s3-winbindd: rework reconnect logic in winbindd_lookup_sids().
[Samba.git] / source3 / lib / ctdbd_conn.c
blob84f26e00d3f4d6048cbe71de08fbea3748c08831
1 /*
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/>.
21 #include "includes.h"
22 #include "util_tdb.h"
23 #include "serverid.h"
24 #include "ctdbd_conn.h"
26 #ifdef CLUSTER_SUPPORT
28 #include "ctdb_packet.h"
29 #include "messages.h"
32 * It is not possible to include ctdb.h and tdb_compat.h (included via
33 * some other include above) without warnings. This fixes those
34 * warnings.
37 #ifdef typesafe_cb
38 #undef typesafe_cb
39 #endif
41 #ifdef typesafe_cb_preargs
42 #undef typesafe_cb_preargs
43 #endif
45 #ifdef typesafe_cb_postargs
46 #undef typesafe_cb_postargs
47 #endif
49 /* paths to these include files come from --with-ctdb= in configure */
51 #include "ctdb.h"
52 #include "ctdb_private.h"
54 struct ctdbd_connection {
55 struct messaging_context *msg_ctx;
56 uint32_t reqid;
57 uint32_t our_vnn;
58 uint64_t rand_srvid;
59 struct ctdb_packet_context *pkt;
60 struct fd_event *fde;
62 void (*release_ip_handler)(const char *ip_addr, void *private_data);
63 void *release_ip_priv;
66 static uint32_t ctdbd_next_reqid(struct ctdbd_connection *conn)
68 conn->reqid += 1;
69 if (conn->reqid == 0) {
70 conn->reqid += 1;
72 return conn->reqid;
75 static NTSTATUS ctdbd_control(struct ctdbd_connection *conn,
76 uint32_t vnn, uint32_t opcode,
77 uint64_t srvid, uint32_t flags, TDB_DATA data,
78 TALLOC_CTX *mem_ctx, TDB_DATA *outdata,
79 int *cstatus);
82 * exit on fatal communications errors with the ctdbd daemon
84 static void cluster_fatal(const char *why)
86 DEBUG(0,("cluster fatal event: %s - exiting immediately\n", why));
87 /* we don't use smb_panic() as we don't want to delay to write
88 a core file. We need to release this process id immediately
89 so that someone else can take over without getting sharing
90 violations */
91 _exit(1);
97 static void ctdb_packet_dump(struct ctdb_req_header *hdr)
99 if (DEBUGLEVEL < 11) {
100 return;
102 DEBUGADD(11, ("len=%d, magic=%x, vers=%d, gen=%d, op=%d, reqid=%d\n",
103 (int)hdr->length, (int)hdr->ctdb_magic,
104 (int)hdr->ctdb_version, (int)hdr->generation,
105 (int)hdr->operation, (int)hdr->reqid));
109 * Register a srvid with ctdbd
111 NTSTATUS register_with_ctdbd(struct ctdbd_connection *conn, uint64_t srvid)
114 int cstatus;
115 return ctdbd_control(conn, CTDB_CURRENT_NODE,
116 CTDB_CONTROL_REGISTER_SRVID, srvid, 0,
117 tdb_null, NULL, NULL, &cstatus);
121 * get our vnn from the cluster
123 static NTSTATUS get_cluster_vnn(struct ctdbd_connection *conn, uint32_t *vnn)
125 int32_t cstatus=-1;
126 NTSTATUS status;
127 status = ctdbd_control(conn,
128 CTDB_CURRENT_NODE, CTDB_CONTROL_GET_PNN, 0, 0,
129 tdb_null, NULL, NULL, &cstatus);
130 if (!NT_STATUS_IS_OK(status)) {
131 cluster_fatal("ctdbd_control failed\n");
133 *vnn = (uint32_t)cstatus;
134 return status;
138 * Are we active (i.e. not banned or stopped?)
140 static bool ctdbd_working(struct ctdbd_connection *conn, uint32_t vnn)
142 int32_t cstatus=-1;
143 NTSTATUS status;
144 TDB_DATA outdata;
145 struct ctdb_node_map *m;
146 uint32_t failure_flags;
147 bool ret = false;
148 int i;
150 status = ctdbd_control(conn, CTDB_CURRENT_NODE,
151 CTDB_CONTROL_GET_NODEMAP, 0, 0,
152 tdb_null, talloc_tos(), &outdata, &cstatus);
153 if (!NT_STATUS_IS_OK(status)) {
154 cluster_fatal("ctdbd_control failed\n");
156 if ((cstatus != 0) || (outdata.dptr == NULL)) {
157 DEBUG(2, ("Received invalid ctdb data\n"));
158 return false;
161 m = (struct ctdb_node_map *)outdata.dptr;
163 for (i=0; i<m->num; i++) {
164 if (vnn == m->nodes[i].pnn) {
165 break;
169 if (i == m->num) {
170 DEBUG(2, ("Did not find ourselves (node %d) in nodemap\n",
171 (int)vnn));
172 goto fail;
175 failure_flags = NODE_FLAGS_BANNED | NODE_FLAGS_DISCONNECTED
176 | NODE_FLAGS_PERMANENTLY_DISABLED | NODE_FLAGS_STOPPED;
178 if ((m->nodes[i].flags & failure_flags) != 0) {
179 DEBUG(2, ("Node has status %x, not active\n",
180 (int)m->nodes[i].flags));
181 goto fail;
184 ret = true;
185 fail:
186 TALLOC_FREE(outdata.dptr);
187 return ret;
190 uint32_t ctdbd_vnn(const struct ctdbd_connection *conn)
192 return conn->our_vnn;
196 * Get us a ctdb connection
199 static NTSTATUS ctdbd_connect(TALLOC_CTX *mem_ctx,
200 struct ctdb_packet_context **presult)
202 struct ctdb_packet_context *result;
203 const char *sockname = lp_ctdbd_socket();
204 struct sockaddr_un addr;
205 int fd;
206 socklen_t salen;
208 fd = socket(AF_UNIX, SOCK_STREAM, 0);
209 if (fd == -1) {
210 DEBUG(3, ("Could not create socket: %s\n", strerror(errno)));
211 return map_nt_error_from_unix(errno);
214 ZERO_STRUCT(addr);
215 addr.sun_family = AF_UNIX;
216 strncpy(addr.sun_path, sockname, sizeof(addr.sun_path));
218 salen = sizeof(struct sockaddr_un);
219 if (connect(fd, (struct sockaddr *)(void *)&addr, salen) == -1) {
220 DEBUG(1, ("connect(%s) failed: %s\n", sockname,
221 strerror(errno)));
222 close(fd);
223 return map_nt_error_from_unix(errno);
226 if (!(result = ctdb_packet_init(mem_ctx, fd))) {
227 close(fd);
228 return NT_STATUS_NO_MEMORY;
231 *presult = result;
232 return NT_STATUS_OK;
236 * Do we have a complete ctdb packet in the queue?
239 static bool ctdb_req_complete(const uint8_t *buf, size_t available,
240 size_t *length,
241 void *private_data)
243 uint32_t msglen;
245 if (available < sizeof(msglen)) {
246 return False;
249 msglen = *((const uint32_t *)buf);
251 DEBUG(11, ("msglen = %d\n", msglen));
253 if (msglen < sizeof(struct ctdb_req_header)) {
254 DEBUG(0, ("Got invalid msglen: %d, expected at least %d for "
255 "the req_header\n", (int)msglen,
256 (int)sizeof(struct ctdb_req_header)));
257 cluster_fatal("ctdbd protocol error\n");
260 if (available < msglen) {
261 return false;
264 *length = msglen;
265 return true;
269 * State necessary to defer an incoming message while we are waiting for a
270 * ctdb reply.
