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 int ctdbd_control(struct ctdbd_connection
*conn
,
67 uint32_t vnn
, uint32_t opcode
,
68 uint64_t srvid
, uint32_t flags
,
70 TALLOC_CTX
*mem_ctx
, TDB_DATA
*outdata
,
74 * exit on fatal communications errors with the ctdbd daemon
76 static void cluster_fatal(const char *why
)
78 DEBUG(0,("cluster fatal event: %s - exiting immediately\n", why
));
79 /* we don't use smb_panic() as we don't want to delay to write
80 a core file. We need to release this process id immediately
81 so that someone else can take over without getting sharing
89 static void ctdb_packet_dump(struct ctdb_req_header
*hdr
)
91 if (DEBUGLEVEL
< 11) {
94 DEBUGADD(11, ("len=%d, magic=%x, vers=%d, gen=%d, op=%d, reqid=%d\n",
95 (int)hdr
->length
, (int)hdr
->ctdb_magic
,
96 (int)hdr
->ctdb_version
, (int)hdr
->generation
,
97 (int)hdr
->operation
, (int)hdr
->reqid
));
101 * Register a srvid with ctdbd
103 int register_with_ctdbd(struct ctdbd_connection
*conn
, uint64_t srvid
,
104 int (*cb
)(uint32_t src_vnn
, uint32_t dst_vnn
,
106 const uint8_t *msg
, size_t msglen
,
112 size_t num_callbacks
;
113 struct ctdbd_srvid_cb
*tmp
;
115 ret
= ctdbd_control(conn
, CTDB_CURRENT_NODE
,
116 CTDB_CONTROL_REGISTER_SRVID
, srvid
, 0,
117 tdb_null
, NULL
, NULL
, &cstatus
);
122 num_callbacks
= talloc_array_length(conn
->callbacks
);
124 tmp
= talloc_realloc(conn
, conn
->callbacks
, struct ctdbd_srvid_cb
,
129 conn
->callbacks
= tmp
;
131 conn
->callbacks
[num_callbacks
] = (struct ctdbd_srvid_cb
) {
132 .srvid
= srvid
, .cb
= cb
, .private_data
= private_data
138 static int ctdbd_msg_call_back(struct ctdbd_connection
*conn
,
139 struct ctdb_req_message
*msg
)
142 size_t i
, num_callbacks
;
144 msg_len
= msg
->hdr
.length
;
145 if (msg_len
< offsetof(struct ctdb_req_message
, data
)) {
146 DEBUG(10, ("%s: len %u too small\n", __func__
,
150 msg_len
-= offsetof(struct ctdb_req_message
, data
);
152 if (msg_len
< msg
->datalen
) {
153 DEBUG(10, ("%s: msg_len=%u < msg->datalen=%u\n", __func__
,
154 (unsigned)msg_len
, (unsigned)msg
->datalen
));
158 num_callbacks
= talloc_array_length(conn
->callbacks
);
160 for (i
=0; i
<num_callbacks
; i
++) {
161 struct ctdbd_srvid_cb
*cb
= &conn
->callbacks
[i
];
163 if ((cb
->srvid
== msg
->srvid
) && (cb
->cb
!= NULL
)) {
166 ret
= cb
->cb(msg
->hdr
.srcnode
, msg
->hdr
.destnode
,
167 msg
->srvid
, msg
->data
, msg
->datalen
,
178 * get our vnn from the cluster
180 static int get_cluster_vnn(struct ctdbd_connection
*conn
, uint32_t *vnn
)
184 ret
= ctdbd_control(conn
,
185 CTDB_CURRENT_NODE
, CTDB_CONTROL_GET_PNN
, 0, 0,
186 tdb_null
, NULL
, NULL
, &cstatus
);
188 DEBUG(1, ("ctdbd_control failed: %s\n", strerror(ret
)));
191 *vnn
= (uint32_t)cstatus
;
196 * Are we active (i.e. not banned or stopped?)
