server/ctdb_call.c

   1 /*
   2    ctdb_call protocol code
   3
   4    Copyright (C) Andrew Tridgell  2006
   5
   6    This program is free software; you can redistribute it and/or modify
   7    it under the terms of the GNU General Public License as published by
   8    the Free Software Foundation; either version 3 of the License, or
   9    (at your option) any later version.
  10
  11    This program is distributed in the hope that it will be useful,
  12    but WITHOUT ANY WARRANTY; without even the implied warranty of
  13    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14    GNU General Public License for more details.
  15
  16    You should have received a copy of the GNU General Public License
  17    along with this program; if not, see <http://www.gnu.org/licenses/>.
  18 */
  19 /*
  20   see http://wiki.samba.org/index.php/Samba_%26_Clustering for
  21   protocol design and packet details
  22 */
  23 #include "includes.h"
  24 #include "tdb.h"
  25 #include "lib/util/dlinklist.h"
  26 #include "system/network.h"
  27 #include "system/filesys.h"
  28 #include "../include/ctdb_private.h"
  29 #include "../common/rb_tree.h"
  30
  31 struct ctdb_sticky_record {
  32         struct ctdb_context *ctdb;
  33         struct ctdb_db_context *ctdb_db;
  34         TDB_CONTEXT *pindown;
  35 };
  36
  37 /*
  38   find the ctdb_db from a db index
  39  */
  40  struct ctdb_db_context *find_ctdb_db(struct ctdb_context *ctdb, uint32_t id)
  41 {
  42         struct ctdb_db_context *ctdb_db;
  43
  44         for (ctdb_db=ctdb->db_list; ctdb_db; ctdb_db=ctdb_db->next) {
  45                 if (ctdb_db->db_id == id) {
  46                         break;
  47                 }
  48         }
  49         return ctdb_db;
  50 }
  51
  52 /*
  53   a varient of input packet that can be used in lock requeue
  54 */
  55 static void ctdb_call_input_pkt(void *p, struct ctdb_req_header *hdr)
  56 {
  57         struct ctdb_context *ctdb = talloc_get_type(p, struct ctdb_context);
  58         ctdb_input_pkt(ctdb, hdr);
  59 }
  60
  61
  62 /*
  63   send an error reply
  64 */
  65 static void ctdb_send_error(struct ctdb_context *ctdb,
  66                             struct ctdb_req_header *hdr, uint32_t status,
  67                             const char *fmt, ...) PRINTF_ATTRIBUTE(4,5);
  68 static void ctdb_send_error(struct ctdb_context *ctdb,
  69                             struct ctdb_req_header *hdr, uint32_t status,
  70                             const char *fmt, ...)
  71 {
  72         va_list ap;
  73         struct ctdb_reply_error *r;
  74         char *msg;
  75         int msglen, len;
  76
  77         if (ctdb->methods == NULL) {
  78                 DEBUG(DEBUG_INFO,(__location__ " Failed to send error. Transport is DOWN\n"));
  79                 return;
  80         }
  81
  82         va_start(ap, fmt);
  83         msg = talloc_vasprintf(ctdb, fmt, ap);
  84         if (msg == NULL) {
  85                 ctdb_fatal(ctdb, "Unable to allocate error in ctdb_send_error\n");
  86         }
  87         va_end(ap);
  88
  89         msglen = strlen(msg)+1;
  90         len = offsetof(struct ctdb_reply_error, msg);
  91         r = ctdb_transport_allocate(ctdb, msg, CTDB_REPLY_ERROR, len + msglen,
  92                                     struct ctdb_reply_error);
  93         CTDB_NO_MEMORY_FATAL(ctdb, r);
  94
  95         r->hdr.destnode  = hdr->srcnode;
  96         r->hdr.reqid     = hdr->reqid;
  97         r->status        = status;
  98         r->msglen        = msglen;
  99         memcpy(&r->msg[0], msg, msglen);
 100
 101         ctdb_queue_packet(ctdb, &r->hdr);
 102
 103         talloc_free(msg);
 104 }
 105
 106
 107 /**
 108  * send a redirect reply
 109  *
 110  * The logic behind this function is this:
 111  *
 112  * A client wants to grab a record and sends a CTDB_REQ_CALL packet
 113  * to its local ctdb (ctdb_request_call). If the node is not itself
 114  * the record's DMASTER, it first redirects the packet to  the
 115  * record's LMASTER. The LMASTER then redirects the call packet to
 116  * the current DMASTER. Note that this works because of this: When
 117  * a record is migrated off a node, then the new DMASTER is stored
 118  * in the record's copy on the former DMASTER.
 119  */
 120 static void ctdb_call_send_redirect(struct ctdb_context *ctdb,
 121                                     struct ctdb_db_context *ctdb_db,
 122                                     TDB_DATA key,
 123                                     struct ctdb_req_call *c,
 124                                     struct ctdb_ltdb_header *header)
 125 {
 126         uint32_t lmaster = ctdb_lmaster(ctdb, &key);
 127
 128         c->hdr.destnode = lmaster;
 129         if (ctdb->pnn == lmaster) {
 130                 c->hdr.destnode = header->dmaster;
 131         }
 132         c->hopcount++;
 133
 134         if (c->hopcount%100 > 95) {
 135                 DEBUG(DEBUG_WARNING,("High hopcount %d dbid:%s "
 136                         "key:0x%08x reqid=%08x pnn:%d src:%d lmaster:%d "
 137                         "header->dmaster:%d dst:%d\n",
 138                         c->hopcount, ctdb_db->db_name, ctdb_hash(&key),
 139                         c->hdr.reqid, ctdb->pnn, c->hdr.srcnode, lmaster,
 140                         header->dmaster, c->hdr.destnode));
 141         }
 142
 143         ctdb_queue_packet(ctdb, &c->hdr);
 144 }
 145
 146
 147 /*
 148   send a dmaster reply
 149
 150   caller must have the chainlock before calling this routine. Caller must be
 151   the lmaster
 152 */
 153 static void ctdb_send_dmaster_reply(struct ctdb_db_context *ctdb_db,
 154                                     struct ctdb_ltdb_header *header,
 155                                     TDB_DATA key, TDB_DATA data,
 156                                     uint32_t new_dmaster,
 157                                     uint32_t reqid)
 158 {
 159         struct ctdb_context *ctdb = ctdb_db->ctdb;
 160         struct ctdb_reply_dmaster *r;
 161         int ret, len;
 162         TALLOC_CTX *tmp_ctx;
 163
 164         if (ctdb->pnn != ctdb_lmaster(ctdb, &key)) {
 165                 DEBUG(DEBUG_ALERT,(__location__ " Caller is not lmaster!\n"));
 166                 return;
 167         }
 168
 169         header->dmaster = new_dmaster;
 170         ret = ctdb_ltdb_store(ctdb_db, key, header, data);
 171         if (ret != 0) {
 172                 ctdb_fatal(ctdb, "ctdb_send_dmaster_reply unable to update dmaster");
 173                 return;
 174         }
 175
 176         if (ctdb->methods == NULL) {
 177                 ctdb_fatal(ctdb, "ctdb_send_dmaster_reply cant update dmaster since transport is down");
 178                 return;
 179         }
 180
 181         /* put the packet on a temporary context, allowing us to safely free
 182            it below even if ctdb_reply_dmaster() has freed it already */
 183         tmp_ctx = talloc_new(ctdb);
 184
 185         /* send the CTDB_REPLY_DMASTER */
 186         len = offsetof(struct ctdb_reply_dmaster, data) + key.dsize + data.dsize + sizeof(uint32_t);
 187         r = ctdb_transport_allocate(ctdb, tmp_ctx, CTDB_REPLY_DMASTER, len,
 188                                     struct ctdb_reply_dmaster);
 189         CTDB_NO_MEMORY_FATAL(ctdb, r);
 190
 191         r->hdr.destnode  = new_dmaster;
 192         r->hdr.reqid     = reqid;
 193         r->rsn           = header->rsn;
 194         r->keylen        = key.dsize;
 195         r->datalen       = data.dsize;
 196         r->db_id         = ctdb_db->db_id;
 197         memcpy(&r->data[0], key.dptr, key.dsize);
 198         memcpy(&r->data[key.dsize], data.dptr, data.dsize);
 199         memcpy(&r->data[key.dsize+data.dsize], &header->flags, sizeof(uint32_t));
 200
 201         ctdb_queue_packet(ctdb, &r->hdr);
 202
 203         talloc_free(tmp_ctx);
 204 }
 205
 206 /*
 207   send a dmaster request (give another node the dmaster for a record)
 208
 209   This is always sent to the lmaster, which ensures that the lmaster
 210   always knows who the dmaster is. The lmaster will then send a
 211   CTDB_REPLY_DMASTER to the new dmaster
 212 */
 213 static void ctdb_call_send_dmaster(struct ctdb_db_context *ctdb_db,
 214                                    struct ctdb_req_call *c,
 215                                    struct ctdb_ltdb_header *header,
 216                                    TDB_DATA *key, TDB_DATA *data)
 217 {
 218         struct ctdb_req_dmaster *r;
 219         struct ctdb_context *ctdb = ctdb_db->ctdb;
 220         int len;
 221         uint32_t lmaster = ctdb_lmaster(ctdb, key);