273 struct deferred_msg_state {
274 struct messaging_context *msg_ctx;
275 struct messaging_rec *rec;
279 * Timed event handler for the deferred message
282 static void deferred_message_dispatch(struct event_context *event_ctx,
283 struct timed_event *te,
284 struct timeval now,
285 void *private_data)
287 struct deferred_msg_state *state = talloc_get_type_abort(
288 private_data, struct deferred_msg_state);
290 messaging_dispatch_rec(state->msg_ctx, state->rec);
291 TALLOC_FREE(state);
292 TALLOC_FREE(te);
295 struct req_pull_state {
296 TALLOC_CTX *mem_ctx;
297 DATA_BLOB req;
301 * Pull a ctdb request out of the incoming ctdb_packet queue
304 static NTSTATUS ctdb_req_pull(uint8_t *buf, size_t length,
305 void *private_data)
307 struct req_pull_state *state = (struct req_pull_state *)private_data;
309 state->req.data = talloc_move(state->mem_ctx, &buf);
310 state->req.length = length;
311 return NT_STATUS_OK;
315 * Fetch a messaging_rec from an incoming ctdb style message
318 static struct messaging_rec *ctdb_pull_messaging_rec(TALLOC_CTX *mem_ctx,
319 size_t overall_length,
320 struct ctdb_req_message *msg)
322 struct messaging_rec *result;
323 DATA_BLOB blob;
324 enum ndr_err_code ndr_err;
326 if ((overall_length < offsetof(struct ctdb_req_message, data))
327 || (overall_length
328 < offsetof(struct ctdb_req_message, data) + msg->datalen)) {
330 cluster_fatal("got invalid msg length");
333 if (!(result = talloc(mem_ctx, struct messaging_rec))) {
334 DEBUG(0, ("talloc failed\n"));
335 return NULL;
338 blob = data_blob_const(msg->data, msg->datalen);
340 ndr_err = ndr_pull_struct_blob(
341 &blob, result, result,
342 (ndr_pull_flags_fn_t)ndr_pull_messaging_rec);
344 if (!NDR_ERR_CODE_IS_SUCCESS(ndr_err)) {
345 DEBUG(0, ("ndr_pull_struct_blob failed: %s\n",
346 ndr_errstr(ndr_err)));
347 TALLOC_FREE(result);
348 return NULL;
351 if (DEBUGLEVEL >= 11) {
352 DEBUG(11, ("ctdb_pull_messaging_rec:\n"));
353 NDR_PRINT_DEBUG(messaging_rec, result);
356 return result;
359 static NTSTATUS ctdb_packet_fd_read_sync(struct ctdb_packet_context *ctx)
361 int timeout = lp_ctdb_timeout();
363 if (timeout == 0) {
364 timeout = -1;
366 return ctdb_packet_fd_read_sync_timeout(ctx, timeout);
370 * Read a full ctdbd request. If we have a messaging context, defer incoming
371 * messages that might come in between.
374 static NTSTATUS ctdb_read_req(struct ctdbd_connection *conn, uint32_t reqid,
375 TALLOC_CTX *mem_ctx, void *result)
377 struct ctdb_req_header *hdr;
378 struct req_pull_state state;
379 NTSTATUS status;
381 next_pkt:
382 ZERO_STRUCT(state);
383 state.mem_ctx = mem_ctx;
385 while (!ctdb_packet_handler(conn->pkt, ctdb_req_complete,
386 ctdb_req_pull, &state, &status)) {
388 * Not enough data
390 status = ctdb_packet_fd_read_sync(conn->pkt);
392 if (NT_STATUS_EQUAL(status, NT_STATUS_NETWORK_BUSY)) {
393 /* EAGAIN */
394 continue;
395 } else if (NT_STATUS_EQUAL(status, NT_STATUS_RETRY)) {
396 /* EAGAIN */
397 continue;
400 if (!NT_STATUS_IS_OK(status)) {
401 DEBUG(0, ("packet_fd_read failed: %s\n", nt_errstr(status)));
402 cluster_fatal("ctdbd died\n");
406 if (!NT_STATUS_IS_OK(status)) {
407 DEBUG(0, ("Could not read ctdb_packet: %s\n", nt_errstr(status)));
408 cluster_fatal("ctdbd died\n");
411 hdr = (struct ctdb_req_header *)state.req.data;
413 DEBUG(11, ("Received ctdb packet\n"));
414 ctdb_packet_dump(hdr);
416 if (hdr->operation == CTDB_REQ_MESSAGE) {
417 struct timed_event *evt;
418 struct deferred_msg_state *msg_state;
419 struct ctdb_req_message *msg = (struct ctdb_req_message *)hdr;
421 if (conn->msg_ctx == NULL) {
422 DEBUG(1, ("Got a message without having a msg ctx, "
423 "dropping msg %llu\n",
424 (long long unsigned)msg->srvid));
425 goto next_pkt;
428 if ((conn->release_ip_handler != NULL)
429 && (msg->srvid == CTDB_SRVID_RELEASE_IP)) {
430 /* must be dispatched immediately */
431 DEBUG(10, ("received CTDB_SRVID_RELEASE_IP\n"));
432 conn->release_ip_handler((const char *)msg->data,
433 conn->release_ip_priv);
434 TALLOC_FREE(hdr);
435 goto next_pkt;
438 if ((msg->srvid == CTDB_SRVID_RECONFIGURE)
439 || (msg->srvid == CTDB_SRVID_SAMBA_NOTIFY)) {
441 DEBUG(1, ("ctdb_read_req: Got %s message\n",
442 (msg->srvid == CTDB_SRVID_RECONFIGURE)
443 ? "cluster reconfigure" : "SAMBA_NOTIFY"));
445 messaging_send(conn->msg_ctx,
446 messaging_server_id(conn->msg_ctx),
447 MSG_SMB_BRL_VALIDATE, &data_blob_null);
448 messaging_send(conn->msg_ctx,
449 messaging_server_id(conn->msg_ctx),
450 MSG_DBWRAP_G_LOCK_RETRY,
451 &data_blob_null);
452 TALLOC_FREE(hdr);
453 goto next_pkt;
456 msg_state = talloc(NULL, struct deferred_msg_state);
457 if (msg_state == NULL) {
458 DEBUG(0, ("talloc failed\n"));
459 TALLOC_FREE(hdr);
460 goto next_pkt;
463 if (!(msg_state->rec = ctdb_pull_messaging_rec(
464 msg_state, state.req.length, msg))) {
465 DEBUG(0, ("ctdbd_pull_messaging_rec failed\n"));
466 TALLOC_FREE(msg_state);
467 TALLOC_FREE(hdr);
468 goto next_pkt;
471 TALLOC_FREE(hdr);
473 msg_state->msg_ctx = conn->msg_ctx;
476 * We're waiting for a call reply, but an async message has
477 * crossed. Defer dispatching to the toplevel event loop.