198 static bool ctdbd_working(struct ctdbd_connection
*conn
, uint32_t vnn
)
202 struct ctdb_node_map
*m
;
203 uint32_t failure_flags
;
207 ret
= ctdbd_control(conn
, CTDB_CURRENT_NODE
,
208 CTDB_CONTROL_GET_NODEMAP
, 0, 0,
209 tdb_null
, talloc_tos(), &outdata
, &cstatus
);
211 DEBUG(1, ("ctdbd_control failed: %s\n", strerror(ret
)));
214 if ((cstatus
!= 0) || (outdata
.dptr
== NULL
)) {
215 DEBUG(2, ("Received invalid ctdb data\n"));
219 m
= (struct ctdb_node_map
*)outdata
.dptr
;
221 for (i
=0; i
<m
->num
; i
++) {
222 if (vnn
== m
->nodes
[i
].pnn
) {
228 DEBUG(2, ("Did not find ourselves (node %d) in nodemap\n",
233 failure_flags
= NODE_FLAGS_BANNED
| NODE_FLAGS_DISCONNECTED
234 | NODE_FLAGS_PERMANENTLY_DISABLED
| NODE_FLAGS_STOPPED
;
236 if ((m
->nodes
[i
].flags
& failure_flags
) != 0) {
237 DEBUG(2, ("Node has status %x, not active\n",
238 (int)m
->nodes
[i
].flags
));
244 TALLOC_FREE(outdata
.dptr
);
248 uint32_t ctdbd_vnn(const struct ctdbd_connection
*conn
)
250 return conn
->our_vnn
;
254 * Get us a ctdb connection
257 static int ctdbd_connect(const char *sockname
, int *pfd
)
259 struct sockaddr_un addr
= { 0, };
264 fd
= socket(AF_UNIX
, SOCK_STREAM
, 0);
267 DEBUG(3, ("Could not create socket: %s\n", strerror(err
)));
271 addr
.sun_family
= AF_UNIX
;
273 namelen
= strlcpy(addr
.sun_path
, sockname
, sizeof(addr
.sun_path
));
274 if (namelen
>= sizeof(addr
.sun_path
)) {
275 DEBUG(3, ("%s: Socket name too long: %s\n", __func__
,
281 salen
= sizeof(struct sockaddr_un
);
283 if (connect(fd
, (struct sockaddr
*)(void *)&addr
, salen
) == -1) {
285 DEBUG(1, ("connect(%s) failed: %s\n", sockname
,
295 static int ctdb_read_packet(int fd
, int timeout
, TALLOC_CTX
*mem_ctx
,
296 struct ctdb_req_header
**result
)
298 struct ctdb_req_header
*req
;
304 ret
= poll_one_fd(fd
, POLLIN
, timeout
, &revents
);
316 nread
= read_data(fd
, &msglen
, sizeof(msglen
));
324 if (msglen
< sizeof(struct ctdb_req_header
)) {
328 req
= talloc_size(mem_ctx
, msglen
);
332 talloc_set_name_const(req
, "struct ctdb_req_header");
334 req
->length
= msglen
;
336 nread
= read_data(fd
, ((char *)req
) + sizeof(msglen
),
337 msglen
- sizeof(msglen
));
352 * Read a full ctdbd request. If we have a messaging context, defer incoming
353 * messages that might come in between.
356 static int ctdb_read_req(struct ctdbd_connection
*conn
, uint32_t reqid
,
357 TALLOC_CTX
*mem_ctx
, struct ctdb_req_header
**result
)
359 struct ctdb_req_header
*hdr
;
364 ret
= ctdb_read_packet(conn
->fd
, conn
->timeout
, mem_ctx
, &hdr
);
366 DEBUG(0, ("ctdb_read_packet failed: %s\n", strerror(ret
)));
367 cluster_fatal("ctdbd died\n");
370 DEBUG(11, ("Received ctdb packet\n"));
371 ctdb_packet_dump(hdr
);
373 if (hdr
->operation
== CTDB_REQ_MESSAGE
) {
374 struct ctdb_req_message
*msg
= (struct ctdb_req_message
*)hdr
;
376 if (conn
->msg_ctx
== NULL
) {
377 DEBUG(1, ("Got a message without having a msg ctx, "
378 "dropping msg %llu\n",
379 (long long unsigned)msg
->srvid
));
384 ret
= ctdbd_msg_call_back(conn
, msg
);
394 if ((reqid
!= 0) && (hdr
->reqid
!= reqid
)) {
395 /* we got the wrong reply */
396 DEBUG(0,("Discarding mismatched ctdb reqid %u should have "
397 "been %u\n", hdr
->reqid
, reqid
));
402 *result
= talloc_move(mem_ctx
, &hdr
);
