 DEBUG(10, ("ctdb_read_req failed: %s\n", strerror(ret)));
699 if (hdr->operation != CTDB_REPLY_CONTROL) {
700 DEBUG(0, ("received invalid reply\n"));
704 reply = (struct ctdb_reply_control *)hdr;
707 if (!(outdata->dptr = (uint8_t *)talloc_memdup(
708 mem_ctx, reply->data, reply->datalen))) {
712 outdata->dsize = reply->datalen;
715 (*cstatus) = reply->status;
722 static NTSTATUS ctdbd_control(struct ctdbd_connection *conn,
723 uint32_t vnn, uint32_t opcode,
724 uint64_t srvid, uint32_t flags,
726 TALLOC_CTX *mem_ctx, TDB_DATA *outdata,
731 ret = ctdbd_control_unix(conn, vnn, opcode, srvid, flags, data,
732 mem_ctx, outdata, cstatus);
734 return map_nt_error_from_unix(ret);
740 * see if a remote process exists
742 bool ctdbd_process_exists(struct ctdbd_connection *conn, uint32_t vnn, pid_t pid)
750 if (!ctdb_processes_exist(conn, &id, 1, &result)) {
751 DEBUG(10, ("ctdb_processes_exist failed\n"));
757 bool ctdb_processes_exist(struct ctdbd_connection *conn,
758 const struct server_id *pids, int num_pids,
761 TALLOC_CTX *frame = talloc_stackframe();
766 reqids = talloc_array(talloc_tos(), uint32_t, num_pids);
767 if (reqids == NULL) {
771 for (i=0; i<num_pids; i++) {
772 struct ctdb_req_control req;
778 reqids[i] = ctdbd_next_reqid(conn);
783 * pids[i].pid is uint64_t, scale down to pid_t which
784 * is the wire protocol towards ctdb.
788 DEBUG(10, ("Requesting PID %d/%d, reqid=%d\n",
789 (int)pids[i].vnn, (int)pid,
792 req.hdr.length = offsetof(struct ctdb_req_control, data);
793 req.hdr.length += sizeof(pid);
794 req.hdr.ctdb_magic = CTDB_MAGIC;
795 req.hdr.ctdb_version = CTDB_PROTOCOL;
796 req.hdr.operation = CTDB_REQ_CONTROL;
797 req.hdr.reqid = reqids[i];
798 req.hdr.destnode = pids[i].vnn;
799 req.opcode = CTDB_CONTROL_PROCESS_EXISTS;
801 req.datalen = sizeof(pid);
804 DEBUG(10, ("ctdbd_control: Sending ctdb packet\n"));
805 ctdb_packet_dump(&req.hdr);
807 iov[0].iov_base = &req;
808 iov[0].iov_len = offsetof(struct ctdb_req_control, data);
809 iov[1].iov_base = &pid;
810 iov[1].iov_len = sizeof(pid);
812 nwritten = write_data_iov(conn->fd, iov, ARRAY_SIZE(iov));
813 if (nwritten == -1) {
814 DEBUG(10, ("write_data_iov failed: %s\n",
822 while (num_received < num_pids) {
823 struct ctdb_req_header *hdr;
824 struct ctdb_reply_control *reply;
828 ret = ctdb_read_req(conn, 0, talloc_tos(), &hdr);
830 DEBUG(10, ("ctdb_read_req failed: %s\n",
835 if (hdr->operation != CTDB_REPLY_CONTROL) {
836 DEBUG(10, ("Received invalid reply\n"));
839 reply = (struct ctdb_reply_control *)hdr;
841 reqid = reply->hdr.reqid;
843 DEBUG(10, ("Received reqid %d\n", (int)reqid));
845 for (i=0; i<num_pids; i++) {
846 if (reqid == reqids[i]) {
851 DEBUG(10, ("Received unknown record number %u\n",
855 results[i] = ((reply->status) == 0);
869 char *ctdbd_dbpath(struct ctdbd_connection *conn,
870 TALLOC_CTX *mem_ctx, uint32_t db_id)
874 TDB_DATA rdata = {0};
877 data.dptr = (uint8_t*)&db_id;
878 data.dsize = sizeof(db_id);
880 ret = ctdbd_control_unix(conn, CTDB_CURRENT_NODE,
881 CTDB_CONTROL_GETDBPATH, 0, 0, data,
882 mem_ctx, &rdata, &cstatus);
883 if ((ret != 0) || cstatus != 0) {
884 DEBUG(0, (__location__ " ctdb_control for getdbpath failed: %s\n",
889 return (char *)rdata.dptr;