479 evt = event_add_timed(conn->msg_ctx->event_ctx,
480 conn->msg_ctx->event_ctx,
481 timeval_zero(),
482 deferred_message_dispatch,
483 msg_state);
484 if (evt == NULL) {
485 DEBUG(0, ("event_add_timed failed\n"));
486 TALLOC_FREE(msg_state);
487 TALLOC_FREE(hdr);
488 goto next_pkt;
491 goto next_pkt;
494 if ((reqid != 0) && (hdr->reqid != reqid)) {
495 /* we got the wrong reply */
496 DEBUG(0,("Discarding mismatched ctdb reqid %u should have "
497 "been %u\n", hdr->reqid, reqid));
498 TALLOC_FREE(hdr);
499 goto next_pkt;
502 *((void **)result) = talloc_move(mem_ctx, &hdr);
504 return NT_STATUS_OK;
508 * Get us a ctdbd connection
511 static NTSTATUS ctdbd_init_connection(TALLOC_CTX *mem_ctx,
512 struct ctdbd_connection **pconn)
514 struct ctdbd_connection *conn;
515 NTSTATUS status;
517 if (!(conn = talloc_zero(mem_ctx, struct ctdbd_connection))) {
518 DEBUG(0, ("talloc failed\n"));
519 return NT_STATUS_NO_MEMORY;
522 status = ctdbd_connect(conn, &conn->pkt);
524 if (!NT_STATUS_IS_OK(status)) {
525 DEBUG(10, ("ctdbd_connect failed: %s\n", nt_errstr(status)));
526 goto fail;
529 status = get_cluster_vnn(conn, &conn->our_vnn);
531 if (!NT_STATUS_IS_OK(status)) {
532 DEBUG(10, ("get_cluster_vnn failed: %s\n", nt_errstr(status)));
533 goto fail;
536 if (!ctdbd_working(conn, conn->our_vnn)) {
537 DEBUG(2, ("Node is not working, can not connect\n"));
538 status = NT_STATUS_INTERNAL_DB_ERROR;
539 goto fail;
542 generate_random_buffer((unsigned char *)&conn->rand_srvid,
543 sizeof(conn->rand_srvid));
545 status = register_with_ctdbd(conn, conn->rand_srvid);
547 if (!NT_STATUS_IS_OK(status)) {
548 DEBUG(5, ("Could not register random srvid: %s\n",
549 nt_errstr(status)));
550 goto fail;
553 *pconn = conn;
554 return NT_STATUS_OK;
556 fail:
557 TALLOC_FREE(conn);
558 return status;
562 * Get us a ctdbd connection and register us as a process
565 NTSTATUS ctdbd_messaging_connection(TALLOC_CTX *mem_ctx,
566 struct ctdbd_connection **pconn)
568 struct ctdbd_connection *conn;
569 NTSTATUS status;
571 status = ctdbd_init_connection(mem_ctx, &conn);
573 if (!NT_STATUS_IS_OK(status)) {
574 return status;
577 status = register_with_ctdbd(conn, (uint64_t)getpid());
578 if (!NT_STATUS_IS_OK(status)) {
579 goto fail;
582 status = register_with_ctdbd(conn, MSG_SRVID_SAMBA);
583 if (!NT_STATUS_IS_OK(status)) {
584 goto fail;
587 status = register_with_ctdbd(conn, CTDB_SRVID_SAMBA_NOTIFY);
588 if (!NT_STATUS_IS_OK(status)) {
589 goto fail;
592 *pconn = conn;
593 return NT_STATUS_OK;
595 fail:
596 TALLOC_FREE(conn);
597 return status;
600 struct messaging_context *ctdb_conn_msg_ctx(struct ctdbd_connection *conn)
602 return conn->msg_ctx;
605 int ctdbd_conn_get_fd(struct ctdbd_connection *conn)
607 return ctdb_packet_get_fd(conn->pkt);
611 * Packet handler to receive and handle a ctdb message
613 static NTSTATUS ctdb_handle_message(uint8_t *buf, size_t length,
614 void *private_data)
616 struct ctdbd_connection *conn = talloc_get_type_abort(
617 private_data, struct ctdbd_connection);
618 struct ctdb_req_message *msg;
619 struct messaging_rec *msg_rec;
621 msg = (struct ctdb_req_message *)buf;
623 if (msg->hdr.operation != CTDB_REQ_MESSAGE) {
624 DEBUG(0, ("Received async msg of type %u, discarding\n",
625 msg->hdr.operation));
626 TALLOC_FREE(buf);
627 return NT_STATUS_INVALID_PARAMETER;
630 if ((conn->release_ip_handler != NULL)
631 && (msg->srvid == CTDB_SRVID_RELEASE_IP)) {
632 /* must be dispatched immediately */
633 DEBUG(10, ("received CTDB_SRVID_RELEASE_IP\n"));
634 conn->release_ip_handler((const char *)msg->data,
635 conn->release_ip_priv);
636 TALLOC_FREE(buf);
637 return NT_STATUS_OK;
640 SMB_ASSERT(conn->msg_ctx != NULL);
642 if ((msg->srvid == CTDB_SRVID_RECONFIGURE)
643 || (msg->srvid == CTDB_SRVID_SAMBA_NOTIFY)){
644 DEBUG(0,("Got cluster reconfigure message\n"));
646 * when the cluster is reconfigured or someone of the
647 * family has passed away (SAMBA_NOTIFY), we need to
648 * clean the brl database
650 messaging_send(conn->msg_ctx,
651 messaging_server_id(conn->msg_ctx),
652 MSG_SMB_BRL_VALIDATE, &data_blob_null);
654 messaging_send(conn->msg_ctx,
655 messaging_server_id(conn->msg_ctx),
656 MSG_DBWRAP_G_LOCK_RETRY,
657 &data_blob_null);
659 TALLOC_FREE(buf);
660 return NT_STATUS_OK;
663 /* only messages to our pid or the broadcast are valid here */
664 if (msg->srvid != getpid() && msg->srvid != MSG_SRVID_SAMBA) {
665 DEBUG(0,("Got unexpected message with srvid=%llu\n",
666 (unsigned long long)msg->srvid));
667 TALLOC_FREE(buf);
668 return NT_STATUS_OK;
671 if (!(msg_rec = ctdb_pull_messaging_rec(NULL, length, msg))) {
672 DEBUG(10, ("ctdb_pull_messaging_rec failed\n"));
673 TALLOC_FREE(buf);
674 return NT_STATUS_NO_MEMORY;
677 messaging_dispatch_rec(conn->msg_ctx, msg_rec);
679 TALLOC_FREE(msg_rec);
680 TALLOC_FREE(buf);
681 return NT_STATUS_OK;
685 * The ctdbd socket is readable asynchronuously
688 static void ctdbd_socket_handler(struct event_context *event_ctx,
689 struct fd_event *event,
690 uint16 flags,
691 void *private_data)
693 struct ctdbd_connection *conn = talloc_get_type_abort(
694 private_data, struct ctdbd_connection);
696 NTSTATUS status;
698 status = ctdb_packet_fd_read(conn->pkt);
700 if (!NT_STATUS_IS_OK(status)) {
701 DEBUG(0, ("packet_fd_read failed: %s\n", nt_errstr(status)));
702 cluster_fatal("ctdbd died\n");
705 while (ctdb_packet_handler(conn->pkt, ctdb_req_complete,
706 ctdb_handle_message, conn, &status)) {
707 if (!NT_STATUS_IS_OK(status)) {
708 DEBUG(10, ("could not handle incoming message: %s\n",
709 nt_errstr(status)));
715 * Prepare a ctdbd connection to receive messages
718 NTSTATUS ctdbd_register_msg_ctx(struct ctdbd_connection *conn,