407 static int ctdbd_connection_destructor(struct ctdbd_connection
*c
)
417 * Get us a ctdbd connection
420 static int ctdbd_init_connection(TALLOC_CTX
*mem_ctx
,
421 const char *sockname
, int timeout
,
422 struct ctdbd_connection
**pconn
)
424 struct ctdbd_connection
*conn
;
427 if (!(conn
= talloc_zero(mem_ctx
, struct ctdbd_connection
))) {
428 DEBUG(0, ("talloc failed\n"));
432 conn
->sockname
= talloc_strdup(conn
, sockname
);
433 if (conn
->sockname
== NULL
) {
434 DBG_ERR("%s: talloc failed\n", __func__
);
439 conn
->timeout
= timeout
;
441 if (conn
->timeout
== 0) {
445 ret
= ctdbd_connect(conn
->sockname
, &conn
->fd
);
447 DEBUG(1, ("ctdbd_connect failed: %s\n", strerror(ret
)));
450 talloc_set_destructor(conn
, ctdbd_connection_destructor
);
452 ret
= get_cluster_vnn(conn
, &conn
->our_vnn
);
455 DEBUG(10, ("get_cluster_vnn failed: %s\n", strerror(ret
)));
459 if (!ctdbd_working(conn
, conn
->our_vnn
)) {
460 DEBUG(2, ("Node is not working, can not connect\n"));
465 generate_random_buffer((unsigned char *)&conn
->rand_srvid
,
466 sizeof(conn
->rand_srvid
));
468 ret
= register_with_ctdbd(conn
, conn
->rand_srvid
, NULL
, NULL
);
471 DEBUG(5, ("Could not register random srvid: %s\n",
485 * Get us a ctdbd connection and register us as a process
488 int ctdbd_messaging_connection(TALLOC_CTX
*mem_ctx
,
489 const char *sockname
, int timeout
,
490 struct ctdbd_connection
**pconn
)
492 struct ctdbd_connection
*conn
;
495 ret
= ctdbd_init_connection(mem_ctx
, sockname
, timeout
, &conn
);
501 ret
= register_with_ctdbd(conn
, MSG_SRVID_SAMBA
, NULL
, NULL
);
514 struct messaging_context
*ctdb_conn_msg_ctx(struct ctdbd_connection
*conn
)
516 return conn
->msg_ctx
;
519 int ctdbd_conn_get_fd(struct ctdbd_connection
*conn
)
525 * Packet handler to receive and handle a ctdb message
527 static int ctdb_handle_message(struct ctdbd_connection
*conn
,
528 struct ctdb_req_header
*hdr
)
530 struct ctdb_req_message
*msg
;
532 if (hdr
->operation
!= CTDB_REQ_MESSAGE
) {
533 DEBUG(0, ("Received async msg of type %u, discarding\n",
538 msg
= (struct ctdb_req_message
*)hdr
;
540 ctdbd_msg_call_back(conn
, msg
);
546 * The ctdbd socket is readable asynchronuously
549 static void ctdbd_socket_handler(struct tevent_context
*event_ctx
,
550 struct tevent_fd
*event
,
554 struct ctdbd_connection
*conn
= talloc_get_type_abort(
555 private_data
, struct ctdbd_connection
);
556 struct ctdb_req_header
*hdr
= NULL
;
559 ret
= ctdb_read_packet(conn
->fd
, conn
->timeout
, talloc_tos(), &hdr
);
561 DEBUG(0, ("ctdb_read_packet failed: %s\n", strerror(ret
)));
562 cluster_fatal("ctdbd died\n");
565 ret
= ctdb_handle_message(conn
, hdr
);
570 DEBUG(10, ("could not handle incoming message: %s\n",
576 * Prepare a ctdbd connection to receive messages
579 int ctdbd_register_msg_ctx(struct ctdbd_connection
*conn
,
580 struct messaging_context
*msg_ctx
,
581 struct tevent_context
*ev
)
583 SMB_ASSERT(conn
->msg_ctx
== NULL
);
584 SMB_ASSERT(conn
->fde
== NULL
);
586 conn
->fde
= tevent_add_fd(ev
, conn
, conn
->fd
, TEVENT_FD_READ
,
587 ctdbd_socket_handler
, conn
);
588 if (conn
->fde
== NULL
) {
589 DEBUG(0, ("event_add_fd failed\n"));
593 conn
->msg_ctx
= msg_ctx
;
598 int ctdbd_messaging_send_iov(struct ctdbd_connection
*conn
,
599 uint32_t dst_vnn