893 * attach to a ctdb database
895 NTSTATUS ctdbd_db_attach(struct ctdbd_connection *conn,
896 const char *name, uint32_t *db_id, int tdb_flags)
901 bool persistent = (tdb_flags & TDB_CLEAR_IF_FIRST) == 0;
903 data = string_term_tdb_data(name);
905 status = ctdbd_control(conn, CTDB_CURRENT_NODE,
907 ? CTDB_CONTROL_DB_ATTACH_PERSISTENT
908 : CTDB_CONTROL_DB_ATTACH,
909 tdb_flags, 0, data, NULL, &data, &cstatus);
910 if (!NT_STATUS_IS_OK(status)) {
911 DEBUG(0, (__location__ " ctdb_control for db_attach "
912 "failed: %s\n", nt_errstr(status)));
916 if (cstatus != 0 || data.dsize != sizeof(uint32_t)) {
917 DEBUG(0,(__location__ " ctdb_control for db_attach failed\n"));
918 return NT_STATUS_INTERNAL_ERROR;
921 *db_id = *(uint32_t *)data.dptr;
922 talloc_free(data.dptr);
924 if (!(tdb_flags & TDB_SEQNUM)) {
928 data.dptr = (uint8_t *)db_id;
929 data.dsize = sizeof(*db_id);
931 status = ctdbd_control(conn, CTDB_CURRENT_NODE,
932 CTDB_CONTROL_ENABLE_SEQNUM, 0, 0, data,
933 NULL, NULL, &cstatus);
934 if (!NT_STATUS_IS_OK(status) || cstatus != 0) {
935 DEBUG(0,(__location__ " ctdb_control for enable seqnum "
937 return NT_STATUS_IS_OK(status) ? NT_STATUS_INTERNAL_ERROR :
945 * force the migration of a record to this node
947 NTSTATUS ctdbd_migrate(struct ctdbd_connection *conn, uint32_t db_id,
950 struct ctdb_req_call req;
951 struct ctdb_req_header *hdr;
959 req.hdr.length = offsetof(struct ctdb_req_call, data) + key.dsize;
960 req.hdr.ctdb_magic = CTDB_MAGIC;
961 req.hdr.ctdb_version = CTDB_PROTOCOL;
962 req.hdr.operation = CTDB_REQ_CALL;
963 req.hdr.reqid = ctdbd_next_reqid(conn);
964 req.flags = CTDB_IMMEDIATE_MIGRATION;
965 req.callid = CTDB_NULL_FUNC;
967 req.keylen = key.dsize;
969 DEBUG(10, ("ctdbd_migrate: Sending ctdb packet\n"));
970 ctdb_packet_dump(&req.hdr);
972 iov[0].iov_base = &req;
973 iov[0].iov_len = offsetof(struct ctdb_req_call, data);
974 iov[1].iov_base = key.dptr;
975 iov[1].iov_len = key.dsize;
977 nwritten = write_data_iov(conn->fd, iov, ARRAY_SIZE(iov));
978 if (nwritten == -1) {
979 DEBUG(3, ("write_data_iov failed: %s\n", strerror(errno)));
980 cluster_fatal("cluster dispatch daemon msg write error\n");
983 ret = ctdb_read_req(conn, req.hdr.reqid, NULL, &hdr);
985 DEBUG(10, ("ctdb_read_req failed: %s\n", strerror(ret)));
986 status = map_nt_error_from_unix(ret);
990 if (hdr->operation != CTDB_REPLY_CALL) {
991 DEBUG(0, ("received invalid reply\n"));
992 status = NT_STATUS_INTERNAL_ERROR;
996 status = NT_STATUS_OK;
1004 * Fetch a record and parse it
1006 NTSTATUS ctdbd_parse(struct ctdbd_connection *conn, uint32_t db_id,
1007 TDB_DATA key, bool local_copy,
1008 void (*parser)(TDB_DATA key, TDB_DATA data,
1009 void *private_data),
1012 struct ctdb_req_call req;
1013 struct ctdb_req_header *hdr = NULL;
1014 struct ctdb_reply_call *reply;
1015 struct iovec iov[2];
1021 flags = local_copy ? CTDB_WANT_READONLY : 0;
1025 req.hdr.length = offsetof(struct ctdb_req_call, data) + key.dsize;
1026 req.hdr.ctdb_magic = CTDB_MAGIC;
1027 req.hdr.ctdb_version = CTDB_PROTOCOL;
1028 req.hdr.operation = CTDB_REQ_CALL;
1029 req.hdr.reqid = ctdbd_next_reqid(conn);
1031 req.callid = CTDB_FETCH_FUNC;
1033 req.keylen = key.dsize;
1035 iov[0].iov_base = &req;
