ctdb/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 #include "common/reqid.h"
  31 #include "common/system.h"
  32
  33 struct ctdb_sticky_record {
  34         struct ctdb_context *ctdb;
  35         struct ctdb_db_context *ctdb_db;
  36         TDB_CONTEXT *pindown;
  37 };
  38
  39 /*
  40   find the ctdb_db from a db index
  41  */
  42  struct ctdb_db_context *find_ctdb_db(struct ctdb_context *ctdb, uint32_t id)
  43 {
  44         struct ctdb_db_context *ctdb_db;
  45
  46         for (ctdb_db=ctdb->db_list; ctdb_db; ctdb_db=ctdb_db->next) {
  47                 if (ctdb_db->db_id == id) {
  48                         break;
  49                 }
  50         }
  51         return ctdb_db;
  52 }
  53
  54 /*
  55   a varient of input packet that can be used in lock requeue
  56 */
  57 static void ctdb_call_input_pkt(void *p, struct ctdb_req_header *hdr)
  58 {
  59         struct ctdb_context *ctdb = talloc_get_type(p, struct ctdb_context);
  60         ctdb_input_pkt(ctdb, hdr);
  61 }
  62
  63
  64 /*
  65   send an error reply
  66 */
  67 static void ctdb_send_error(struct ctdb_context *ctdb,
  68                             struct ctdb_req_header *hdr, uint32_t status,
  69                             const char *fmt, ...) PRINTF_ATTRIBUTE(4,5);
  70 static void ctdb_send_error(struct ctdb_context *ctdb,
  71                             struct ctdb_req_header *hdr, uint32_t status,
  72                             const char *fmt, ...)
  73 {
  74         va_list ap;
  75         struct ctdb_reply_error *r;
  76         char *msg;
  77         int msglen, len;
  78
  79         if (ctdb->methods == NULL) {
  80                 DEBUG(DEBUG_INFO,(__location__ " Failed to send error. Transport is DOWN\n"));
  81                 return;
  82         }
  83
  84         va_start(ap, fmt);
  85         msg = talloc_vasprintf(ctdb, fmt, ap);
  86         if (msg == NULL) {
  87                 ctdb_fatal(ctdb, "Unable to allocate error in ctdb_send_error\n");
  88         }
  89         va_end(ap);
  90
  91         msglen = strlen(msg)+1;
  92         len = offsetof(struct ctdb_reply_error, msg);
  93         r = ctdb_transport_allocate(ctdb, msg, CTDB_REPLY_ERROR, len + msglen,
  94                                     struct ctdb_reply_error);
  95         CTDB_NO_MEMORY_FATAL(ctdb, r);
  96
  97         r->hdr.destnode  = hdr->srcnode;
  98         r->hdr.reqid     = hdr->reqid;
  99         r->status        = status;
 100         r->msglen        = msglen;
 101         memcpy(&r->msg[0], msg, msglen);
 102
 103         ctdb_queue_packet(ctdb, &r->hdr);
 104
 105         talloc_free(msg);
 106 }
 107
 108
 109 /**
 110  * send a redirect reply
 111  *
 112  * The logic behind this function is this:
 113  *
 114  * A client wants to grab a record and sends a CTDB_REQ_CALL packet
 115  * to its local ctdb (ctdb_request_call). If the node is not itself
 116  * the record's DMASTER, it first redirects the packet to  the
 117  * record's LMASTER. The LMASTER then redirects the call packet to
 118  * the current DMASTER. Note that this works because of this: When
 119  * a record is migrated off a node, then the new DMASTER is stored
 120  * in the record's copy on the former DMASTER.