, uint64_t dst_srvid
,
600 const struct iovec
*iov
, int iovlen
)
602 struct ctdb_req_message r
;
603 struct iovec iov2
[iovlen
+1];
604 size_t buflen
= iov_buflen(iov
, iovlen
);
607 r
.hdr
.length
= offsetof(struct ctdb_req_message
, data
) + buflen
;
608 r
.hdr
.ctdb_magic
= CTDB_MAGIC
;
609 r
.hdr
.ctdb_version
= CTDB_PROTOCOL
;
610 r
.hdr
.generation
= 1;
611 r
.hdr
.operation
= CTDB_REQ_MESSAGE
;
612 r
.hdr
.destnode
= dst_vnn
;
613 r
.hdr
.srcnode
= conn
->our_vnn
;
618 DEBUG(10, ("ctdbd_messaging_send: Sending ctdb packet\n"));
619 ctdb_packet_dump(&r
.hdr
);
621 iov2
[0].iov_base
= &r
;
622 iov2
[0].iov_len
= offsetof(struct ctdb_req_message
, data
);
623 memcpy(&iov2
[1], iov
, iovlen
* sizeof(struct iovec
));
625 nwritten
= write_data_iov(conn
->fd
, iov2
, iovlen
+1);
626 if (nwritten
== -1) {
627 DEBUG(3, ("write_data_iov failed: %s\n", strerror(errno
)));
628 cluster_fatal("cluster dispatch daemon msg write error\n");
635 * send/recv a generic ctdb control message
637 static int ctdbd_control(struct ctdbd_connection
*conn
,
638 uint32_t vnn
, uint32_t opcode
,
639 uint64_t srvid
, uint32_t flags
,
641 TALLOC_CTX
*mem_ctx
, TDB_DATA
*outdata
,
644 struct ctdb_req_control req
;
645 struct ctdb_req_header
*hdr
;
646 struct ctdb_reply_control
*reply
= NULL
;
652 req
.hdr
.length
= offsetof(struct ctdb_req_control
, data
) + data
.dsize
;
653 req
.hdr
.ctdb_magic
= CTDB_MAGIC
;
654 req
.hdr
.ctdb_version
= CTDB_PROTOCOL
;
655 req
.hdr
.operation
= CTDB_REQ_CONTROL
;
656 req
.hdr
.reqid
= ctdbd_next_reqid(conn
);
657 req
.hdr
.destnode
= vnn
;
660 req
.datalen
= data
.dsize
;
663 DEBUG(10, ("ctdbd_control: Sending ctdb packet\n"));
664 ctdb_packet_dump(&req
.hdr
);
666 iov
[0].iov_base
= &req
;
667 iov
[0].iov_len
= offsetof(struct ctdb_req_control
, data
);
668 iov
[1].iov_base
= data
.dptr
;
669 iov
[1].iov_len
= data
.dsize
;
671 nwritten
= write_data_iov(conn
->fd
, iov
, ARRAY_SIZE(iov
));
672 if (nwritten
== -1) {
673 DEBUG(3, ("write_data_iov failed: %s\n", strerror(errno
)));
674 cluster_fatal("cluster dispatch daemon msg write error\n");
677 if (flags
& CTDB_CTRL_FLAG_NOREPLY
) {
684 ret
= ctdb_read_req(conn
, req
.hdr
.reqid
, NULL
, &hdr
);
686 DEBUG(10, ("ctdb_read_req failed: %s\n", strerror(ret
)));
690 if (hdr
->operation
!= CTDB_REPLY_CONTROL
) {
691 DEBUG(0, ("received invalid reply\n"));
695 reply
= (struct ctdb_reply_control
*)hdr
;
698 if (!(outdata
->dptr
= (uint8_t *)talloc_memdup(
699 mem_ctx
, reply
->data
, reply
->datalen
))) {
703 outdata
->dsize
= reply
->datalen
;
706 (*cstatus
) = reply
->status
;
714 * see if a remote process exists
716 bool ctdbd_process_exists(struct ctdbd_connection
*conn
, uint32_t vnn
, pid_t pid
)
724 if (!ctdb_processes_exist(conn
, &id
, 1, &result
)) {
725 DEBUG(10, ("ctdb_processes_exist failed\n"));
731 bool ctdb_processes_exist(struct ctdbd_connection
*conn
,
732 const struct server_id
*pids
, int num_pids
,
735 TALLOC_CTX
*frame
= talloc_stackframe();
740 reqids
= talloc_array(talloc_tos(), uint32_t, num_pids
);
741 if (reqids
== NULL
) {
745 for (i
=0; i
<num_pids
; i
++) {
746 struct ctdb_req_control req
;
752 reqids
[i
] = ctdbd_next_reqid(conn
);
757 * pids[i].pid is uint64_t, scale down to pid_t which
758 * is the wire protocol towards ctdb.