1036 iov[0].iov_len = offsetof(struct ctdb_req_call, data);
1037 iov[1].iov_base = key.dptr;
1038 iov[1].iov_len = key.dsize;
1040 nwritten = write_data_iov(conn->fd, iov, ARRAY_SIZE(iov));
1041 if (nwritten == -1) {
1042 DEBUG(3, ("write_data_iov failed: %s\n", strerror(errno)));
1043 cluster_fatal("cluster dispatch daemon msg write error\n");
1046 ret = ctdb_read_req(conn, req.hdr.reqid, NULL, &hdr);
1048 DEBUG(10, ("ctdb_read_req failed: %s\n", strerror(ret)));
1049 status = map_nt_error_from_unix(ret);
1053 if ((hdr == NULL) || (hdr->operation != CTDB_REPLY_CALL)) {
1054 DEBUG(0, ("received invalid reply\n"));
1055 status = NT_STATUS_INTERNAL_ERROR;
1058 reply = (struct ctdb_reply_call *)hdr;
1060 if (reply->datalen == 0) {
1062 * Treat an empty record as non-existing
1064 status = NT_STATUS_NOT_FOUND;
1068 parser(key, make_tdb_data(&reply->data[0], reply->datalen),
1071 status = NT_STATUS_OK;
1078 Traverse a ctdb database. This uses a kind-of hackish way to open a second
1079 connection to ctdbd to avoid the hairy recursive and async problems with
1083 NTSTATUS ctdbd_traverse(struct ctdbd_connection *master, uint32_t db_id,
1084 void (*fn)(TDB_DATA key, TDB_DATA data,
1085 void *private_data),
1088 struct ctdbd_connection *conn;
1092 struct ctdb_traverse_start t;
1096 status = ctdbd_init_connection(NULL, master->sockname, master->timeout,
1099 if (!NT_STATUS_IS_OK(status)) {
1100 DEBUG(0, ("ctdbd_init_connection failed: %s\n",
1101 nt_errstr(status)));
1106 t.srvid = conn->rand_srvid;
1107 t.reqid = ctdbd_next_reqid(conn);
1109 data.dptr = (uint8_t *)&t;
1110 data.dsize = sizeof(t);
1112 status = ctdbd_control(conn, CTDB_CURRENT_NODE,
1113 CTDB_CONTROL_TRAVERSE_START, conn->rand_srvid, 0,
1114 data, NULL, NULL, &cstatus);
1116 if (!NT_STATUS_IS_OK(status) || (cstatus != 0)) {
1118 DEBUG(0,("ctdbd_control failed: %s, %d\n", nt_errstr(status),
1121 if (NT_STATUS_IS_OK(status)) {
1123 * We need a mapping here
1125 status = NT_STATUS_UNSUCCESSFUL;
1132 struct ctdb_req_header *hdr = NULL;
1133 struct ctdb_req_message *m;
1134 struct ctdb_rec_data *d;
1137 ret = ctdb_read_packet(conn->fd, conn->timeout, conn, &hdr);
1139 DEBUG(0, ("ctdb_read_packet failed: %s\n",
1141 cluster_fatal("ctdbd died\n");
1144 if (hdr->operation != CTDB_REQ_MESSAGE) {
1145 DEBUG(0, ("Got operation %u, expected a message\n",
1146 (unsigned)hdr->operation));
1148 return NT_STATUS_UNEXPECTED_IO_ERROR;
1151 m = (struct ctdb_req_message *)hdr;
1152 d = (struct ctdb_rec_data *)&m->data[0];
1153 if (m->datalen < sizeof(uint32_t) || m->datalen != d->length) {
1154 DEBUG(0, ("Got invalid traverse data of length %d\n",
1157 return NT_STATUS_UNEXPECTED_IO_ERROR;
1160 key.dsize = d->keylen;
1161 key.dptr = &d->data[0];
1162 data.dsize = d->datalen;
1163 data.dptr = &d->data[d->keylen];
1165 if (key.dsize == 0 && data.dsize == 0) {
1166 /* end of traverse */
1168 return NT_STATUS_OK;
1171 if (data.dsize < sizeof(struct ctdb_ltdb_header)) {
1172 DEBUG(0, ("Got invalid ltdb header length %d\n",
1175 return NT_STATUS_UNEXPECTED_IO_ERROR;
1177 data.dsize -= sizeof(struct ctdb_ltdb_header);
1178 data.dptr += sizeof(struct ctdb_ltdb_header);
1181 fn(key, data, private_data);
1184 return NT_STATUS_OK;