719 struct messaging_context *msg_ctx)
721 SMB_ASSERT(conn->msg_ctx == NULL);
722 SMB_ASSERT(conn->fde == NULL);
724 if (!(conn->fde = event_add_fd(msg_ctx->event_ctx, conn,
725 ctdb_packet_get_fd(conn->pkt),
726 EVENT_FD_READ,
727 ctdbd_socket_handler,
728 conn))) {
729 DEBUG(0, ("event_add_fd failed\n"));
730 return NT_STATUS_NO_MEMORY;
733 conn->msg_ctx = msg_ctx;
735 return NT_STATUS_OK;
739 * Send a messaging message across a ctdbd
742 NTSTATUS ctdbd_messaging_send(struct ctdbd_connection *conn,
743 uint32_t dst_vnn, uint64_t dst_srvid,
744 struct messaging_rec *msg)
746 DATA_BLOB blob;
747 NTSTATUS status;
748 enum ndr_err_code ndr_err;
750 ndr_err = ndr_push_struct_blob(
751 &blob, talloc_tos(), msg,
752 (ndr_push_flags_fn_t)ndr_push_messaging_rec);
754 if (!NDR_ERR_CODE_IS_SUCCESS(ndr_err)) {
755 DEBUG(0, ("ndr_push_struct_blob failed: %s\n",
756 ndr_errstr(ndr_err)));
757 return ndr_map_error2ntstatus(ndr_err);
760 status = ctdbd_messaging_send_blob(conn, dst_vnn, dst_srvid,
761 blob.data, blob.length);
762 TALLOC_FREE(blob.data);
763 return status;
766 NTSTATUS ctdbd_messaging_send_blob(struct ctdbd_connection *conn,
767 uint32_t dst_vnn, uint64_t dst_srvid,
768 const uint8_t *buf, size_t buflen)
770 struct ctdb_req_message r;
771 NTSTATUS status;
773 r.hdr.length = offsetof(struct ctdb_req_message, data) + buflen;
774 r.hdr.ctdb_magic = CTDB_MAGIC;
775 r.hdr.ctdb_version = CTDB_VERSION;
776 r.hdr.generation = 1;
777 r.hdr.operation = CTDB_REQ_MESSAGE;
778 r.hdr.destnode = dst_vnn;
779 r.hdr.srcnode = conn->our_vnn;
780 r.hdr.reqid = 0;
781 r.srvid = dst_srvid;
782 r.datalen = buflen;
784 DEBUG(10, ("ctdbd_messaging_send: Sending ctdb packet\n"));
785 ctdb_packet_dump(&r.hdr);
787 status = ctdb_packet_send(
788 conn->pkt, 2,
789 data_blob_const(&r, offsetof(struct ctdb_req_message, data)),
790 data_blob_const(buf, buflen));
792 if (!NT_STATUS_IS_OK(status)) {
793 DEBUG(0, ("ctdb_packet_send failed: %s\n", nt_errstr(status)));
794 return status;
797 status = ctdb_packet_flush(conn->pkt);
798 if (!NT_STATUS_IS_OK(status)) {
799 DEBUG(3, ("write to ctdbd failed: %s\n", nt_errstr(status)));
800 cluster_fatal("cluster dispatch daemon msg write error\n");
802 return NT_STATUS_OK;
806 * send/recv a generic ctdb control message
808 static NTSTATUS ctdbd_control(struct ctdbd_connection *conn,
809 uint32_t vnn, uint32_t opcode,
810 uint64_t srvid, uint32_t flags,
811 TDB_DATA data,
812 TALLOC_CTX *mem_ctx, TDB_DATA *outdata,
813 int *cstatus)
815 struct ctdb_req_control req;
816 struct ctdb_reply_control *reply = NULL;
817 struct ctdbd_connection *new_conn = NULL;
818 NTSTATUS status;
820 if (conn == NULL) {
821 status = ctdbd_init_connection(NULL, &new_conn);
823 if (!NT_STATUS_IS_OK(status)) {
824 DEBUG(10, ("Could not init temp connection: %s\n",
825 nt_errstr(status)));
826 goto fail;
829 conn = new_conn;
832 ZERO_STRUCT(req);
833 req.hdr.length = offsetof(struct ctdb_req_control, data) + data.dsize;
834 req.hdr.ctdb_magic = CTDB_MAGIC;
835 req.hdr.ctdb_version = CTDB_VERSION;
836 req.hdr.operation = CTDB_REQ_CONTROL;
837 req.hdr.reqid = ctdbd_next_reqid(conn);
838 req.hdr.destnode = vnn;
839 req.opcode = opcode;
840 req.srvid = srvid;
841 req.datalen = data.dsize;
842 req.flags = flags;
844 DEBUG(10, ("ctdbd_control: Sending ctdb packet\n"));
845 ctdb_packet_dump(&req.hdr);
847 status = ctdb_packet_send(
848 conn->pkt, 2,
849 data_blob_const(&req, offsetof(struct ctdb_req_control, data)),
850 data_blob_const(data.dptr, data.dsize));
852 if (!NT_STATUS_IS_OK(status)) {
853 DEBUG(3, ("ctdb_packet_send failed: %s\n", nt_errstr(status)));
854 goto fail;
857 status = ctdb_packet_flush(conn->pkt);
859 if (!NT_STATUS_IS_OK(status)) {
860 DEBUG(3, ("write to ctdbd failed: %s\n", nt_errstr(status)));
861 cluster_fatal("cluster dispatch daemon control write error\n");
864 if (flags & CTDB_CTRL_FLAG_NOREPLY) {
865 TALLOC_FREE(new_conn);
866 if (cstatus) {
867 *cstatus = 0;
869 return NT_STATUS_OK;
872 status = ctdb_read_req(conn, req.hdr.reqid, NULL, (void *)&reply);
874 if (!NT_STATUS_IS_OK(status)) {
875 DEBUG(10, ("ctdb_read_req failed: %s\n", nt_errstr(status)));
876 goto fail;
879 if (reply->hdr.operation != CTDB_REPLY_CONTROL) {
880 DEBUG(0, ("received invalid reply\n"));
881 goto fail;
884 if (outdata) {
885 if (!(outdata->dptr = (uint8 *)talloc_memdup(
886 mem_ctx, reply->data, reply->datalen))) {
887 TALLOC_FREE(reply);
888 return NT_STATUS_NO_MEMORY;
890 outdata->dsize = reply->datalen;
892 if (cstatus) {
893 (*cstatus) = reply->status;
896 status = NT_STATUS_OK;
898 fail:
899 TALLOC_FREE(new_conn);
900 TALLOC_FREE(reply);
901 return status;
905 * see if a remote process exists
907 bool ctdbd_process_exists(struct ctdbd_connection *conn, uint32_t vnn, pid_t pid)
909 struct server_id id;
910 bool result;
912 id.pid = pid;
913 id.vnn = vnn;
915 if (!ctdb_processes_exist(conn, &id, 1, &result)) {
916 DEBUG(10, ("ctdb_processes_exist failed\n"));
917 return false;
919 return result;
922 bool ctdb_processes_exist(struct ctdbd_connection *conn,
923 const struct server_id *pids, int num_pids,
924 bool *results)
926 TALLOC_CTX *frame = talloc_stackframe();
927 int i, num_received;
928 NTSTATUS status;
929 uint32_t *reqids;
930 bool result = false;
932 reqids = talloc_array(talloc_tos(), uint32_t, num_pids);
933 if (reqids == NULL) {
934 goto fail;
937 for (i=0; i<num_pids; i++) {
938 struct ctdb_req_control req;
939 pid_t pid;
941 results[i] = false;
942 reqids[i] = ctdbd_next_reqid(conn);
944 ZERO_STRUCT(req);
947 * pids[i].pid is uint64_t, scale down to pid_t which
948 * is the wire protocol towards ctdb.