 222
 223         if (ctdb->methods == NULL) {
 224                 ctdb_fatal(ctdb, "Failed ctdb_call_send_dmaster since transport is down");
 225                 return;
 226         }
 227
 228         if (data->dsize != 0) {
 229                 header->flags |= CTDB_REC_FLAG_MIGRATED_WITH_DATA;
 230         }
 231
 232         if (lmaster == ctdb->pnn) {
 233                 ctdb_send_dmaster_reply(ctdb_db, header, *key, *data,
 234                                         c->hdr.srcnode, c->hdr.reqid);
 235                 return;
 236         }
 237
 238         len = offsetof(struct ctdb_req_dmaster, data) + key->dsize + data->dsize
 239                         + sizeof(uint32_t);
 240         r = ctdb_transport_allocate(ctdb, ctdb, CTDB_REQ_DMASTER, len,
 241                                     struct ctdb_req_dmaster);
 242         CTDB_NO_MEMORY_FATAL(ctdb, r);
 243         r->hdr.destnode  = lmaster;
 244         r->hdr.reqid     = c->hdr.reqid;
 245         r->db_id         = c->db_id;
 246         r->rsn           = header->rsn;
 247         r->dmaster       = c->hdr.srcnode;
 248         r->keylen        = key->dsize;
 249         r->datalen       = data->dsize;
 250         memcpy(&r->data[0], key->dptr, key->dsize);
 251         memcpy(&r->data[key->dsize], data->dptr, data->dsize);
 252         memcpy(&r->data[key->dsize + data->dsize], &header->flags, sizeof(uint32_t));
 253
 254         header->dmaster = c->hdr.srcnode;
 255         if (ctdb_ltdb_store(ctdb_db, *key, header, *data) != 0) {
 256                 ctdb_fatal(ctdb, "Failed to store record in ctdb_call_send_dmaster");
 257         }
 258
 259         ctdb_queue_packet(ctdb, &r->hdr);
 260
 261         talloc_free(r);
 262 }
 263
 264 static void ctdb_sticky_pindown_timeout(struct event_context *ev, struct timed_event *te,
 265                                        struct timeval t, void *private_data)
 266 {
 267         struct ctdb_sticky_record *sr = talloc_get_type(private_data,
 268                                                        struct ctdb_sticky_record);
 269
 270         DEBUG(DEBUG_ERR,("Pindown timeout db:%s  unstick record\n", sr->ctdb_db->db_name));
 271         if (sr->pindown != NULL) {
 272                 talloc_free(sr->pindown);
 273                 sr->pindown = NULL;
 274         }
 275 }
 276
 277 static int
 278 ctdb_set_sticky_pindown(struct ctdb_context *ctdb, struct ctdb_db_context *ctdb_db, TDB_DATA key)
 279 {
 280         TALLOC_CTX *tmp_ctx = talloc_new(NULL);
 281         uint32_t *k;
 282         struct ctdb_sticky_record *sr;
 283
 284         k = talloc_zero_size(tmp_ctx, ((key.dsize + 3) & 0xfffffffc) + 4);
 285         if (k == NULL) {
 286                 DEBUG(DEBUG_ERR,("Failed to allocate key for sticky record\n"));
 287                 talloc_free(tmp_ctx);
 288                 return -1;
 289         }
 290
 291         k[0] = (key.dsize + 3) / 4 + 1;
 292         memcpy(&k[1], key.dptr, key.dsize);
 293
 294         sr = trbt_lookuparray32(ctdb_db->sticky_records, k[0], &k[0]);
 295         if (sr == NULL) {
 296                 talloc_free(tmp_ctx);
 297                 return 0;
 298         }
 299
 300         talloc_free(tmp_ctx);
 301
 302         if (sr->pindown == NULL) {
 303                 DEBUG(DEBUG_ERR,("Pinning down record in %s for %d ms\n", ctdb_db->db_name, ctdb->tunable.sticky_pindown));
 304                 sr->pindown = talloc_new(sr);
 305                 if (sr->pindown == NULL) {
 306                         DEBUG(DEBUG_ERR,("Failed to allocate pindown context for sticky record\n"));
 307                         return -1;
 308                 }
 309                 event_add_timed(ctdb->ev, sr->pindown, timeval_current_ofs(ctdb->tunable.sticky_pindown / 1000, (ctdb->tunable.sticky_pindown * 1000) % 1000000), ctdb_sticky_pindown_timeout, sr);
 310         }
 311
 312         return 0;
 313 }
 314
 315 /*
 316   called when a CTDB_REPLY_DMASTER packet comes in, or when the lmaster
 317   gets a CTDB_REQUEST_DMASTER for itself. We become the dmaster.
 318
 319   must be called with the chainlock held. This function releases the chainlock
 320 */
 321 static void ctdb_become_dmaster(struct ctdb_db_context *ctdb_db,
 322                                 struct ctdb_req_header *hdr,
 323                                 TDB_DATA key, TDB_DATA data,
 324                                 uint64_t rsn, uint32_t record_flags)
 325 {
 326         struct ctdb_call_state *state;
 327         struct ctdb_context *ctdb = ctdb_db->ctdb;
 328         struct ctdb_ltdb_header header;
 329         int ret;
 330
 331         DEBUG(DEBUG_DEBUG,("pnn %u dmaster response %08x\n", ctdb->pnn, ctdb_hash(&key)));
 332
 333         ZERO_STRUCT(header);
 334         header.rsn = rsn;
 335         header.dmaster = ctdb->pnn;
 336         header.flags = record_flags;
 337
 338         state = ctdb_reqid_find(ctdb, hdr->reqid, struct ctdb_call_state);
 339
 340         if (state) {
 341                 if (state->call->flags & CTDB_CALL_FLAG_VACUUM_MIGRATION) {
 342                         /*
 343                          * We temporarily add the VACUUM_MIGRATED flag to
 344                          * the record flags, so that ctdb_ltdb_store can
 345                          * decide whether the record should be stored or
 346                          * deleted.
 347                          */
 348                         header.flags |= CTDB_REC_FLAG_VACUUM_MIGRATED;
 349                 }
 350         }
 351
 352         if (ctdb_ltdb_store(ctdb_db, key, &header, data) != 0) {
 353                 ctdb_fatal(ctdb, "ctdb_reply_dmaster store failed\n");
 354
 355                 ret = ctdb_ltdb_unlock(ctdb_db, key);
 356                 if (ret != 0) {
 357                         DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 358                 }
 359                 return;
 360         }
 361
 362         /* we just became DMASTER and this database is "sticky",
 363            see if the record is flagged as "hot" and set up a pin-down
 364            context to stop migrations for a little while if so
 365         */
 366         if (ctdb_db->sticky) {
 367                 ctdb_set_sticky_pindown(ctdb, ctdb_db, key);
 368         }
 369
 370         if (state == NULL) {
 371                 DEBUG(DEBUG_ERR,("pnn %u Invalid reqid %u in ctdb_become_dmaster from node %u\n",
 372                          ctdb->pnn, hdr->reqid, hdr->srcnode));
 373
 374                 ret = ctdb_ltdb_unlock(ctdb_db, key);
 375                 if (ret != 0) {
 376                         DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 377                 }
 378                 return;
 379         }
 380
 381         if (key.dsize != state->call->key.dsize || memcmp(key.dptr, state->call->key.dptr, key.dsize)) {
 382                 DEBUG(DEBUG_ERR, ("Got bogus DMASTER packet reqid:%u from node %u. Key does not match key held in matching idr.\n", hdr->reqid, hdr->srcnode));
 383
 384                 ret = ctdb_ltdb_unlock(ctdb_db, key);
 385                 if (ret != 0) {
 386                         DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 387                 }
 388                 return;
 389         }
 390
 391         if (hdr->reqid != state->reqid) {
 392                 /* we found a record  but it was the wrong one */
 393                 DEBUG(DEBUG_ERR, ("Dropped orphan in ctdb_become_dmaster with reqid:%u\n from node %u", hdr->reqid, hdr->srcnode));
 394
 395                 ret = ctdb_ltdb_unlock(ctdb_db, key);
 396                 if (ret != 0) {
 397                         DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 398                 }
 399                 return;
 400         }
 401
 402         ctdb_call_local(ctdb_db, state->call, &header, state, &data, true);
 403
 404         ret = ctdb_ltdb_unlock(ctdb_db, state->call->key);
 405         if (ret != 0) {
 406                 DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 407         }
 408
 409         state->state = CTDB_CALL_DONE;
 410         if (state->async.fn) {
 411                 state->async.fn(state);
 412         }
 413 }
 414
 415
 416
 417 /*
 418   called when a CTDB_REQ_DMASTER packet comes in
 419
 420   this comes into the lmaster for a record when the current dmaster
 421   wants to give up the dmaster role and give it to someone else
 422 */
 423 void ctdb_request_dmaster(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
 424 {
 425         struct ctdb_req_dmaster *c = (struct ctdb_req_dmaster *)hdr;
 426         TDB_DATA key, data, data2;