 121  */
 122 static void ctdb_call_send_redirect(struct ctdb_context *ctdb,
 123                                     struct ctdb_db_context *ctdb_db,
 124                                     TDB_DATA key,
 125                                     struct ctdb_req_call *c,
 126                                     struct ctdb_ltdb_header *header)
 127 {
 128         uint32_t lmaster = ctdb_lmaster(ctdb, &key);
 129
 130         c->hdr.destnode = lmaster;
 131         if (ctdb->pnn == lmaster) {
 132                 c->hdr.destnode = header->dmaster;
 133         }
 134         c->hopcount++;
 135
 136         if (c->hopcount%100 > 95) {
 137                 DEBUG(DEBUG_WARNING,("High hopcount %d dbid:%s "
 138                         "key:0x%08x reqid=%08x pnn:%d src:%d lmaster:%d "
 139                         "header->dmaster:%d dst:%d\n",
 140                         c->hopcount, ctdb_db->db_name, ctdb_hash(&key),
 141                         c->hdr.reqid, ctdb->pnn, c->hdr.srcnode, lmaster,
 142                         header->dmaster, c->hdr.destnode));
 143         }
 144
 145         ctdb_queue_packet(ctdb, &c->hdr);
 146 }
 147
 148
 149 /*
 150   send a dmaster reply
 151
 152   caller must have the chainlock before calling this routine. Caller must be
 153   the lmaster
 154 */
 155 static void ctdb_send_dmaster_reply(struct ctdb_db_context *ctdb_db,
 156                                     struct ctdb_ltdb_header *header,
 157                                     TDB_DATA key, TDB_DATA data,
 158                                     uint32_t new_dmaster,
 159                                     uint32_t reqid)
 160 {
 161         struct ctdb_context *ctdb = ctdb_db->ctdb;
 162         struct ctdb_reply_dmaster *r;
 163         int ret, len;
 164         TALLOC_CTX *tmp_ctx;
 165
 166         if (ctdb->pnn != ctdb_lmaster(ctdb, &key)) {
 167                 DEBUG(DEBUG_ALERT,(__location__ " Caller is not lmaster!\n"));
 168                 return;
 169         }
 170
 171         header->dmaster = new_dmaster;
 172         ret = ctdb_ltdb_store(ctdb_db, key, header, data);
 173         if (ret != 0) {
 174                 ctdb_fatal(ctdb, "ctdb_send_dmaster_reply unable to update dmaster");
 175                 return;
 176         }
 177
 178         if (ctdb->methods == NULL) {
 179                 ctdb_fatal(ctdb, "ctdb_send_dmaster_reply cant update dmaster since transport is down");
 180                 return;
 181         }
 182
 183         /* put the packet on a temporary context, allowing us to safely free
 184            it below even if ctdb_reply_dmaster() has freed it already */
 185         tmp_ctx = talloc_new(ctdb);
 186
 187         /* send the CTDB_REPLY_DMASTER */
 188         len = offsetof(struct ctdb_reply_dmaster, data) + key.dsize + data.dsize + sizeof(uint32_t);
 189         r = ctdb_transport_allocate(ctdb, tmp_ctx, CTDB_REPLY_DMASTER, len,
 190                                     struct ctdb_reply_dmaster);
 191         CTDB_NO_MEMORY_FATAL(ctdb, r);
 192
 193         r->hdr.destnode  = new_dmaster;
 194         r->hdr.reqid     = reqid;
 195         r->hdr.generation = ctdb_db->generation;
 196         r->rsn           = header->rsn;
 197         r->keylen        = key.dsize;
 198         r->datalen       = data.dsize;
 199         r->db_id         = ctdb_db->db_id;
 200         memcpy(&r->data[0], key.dptr, key.dsize);
 201         memcpy(&r->data[key.dsize], data.dptr, data.dsize);
 202         memcpy(&r->data[key.dsize+data.dsize], &header->flags, sizeof(uint32_t));
 203
 204         ctdb_queue_packet(ctdb, &r->hdr);
 205
 206         talloc_free(tmp_ctx);
 207 }
 208
 209 /*
 210   send a dmaster request (give another node the dmaster for a record)
 211
 212   This is always sent to the lmaster, which ensures that the lmaster
 213   always knows who the dmaster is. The lmaster will then send a
 214   CTDB_REPLY_DMASTER to the new dmaster
 215 */
 216 static void ctdb_call_send_dmaster(struct ctdb_db_context *ctdb_db,
 217                                    struct ctdb_req_call *c,
 218                                    struct ctdb_ltdb_header *header,