950 pid = pids[i].pid;
952 DEBUG(10, ("Requesting PID %d/%d, reqid=%d\n",
953 (int)pids[i].vnn, (int)pid,
954 (int)reqids[i]));
956 req.hdr.length = offsetof(struct ctdb_req_control, data);
957 req.hdr.length += sizeof(pid);
958 req.hdr.ctdb_magic = CTDB_MAGIC;
959 req.hdr.ctdb_version = CTDB_VERSION;
960 req.hdr.operation = CTDB_REQ_CONTROL;
961 req.hdr.reqid = reqids[i];
962 req.hdr.destnode = pids[i].vnn;
963 req.opcode = CTDB_CONTROL_PROCESS_EXISTS;
964 req.srvid = 0;
965 req.datalen = sizeof(pid);
966 req.flags = 0;
968 DEBUG(10, ("ctdbd_control: Sending ctdb packet\n"));
969 ctdb_packet_dump(&req.hdr);
971 status = ctdb_packet_send(
972 conn->pkt, 2,
973 data_blob_const(
974 &req, offsetof(struct ctdb_req_control, data)),
975 data_blob_const(&pid, sizeof(pid)));
976 if (!NT_STATUS_IS_OK(status)) {
977 DEBUG(10, ("ctdb_packet_send failed: %s\n",
978 nt_errstr(status)));
979 goto fail;
983 status = ctdb_packet_flush(conn->pkt);
984 if (!NT_STATUS_IS_OK(status)) {
985 DEBUG(10, ("ctdb_packet_flush failed: %s\n",
986 nt_errstr(status)));
987 goto fail;
990 num_received = 0;
992 while (num_received < num_pids) {
993 struct ctdb_reply_control *reply = NULL;
994 uint32_t reqid;
996 status = ctdb_read_req(conn, 0, talloc_tos(), (void *)&reply);
997 if (!NT_STATUS_IS_OK(status)) {
998 DEBUG(10, ("ctdb_read_req failed: %s\n",
999 nt_errstr(status)));
1000 goto fail;
1003 if (reply->hdr.operation != CTDB_REPLY_CONTROL) {
1004 DEBUG(10, ("Received invalid reply\n"));
1005 goto fail;
1008 reqid = reply->hdr.reqid;
1010 DEBUG(10, ("Received reqid %d\n", (int)reqid));
1012 for (i=0; i<num_pids; i++) {
1013 if (reqid == reqids[i]) {
1014 break;
1017 if (i == num_pids) {
1018 DEBUG(10, ("Received unknown record number %u\n",
1019 (unsigned)reqid));
1020 goto fail;
1022 results[i] = ((reply->status) == 0);
1023 TALLOC_FREE(reply);
1024 num_received += 1;
1027 result = true;
1028 fail:
1029 TALLOC_FREE(frame);
1030 return result;
1033 struct ctdb_vnn_list {
1034 uint32_t vnn;
1035 uint32_t reqid;
1036 unsigned num_srvids;
1037 unsigned num_filled;
1038 uint64_t *srvids;
1039 unsigned *pid_indexes;
1043 * Get a list of all vnns mentioned in a list of
1044 * server_ids. vnn_indexes tells where in the vnns array we have to
1045 * place the pids.
1047 static bool ctdb_collect_vnns(TALLOC_CTX *mem_ctx,
1048 const struct server_id *pids, unsigned num_pids,
1049 struct ctdb_vnn_list **pvnns,
1050 unsigned *pnum_vnns)
1052 struct ctdb_vnn_list *vnns = NULL;
1053 unsigned *vnn_indexes = NULL;
1054 unsigned i, num_vnns = 0;
1056 vnn_indexes = talloc_array(mem_ctx, unsigned, num_pids);
1057 if (vnn_indexes == NULL) {
1058 DEBUG(1, ("talloc_array failed\n"));
1059 goto fail;
1062 for (i=0; i<num_pids; i++) {
1063 unsigned j;
1064 uint32_t vnn = pids[i].vnn;
1066 for (j=0; j<num_vnns; j++) {
1067 if (vnn == vnns[j].vnn) {
1068 break;
1071 vnn_indexes[i] = j;
1073 if (j < num_vnns) {
1075 * Already in the array
1077 vnns[j].num_srvids += 1;
1078 continue;
1080 vnns = talloc_realloc(mem_ctx, vnns, struct ctdb_vnn_list,
1081 num_vnns+1);
1082 if (vnns == NULL) {
1083 DEBUG(1, ("talloc_realloc failed\n"));
1084 goto fail;
1086 vnns[num_vnns].vnn = vnn;
1087 vnns[num_vnns].num_srvids = 1;
1088 vnns[num_vnns].num_filled = 0;
1089 num_vnns += 1;
1091 for (i=0; i<num_vnns; i++) {
1092 struct ctdb_vnn_list *vnn = &vnns[i];
1094 vnn->srvids = talloc_array(vnns, uint64_t, vnn->num_srvids);
1095 if (vnn->srvids == NULL) {
1096 DEBUG(1, ("talloc_array failed\n"));
1097 goto fail;
1099 vnn->pid_indexes = talloc_array(vnns, unsigned,
1100 vnn->num_srvids);
1101 if (vnn->pid_indexes == NULL) {
1102 DEBUG(1, ("talloc_array failed\n"));
1103 goto fail;
1106 for (i=0; i<num_pids; i++) {
1107 struct ctdb_vnn_list *vnn = &vnns[vnn_indexes[i]];
1108 vnn->srvids[vnn->num_filled] = pids[i].unique_id;
1109 vnn->pid_indexes[vnn->num_filled] = i;
1110 vnn->num_filled += 1;
1113 TALLOC_FREE(vnn_indexes);
1114 *pvnns = vnns;
1115 *pnum_vnns = num_vnns;
1116 return true;
1117 fail:
1118 TALLOC_FREE(vnns);
1119 TALLOC_FREE(vnn_indexes);
1120 return false;
1123 #ifdef HAVE_CTDB_CONTROL_CHECK_SRVIDS_DECL
1125 bool ctdb_serverids_exist(struct ctdbd_connection *conn,
1126 const struct server_id *pids, unsigned num_pids,
1127 bool *results)
1129 unsigned i, num_received;
1130 NTSTATUS status;
1131 struct ctdb_vnn_list *vnns = NULL;
1132 unsigned num_vnns;
1133 bool result = false;
1135 if (!ctdb_collect_vnns(talloc_tos(), pids, num_pids,
1136 &vnns, &num_vnns)) {
1137 DEBUG(1, ("ctdb_collect_vnns failed\n"));
1138 goto fail;
1141 for (i=0; i<num_vnns; i++) {
1142 struct ctdb_vnn_list *vnn = &vnns[i];
1143 struct ctdb_req_control req;
1145 vnn->reqid = ctdbd_next_reqid(conn);
1147 ZERO_STRUCT(req);
1149 DEBUG(10, ("Requesting VNN %d, reqid=%d, num_srvids=%u\n",
1150 (int)vnn->vnn, (int)vnn->reqid, vnn->num_srvids));
1152 req.hdr.length = offsetof(struct ctdb_req_control, data);
1153 req.hdr.ctdb_magic = CTDB_MAGIC;
1154 req.hdr.ctdb_version = CTDB_VERSION;
1155 req.hdr.operation = CTDB_REQ_CONTROL;
1156 req.hdr.reqid = vnn->reqid;
1157 req.hdr.destnode = vnn->vnn;
1158 req.opcode = CTDB_CONTROL_CHECK_SRVIDS;
1159 req.srvid = 0;
1160 req.datalen = sizeof(uint64_t) * vnn->num_srvids;
1161 req.hdr.length += req.datalen;
1162 req.flags = 0;
1164 DEBUG(10, ("ctdbd_control: Sending ctdb packet\n"));
1165 ctdb_packet_dump(&req.hdr);
1167 status = ctdb_packet_send(
1168 conn->pkt, 2,
1169 data_blob_const(
1170 &req, offsetof(struct ctdb_req_control,
1171 data)),
1172 data_blob_const(vnn->srvids, req.datalen));
1173 if (!NT_STATUS_IS_OK(status)) {
1174 DEBUG(1, ("ctdb_packet_send failed: %s\n",
1175 nt_errstr(status)));
1176 goto fail;
1180 status = ctdb_packet_flush(conn->pkt);
1181 if (!NT_STATUS_IS_OK(status)) {