 427         struct ctdb_ltdb_header header;
 428         struct ctdb_db_context *ctdb_db;
 429         uint32_t record_flags = 0;
 430         size_t len;
 431         int ret;
 432
 433         key.dptr = c->data;
 434         key.dsize = c->keylen;
 435         data.dptr = c->data + c->keylen;
 436         data.dsize = c->datalen;
 437         len = offsetof(struct ctdb_req_dmaster, data) + key.dsize + data.dsize
 438                         + sizeof(uint32_t);
 439         if (len <= c->hdr.length) {
 440                 record_flags = *(uint32_t *)&c->data[c->keylen + c->datalen];
 441         }
 442
 443         ctdb_db = find_ctdb_db(ctdb, c->db_id);
 444         if (!ctdb_db) {
 445                 ctdb_send_error(ctdb, hdr, -1,
 446                                 "Unknown database in request. db_id==0x%08x",
 447                                 c->db_id);
 448                 return;
 449         }
 450
 451         /* fetch the current record */
 452         ret = ctdb_ltdb_lock_fetch_requeue(ctdb_db, key, &header, hdr, &data2,
 453                                            ctdb_call_input_pkt, ctdb, false);
 454         if (ret == -1) {
 455                 ctdb_fatal(ctdb, "ctdb_req_dmaster failed to fetch record");
 456                 return;
 457         }
 458         if (ret == -2) {
 459                 DEBUG(DEBUG_INFO,(__location__ " deferring ctdb_request_dmaster\n"));
 460                 return;
 461         }
 462
 463         if (ctdb_lmaster(ctdb, &key) != ctdb->pnn) {
 464                 DEBUG(DEBUG_ALERT,("pnn %u dmaster request to non-lmaster lmaster=%u gen=%u curgen=%u\n",
 465                          ctdb->pnn, ctdb_lmaster(ctdb, &key),
 466                          hdr->generation, ctdb->vnn_map->generation));
 467                 ctdb_fatal(ctdb, "ctdb_req_dmaster to non-lmaster");
 468         }
 469
 470         DEBUG(DEBUG_DEBUG,("pnn %u dmaster request on %08x for %u from %u\n",
 471                  ctdb->pnn, ctdb_hash(&key), c->dmaster, c->hdr.srcnode));
 472
 473         /* its a protocol error if the sending node is not the current dmaster */
 474         if (header.dmaster != hdr->srcnode) {
 475                 DEBUG(DEBUG_ALERT,("pnn %u dmaster request for new-dmaster %u from non-master %u real-dmaster=%u key %08x dbid 0x%08x gen=%u curgen=%u c->rsn=%llu header.rsn=%llu reqid=%u keyval=0x%08x\n",
 476                          ctdb->pnn, c->dmaster, hdr->srcnode, header.dmaster, ctdb_hash(&key),
 477                          ctdb_db->db_id, hdr->generation, ctdb->vnn_map->generation,
 478                          (unsigned long long)c->rsn, (unsigned long long)header.rsn, c->hdr.reqid,
 479                          (key.dsize >= 4)?(*(uint32_t *)key.dptr):0));
 480                 if (header.rsn != 0 || header.dmaster != ctdb->pnn) {
 481                         DEBUG(DEBUG_ERR,("ctdb_req_dmaster from non-master. Force a recovery.\n"));
 482
 483                         ctdb->recovery_mode = CTDB_RECOVERY_ACTIVE;
 484                         ctdb_ltdb_unlock(ctdb_db, key);
 485                         return;
 486                 }
 487         }
 488
 489         if (header.rsn > c->rsn) {
 490                 DEBUG(DEBUG_ALERT,("pnn %u dmaster request with older RSN new-dmaster %u from %u real-dmaster=%u key %08x dbid 0x%08x gen=%u curgen=%u c->rsn=%llu header.rsn=%llu reqid=%u\n",
 491                          ctdb->pnn, c->dmaster, hdr->srcnode, header.dmaster, ctdb_hash(&key),
 492                          ctdb_db->db_id, hdr->generation, ctdb->vnn_map->generation,
 493                          (unsigned long long)c->rsn, (unsigned long long)header.rsn, c->hdr.reqid));
 494         }
 495
 496         /* use the rsn from the sending node */
 497         header.rsn = c->rsn;
 498
 499         /* store the record flags from the sending node */
 500         header.flags = record_flags;
 501
 502         /* check if the new dmaster is the lmaster, in which case we
 503            skip the dmaster reply */
 504         if (c->dmaster == ctdb->pnn) {
 505                 ctdb_become_dmaster(ctdb_db, hdr, key, data, c->rsn, record_flags);
 506         } else {
 507                 ctdb_send_dmaster_reply(ctdb_db, &header, key, data, c->dmaster, hdr->reqid);
 508
 509                 ret = ctdb_ltdb_unlock(ctdb_db, key);
 510                 if (ret != 0) {
 511                         DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 512                 }
 513         }
 514 }
 515
 516 static void ctdb_sticky_record_timeout(struct event_context *ev, struct timed_event *te,
 517                                        struct timeval t, void *private_data)
 518 {
 519         struct ctdb_sticky_record *sr = talloc_get_type(private_data,
 520                                                        struct ctdb_sticky_record);
 521         talloc_free(sr);
 522 }
 523
 524 static void *ctdb_make_sticky_record_callback(void *parm, void *data)
 525 {
 526         if (data) {
 527                 DEBUG(DEBUG_ERR,("Already have sticky record registered. Free old %p and create new %p\n", data, parm));
 528                 talloc_free(data);
 529         }
 530         return parm;
 531 }
 532
 533 static int
 534 ctdb_make_record_sticky(struct ctdb_context *ctdb, struct ctdb_db_context *ctdb_db, TDB_DATA key)
 535 {
 536         TALLOC_CTX *tmp_ctx = talloc_new(NULL);
 537         uint32_t *k;
 538         struct ctdb_sticky_record *sr;
 539
 540         k = talloc_zero_size(tmp_ctx, ((key.dsize + 3) & 0xfffffffc) + 4);
 541         if (k == NULL) {
 542                 DEBUG(DEBUG_ERR,("Failed to allocate key for sticky record\n"));
 543                 talloc_free(tmp_ctx);
 544                 return -1;
 545         }
 546
 547         k[0] = (key.dsize + 3) / 4 + 1;
 548         memcpy(&k[1], key.dptr, key.dsize);
 549
 550         sr = trbt_lookuparray32(ctdb_db->sticky_records, k[0], &k[0]);
 551         if (sr != NULL) {
 552                 talloc_free(tmp_ctx);
 553                 return 0;
 554         }
 555
 556         sr = talloc(ctdb_db->sticky_records, struct ctdb_sticky_record);
 557         if (sr == NULL) {
 558                 talloc_free(tmp_ctx);
 559                 DEBUG(DEBUG_ERR,("Failed to allocate sticky record structure\n"));
 560                 return -1;
 561         }
 562
 563         sr->ctdb    = ctdb;
 564         sr->ctdb_db = ctdb_db;
 565         sr->pindown = NULL;
 566
 567         DEBUG(DEBUG_ERR,("Make record sticky for %d seconds in db %s key:0x%08x.\n",
 568                          ctdb->tunable.sticky_duration,
 569                          ctdb_db->db_name, ctdb_hash(&key)));
 570
 571         trbt_insertarray32_callback(ctdb_db->sticky_records, k[0], &k[0], ctdb_make_sticky_record_callback, sr);
 572
 573         event_add_timed(ctdb->ev, sr, timeval_current_ofs(ctdb->tunable.sticky_duration, 0), ctdb_sticky_record_timeout, sr);
 574
 575         talloc_free(tmp_ctx);
 576         return 0;
 577 }
 578
 579 struct pinned_down_requeue_handle {
 580         struct ctdb_context *ctdb;
 581         struct ctdb_req_header *hdr;
 582 };
 583
 584 struct pinned_down_deferred_call {
 585         struct ctdb_context *ctdb;
 586         struct ctdb_req_header *hdr;
 587 };
 588
 589 static void pinned_down_requeue(struct event_context *ev, struct timed_event *te,
 590                        struct timeval t, void *private_data)
 591 {
 592         struct pinned_down_requeue_handle *handle = talloc_get_type(private_data, struct pinned_down_requeue_handle);
 593         struct ctdb_context *ctdb = handle->ctdb;
 594
 595         talloc_steal(ctdb, handle->hdr);
 596         ctdb_call_input_pkt(ctdb, handle->hdr);
 597
 598         talloc_free(handle);
 599 }
 600
 601 static int pinned_down_destructor(struct pinned_down_deferred_call *pinned_down)
 602 {
 603         struct ctdb_context *ctdb = pinned_down->ctdb;
 604         struct pinned_down_requeue_handle *handle = talloc(ctdb, struct pinned_down_requeue_handle);
 605
 606         handle->ctdb = pinned_down->ctdb;
 607         handle->hdr  = pinned_down->hdr;
 608         talloc_steal(handle, handle->hdr);
 609
 610         event_add_timed(ctdb->ev, handle, timeval_zero(), pinned_down_requeue, handle);
 611
 612         return 0;
 613 }
 614
 615 static int
 616 ctdb_defer_pinned_down_request(struct ctdb_context *ctdb, struct ctdb_db_context *ctdb_db, TDB_DATA key, struct ctdb_req_header *hdr)
 617 {
 618         TALLOC_CTX *tmp_ctx = talloc_new(NULL);
 619         uint32_t *k;
 620         struct ctdb_sticky_record *sr;