 219                                    TDB_DATA *key, TDB_DATA *data)
 220 {
 221         struct ctdb_req_dmaster *r;
 222         struct ctdb_context *ctdb = ctdb_db->ctdb;
 223         int len;
 224         uint32_t lmaster = ctdb_lmaster(ctdb, key);
 225
 226         if (ctdb->methods == NULL) {
 227                 ctdb_fatal(ctdb, "Failed ctdb_call_send_dmaster since transport is down");
 228                 return;
 229         }
 230
 231         if (data->dsize != 0) {
 232                 header->flags |= CTDB_REC_FLAG_MIGRATED_WITH_DATA;
 233         }
 234
 235         if (lmaster == ctdb->pnn) {
 236                 ctdb_send_dmaster_reply(ctdb_db, header, *key, *data,
 237                                         c->hdr.srcnode, c->hdr.reqid);
 238                 return;
 239         }
 240
 241         len = offsetof(struct ctdb_req_dmaster, data) + key->dsize + data->dsize
 242                         + sizeof(uint32_t);
 243         r = ctdb_transport_allocate(ctdb, ctdb, CTDB_REQ_DMASTER, len,
 244                                     struct ctdb_req_dmaster);
 245         CTDB_NO_MEMORY_FATAL(ctdb, r);
 246         r->hdr.destnode  = lmaster;
 247         r->hdr.reqid     = c->hdr.reqid;
 248         r->hdr.generation = ctdb_db->generation;
 249         r->db_id         = c->db_id;
 250         r->rsn           = header->rsn;
 251         r->dmaster       = c->hdr.srcnode;
 252         r->keylen        = key->dsize;
 253         r->datalen       = data->dsize;
 254         memcpy(&r->data[0], key->dptr, key->dsize);
 255         memcpy(&r->data[key->dsize], data->dptr, data->dsize);
 256         memcpy(&r->data[key->dsize + data->dsize], &header->flags, sizeof(uint32_t));
 257
 258         header->dmaster = c->hdr.srcnode;
 259         if (ctdb_ltdb_store(ctdb_db, *key, header, *data) != 0) {
 260                 ctdb_fatal(ctdb, "Failed to store record in ctdb_call_send_dmaster");
 261         }
 262
 263         ctdb_queue_packet(ctdb, &r->hdr);
 264
 265         talloc_free(r);
 266 }
 267
 268 static void ctdb_sticky_pindown_timeout(struct event_context *ev, struct timed_event *te,
 269                                        struct timeval t, void *private_data)
 270 {
 271         struct ctdb_sticky_record *sr = talloc_get_type(private_data,
 272                                                        struct ctdb_sticky_record);
 273
 274         DEBUG(DEBUG_ERR,("Pindown timeout db:%s  unstick record\n", sr->ctdb_db->db_name));
 275         if (sr->pindown != NULL) {
 276                 talloc_free(sr->pindown);
 277                 sr->pindown = NULL;
 278         }
 279 }
 280
 281 static int
 282 ctdb_set_sticky_pindown(struct ctdb_context *ctdb, struct ctdb_db_context *ctdb_db, TDB_DATA key)
 283 {
 284         TALLOC_CTX *tmp_ctx = talloc_new(NULL);
 285         uint32_t *k;
 286         struct ctdb_sticky_record *sr;
 287
 288         k = ctdb_key_to_idkey(tmp_ctx, key);
 289         if (k == NULL) {
 290                 DEBUG(DEBUG_ERR,("Failed to allocate key for sticky record\n"));
 291                 talloc_free(tmp_ctx);
 292                 return -1;
 293         }
 294
 295         sr = trbt_lookuparray32(ctdb_db->sticky_records, k[0], &k[0]);
 296         if (sr == NULL) {
 297                 talloc_free(tmp_ctx);
 298                 return 0;
 299         }
 300
 301         talloc_free(tmp_ctx);
 302
 303         if (sr->pindown == NULL) {
 304                 DEBUG(DEBUG_ERR,("Pinning down record in %s for %d ms\n", ctdb_db->db_name, ctdb->tunable.sticky_pindown));
 305                 sr->pindown = talloc_new(sr);
 306                 if (sr->pindown == NULL) {
 307                         DEBUG(DEBUG_ERR,("Failed to allocate pindown context for sticky record\n"));
 308                         return -1;
 309                 }
 310                 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);
 311         }
 312
 313         return 0;
 314 }
 315
 316 /*
 317   called when a CTDB_REPLY_DMASTER packet comes in, or when the lmaster
 318   gets a CTDB_REQUEST_DMASTER for itself. We become the dmaster.