1182 DEBUG(1, ("ctdb_packet_flush failed: %s\n",
1183 nt_errstr(status)));
1184 goto fail;
1187 num_received = 0;
1189 while (num_received < num_vnns) {
1190 struct ctdb_reply_control *reply = NULL;
1191 struct ctdb_vnn_list *vnn;
1192 uint32_t reqid;
1193 uint8_t *reply_data;
1195 status = ctdb_read_req(conn, 0, talloc_tos(), (void *)&reply);
1196 if (!NT_STATUS_IS_OK(status)) {
1197 DEBUG(1, ("ctdb_read_req failed: %s\n",
1198 nt_errstr(status)));
1199 goto fail;
1202 if (reply->hdr.operation != CTDB_REPLY_CONTROL) {
1203 DEBUG(1, ("Received invalid reply %u\n",
1204 (unsigned)reply->hdr.operation));
1205 goto fail;
1208 reqid = reply->hdr.reqid;
1210 DEBUG(10, ("Received reqid %d\n", (int)reqid));
1212 for (i=0; i<num_vnns; i++) {
1213 if (reqid == vnns[i].reqid) {
1214 break;
1217 if (i == num_vnns) {
1218 DEBUG(1, ("Received unknown reqid number %u\n",
1219 (unsigned)reqid));
1220 goto fail;
1223 DEBUG(10, ("Found index %u\n", i));
1225 vnn = &vnns[i];
1227 DEBUG(10, ("Received vnn %u, vnn->num_srvids %u, datalen %u\n",
1228 (unsigned)vnn->vnn, vnn->num_srvids,
1229 (unsigned)reply->datalen));
1231 if (reply->datalen >= ((vnn->num_srvids+7)/8)) {
1233 * Got a real reply
1235 reply_data = reply->data;
1236 } else {
1238 * Got an error reply
1240 DEBUG(5, ("Received short reply len %d, status %u, "
1241 "errorlen %u\n",
1242 (unsigned)reply->datalen,
1243 (unsigned)reply->status,
1244 (unsigned)reply->errorlen));
1245 dump_data(5, reply->data, reply->errorlen);
1248 * This will trigger everything set to false
1250 reply_data = NULL;
1253 for (i=0; i<vnn->num_srvids; i++) {
1254 int idx = vnn->pid_indexes[i];
1256 if (pids[i].unique_id ==
1257 SERVERID_UNIQUE_ID_NOT_TO_VERIFY) {
1258 results[idx] = true;
1259 continue;
1261 results[idx] =
1262 (reply_data != NULL) &&
1263 ((reply_data[i/8] & (1<<(i%8))) != 0);
1266 TALLOC_FREE(reply);
1267 num_received += 1;
1270 result = true;
1271 fail:
1272 TALLOC_FREE(vnns);
1273 return result;
1276 #endif /* HAVE_CTDB_CONTROL_CHECK_SRVIDS_DECL */
1279 * Get a db path
1281 char *ctdbd_dbpath(struct ctdbd_connection *conn,
1282 TALLOC_CTX *mem_ctx, uint32_t db_id)
1284 NTSTATUS status;
1285 TDB_DATA data;
1286 int32_t cstatus;
1288 data.dptr = (uint8_t*)&db_id;
1289 data.dsize = sizeof(db_id);
1291 status = ctdbd_control(conn, CTDB_CURRENT_NODE,
1292 CTDB_CONTROL_GETDBPATH, 0, 0, data,
1293 mem_ctx, &data, &cstatus);
1294 if (!NT_STATUS_IS_OK(status) || cstatus != 0) {
1295 DEBUG(0,(__location__ " ctdb_control for getdbpath failed\n"));
1296 return NULL;
1299 return (char *)data.dptr;
1303 * attach to a ctdb database
1305 NTSTATUS ctdbd_db_attach(struct ctdbd_connection *conn,
1306 const char *name, uint32_t *db_id, int tdb_flags)
1308 NTSTATUS status;
1309 TDB_DATA data;
1310 int32_t cstatus;
1311 bool persistent = (tdb_flags & TDB_CLEAR_IF_FIRST) == 0;
1313 data = string_term_tdb_data(name);
1315 status = ctdbd_control(conn, CTDB_CURRENT_NODE,
1316 persistent
1317 ? CTDB_CONTROL_DB_ATTACH_PERSISTENT
1318 : CTDB_CONTROL_DB_ATTACH,
1319 tdb_flags, 0, data, NULL, &data, &cstatus);
1320 if (!NT_STATUS_IS_OK(status)) {
1321 DEBUG(0, (__location__ " ctdb_control for db_attach "
1322 "failed: %s\n", nt_errstr(status)));
1323 return status;
1326 if (cstatus != 0 || data.dsize != sizeof(uint32_t)) {
1327 DEBUG(0,(__location__ " ctdb_control for db_attach failed\n"));
1328 return NT_STATUS_INTERNAL_ERROR;
1331 *db_id = *(uint32_t *)data.dptr;
1332 talloc_free(data.dptr);
1334 if (!(tdb_flags & TDB_SEQNUM)) {
1335 return NT_STATUS_OK;
1338 data.dptr = (uint8_t *)db_id;
1339 data.dsize = sizeof(*db_id);
1341 status = ctdbd_control(conn, CTDB_CURRENT_NODE,
1342 CTDB_CONTROL_ENABLE_SEQNUM, 0, 0, data,
1343 NULL, NULL, &cstatus);
1344 if (!NT_STATUS_IS_OK(status) || cstatus != 0) {
1345 DEBUG(0,(__location__ " ctdb_control for enable seqnum "
1346 "failed\n"));
1347 return NT_STATUS_IS_OK(status) ? NT_STATUS_INTERNAL_ERROR :
1348 status;
1351 return NT_STATUS_OK;
1355 * force the migration of a record to this node
1357 NTSTATUS ctdbd_migrate(struct ctdbd_connection *conn, uint32_t db_id,
1358 TDB_DATA key)
1360 struct ctdb_req_call req;
1361 struct ctdb_reply_call *reply;
1362 NTSTATUS status;
1364 ZERO_STRUCT(req);
1366 req.hdr.length = offsetof(struct ctdb_req_call, data) + key.dsize;
1367 req.hdr.ctdb_magic = CTDB_MAGIC;
1368 req.hdr.ctdb_version = CTDB_VERSION;
1369 req.hdr.operation = CTDB_REQ_CALL;
1370 req.hdr.reqid = ctdbd_next_reqid(conn);
1371 req.flags = CTDB_IMMEDIATE_MIGRATION;
1372 req.callid = CTDB_NULL_FUNC;
1373 req.db_id = db_id;
1374 req.keylen = key.dsize;
1376 DEBUG(10, ("ctdbd_migrate: Sending ctdb packet\n"));
1377 ctdb_packet_dump(&req.hdr);
1379 status = ctdb_packet_send(
1380 conn->pkt, 2,
1381 data_blob_const(&req, offsetof(struct ctdb_req_call, data)),
1382 data_blob_const(key.dptr, key.dsize));
1384 if (!NT_STATUS_IS_OK(status)) {
1385 DEBUG(3, ("ctdb_packet_send failed: %s\n", nt_errstr(status)));
1386 return status;
1389 status = ctdb_packet_flush(conn->pkt);
1391 if (!NT_STATUS_IS_OK(status)) {
1392 DEBUG(3, ("write to ctdbd failed: %s\n", nt_errstr(status)));
1393 cluster_fatal("cluster dispatch daemon control write error\n");
1396 status = ctdb_read_req(conn, req.hdr.reqid, NULL, (void *)&reply);
1398 if (!NT_STATUS_IS_OK(status)) {
1399 DEBUG(0, ("ctdb_read_req failed: %s\n", nt_errstr(status)));
1400 goto fail;
1403 if (reply->hdr.operation != CTDB_REPLY_CALL) {
1404 DEBUG(0, ("received invalid reply\n"));
1405 status = NT_STATUS_INTERNAL_ERROR;
1406 goto fail;
1409 status = NT_STATUS_OK;
1410 fail:
1412 TALLOC_FREE(reply);
1413 return status;
1417 * remotely fetch a record (read-only)
1419 NTSTATUS ctdbd_fetch(struct ctdbd_connection *conn, uint32_t db_id,