 621         struct pinned_down_deferred_call *pinned_down;
 622
 623         k = talloc_zero_size(tmp_ctx, ((key.dsize + 3) & 0xfffffffc) + 4);
 624         if (k == NULL) {
 625                 DEBUG(DEBUG_ERR,("Failed to allocate key for sticky record\n"));
 626                 talloc_free(tmp_ctx);
 627                 return -1;
 628         }
 629
 630         k[0] = (key.dsize + 3) / 4 + 1;
 631         memcpy(&k[1], key.dptr, key.dsize);
 632
 633         sr = trbt_lookuparray32(ctdb_db->sticky_records, k[0], &k[0]);
 634         if (sr == NULL) {
 635                 talloc_free(tmp_ctx);
 636                 return -1;
 637         }
 638
 639         talloc_free(tmp_ctx);
 640
 641         if (sr->pindown == NULL) {
 642                 return -1;
 643         }
 644
 645         pinned_down = talloc(sr->pindown, struct pinned_down_deferred_call);
 646         if (pinned_down == NULL) {
 647                 DEBUG(DEBUG_ERR,("Failed to allocate structure for deferred pinned down request\n"));
 648                 return -1;
 649         }
 650
 651         pinned_down->ctdb = ctdb;
 652         pinned_down->hdr  = hdr;
 653
 654         talloc_set_destructor(pinned_down, pinned_down_destructor);
 655         talloc_steal(pinned_down, hdr);
 656
 657         return 0;
 658 }
 659
 660 static void
 661 ctdb_update_db_stat_hot_keys(struct ctdb_db_context *ctdb_db, TDB_DATA key, int hopcount)
 662 {
 663         int i, id;
 664
 665         /* smallest value is always at index 0 */
 666         if (hopcount <= ctdb_db->statistics.hot_keys[0].count) {
 667                 return;
 668         }
 669
 670         /* see if we already know this key */
 671         for (i = 0; i < MAX_HOT_KEYS; i++) {
 672                 if (key.dsize != ctdb_db->statistics.hot_keys[i].key.dsize) {
 673                         continue;
 674                 }
 675                 if (memcmp(key.dptr, ctdb_db->statistics.hot_keys[i].key.dptr, key.dsize)) {
 676                         continue;
 677                 }
 678                 /* found an entry for this key */
 679                 if (hopcount <= ctdb_db->statistics.hot_keys[i].count) {
 680                         return;
 681                 }
 682                 ctdb_db->statistics.hot_keys[i].count = hopcount;
 683                 goto sort_keys;
 684         }
 685
 686         if (ctdb_db->statistics.num_hot_keys < MAX_HOT_KEYS) {
 687                 id = ctdb_db->statistics.num_hot_keys;
 688                 ctdb_db->statistics.num_hot_keys++;
 689         } else {
 690                 id = 0;
 691         }
 692
 693         if (ctdb_db->statistics.hot_keys[id].key.dptr != NULL) {
 694                 talloc_free(ctdb_db->statistics.hot_keys[id].key.dptr);
 695         }
 696         ctdb_db->statistics.hot_keys[id].key.dsize = key.dsize;
 697         ctdb_db->statistics.hot_keys[id].key.dptr  = talloc_memdup(ctdb_db, key.dptr, key.dsize);
 698         ctdb_db->statistics.hot_keys[id].count = hopcount;
 699         DEBUG(DEBUG_NOTICE,("Updated hot key database=%s key=0x%08x id=%d hop_count=%d\n",
 700                             ctdb_db->db_name, ctdb_hash(&key), id, hopcount));
 701
 702 sort_keys:
 703         for (i = 1; i < MAX_HOT_KEYS; i++) {
 704                 if (ctdb_db->statistics.hot_keys[i].count == 0) {
 705                         continue;
 706                 }
 707                 if (ctdb_db->statistics.hot_keys[i].count < ctdb_db->statistics.hot_keys[0].count) {
 708                         hopcount = ctdb_db->statistics.hot_keys[i].count;
 709                         ctdb_db->statistics.hot_keys[i].count = ctdb_db->statistics.hot_keys[0].count;
 710                         ctdb_db->statistics.hot_keys[0].count = hopcount;
 711
 712                         key = ctdb_db->statistics.hot_keys[i].key;
 713                         ctdb_db->statistics.hot_keys[i].key = ctdb_db->statistics.hot_keys[0].key;
 714                         ctdb_db->statistics.hot_keys[0].key = key;
 715                 }
 716         }
 717 }
 718
 719 /*
 720   called when a CTDB_REQ_CALL packet comes in
 721 */
 722 void ctdb_request_call(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
 723 {
 724         struct ctdb_req_call *c = (struct ctdb_req_call *)hdr;
 725         TDB_DATA data;
 726         struct ctdb_reply_call *r;
 727         int ret, len;
 728         struct ctdb_ltdb_header header;
 729         struct ctdb_call *call;
 730         struct ctdb_db_context *ctdb_db;
 731         int tmp_count, bucket;
 732
 733         if (ctdb->methods == NULL) {
 734                 DEBUG(DEBUG_INFO,(__location__ " Failed ctdb_request_call. Transport is DOWN\n"));
 735                 return;
 736         }
 737
 738
 739         ctdb_db = find_ctdb_db(ctdb, c->db_id);
 740         if (!ctdb_db) {
 741                 ctdb_send_error(ctdb, hdr, -1,
 742                                 "Unknown database in request. db_id==0x%08x",
 743                                 c->db_id);
 744                 return;
 745         }
 746
 747         call = talloc(hdr, struct ctdb_call);
 748         CTDB_NO_MEMORY_FATAL(ctdb, call);
 749
 750         call->call_id  = c->callid;
 751         call->key.dptr = c->data;
 752         call->key.dsize = c->keylen;
 753         call->call_data.dptr = c->data + c->keylen;
 754         call->call_data.dsize = c->calldatalen;
 755         call->reply_data.dptr  = NULL;
 756         call->reply_data.dsize = 0;
 757
 758
 759         /* If this record is pinned down we should defer the
 760            request until the pindown times out
 761         */
 762         if (ctdb_db->sticky) {
 763                 if (ctdb_defer_pinned_down_request(ctdb, ctdb_db, call->key, hdr) == 0) {
 764                         DEBUG(DEBUG_WARNING,
 765                               ("Defer request for pinned down record in %s\n", ctdb_db->db_name));
 766                         talloc_free(call);
 767                         return;
 768                 }
 769         }
 770
 771
 772         /* determine if we are the dmaster for this key. This also
 773            fetches the record data (if any), thus avoiding a 2nd fetch of the data
 774            if the call will be answered locally */
 775
 776         ret = ctdb_ltdb_lock_fetch_requeue(ctdb_db, call->key, &header, hdr, &data,
 777                                            ctdb_call_input_pkt, ctdb, false);
 778         if (ret == -1) {
 779                 ctdb_send_error(ctdb, hdr, ret, "ltdb fetch failed in ctdb_request_call");
 780                 talloc_free(call);
 781                 return;
 782         }
 783         if (ret == -2) {
 784                 DEBUG(DEBUG_INFO,(__location__ " deferred ctdb_request_call\n"));
 785                 talloc_free(call);
 786                 return;
 787         }
 788
 789         /* Dont do READONLY if we dont have a tracking database */
 790         if ((c->flags & CTDB_WANT_READONLY) && !ctdb_db->readonly) {
 791                 c->flags &= ~CTDB_WANT_READONLY;
 792         }
 793
 794         if (header.flags & CTDB_REC_RO_REVOKE_COMPLETE) {
 795                 header.flags &= ~CTDB_REC_RO_FLAGS;
 796                 CTDB_INCREMENT_STAT(ctdb, total_ro_revokes);
 797                 CTDB_INCREMENT_DB_STAT(ctdb_db, db_ro_revokes);
 798                 if (ctdb_ltdb_store(ctdb_db, call->key, &header, data) != 0) {
 799                         ctdb_fatal(ctdb, "Failed to write header with cleared REVOKE flag");
 800                 }
 801                 /* and clear out the tracking data */
 802                 if (tdb_delete(ctdb_db->rottdb, call->key) != 0) {
 803                         DEBUG(DEBUG_ERR,(__location__ " Failed to clear out trackingdb record\n"));
 804                 }
 805         }
 806
 807         /* if we are revoking, we must defer all other calls until the revoke
 808          * had completed.
 809          */
 810         if (header.flags & CTDB_REC_RO_REVOKING_READONLY) {
 811                 talloc_free(data.dptr);
 812                 ret = ctdb_ltdb_unlock(ctdb_db, call->key);
 813
 814                 if (ctdb_add_revoke_deferred_call(ctdb, ctdb_db, call->key, hdr, ctdb_call_input_pkt, ctdb) != 0) {
 815                         ctdb_fatal(ctdb, "Failed to add deferred call for revoke child");
 816                 }
 817                 talloc_free(call);
 818                 return;
 819         }
 820
 821         /*
 822          * If we are not the dmaster and are not hosting any delegations,
 823          * then we redirect the request to the node than can answer it
 824          * (the lmaster or the dmaster).