1188 This is used to canonicalize a ctdb_sock_addr structure.
1190 static void smbd_ctdb_canonicalize_ip(const struct sockaddr_storage *in,
1191 struct sockaddr_storage *out)
1193 memcpy(out, in, sizeof (*out));
1196 if (in->ss_family == AF_INET6) {
1197 const char prefix[12] = { 0,0,0,0,0,0,0,0,0,0,0xff,0xff };
1198 const struct sockaddr_in6 *in6 =
1199 (const struct sockaddr_in6 *)in;
1200 struct sockaddr_in *out4 = (struct sockaddr_in *)out;
1201 if (memcmp(&in6->sin6_addr, prefix, 12) == 0) {
1202 memset(out, 0, sizeof(*out));
1203 #ifdef HAVE_SOCK_SIN_LEN
1204 out4->sin_len = sizeof(*out);
1206 out4->sin_family = AF_INET;
1207 out4->sin_port = in6->sin6_port;
1208 memcpy(&out4->sin_addr, &in6->sin6_addr.s6_addr[12], 4);
1215 * Register us as a server for a particular tcp connection
1218 NTSTATUS ctdbd_register_ips(struct ctdbd_connection *conn,
1219 const struct sockaddr_storage *_server,
1220 const struct sockaddr_storage *_client,
1221 int (*cb)(uint32_t src_vnn, uint32_t dst_vnn,
1223 const uint8_t *msg, size_t msglen,
1224 void *private_data),
1227 struct ctdb_control_tcp_addr p;
1228 TDB_DATA data = { .dptr = (uint8_t *)&p, .dsize = sizeof(p) };
1230 struct sockaddr_storage client;
1231 struct sockaddr_storage server;
1234 * Only one connection so far
1237 smbd_ctdb_canonicalize_ip(_client, &client);
1238 smbd_ctdb_canonicalize_ip(_server, &server);
1240 switch (client.ss_family) {
1242 memcpy(&p.dest.ip, &server, sizeof(p.dest.ip));
1243 memcpy(&p.src.ip, &client, sizeof(p.src.ip));
1246 memcpy(&p.dest.ip6, &server, sizeof(p.dest.ip6));
1247 memcpy(&p.src.ip6, &client, sizeof(p.src.ip6));
1250 return NT_STATUS_INTERNAL_ERROR;
1254 * We want to be told about IP releases
1257 status = register_with_ctdbd(conn, CTDB_SRVID_RELEASE_IP,
1259 if (!NT_STATUS_IS_OK(status)) {
1264 * inform ctdb of our tcp connection, so if IP takeover happens ctdb
1265 * can send an extra ack to trigger a reset for our client, so it
1266 * immediately reconnects
1268 return ctdbd_control(conn, CTDB_CURRENT_NODE,
1269 CTDB_CONTROL_TCP_CLIENT, 0,
1270 CTDB_CTRL_FLAG_NOREPLY, data, NULL, NULL, NULL);
1274 call a control on the local node
1276 NTSTATUS ctdbd_control_local(struct ctdbd_connection *conn, uint32_t opcode,
1277 uint64_t srvid, uint32_t flags, TDB_DATA data,
1278 TALLOC_CTX *mem_ctx, TDB_DATA *outdata,
1281 return ctdbd_control(conn, CTDB_CURRENT_NODE, opcode, srvid, flags, data, mem_ctx, outdata, cstatus);
1284 NTSTATUS ctdb_watch_us(struct ctdbd_connection *conn)
1286 struct ctdb_client_notify_register reg_data;
1291 reg_data.srvid = CTDB_SRVID_SAMBA_NOTIFY;
1293 reg_data.notify_data[0] = 0;
1295 struct_len = offsetof(struct ctdb_client_notify_register,
1296 notify_data) + reg_data.len;
1298 status = ctdbd_control_local(
1299 conn, CTDB_CONTROL_REGISTER_NOTIFY, conn->rand_srvid, 0,
1300 make_tdb_data((uint8_t *)®_data, struct_len),
1301 NULL, NULL, &cstatus);
1302 if (!NT_STATUS_IS_OK(status)) {
1303 DEBUG(1, ("ctdbd_control_local failed: %s\n",
1304 nt_errstr(status)));
1309 NTSTATUS ctdb_unwatch(struct ctdbd_connection *conn)
1311 struct ctdb_client_notify_deregister dereg_data;
1315 dereg_data.srvid = CTDB_SRVID_SAMBA_NOTIFY;
1317 status = ctdbd_control_local(
1318 conn, CTDB_CONTROL_DEREGISTER_NOTIFY, conn->rand_srvid, 0,
1319 make_tdb_data((uint8_t *)&dereg_data, sizeof(dereg_data)),
1320 NULL, NULL, &cstatus);
1321 if (!NT_STATUS_IS_OK(status)) {
1322 DEBUG(1, ("ctdbd_control_local failed: %s\n",
1323 nt_errstr(status)));
1328 NTSTATUS ctdbd_probe(const char *sockname, int timeout)
1331 * Do a very early check if ctdbd is around to avoid an abort and core
1334 struct ctdbd_connection *conn = NULL;
1337 status = ctdbd_messaging_connection(talloc_tos(), sockname, timeout,
1341 * We only care if we can connect.