1420 TDB_DATA key, TALLOC_CTX *mem_ctx, TDB_DATA *data,
1421 bool local_copy)
1423 struct ctdb_req_call req;
1424 struct ctdb_reply_call *reply;
1425 NTSTATUS status;
1426 uint32_t flags;
1428 #ifdef HAVE_CTDB_WANT_READONLY_DECL
1429 flags = local_copy ? CTDB_WANT_READONLY : 0;
1430 #else
1431 flags = 0;
1432 #endif
1434 ZERO_STRUCT(req);
1436 req.hdr.length = offsetof(struct ctdb_req_call, data) + key.dsize;
1437 req.hdr.ctdb_magic = CTDB_MAGIC;
1438 req.hdr.ctdb_version = CTDB_VERSION;
1439 req.hdr.operation = CTDB_REQ_CALL;
1440 req.hdr.reqid = ctdbd_next_reqid(conn);
1441 req.flags = flags;
1442 req.callid = CTDB_FETCH_FUNC;
1443 req.db_id = db_id;
1444 req.keylen = key.dsize;
1446 status = ctdb_packet_send(
1447 conn->pkt, 2,
1448 data_blob_const(&req, offsetof(struct ctdb_req_call, data)),
1449 data_blob_const(key.dptr, key.dsize));
1451 if (!NT_STATUS_IS_OK(status)) {
1452 DEBUG(3, ("ctdb_packet_send failed: %s\n", nt_errstr(status)));
1453 return status;
1456 status = ctdb_packet_flush(conn->pkt);
1458 if (!NT_STATUS_IS_OK(status)) {
1459 DEBUG(3, ("write to ctdbd failed: %s\n", nt_errstr(status)));
1460 cluster_fatal("cluster dispatch daemon control write error\n");
1463 status = ctdb_read_req(conn, req.hdr.reqid, NULL, (void *)&reply);
1465 if (!NT_STATUS_IS_OK(status)) {
1466 DEBUG(0, ("ctdb_read_req failed: %s\n", nt_errstr(status)));
1467 goto fail;
1470 if (reply->hdr.operation != CTDB_REPLY_CALL) {
1471 DEBUG(0, ("received invalid reply\n"));
1472 status = NT_STATUS_INTERNAL_ERROR;
1473 goto fail;
1476 data->dsize = reply->datalen;
1477 if (data->dsize == 0) {
1478 data->dptr = NULL;
1479 goto done;
1482 data->dptr = (uint8 *)talloc_memdup(mem_ctx, &reply->data[0],
1483 reply->datalen);
1484 if (data->dptr == NULL) {
1485 DEBUG(0, ("talloc failed\n"));
1486 status = NT_STATUS_NO_MEMORY;
1487 goto fail;
1490 done:
1491 status = NT_STATUS_OK;
1492 fail:
1493 TALLOC_FREE(reply);
1494 return status;
1497 struct ctdbd_traverse_state {
1498 void (*fn)(TDB_DATA key, TDB_DATA data, void *private_data);
1499 void *private_data;
1503 * Handle a traverse record coming in on the ctdbd connection
1506 static NTSTATUS ctdb_traverse_handler(uint8_t *buf, size_t length,
1507 void *private_data)
1509 struct ctdbd_traverse_state *state =
1510 (struct ctdbd_traverse_state *)private_data;
1512 struct ctdb_req_message *m;
1513 struct ctdb_rec_data *d;
1514 TDB_DATA key, data;
1516 m = (struct ctdb_req_message *)buf;
1518 if (length < sizeof(*m) || m->hdr.length != length) {
1519 DEBUG(0, ("Got invalid message of length %d\n", (int)length));
1520 TALLOC_FREE(buf);
1521 return NT_STATUS_UNEXPECTED_IO_ERROR;
1524 d = (struct ctdb_rec_data *)&m->data[0];
1525 if (m->datalen < sizeof(uint32_t) || m->datalen != d->length) {
1526 DEBUG(0, ("Got invalid traverse data of length %d\n",
1527 (int)m->datalen));
1528 TALLOC_FREE(buf);
1529 return NT_STATUS_UNEXPECTED_IO_ERROR;
1532 key.dsize = d->keylen;
1533 key.dptr = &d->data[0];
1534 data.dsize = d->datalen;
1535 data.dptr = &d->data[d->keylen];
1537 if (key.dsize == 0 && data.dsize == 0) {
1538 /* end of traverse */
1539 return NT_STATUS_END_OF_FILE;
1542 if (data.dsize < sizeof(struct ctdb_ltdb_header)) {
1543 DEBUG(0, ("Got invalid ltdb header length %d\n",
1544 (int)data.dsize));
1545 TALLOC_FREE(buf);
1546 return NT_STATUS_UNEXPECTED_IO_ERROR;
1548 data.dsize -= sizeof(struct ctdb_ltdb_header);
1549 data.dptr += sizeof(struct ctdb_ltdb_header);
1551 if (state->fn) {
1552 state->fn(key, data, state->private_data);
1555 TALLOC_FREE(buf);
1556 return NT_STATUS_OK;
1560 Traverse a ctdb database. This uses a kind-of hackish way to open a second
1561 connection to ctdbd to avoid the hairy recursive and async problems with
1562 everything in-line.
1565 NTSTATUS ctdbd_traverse(uint32_t db_id,
1566 void (*fn)(TDB_DATA key, TDB_DATA data,
1567 void *private_data),
1568 void *private_data)
1570 struct ctdbd_connection *conn;
1571 NTSTATUS status;
1573 TDB_DATA data;
1574 struct ctdb_traverse_start t;
1575 int cstatus;
1576 struct ctdbd_traverse_state state;
1578 become_root();
1579 status = ctdbd_init_connection(NULL, &conn);
1580 unbecome_root();
1581 if (!NT_STATUS_IS_OK(status)) {
1582 DEBUG(0, ("ctdbd_init_connection failed: %s\n",
1583 nt_errstr(status)));
1584 return status;
1587 t.db_id = db_id;
1588 t.srvid = conn->rand_srvid;
1589 t.reqid = ctdbd_next_reqid(conn);
1591 data.dptr = (uint8_t *)&t;
1592 data.dsize = sizeof(t);
1594 status = ctdbd_control(conn, CTDB_CURRENT_NODE,
1595 CTDB_CONTROL_TRAVERSE_START, conn->rand_srvid, 0,
1596 data, NULL, NULL, &cstatus);
1598 if (!NT_STATUS_IS_OK(status) || (cstatus != 0)) {
1600 DEBUG(0,("ctdbd_control failed: %s, %d\n", nt_errstr(status),
1601 cstatus));
1603 if (NT_STATUS_IS_OK(status)) {
1605 * We need a mapping here
1607 status = NT_STATUS_UNSUCCESSFUL;
1609 goto done;
1612 state.fn = fn;
1613 state.private_data = private_data;
1615 while (True) {
1617 status = NT_STATUS_OK;
1619 if (ctdb_packet_handler(conn->pkt, ctdb_req_complete,
1620 ctdb_traverse_handler, &state, &status)) {
1622 if (NT_STATUS_EQUAL(status, NT_STATUS_END_OF_FILE)) {
1623 status = NT_STATUS_OK;
1624 break;
1628 * There might be more in the queue
1630 continue;
1633 if (!NT_STATUS_IS_OK(status)) {
1634 break;
1637 status = ctdb_packet_fd_read_sync(conn->pkt);
1639 if (NT_STATUS_EQUAL(status, NT_STATUS_RETRY)) {
1641 * There might be more in the queue
1643 continue;
1646 if (NT_STATUS_EQUAL(status, NT_STATUS_END_OF_FILE)) {
1647 status = NT_STATUS_OK;
1648 break;
1651 if (!NT_STATUS_IS_OK(status)) {
1652 DEBUG(0, ("ctdb_packet_fd_read_sync failed: %s\n", nt_errstr(status)));
1653 cluster_fatal("ctdbd died\n");
1657 done:
1658 TALLOC_FREE(conn);
1659 return status;