 825          */
 826         if ((header.dmaster != ctdb->pnn)
 827             && (!(header.flags & CTDB_REC_RO_HAVE_DELEGATIONS)) ) {
 828                 talloc_free(data.dptr);
 829                 ctdb_call_send_redirect(ctdb, ctdb_db, call->key, c, &header);
 830
 831                 ret = ctdb_ltdb_unlock(ctdb_db, call->key);
 832                 if (ret != 0) {
 833                         DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 834                 }
 835                 talloc_free(call);
 836                 return;
 837         }
 838
 839         if ( (!(c->flags & CTDB_WANT_READONLY))
 840         && (header.flags & (CTDB_REC_RO_HAVE_DELEGATIONS|CTDB_REC_RO_HAVE_READONLY)) ) {
 841                 header.flags   |= CTDB_REC_RO_REVOKING_READONLY;
 842                 if (ctdb_ltdb_store(ctdb_db, call->key, &header, data) != 0) {
 843                         ctdb_fatal(ctdb, "Failed to store record with HAVE_DELEGATIONS set");
 844                 }
 845                 ret = ctdb_ltdb_unlock(ctdb_db, call->key);
 846
 847                 if (ctdb_start_revoke_ro_record(ctdb, ctdb_db, call->key, &header, data) != 0) {
 848                         ctdb_fatal(ctdb, "Failed to start record revoke");
 849                 }
 850                 talloc_free(data.dptr);
 851
 852                 if (ctdb_add_revoke_deferred_call(ctdb, ctdb_db, call->key, hdr, ctdb_call_input_pkt, ctdb) != 0) {
 853                         ctdb_fatal(ctdb, "Failed to add deferred call for revoke child");
 854                 }
 855                 talloc_free(call);
 856
 857                 return;
 858         }
 859
 860         /* If this is the first request for delegation. bump rsn and set
 861          * the delegations flag
 862          */
 863         if ((c->flags & CTDB_WANT_READONLY)
 864         &&  (c->callid == CTDB_FETCH_WITH_HEADER_FUNC)
 865         &&  (!(header.flags & CTDB_REC_RO_HAVE_DELEGATIONS))) {
 866                 header.rsn     += 3;
 867                 header.flags   |= CTDB_REC_RO_HAVE_DELEGATIONS;
 868                 if (ctdb_ltdb_store(ctdb_db, call->key, &header, data) != 0) {
 869                         ctdb_fatal(ctdb, "Failed to store record with HAVE_DELEGATIONS set");
 870                 }
 871         }
 872         if ((c->flags & CTDB_WANT_READONLY)
 873         &&  (call->call_id == CTDB_FETCH_WITH_HEADER_FUNC)) {
 874                 TDB_DATA tdata;
 875
 876                 tdata = tdb_fetch(ctdb_db->rottdb, call->key);
 877                 if (ctdb_trackingdb_add_pnn(ctdb, &tdata, c->hdr.srcnode) != 0) {
 878                         ctdb_fatal(ctdb, "Failed to add node to trackingdb");
 879                 }
 880                 if (tdb_store(ctdb_db->rottdb, call->key, tdata, TDB_REPLACE) != 0) {
 881                         ctdb_fatal(ctdb, "Failed to store trackingdb data");
 882                 }
 883                 free(tdata.dptr);
 884
 885                 ret = ctdb_ltdb_unlock(ctdb_db, call->key);
 886                 if (ret != 0) {
 887                         DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 888                 }
 889
 890                 len = offsetof(struct ctdb_reply_call, data) + data.dsize + sizeof(struct ctdb_ltdb_header);
 891                 r = ctdb_transport_allocate(ctdb, ctdb, CTDB_REPLY_CALL, len,
 892                                             struct ctdb_reply_call);
 893                 CTDB_NO_MEMORY_FATAL(ctdb, r);
 894                 r->hdr.destnode  = c->hdr.srcnode;
 895                 r->hdr.reqid     = c->hdr.reqid;
 896                 r->status        = 0;
 897                 r->datalen       = data.dsize + sizeof(struct ctdb_ltdb_header);
 898                 header.rsn      -= 2;
 899                 header.flags   |= CTDB_REC_RO_HAVE_READONLY;
 900                 header.flags   &= ~CTDB_REC_RO_HAVE_DELEGATIONS;
 901                 memcpy(&r->data[0], &header, sizeof(struct ctdb_ltdb_header));
 902
 903                 if (data.dsize) {
 904                         memcpy(&r->data[sizeof(struct ctdb_ltdb_header)], data.dptr, data.dsize);
 905                 }
 906
 907                 ctdb_queue_packet(ctdb, &r->hdr);
 908                 CTDB_INCREMENT_STAT(ctdb, total_ro_delegations);
 909                 CTDB_INCREMENT_DB_STAT(ctdb_db, db_ro_delegations);
 910
 911                 talloc_free(r);
 912                 talloc_free(call);
 913                 return;
 914         }
 915
 916         CTDB_UPDATE_STAT(ctdb, max_hop_count, c->hopcount);
 917         tmp_count = c->hopcount;
 918         bucket = 0;
 919         while (tmp_count) {
 920                 tmp_count >>= 2;
 921                 bucket++;
 922         }
 923         if (bucket >= MAX_COUNT_BUCKETS) {
 924                 bucket = MAX_COUNT_BUCKETS - 1;
 925         }
 926         CTDB_INCREMENT_STAT(ctdb, hop_count_bucket[bucket]);
 927         CTDB_INCREMENT_DB_STAT(ctdb_db, hop_count_bucket[bucket]);
 928         ctdb_update_db_stat_hot_keys(ctdb_db, call->key, c->hopcount);
 929
 930         /* If this database supports sticky records, then check if the
 931            hopcount is big. If it is it means the record is hot and we
 932            should make it sticky.
 933         */
 934         if (ctdb_db->sticky && c->hopcount >= ctdb->tunable.hopcount_make_sticky) {
 935                 ctdb_make_record_sticky(ctdb, ctdb_db, call->key);
 936         }
 937
 938
 939         /* Try if possible to migrate the record off to the caller node.
 940          * From the clients perspective a fetch of the data is just as
 941          * expensive as a migration.
 942          */
 943         if (c->hdr.srcnode != ctdb->pnn) {
 944                 if (ctdb_db->persistent_state) {
 945                         DEBUG(DEBUG_INFO, (__location__ " refusing migration"
 946                               " of key %s while transaction is active\n",
 947                               (char *)call->key.dptr));
 948                 } else {
 949                         DEBUG(DEBUG_DEBUG,("pnn %u starting migration of %08x to %u\n",
 950                                  ctdb->pnn, ctdb_hash(&(call->key)), c->hdr.srcnode));
 951                         ctdb_call_send_dmaster(ctdb_db, c, &header, &(call->key), &data);
 952                         talloc_free(data.dptr);
 953
 954                         ret = ctdb_ltdb_unlock(ctdb_db, call->key);
 955                         if (ret != 0) {
 956                                 DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 957                         }
 958                 }
 959                 talloc_free(call);
 960                 return;
 961         }
 962
 963         ret = ctdb_call_local(ctdb_db, call, &header, hdr, &data, true);
 964         if (ret != 0) {
 965                 DEBUG(DEBUG_ERR,(__location__ " ctdb_call_local failed\n"));
 966                 call->status = -1;
 967         }
 968
 969         ret = ctdb_ltdb_unlock(ctdb_db, call->key);
 970         if (ret != 0) {
 971                 DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 972         }
 973
 974         len = offsetof(struct ctdb_reply_call, data) + call->reply_data.dsize;
 975         r = ctdb_transport_allocate(ctdb, ctdb, CTDB_REPLY_CALL, len,
 976                                     struct ctdb_reply_call);
 977         CTDB_NO_MEMORY_FATAL(ctdb, r);
 978         r->hdr.destnode  = hdr->srcnode;
 979         r->hdr.reqid     = hdr->reqid;
 980         r->status        = call->status;
 981         r->datalen       = call->reply_data.dsize;
 982         if (call->reply_data.dsize) {
 983                 memcpy(&r->data[0], call->reply_data.dptr, call->reply_data.dsize);
 984         }
 985
 986         ctdb_queue_packet(ctdb, &r->hdr);
 987
 988         talloc_free(r);
 989         talloc_free(call);
 990 }
 991
 992 /**
 993  * called when a CTDB_REPLY_CALL packet comes in
 994  *
 995  * This packet comes in response to a CTDB_REQ_CALL request packet. It
 996  * contains any reply data from the call
 997  */
 998 void ctdb_reply_call(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
 999 {
1000         struct ctdb_reply_call *c = (struct ctdb_reply_call *)hdr;
1001         struct ctdb_call_state *state;
1002
1003         state = ctdb_reqid_find(ctdb, hdr->reqid, struct ctdb_call_state);
1004         if (state == NULL) {
1005                 DEBUG(DEBUG_ERR, (__location__ " reqid %u not found\n", hdr->reqid));
1006                 return;
1007         }
1008
1009         if (hdr->reqid != state->reqid) {
1010                 /* we found a record  but it was the wrong one */
1011                 DEBUG(DEBUG_ERR, ("Dropped orphaned call reply with reqid:%u\n",hdr->reqid));
1012                 return;
1013         }
1014
1015
1016         /* read only delegation processing */
1017         /* If we got a FETCH_WITH_HEADER we should check if this is a ro
1018          * delegation since we may need to update the record header
1019          */
1020         if (state->c->callid == CTDB_FETCH_WITH_HEADER_FUNC) {
1021                 struct ctdb_db_context *ctdb_db = state->ctdb_db;
1022                 struct ctdb_ltdb_header *header = (struct ctdb_ltdb_header *)&c->data[0];
1023                 struct ctdb_ltdb_header oldheader;
1024                 TDB_DATA key, data, olddata;
1025                 int ret;
1026
1027                 if (!(header->flags & CTDB_REC_RO_HAVE_READONLY)) {
1028                         goto finished_ro;
1029                         return;
1030                 }
1031
1032                 key.dsize = state->c->keylen;
1033                 key.dptr  = state->c->data;
1034                 ret = ctdb_ltdb_lock_requeue(ctdb_db, key, hdr,
1035                                      ctdb_call_input_pkt, ctdb, false);
1036                 if (ret == -2) {
1037                         return;
1038                 }
1039                 if (ret != 0) {
1040                         DEBUG(DEBUG_ERR,(__location__ " Failed to get lock in ctdb_reply_call\n"));
1041                         return;
1042                 }
1043
1044                 ret = ctdb_ltdb_fetch(ctdb_db, key, &oldheader, state, &olddata);
1045                 if (ret != 0) {
1046                         DEBUG(DEBUG_ERR, ("Failed to fetch old record in ctdb_reply_call\n"));
1047                         ctdb_ltdb_unlock(ctdb_db, key);
1048                         goto finished_ro;
1049                 }
1050
1051                 if (header->rsn <= oldheader.rsn) {
1052                         ctdb_ltdb_unlock(ctdb_db, key);
1053                         goto finished_ro;
1054                 }
1055
1056                 if (c->datalen < sizeof(struct ctdb_ltdb_header)) {
1057                         DEBUG(DEBUG_ERR,(__location__ " Got FETCH_WITH_HEADER reply with too little data: %d bytes\n", c->datalen));
1058                         ctdb_ltdb_unlock(ctdb_db, key);
1059                         goto finished_ro;
1060                 }
1061
1062                 data.dsize = c->datalen - sizeof(struct ctdb_ltdb_header);
1063                 data.dptr  = &c->data[sizeof(struct ctdb_ltdb_header)];
1064                 ret = ctdb_ltdb_store(ctdb_db, key, header, data);
1065                 if (ret != 0) {
1066                         DEBUG(DEBUG_ERR, ("Failed to store new record in ctdb_reply_call\n"));
1067                         ctdb_ltdb_unlock(ctdb_db, key);
1068                         goto finished_ro;
1069                 }
1070
1071                 ctdb_ltdb_unlock(ctdb_db, key);
1072         }
1073 finished_ro:
1074
1075         state->call->reply_data.dptr = c->data;
1076         state->call->reply_data.dsize = c->datalen;
1077         state->call->status = c->status;
1078
1079         talloc_steal(state, c);
1080
1081         state->state = CTDB_CALL_DONE;
1082         if (state->async.fn) {
1083                 state->async.fn(state);
1084         }
1085 }
1086
1087
1088 /**
1089  * called when a CTDB_REPLY_DMASTER packet comes in
1090  *
1091  * This packet comes in from the lmaster in response to a CTDB_REQ_CALL
1092  * request packet. It means that the current dmaster wants to give us
1093  * the dmaster role.