762 DEBUG(10, ("Requesting PID %d/%d, reqid=%d\n",
763 (int)pids
[i
].vnn
, (int)pid
,
766 req
.hdr
.length
= offsetof(struct ctdb_req_control
, data
);
767 req
.hdr
.length
+= sizeof(pid
);
768 req
.hdr
.ctdb_magic
= CTDB_MAGIC
;
769 req
.hdr
.ctdb_version
= CTDB_PROTOCOL
;
770 req
.hdr
.operation
= CTDB_REQ_CONTROL
;
771 req
.hdr
.reqid
= reqids
[i
];
772 req
.hdr
.destnode
= pids
[i
].vnn
;
773 req
.opcode
= CTDB_CONTROL_PROCESS_EXISTS
;
775 req
.datalen
= sizeof(pid
);
778 DEBUG(10, ("ctdbd_control: Sending ctdb packet\n"));
779 ctdb_packet_dump(&req
.hdr
);
781 iov
[0].iov_base
= &req
;
782 iov
[0].iov_len
= offsetof(struct ctdb_req_control
, data
);
783 iov
[1].iov_base
= &pid
;
784 iov
[1].iov_len
= sizeof(pid
);
786 nwritten
= write_data_iov(conn
->fd
, iov
, ARRAY_SIZE(iov
));
787 if (nwritten
== -1) {
788 DEBUG(10, ("write_data_iov failed: %s\n",
796 while (num_received
< num_pids
) {
797 struct ctdb_req_header
*hdr
;
798 struct ctdb_reply_control
*reply
;
802 ret
= ctdb_read_req(conn
, 0, talloc_tos(), &hdr
);
804 DEBUG(10, ("ctdb_read_req failed: %s\n",
809 if (hdr
->operation
!= CTDB_REPLY_CONTROL
) {
810 DEBUG(10, ("Received invalid reply\n"));
813 reply
= (struct ctdb_reply_control
*)hdr
;
815 reqid
= reply
->hdr
.reqid
;
817 DEBUG(10, ("Received reqid %d\n", (int)reqid
));
819 for (i
=0; i
<num_pids
; i
++) {
820 if (reqid
== reqids
[i
]) {
825 DEBUG(10, ("Received unknown record number %u\n",
829 results
[i
] = ((reply
->status
) == 0);
843 char *ctdbd_dbpath(struct ctdbd_connection
*conn
,
844 TALLOC_CTX
*mem_ctx
, uint32_t db_id
)
848 TDB_DATA rdata
= {0};
851 data
.dptr
= (uint8_t*)&db_id
;
852 data
.dsize
= sizeof(db_id
);
854 ret
= ctdbd_control(conn
, CTDB_CURRENT_NODE
,
855 CTDB_CONTROL_GETDBPATH
, 0, 0, data
,
856 mem_ctx
, &rdata
, &cstatus
);
857 if ((ret
!= 0) || cstatus
!= 0) {
858 DEBUG(0, (__location__
" ctdb_control for getdbpath failed: %s\n",
863 return (char *)rdata
.dptr
;
867 * attach to a ctdb database
869 int ctdbd_db_attach(struct ctdbd_connection
*conn
,
870 const char *name
, uint32_t *db_id
, int tdb_flags
)
875 bool persistent
= (tdb_flags
& TDB_CLEAR_IF_FIRST
) == 0;
877 data
= string_term_tdb_data(name
);
879 ret
= ctdbd_control(conn
, CTDB_CURRENT_NODE
,
881 ? CTDB_CONTROL_DB_ATTACH_PERSISTENT
882 : CTDB_CONTROL_DB_ATTACH
,
883 tdb_flags
, 0, data
, NULL
, &data
, &cstatus
);
885 DEBUG(0, (__location__
" ctdb_control for db_attach "
886 "failed: %s\n", strerror(ret
)));
890 if (cstatus
!= 0 || data
.dsize
!= sizeof(uint32_t)) {
891 DEBUG(0,(__location__
" ctdb_control for db_attach failed\n"));
895 *db_id
= *(uint32_t *)data