1663 This is used to canonicalize a ctdb_sock_addr structure.
1665 static void smbd_ctdb_canonicalize_ip(const struct sockaddr_storage *in,
1666 struct sockaddr_storage *out)
1668 memcpy(out, in, sizeof (*out));
1670 #ifdef HAVE_IPV6
1671 if (in->ss_family == AF_INET6) {
1672 const char prefix[12] = { 0,0,0,0,0,0,0,0,0,0,0xff,0xff };
1673 const struct sockaddr_in6 *in6 =
1674 (const struct sockaddr_in6 *)in;
1675 struct sockaddr_in *out4 = (struct sockaddr_in *)out;
1676 if (memcmp(&in6->sin6_addr, prefix, 12) == 0) {
1677 memset(out, 0, sizeof(*out));
1678 #ifdef HAVE_SOCK_SIN_LEN
1679 out4->sin_len = sizeof(*out);
1680 #endif
1681 out4->sin_family = AF_INET;
1682 out4->sin_port = in6->sin6_port;
1683 memcpy(&out4->sin_addr, &in6->sin6_addr.s6_addr[12], 4);
1686 #endif
1690 * Register us as a server for a particular tcp connection
1693 NTSTATUS ctdbd_register_ips(struct ctdbd_connection *conn,
1694 const struct sockaddr_storage *_server,
1695 const struct sockaddr_storage *_client,
1696 void (*release_ip_handler)(const char *ip_addr,
1697 void *private_data),
1698 void *private_data)
1701 * we still use ctdb_control_tcp for ipv4
1702 * because we want to work against older ctdb
1703 * versions at runtime
1705 struct ctdb_control_tcp p4;
1706 #ifdef HAVE_STRUCT_CTDB_CONTROL_TCP_ADDR
1707 struct ctdb_control_tcp_addr p;
1708 #endif
1709 TDB_DATA data;
1710 NTSTATUS status;
1711 struct sockaddr_storage client;
1712 struct sockaddr_storage server;
1715 * Only one connection so far
1717 SMB_ASSERT(conn->release_ip_handler == NULL);
1719 smbd_ctdb_canonicalize_ip(_client, &client);
1720 smbd_ctdb_canonicalize_ip(_server, &server);
1722 switch (client.ss_family) {
1723 case AF_INET:
1724 memcpy(&p4.dest, &server, sizeof(p4.dest));
1725 memcpy(&p4.src, &client, sizeof(p4.src));
1726 data.dptr = (uint8_t *)&p4;
1727 data.dsize = sizeof(p4);
1728 break;
1729 #ifdef HAVE_STRUCT_CTDB_CONTROL_TCP_ADDR
1730 case AF_INET6:
1731 memcpy(&p.dest.ip6, &server, sizeof(p.dest.ip6));
1732 memcpy(&p.src.ip6, &client, sizeof(p.src.ip6));
1733 data.dptr = (uint8_t *)&p;
1734 data.dsize = sizeof(p);
1735 break;
1736 #endif
1737 default:
1738 return NT_STATUS_INTERNAL_ERROR;
1741 conn->release_ip_handler = release_ip_handler;
1742 conn->release_ip_priv = private_data;
1745 * We want to be told about IP releases
1748 status = register_with_ctdbd(conn, CTDB_SRVID_RELEASE_IP);
1749 if (!NT_STATUS_IS_OK(status)) {
1750 return status;
1754 * inform ctdb of our tcp connection, so if IP takeover happens ctdb
1755 * can send an extra ack to trigger a reset for our client, so it
1756 * immediately reconnects
1758 return ctdbd_control(conn, CTDB_CURRENT_NODE,
1759 CTDB_CONTROL_TCP_CLIENT, 0,
1760 CTDB_CTRL_FLAG_NOREPLY, data, NULL, NULL, NULL);
1764 * We want to handle reconfigure events
1766 NTSTATUS ctdbd_register_reconfigure(struct ctdbd_connection *conn)
1768 return register_with_ctdbd(conn, CTDB_SRVID_RECONFIGURE);
1772 call a control on the local node
1774 NTSTATUS ctdbd_control_local(struct ctdbd_connection *conn, uint32_t opcode,
1775 uint64_t srvid, uint32_t flags, TDB_DATA data,
1776 TALLOC_CTX *mem_ctx, TDB_DATA *outdata,
1777 int *cstatus)
1779 return ctdbd_control(conn, CTDB_CURRENT_NODE, opcode, srvid, flags, data, mem_ctx, outdata, cstatus);
1782 NTSTATUS ctdb_watch_us(struct ctdbd_connection *conn)
1784 struct ctdb_client_notify_register reg_data;
1785 size_t struct_len;
1786 NTSTATUS status;
1787 int cstatus;
1789 reg_data.srvid = CTDB_SRVID_SAMBA_NOTIFY;
1790 reg_data.len = 1;
1791 reg_data.notify_data[0] = 0;
1793 struct_len = offsetof(struct ctdb_client_notify_register,
1794 notify_data) + reg_data.len;
1796 status = ctdbd_control_local(
1797 conn, CTDB_CONTROL_REGISTER_NOTIFY, conn->rand_srvid, 0,
1798 make_tdb_data((uint8_t *)&reg_data, struct_len),
1799 NULL, NULL, &cstatus);
1800 if (!NT_STATUS_IS_OK(status)) {
1801 DEBUG(1, ("ctdbd_control_local failed: %s\n",
1802 nt_errstr(status)));
1804 return status;
1807 NTSTATUS ctdb_unwatch(struct ctdbd_connection *conn)
1809 struct ctdb_client_notify_deregister dereg_data;
1810 NTSTATUS status;
1811 int cstatus;
1813 dereg_data.srvid = CTDB_SRVID_SAMBA_NOTIFY;
1815 status = ctdbd_control_local(
1816 conn, CTDB_CONTROL_DEREGISTER_NOTIFY, conn->rand_srvid, 0,
1817 make_tdb_data((uint8_t *)&dereg_data, sizeof(dereg_data)),
1818 NULL, NULL, &cstatus);
1819 if (!NT_STATUS_IS_OK(status)) {
1820 DEBUG(1, ("ctdbd_control_local failed: %s\n",
1821 nt_errstr(status)));
1823 return status;
1826 #else
1828 NTSTATUS ctdbd_messaging_send_blob(struct ctdbd_connection *conn,
1829 uint32_t dst_vnn, uint64_t dst_srvid,
1830 const uint8_t *buf, size_t buflen)
1832 return NT_STATUS_NOT_IMPLEMENTED;
1835 #endif