1094  */
1095 void ctdb_reply_dmaster(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
1096 {
1097         struct ctdb_reply_dmaster *c = (struct ctdb_reply_dmaster *)hdr;
1098         struct ctdb_db_context *ctdb_db;
1099         TDB_DATA key, data;
1100         uint32_t record_flags = 0;
1101         size_t len;
1102         int ret;
1103
1104         ctdb_db = find_ctdb_db(ctdb, c->db_id);
1105         if (ctdb_db == NULL) {
1106                 DEBUG(DEBUG_ERR,("Unknown db_id 0x%x in ctdb_reply_dmaster\n", c->db_id));
1107                 return;
1108         }
1109
1110         key.dptr = c->data;
1111         key.dsize = c->keylen;
1112         data.dptr = &c->data[key.dsize];
1113         data.dsize = c->datalen;
1114         len = offsetof(struct ctdb_reply_dmaster, data) + key.dsize + data.dsize
1115                 + sizeof(uint32_t);
1116         if (len <= c->hdr.length) {
1117                 record_flags = *(uint32_t *)&c->data[c->keylen + c->datalen];
1118         }
1119
1120         ret = ctdb_ltdb_lock_requeue(ctdb_db, key, hdr,
1121                                      ctdb_call_input_pkt, ctdb, false);
1122         if (ret == -2) {
1123                 return;
1124         }
1125         if (ret != 0) {
1126                 DEBUG(DEBUG_ERR,(__location__ " Failed to get lock in ctdb_reply_dmaster\n"));
1127                 return;
1128         }
1129
1130         ctdb_become_dmaster(ctdb_db, hdr, key, data, c->rsn, record_flags);
1131 }
1132
1133
1134 /*
1135   called when a CTDB_REPLY_ERROR packet comes in
1136 */
1137 void ctdb_reply_error(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
1138 {
1139         struct ctdb_reply_error *c = (struct ctdb_reply_error *)hdr;
1140         struct ctdb_call_state *state;
1141
1142         state = ctdb_reqid_find(ctdb, hdr->reqid, struct ctdb_call_state);
1143         if (state == NULL) {
1144                 DEBUG(DEBUG_ERR,("pnn %u Invalid reqid %u in ctdb_reply_error\n",
1145                          ctdb->pnn, hdr->reqid));
1146                 return;
1147         }
1148
1149         if (hdr->reqid != state->reqid) {
1150                 /* we found a record  but it was the wrong one */
1151                 DEBUG(DEBUG_ERR, ("Dropped orphaned error reply with reqid:%u\n",hdr->reqid));
1152                 return;
1153         }
1154
1155         talloc_steal(state, c);
1156
1157         state->state  = CTDB_CALL_ERROR;
1158         state->errmsg = (char *)c->msg;
1159         if (state->async.fn) {
1160                 state->async.fn(state);
1161         }
1162 }
1163
1164
1165 /*
1166   destroy a ctdb_call
1167 */
1168 static int ctdb_call_destructor(struct ctdb_call_state *state)
1169 {
1170         DLIST_REMOVE(state->ctdb_db->ctdb->pending_calls, state);
1171         ctdb_reqid_remove(state->ctdb_db->ctdb, state->reqid);
1172         return 0;
1173 }
1174
1175
1176 /*
1177   called when a ctdb_call needs to be resent after a reconfigure event
1178 */
1179 static void ctdb_call_resend(struct ctdb_call_state *state)
1180 {
1181         struct ctdb_context *ctdb = state->ctdb_db->ctdb;
1182
1183         state->generation = ctdb->vnn_map->generation;
1184
1185         /* use a new reqid, in case the old reply does eventually come in */
1186         ctdb_reqid_remove(ctdb, state->reqid);
1187         state->reqid = ctdb_reqid_new(ctdb, state);
1188         state->c->hdr.reqid = state->reqid;
1189
1190         /* update the generation count for this request, so its valid with the new vnn_map */
1191         state->c->hdr.generation = state->generation;
1192
1193         /* send the packet to ourselves, it will be redirected appropriately */
1194         state->c->hdr.destnode = ctdb->pnn;
1195
1196         ctdb_queue_packet(ctdb, &state->c->hdr);
1197         DEBUG(DEBUG_NOTICE,("resent ctdb_call\n"));
1198 }
1199
1200 /*
1201   resend all pending calls on recovery
1202  */
1203 void ctdb_call_resend_all(struct ctdb_context *ctdb)
1204 {
1205         struct ctdb_call_state *state, *next;
1206         for (state=ctdb->pending_calls;state;state=next) {
1207                 next = state->next;
1208                 ctdb_call_resend(state);
1209         }
1210 }
1211
1212 /*
1213   this allows the caller to setup a async.fn
1214 */
1215 static void call_local_trigger(struct event_context *ev, struct timed_event *te,
1216                        struct timeval t, void *private_data)
1217 {
1218         struct ctdb_call_state *state = talloc_get_type(private_data, struct ctdb_call_state);
1219         if (state->async.fn) {
1220                 state->async.fn(state);
1221         }
1222 }
1223
1224
1225 /*
1226   construct an event driven local ctdb_call
1227
1228   this is used so that locally processed ctdb_call requests are processed
1229   in an event driven manner
1230 */
1231 struct ctdb_call_state *ctdb_call_local_send(struct ctdb_db_context *ctdb_db,
1232                                              struct ctdb_call *call,
1233                                              struct ctdb_ltdb_header *header,
1234                                              TDB_DATA *data)
1235 {
1236         struct ctdb_call_state *state;
1237         struct ctdb_context *ctdb = ctdb_db->ctdb;
1238         int ret;
1239
1240         state = talloc_zero(ctdb_db, struct ctdb_call_state);
1241         CTDB_NO_MEMORY_NULL(ctdb, state);
1242
1243         talloc_steal(state, data->dptr);
1244
1245         state->state = CTDB_CALL_DONE;
1246         state->call  = talloc(state, struct ctdb_call);
1247         CTDB_NO_MEMORY_NULL(ctdb, state->call);
1248         *(state->call) = *call;
1249         state->ctdb_db = ctdb_db;
1250
1251         ret = ctdb_call_local(ctdb_db, state->call, header, state, data, true);
1252         if (ret != 0) {
1253                 DEBUG(DEBUG_DEBUG,("ctdb_call_local() failed, ignoring return code %d\n", ret));
1254         }
1255
1256         event_add_timed(ctdb->ev, state, timeval_zero(), call_local_trigger, state);
1257
1258         return state;
1259 }
1260
1261
1262 /*
1263   make a remote ctdb call - async send. Called in daemon context.