.dptr
;
896 talloc_free(data
.dptr
);
898 if (!(tdb_flags
& TDB_SEQNUM
)) {
902 data
.dptr
= (uint8_t *)db_id
;
903 data
.dsize
= sizeof(*db_id
);
905 ret
= ctdbd_control(conn
, CTDB_CURRENT_NODE
,
906 CTDB_CONTROL_ENABLE_SEQNUM
, 0, 0, data
,
907 NULL
, NULL
, &cstatus
);
908 if ((ret
!= 0) || cstatus
!= 0) {
909 DEBUG(0, (__location__
" ctdb_control for enable seqnum "
910 "failed: %s\n", strerror(ret
)));
911 return (ret
== 0) ? EIO
: ret
;
918 * force the migration of a record to this node
920 int ctdbd_migrate(struct ctdbd_connection
*conn
, uint32_t db_id
, TDB_DATA key
)
922 struct ctdb_req_call req
;
923 struct ctdb_req_header
*hdr
;
930 req
.hdr
.length
= offsetof(struct ctdb_req_call
, data
) + key
.dsize
;
931 req
.hdr
.ctdb_magic
= CTDB_MAGIC
;
932 req
.hdr
.ctdb_version
= CTDB_PROTOCOL
;
933 req
.hdr
.operation
= CTDB_REQ_CALL
;
934 req
.hdr
.reqid
= ctdbd_next_reqid(conn
);
935 req
.flags
= CTDB_IMMEDIATE_MIGRATION
;
936 req
.callid
= CTDB_NULL_FUNC
;
938 req
.keylen
= key
.dsize
;
940 DEBUG(10, ("ctdbd_migrate: Sending ctdb packet\n"));
941 ctdb_packet_dump(&req
.hdr
);
943 iov
[0].iov_base
= &req
;
944 iov
[0].iov_len
= offsetof(struct ctdb_req_call
, data
);
945 iov
[1].iov_base
= key
.dptr
;
946 iov
[1].iov_len
= key
.dsize
;
948 nwritten
= write_data_iov(conn
->fd
, iov
, ARRAY_SIZE(iov
));
949 if (nwritten
== -1) {
950 DEBUG(3, ("write_data_iov failed: %s\n", strerror(errno
)));
951 cluster_fatal("cluster dispatch daemon msg write error\n");
954 ret
= ctdb_read_req(conn
, req
.hdr
.reqid
, NULL
, &hdr
);
956 DEBUG(10, ("ctdb_read_req failed: %s\n", strerror(ret
)));
960 if (hdr
->operation
!= CTDB_REPLY_CALL
) {
961 DEBUG(0, ("received invalid reply\n"));
972 * Fetch a record and parse it
974 int ctdbd_parse(struct ctdbd_connection
*conn
, uint32_t db_id
,
975 TDB_DATA key
, bool local_copy
,
976 void (*parser
)(TDB_DATA key
, TDB_DATA data
,
980 struct ctdb_req_call req
;
981 struct ctdb_req_header
*hdr
= NULL
;
982 struct ctdb_reply_call
*reply
;
988 flags
= local_copy
? CTDB_WANT_READONLY
: 0;
992 req
.hdr
.length
= offsetof(struct ctdb_req_call
, data
) + key
.dsize
;
993 req
.hdr
.ctdb_magic
= CTDB_MAGIC
;
994 req
.hdr
.ctdb_version
= CTDB_PROTOCOL
;
995 req
.hdr
.operation
= CTDB_REQ_CALL
;
996 req
.hdr
.reqid
= ctdbd_next_reqid(conn
);
998 req
.callid
= CTDB_FETCH_FUNC
;
1000 req
.keylen
= key
.dsize
;
1002 iov
[0].iov_base
= &req
;
1003 iov
[0].iov_len
= offsetof(struct ctdb_req_call
, data
);
1004 iov
[1].iov_base
= key
.dptr
;
1005 iov
[1].iov_len
= key
.dsize
;
1007 nwritten
= write_data_iov(conn
->fd
, iov
, ARRAY_SIZE(iov
));
1008 if (nwritten