1264
1265   This constructs a ctdb_call request and queues it for processing.
1266   This call never blocks.
1267 */
1268 struct ctdb_call_state *ctdb_daemon_call_send_remote(struct ctdb_db_context *ctdb_db,
1269                                                      struct ctdb_call *call,
1270                                                      struct ctdb_ltdb_header *header)
1271 {
1272         uint32_t len;
1273         struct ctdb_call_state *state;
1274         struct ctdb_context *ctdb = ctdb_db->ctdb;
1275
1276         if (ctdb->methods == NULL) {
1277                 DEBUG(DEBUG_INFO,(__location__ " Failed send packet. Transport is down\n"));
1278                 return NULL;
1279         }
1280
1281         state = talloc_zero(ctdb_db, struct ctdb_call_state);
1282         CTDB_NO_MEMORY_NULL(ctdb, state);
1283         state->call = talloc(state, struct ctdb_call);
1284         CTDB_NO_MEMORY_NULL(ctdb, state->call);
1285
1286         state->reqid = ctdb_reqid_new(ctdb, state);
1287         state->ctdb_db = ctdb_db;
1288         talloc_set_destructor(state, ctdb_call_destructor);
1289
1290         len = offsetof(struct ctdb_req_call, data) + call->key.dsize + call->call_data.dsize;
1291         state->c = ctdb_transport_allocate(ctdb, state, CTDB_REQ_CALL, len,
1292                                            struct ctdb_req_call);
1293         CTDB_NO_MEMORY_NULL(ctdb, state->c);
1294         state->c->hdr.destnode  = header->dmaster;
1295
1296         /* this limits us to 16k outstanding messages - not unreasonable */
1297         state->c->hdr.reqid     = state->reqid;
1298         state->c->flags         = call->flags;
1299         state->c->db_id         = ctdb_db->db_id;
1300         state->c->callid        = call->call_id;
1301         state->c->hopcount      = 0;
1302         state->c->keylen        = call->key.dsize;
1303         state->c->calldatalen   = call->call_data.dsize;
1304         memcpy(&state->c->data[0], call->key.dptr, call->key.dsize);
1305         memcpy(&state->c->data[call->key.dsize],
1306                call->call_data.dptr, call->call_data.dsize);
1307         *(state->call)              = *call;
1308         state->call->call_data.dptr = &state->c->data[call->key.dsize];
1309         state->call->key.dptr       = &state->c->data[0];
1310
1311         state->state  = CTDB_CALL_WAIT;
1312         state->generation = ctdb->vnn_map->generation;
1313
1314         DLIST_ADD(ctdb->pending_calls, state);
1315
1316         ctdb_queue_packet(ctdb, &state->c->hdr);
1317
1318         return state;
1319 }
1320
1321 /*
1322   make a remote ctdb call - async recv - called in daemon context
1323
1324   This is called when the program wants to wait for a ctdb_call to complete and get the
1325   results. This call will block unless the call has already completed.
1326 */
1327 int ctdb_daemon_call_recv(struct ctdb_call_state *state, struct ctdb_call *call)
1328 {
1329         while (state->state < CTDB_CALL_DONE) {
1330                 event_loop_once(state->ctdb_db->ctdb->ev);
1331         }
1332         if (state->state != CTDB_CALL_DONE) {
1333                 ctdb_set_error(state->ctdb_db->ctdb, "%s", state->errmsg);
1334                 talloc_free(state);
1335                 return -1;
1336         }
1337
1338         if (state->call->reply_data.dsize) {
1339                 call->reply_data.dptr = talloc_memdup(call,
1340                                                       state->call->reply_data.dptr,
1341                                                       state->call->reply_data.dsize);
1342                 call->reply_data.dsize = state->call->reply_data.dsize;
1343         } else {
1344                 call->reply_data.dptr = NULL;
1345                 call->reply_data.dsize = 0;
1346         }
1347         call->status = state->call->status;
1348         talloc_free(state);
1349         return 0;
1350 }
1351
1352
1353 /*
1354    send a keepalive packet to the other node
1355 */
1356 void ctdb_send_keepalive(struct ctdb_context *ctdb, uint32_t destnode)
1357 {
1358         struct ctdb_req_keepalive *r;
1359
1360         if (ctdb->methods == NULL) {
1361                 DEBUG(DEBUG_INFO,(__location__ " Failed to send keepalive. Transport is DOWN\n"));
1362                 return;
1363         }
1364
1365         r = ctdb_transport_allocate(ctdb, ctdb, CTDB_REQ_KEEPALIVE,
1366                                     sizeof(struct ctdb_req_keepalive),
1367                                     struct ctdb_req_keepalive);
1368         CTDB_NO_MEMORY_FATAL(ctdb, r);
1369         r->hdr.destnode  = destnode;
1370         r->hdr.reqid     = 0;
1371
1372         CTDB_INCREMENT_STAT(ctdb, keepalive_packets_sent);
1373
1374         ctdb_queue_packet(ctdb, &r->hdr);
1375
1376         talloc_free(r);
1377 }
1378
1379
1380
1381 struct revokechild_deferred_call {
1382         struct ctdb_context *ctdb;
1383         struct ctdb_req_header *hdr;
1384         deferred_requeue_fn fn;
1385         void *ctx;
1386 };
1387
1388 struct revokechild_handle {
1389         struct revokechild_handle *next, *prev;
1390         struct ctdb_context *ctdb;
1391         struct ctdb_db_context *ctdb_db;
1392         struct fd_event *fde;
1393         int status;
1394         int fd[2];
1395         pid_t child;
1396         TDB_DATA key;
1397 };
1398
1399 struct revokechild_requeue_handle {
1400         struct ctdb_context *ctdb;
1401         struct ctdb_req_header *hdr;
1402         deferred_requeue_fn fn;
1403         void *ctx;
1404 };
1405
1406 static void deferred_call_requeue(struct event_context *ev, struct timed_event *te,
1407                        struct timeval t, void *private_data)
1408 {
1409         struct revokechild_requeue_handle *requeue_handle = talloc_get_type(private_data, struct revokechild_requeue_handle);
1410
1411         requeue_handle->fn(requeue_handle->ctx, requeue_handle->hdr);
1412         talloc_free(requeue_handle);
1413 }
1414
1415 static int deferred_call_destructor(struct revokechild_deferred_call *deferred_call)
1416 {
1417         struct ctdb_context *ctdb = deferred_call->ctdb;
1418         struct revokechild_requeue_handle *requeue_handle = talloc(ctdb, struct revokechild_requeue_handle);
1419         struct ctdb_req_call *c = (struct ctdb_req_call *)deferred_call->hdr;
1420
1421         requeue_handle->ctdb = ctdb;
1422         requeue_handle->hdr  = deferred_call->hdr;
1423         requeue_handle->fn   = deferred_call->fn;
1424         requeue_handle->ctx  = deferred_call->ctx;
1425         talloc_steal(requeue_handle, requeue_handle->hdr);
1426
1427         /* when revoking, any READONLY requests have 1 second grace to let read/write finish first */
1428         event_add_timed(ctdb->ev, requeue_handle, timeval_current_ofs(c->flags & CTDB_WANT_READONLY ? 1 : 0, 0), deferred_call_requeue, requeue_handle);
1429
1430         return 0;
1431 }
1432
1433
1434 static int revokechild_destructor(struct revokechild_handle *rc)
1435 {
1436         if (rc->fde != NULL) {
1437                 talloc_free(rc->fde);
1438         }
1439
1440         if (rc->fd[0] != -1) {
1441                 close(rc->fd[0]);
1442         }
1443         if (rc->fd[1] != -1) {
1444                 close(rc->fd[1]);
1445         }
1446         ctdb_kill(rc->ctdb, rc->child, SIGKILL);
1447
1448         DLIST_REMOVE(rc->ctdb_db->revokechild_active, rc);
1449         return 0;
1450 }
1451
1452 static void revokechild_handler(struct event_context *ev, struct fd_event *fde,
1453                              uint16_t flags, void *private_data)
1454 {
1455         struct revokechild_handle *rc = talloc_get_type(private_data,
1456                                                      struct revokechild_handle);
1457         int ret;
1458         char c;
1459
1460         ret = read(rc->fd[0], &c, 1);
1461         if (ret != 1) {
1462                 DEBUG(DEBUG_ERR,("Failed to read status from revokechild. errno:%d\n", errno));
1463                 rc->status = -1;
1464                 talloc_free(rc);
1465                 return;
1466         }
1467         if (c != 0) {
1468                 DEBUG(DEBUG_ERR,("revokechild returned failure. status:%d\n", c));
1469                 rc->status = -1;
1470                 talloc_free(rc);
1471                 return;
1472         }
1473
1474         talloc_free(rc);
1475 }
1476
1477 struct ctdb_revoke_state {
1478         struct ctdb_db_context *ctdb_db;
1479         TDB_DATA key;
1480         struct ctdb_ltdb_header *header;
1481         TDB_DATA data;
1482         int count;
1483         int status;
1484         int finished;
1485 };
1486
1487 static void update_record_cb(struct ctdb_client_control_state *state)
1488 {
1489         struct ctdb_revoke_state *revoke_state;
1490         int ret;
1491         int32_t res;
1492
1493         if (state == NULL) {
1494                 return;
1495         }
1496         revoke_state = state->async.private_data;
1497
1498         state->async.fn = NULL;
1499         ret = ctdb_control_recv(state->ctdb, state, state, NULL, &res, NULL);
1500         if ((ret != 0) || (res != 0)) {
1501                 DEBUG(DEBUG_ERR,("Recv for revoke update record failed ret:%d res:%d\n", ret, res));
1502                 revoke_state->status = -1;
1503         }
1504
1505         revoke_state->count--;
1506         if (revoke_state->count <= 0) {
1507                 revoke_state->finished = 1;
1508         }
1509 }
1510
1511 static void revoke_send_cb(struct ctdb_context *ctdb, uint32_t pnn, void *private_data)
1512 {
1513         struct ctdb_revoke_state *revoke_state = private_data;
1514         struct ctdb_client_control_state *state;
1515
1516         state = ctdb_ctrl_updaterecord_send(ctdb, revoke_state, timeval_current_ofs(5,0), pnn, revoke_state->ctdb_db, revoke_state->key, revoke_state->header, revoke_state->data);
1517         if (state == NULL) {
1518                 DEBUG(DEBUG_ERR,("Failure to send update record to revoke readonly delegation\n"));
1519                 revoke_state->status = -1;
1520                 return;
1521         }
1522         state->async.fn           = update_record_cb;
1523         state->async.private_data = revoke_state;
1524
1525         revoke_state->count++;
1526
1527 }
1528
1529 static void ctdb_revoke_timeout_handler(struct event_context *ev, struct timed_event *te,
1530                               struct timeval yt, void *private_data)
1531 {
1532         struct ctdb_revoke_state *state = private_data;
1533
1534         DEBUG(DEBUG_ERR,("Timed out waiting for revoke to finish\n"));
1535         state->finished = 1;
1536         state->status   = -1;
1537 }
1538
1539 static int ctdb_revoke_all_delegations(struct ctdb_context *ctdb, struct ctdb_db_context *ctdb_db, TDB_DATA tdata, TDB_DATA key, struct ctdb_ltdb_header *header, TDB_DATA data)
1540 {
1541         struct ctdb_revoke_state *state = talloc_zero(ctdb, struct ctdb_revoke_state);
1542         int status;
1543
1544         state->ctdb_db = ctdb_db;
1545         state->key     = key;
1546         state->header  = header;
1547         state->data    = data;
1548
1549         ctdb_trackingdb_traverse(ctdb, tdata, revoke_send_cb, state);
1550
1551         event_add_timed(ctdb->ev, state, timeval_current_ofs(5, 0), ctdb_revoke_timeout_handler, state);
1552
1553         while (state->finished == 0) {
1554                 event_loop_once(ctdb->ev);
1555         }
1556
1557         status = state->status;
1558
1559         if (status == 0) {
1560                 struct ctdb_ltdb_header new_header;
1561                 TDB_DATA new_data;
1562
1563                 if (ctdb_ltdb_lock(ctdb_db, key) != 0) {
1564                         DEBUG(DEBUG_ERR,("Failed to chainlock the database in revokechild\n"));
1565                         talloc_free(state);
1566                         return -1;
1567                 }
1568                 if (ctdb_ltdb_fetch(ctdb_db, key, &new_header, state, &new_data) != 0) {
1569                         ctdb_ltdb_unlock(ctdb_db, key);
1570                         DEBUG(DEBUG_ERR,("Failed for fetch tdb record in revokechild\n"));
1571                         talloc_free(state);
1572                         return -1;
1573                 }
1574                 header->rsn++;
1575                 if (new_header.rsn > header->rsn) {
1576                         ctdb_ltdb_unlock(ctdb_db, key);
1577                         DEBUG(DEBUG_ERR,("RSN too high in tdb record in revokechild\n"));
1578                         talloc_free(state);
1579                         return -1;
1580                 }
1581                 if ( (new_header.flags & (CTDB_REC_RO_REVOKING_READONLY|CTDB_REC_RO_HAVE_DELEGATIONS)) != (CTDB_REC_RO_REVOKING_READONLY|CTDB_REC_RO_HAVE_DELEGATIONS) ) {
1582                         ctdb_ltdb_unlock(ctdb_db, key);
1583                         DEBUG(DEBUG_ERR,("Flags are wrong in tdb record in revokechild\n"));
1584                         talloc_free(state);
1585                         return -1;
1586                 }
1587                 new_header.rsn++;
1588                 new_header.flags |= CTDB_REC_RO_REVOKE_COMPLETE;
1589                 if (ctdb_ltdb_store(ctdb_db, key, &new_header, new_data) != 0) {
1590                         ctdb_ltdb_unlock(ctdb_db, key);
1591                         DEBUG(DEBUG_ERR,("Failed to write new record in revokechild\n"));
1592                         talloc_free(state);
1593                         return -1;
1594                 }
1595                 ctdb_ltdb_unlock(ctdb_db, key);
1596         }
1597
1598         talloc_free(state);
1599         return status;
1600 }
1601
1602
1603 int ctdb_start_revoke_ro_record(struct ctdb_context *ctdb, struct ctdb_db_context *ctdb_db, TDB_DATA key, struct ctdb_ltdb_header *header, TDB_DATA data)
1604 {
1605         TDB_DATA tdata;
1606         struct revokechild_handle *rc;
1607         pid_t parent = getpid();
1608         int ret;
1609
1610         header->flags &= ~(CTDB_REC_RO_REVOKING_READONLY|CTDB_REC_RO_HAVE_DELEGATIONS|CTDB_REC_RO_HAVE_READONLY);
1611         header->flags |= CTDB_REC_FLAG_MIGRATED_WITH_DATA;
1612         header->rsn   -= 1;
1613
1614         if ((rc = talloc_zero(ctdb_db, struct revokechild_handle)) == NULL) {
1615                 DEBUG(DEBUG_ERR,("Failed to allocate revokechild_handle\n"));
1616                 return -1;
1617         }
1618
1619         tdata = tdb_fetch(ctdb_db->rottdb, key);
1620         if (tdata.dsize > 0) {
1621                 uint8_t *tmp;
1622
1623                 tmp = tdata.dptr;
1624                 tdata.dptr = talloc_memdup(rc, tdata.dptr, tdata.dsize);
1625                 free(tmp);
1626         }
1627
1628         rc->status    = 0;
1629         rc->ctdb      = ctdb;
1630         rc->ctdb_db   = ctdb_db;
1631         rc->fd[0]     = -1;
1632         rc->fd[1]     = -1;
1633
1634         talloc_set_destructor(rc, revokechild_destructor);
1635
1636         rc->key.dsize = key.dsize;
1637         rc->key.dptr  = talloc_memdup(rc, key.dptr, key.dsize);
1638         if (rc->key.dptr == NULL) {
1639                 DEBUG(DEBUG_ERR,("Failed to allocate key for revokechild_handle\n"));
1640                 talloc_free(rc);
1641                 return -1;
1642         }
1643
1644         ret = pipe(rc->fd);
1645         if (ret != 0) {
1646                 DEBUG(DEBUG_ERR,("Failed to allocate key for revokechild_handle\n"));
1647                 talloc_free(rc);
1648                 return -1;
1649         }
1650
1651
1652         rc->child = ctdb_fork(ctdb);
1653         if (rc->child == (pid_t)-1) {
1654                 DEBUG(DEBUG_ERR,("Failed to fork child for revokechild\n"));
1655                 talloc_free(rc);
1656                 return -1;
1657         }
1658
1659         if (rc->child == 0) {
1660                 char c = 0;
1661                 close(rc->fd[0]);
1662                 debug_extra = talloc_asprintf(NULL, "revokechild-%s:", ctdb_db->db_name);
1663
1664                 ctdb_set_process_name("ctdb_revokechild");
1665                 if (switch_from_server_to_client(ctdb, "revokechild-%s", ctdb_db->db_name) != 0) {
1666                         DEBUG(DEBUG_ERR,("Failed to switch from server to client for revokechild process\n"));
1667                         c = 1;
1668                         goto child_finished;
1669                 }
1670
1671                 c = ctdb_revoke_all_delegations(ctdb, ctdb_db, tdata, key, header, data);
1672
1673 child_finished:
1674                 write(rc->fd[1], &c, 1);
1675                 /* make sure we die when our parent dies */
1676                 while (ctdb_kill(ctdb, parent, 0) == 0 || errno != ESRCH) {
1677                         sleep(5);
1678                 }
1679                 _exit(0);
1680         }
1681
1682         close(rc->fd[1]);
1683         rc->fd[1] = -1;
1684         set_close_on_exec(rc->fd[0]);
1685
1686         /* This is an active revokechild child process */
1687         DLIST_ADD_END(ctdb_db->revokechild_active, rc, NULL);
1688
1689         rc->fde = event_add_fd(ctdb->ev, rc, rc->fd[0],
1690                                    EVENT_FD_READ, revokechild_handler,
1691                                    (void *)rc);
1692         if (rc->fde == NULL) {
1693                 DEBUG(DEBUG_ERR,("Failed to set up fd event for revokechild process\n"));
1694                 talloc_free(rc);
1695         }
1696         tevent_fd_set_auto_close(rc->fde);
1697
1698         return 0;
1699 }
1700
1701 int ctdb_add_revoke_deferred_call(struct ctdb_context *ctdb, struct ctdb_db_context *ctdb_db, TDB_DATA key, struct ctdb_req_header *hdr, deferred_requeue_fn fn, void *call_context)
1702 {
1703         struct revokechild_handle *rc;
1704         struct revokechild_deferred_call *deferred_call;
1705
1706         for (rc = ctdb_db->revokechild_active; rc; rc = rc->next) {
1707                 if (rc->key.dsize == 0) {
1708                         continue;
1709                 }
1710                 if (rc->key.dsize != key.dsize) {
1711                         continue;
1712                 }
1713                 if (!memcmp(rc->key.dptr, key.dptr, key.dsize)) {
1714                         break;
1715                 }
1716         }
1717
1718         if (rc == NULL) {
1719                 DEBUG(DEBUG_ERR,("Failed to add deferred call to revoke list. revoke structure not found\n"));
1720                 return -1;
1721         }
1722
1723         deferred_call = talloc(rc, struct revokechild_deferred_call);
1724         if (deferred_call == NULL) {
1725                 DEBUG(DEBUG_ERR,("Failed to allocate deferred call structure for revoking record\n"));
1726                 return -1;
1727         }
1728
1729         deferred_call->ctdb = ctdb;
1730         deferred_call->hdr  = hdr;
1731         deferred_call->fn   = fn;
1732         deferred_call->ctx  = call_context;
1733
1734         talloc_set_destructor(deferred_call, deferred_call_destructor);
1735         talloc_steal(deferred_call, hdr);
1736
1737         return 0;
1738 }