== -1) {
1009 DEBUG(3, ("write_data_iov failed: %s\n", strerror(errno
)));
1010 cluster_fatal("cluster dispatch daemon msg write error\n");
1013 ret
= ctdb_read_req(conn
, req
.hdr
.reqid
, NULL
, &hdr
);
1015 DEBUG(10, ("ctdb_read_req failed: %s\n", strerror(ret
)));
1019 if ((hdr
== NULL
) || (hdr
->operation
!= CTDB_REPLY_CALL
)) {
1020 DEBUG(0, ("received invalid reply\n"));
1024 reply
= (struct ctdb_reply_call
*)hdr
;
1026 if (reply
->datalen
== 0) {
1028 * Treat an empty record as non-existing
1034 parser(key
, make_tdb_data(&reply
->data
[0], reply
->datalen
),
1044 Traverse a ctdb database. This uses a kind-of hackish way to open a second
1045 connection to ctdbd to avoid the hairy recursive and async problems with
1049 int ctdbd_traverse(struct ctdbd_connection
*master
, uint32_t db_id
,
1050 void (*fn
)(TDB_DATA key
, TDB_DATA data
,
1051 void *private_data
),
1054 struct ctdbd_connection
*conn
;
1057 struct ctdb_traverse_start t
;
1061 ret
= ctdbd_init_connection(NULL
, master
->sockname
, master
->timeout
,
1065 DEBUG(0, ("ctdbd_init_connection failed: %s\n",
1071 t
.srvid
= conn
->rand_srvid
;
1072 t
.reqid
= ctdbd_next_reqid(conn
);
1074 data
.dptr
= (uint8_t *)&t
;
1075 data
.dsize
= sizeof(t
);
1077 ret
= ctdbd_control(conn
, CTDB_CURRENT_NODE
,
1078 CTDB_CONTROL_TRAVERSE_START
, conn
->rand_srvid
,
1079 0, data
, NULL
, NULL
, &cstatus
);
1081 if ((ret
!= 0) || (cstatus
!= 0)) {
1082 DEBUG(0,("ctdbd_control failed: %s, %d\n", strerror(ret
),
1087 * We need a mapping here
1096 struct ctdb_req_header
*hdr
= NULL
;
1097 struct ctdb_req_message
*m
;
1098 struct ctdb_rec_data
*d
;
1100 ret
= ctdb_read_packet(conn
->fd
, conn
->timeout
, conn
, &hdr
);
1102 DEBUG(0, ("ctdb_read_packet failed: %s\n",
1104 cluster_fatal("ctdbd died\n");
1107 if (hdr
->operation
!= CTDB_REQ_MESSAGE
) {
1108 DEBUG(0, ("Got operation %u, expected a message\n",
1109 (unsigned)hdr
->operation
));
1114 m
= (struct ctdb_req_message
*)hdr
;
1115 d
= (struct ctdb_rec_data
*)&m
->data
[0];
1116 if (m
->datalen
< sizeof(uint32_t) || m
->datalen
!= d
->length
) {
1117 DEBUG(0, ("Got invalid traverse data of length %d\n",
1123 key
.dsize
= d
->keylen
;
1124 key
.dptr
= &d
->data
[0];
1125 data
.dsize
= d
->datalen
;
1126 data
.dptr
= &d
->data
[d
->keylen
];
1128 if (key
.dsize
== 0 && data
.dsize
== 0) {
1129 /* end of traverse */
1134 if (data
.dsize
< sizeof(struct ctdb_ltdb_header
)) {
1135 DEBUG(0, ("Got invalid ltdb header length %d\n",
1140 data
.dsize
-= sizeof(struct ctdb_ltdb_header
);
1141 data
.dptr
+= sizeof(struct ctdb_ltdb_header
);
1144 fn(key
, data
, private_data
);
1151 This is used to canonicalize a ctdb_sock_addr structure.
1153 static void smbd_ctdb_canonicalize_ip(const struct sockaddr_storage
*in
,
1154 struct sockaddr_storage
*out
)
1156 memcpy(out
, in
, sizeof (*out
));
1159 if (in
->ss_family
== AF_INET6
) {
1160 const char prefix
[12] = { 0,0,0,0,0,0,0,0,0,0,0xff,0xff };
1161 const struct sockaddr_in6
*in6
=
1162 (const struct sockaddr_in6
*)in
;
1163 struct sockaddr_in
*out4
= (struct sockaddr_in
*)out
;
1164 if (memcmp(&in6
->sin6_addr
, prefix
, 12) == 0) {
1165 memset(out
, 0, sizeof(*out
));
1166 #ifdef HAVE_SOCK_SIN_LEN
1167 out4
->sin_len
= sizeof(*out
);
1169 out4
->sin_family
= AF_INET
;
1170 out4
->sin_port
= in6
->sin6_port
;
1171 memcpy(&out4
->sin_addr
, &in6
->sin6_addr
.s6_addr
[12], 4);
1178 * Register us as a server for a particular tcp connection
1181 int ctdbd_register_ips(struct ctdbd_connection
*conn
,
1182 const struct sockaddr_storage
*_server
,
1183 const struct sockaddr_storage
*_client
,
1184 int (*cb
)(uint32_t src_vnn
, uint32_t dst_vnn
,
1186 const uint8_t *msg
, size_t msglen
,
1187 void *private_data
),
1190 struct ctdb_control_tcp_addr p
;
1191 TDB_DATA data
= { .dptr
= (uint8_t *)&p
, .dsize
= sizeof(p
) };
1193 struct sockaddr_storage client
;
1194 struct sockaddr_storage server
;
1197 * Only one connection so far
1200 smbd_ctdb_canonicalize_ip(_client
, &client
);
1201 smbd_ctdb_canonicalize_ip(_server
, &server
);
1203 switch (client
.ss_family
) {
1205 memcpy(&p
.dest
.ip
, &server
, sizeof(p
.dest
.ip
));
1206 memcpy(&p
.src
.ip
, &client
, sizeof(p
.src
.ip
));
1209 memcpy(&p
.dest
.ip6
, &server
, sizeof(p
.dest
.ip6
));
1210 memcpy(&p
.src
.ip6
, &client
, sizeof(p
.src
.ip6
));
1217 * We want to be told about IP releases
1220 ret
= register_with_ctdbd(conn
, CTDB_SRVID_RELEASE_IP
,
1227 * inform ctdb of our tcp connection, so if IP takeover happens ctdb
1228 * can send an extra ack to trigger a reset for our client, so it
1229 * immediately reconnects
1231 ret
= ctdbd_control(conn
, CTDB_CURRENT_NODE
,
1232 CTDB_CONTROL_TCP_CLIENT
, 0,
1233 CTDB_CTRL_FLAG_NOREPLY
, data
, NULL
, NULL
,
1242 call a control on the local node
1244 int ctdbd_control_local(struct ctdbd_connection
*conn
, uint32_t opcode
,
1245 uint64_t srvid
, uint32_t flags
, TDB_DATA data
,
1246 TALLOC_CTX
*mem_ctx
, TDB_DATA
*outdata
,
1249 return ctdbd_control(conn
, CTDB_CURRENT_NODE
, opcode
, srvid
, flags
, data
,
1250 mem_ctx
, outdata
, cstatus
);
1253 int ctdb_watch_us(struct ctdbd_connection
*conn
)
1255 struct ctdb_client_notify_register reg_data
;
1260 reg_data
.srvid
= CTDB_SRVID_SAMBA_NOTIFY
;
1262 reg_data
.notify_data
[0] = 0;
1264 struct_len
= offsetof(struct ctdb_client_notify_register
,
1265 notify_data
) + reg_data
.len
;
1267 ret
= ctdbd_control_local(
1268 conn
, CTDB_CONTROL_REGISTER_NOTIFY
, conn
->rand_srvid
, 0,
1269 make_tdb_data((uint8_t *)®_data
, struct_len
),
1270 NULL
, NULL
, &cstatus
);
1272 DEBUG(1, ("ctdbd_control_local failed: %s\n",
1278 int ctdb_unwatch(struct ctdbd_connection
*conn
)
1280 struct ctdb_client_notify_deregister dereg_data
;
1284 dereg_data
.srvid
= CTDB_SRVID_SAMBA_NOTIFY
;
1286 ret
= ctdbd_control_local(
1287 conn
, CTDB_CONTROL_DEREGISTER_NOTIFY
, conn
->rand_srvid
, 0,
1288 make_tdb_data((uint8_t *)&dereg_data
, sizeof(dereg_data
)),
1289 NULL
, NULL
, &cstatus
);
1291 DEBUG(1, ("ctdbd_control_local failed: %s\n",
1297 int ctdbd_probe(const char *sockname
, int timeout
)
1300 * Do a very early check if ctdbd is around to avoid an abort and core
1303 struct ctdbd_connection
*conn
= NULL
;
1306 ret
= ctdbd_messaging_connection(talloc_tos(), sockname
, timeout
,
1310 * We only care if we can connect.