 319
 320   must be called with the chainlock held. This function releases the chainlock
 321 */
 322 static void ctdb_become_dmaster(struct ctdb_db_context *ctdb_db,
 323                                 struct ctdb_req_header *hdr,
 324                                 TDB_DATA key, TDB_DATA data,
 325                                 uint64_t rsn, uint32_t record_flags)
 326 {
 327         struct ctdb_call_state *state;
 328         struct ctdb_context *ctdb = ctdb_db->ctdb;
 329         struct ctdb_ltdb_header header;
 330         int ret;
 331
 332         DEBUG(DEBUG_DEBUG,("pnn %u dmaster response %08x\n", ctdb->pnn, ctdb_hash(&key)));
 333
 334         ZERO_STRUCT(header);
 335         header.rsn = rsn;
 336         header.dmaster = ctdb->pnn;
 337         header.flags = record_flags;
 338
 339         state = reqid_find(ctdb->idr, hdr->reqid, struct ctdb_call_state);
 340
 341         if (state) {
 342                 if (state->call->flags & CTDB_CALL_FLAG_VACUUM_MIGRATION) {
 343                         /*
 344                          * We temporarily add the VACUUM_MIGRATED flag to
 345                          * the record flags, so that ctdb_ltdb_store can
 346                          * decide whether the record should be stored or
 347                          * deleted.
 348                          */
 349                         header.flags |= CTDB_REC_FLAG_VACUUM_MIGRATED;
 350                 }
 351         }
 352
 353         if (ctdb_ltdb_store(ctdb_db, key, &header, data) != 0) {
 354                 ctdb_fatal(ctdb, "ctdb_reply_dmaster store failed\n");
 355
 356                 ret = ctdb_ltdb_unlock(ctdb_db, key);
 357                 if (ret != 0) {
 358                         DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 359                 }
 360                 return;
 361         }
 362
 363         /* we just became DMASTER and this database is "sticky",
 364            see if the record is flagged as "hot" and set up a pin-down
 365            context to stop migrations for a little while if so
 366         */
 367         if (ctdb_db->sticky) {
 368                 ctdb_set_sticky_pindown(ctdb, ctdb_db, key);
 369         }
 370
 371         if (state == NULL) {
 372                 DEBUG(DEBUG_ERR,("pnn %u Invalid reqid %u in ctdb_become_dmaster from node %u\n",
 373                          ctdb->pnn, hdr->reqid, hdr->srcnode));
 374
 375                 ret = ctdb_ltdb_unlock(ctdb_db, key);
 376                 if (ret != 0) {
 377                         DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 378                 }
 379                 return;
 380         }
 381
 382         if (key.dsize != state->call->key.dsize || memcmp(key.dptr, state->call->key.dptr, key.dsize)) {
 383                 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));
 384
 385                 ret = ctdb_ltdb_unlock(ctdb_db, key);
 386                 if (ret != 0) {
 387                         DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 388                 }
 389                 return;
 390         }
 391
 392         if (hdr->reqid != state->reqid) {
 393                 /* we found a record  but it was the wrong one */
 394                 DEBUG(DEBUG_ERR, ("Dropped orphan in ctdb_become_dmaster with reqid:%u\n from node %u", hdr->reqid, hdr->srcnode));
 395
 396                 ret = ctdb_ltdb_unlock(ctdb_db, key);
 397                 if (ret != 0) {
 398                         DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 399                 }
 400                 return;
 401         }
 402
 403         ctdb_call_local(ctdb_db, state->call, &header, state, &data, true);
 404
 405         ret = ctdb_ltdb_unlock(ctdb_db, state->call->key);
 406         if (ret != 0) {
 407                 DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 408         }
 409
 410         state->state = CTDB_CALL_DONE;
 411         if (state->async.fn) {
 412                 state->async.fn(state);
 413         }
 414 }
 415
 416 struct dmaster_defer_call {
 417         struct dmaster_defer_call *next, *prev;
 418         struct ctdb_context *ctdb;
 419         struct ctdb_req_header *hdr;
 420 };
 421
 422 struct dmaster_defer_queue {
 423         struct ctdb_db_context *ctdb_db;
 424         uint32_t generation;
 425         struct dmaster_defer_call *deferred_calls;
 426 };
 427
 428 static void dmaster_defer_reprocess(struct tevent_context *ev,
 429                                     struct tevent_timer *te,
 430                                     struct timeval t,
 431                                     void *private_data)
 432 {
 433         struct dmaster_defer_call *call = talloc_get_type(
 434                 private_data, struct dmaster_defer_call);
 435
 436         ctdb_input_pkt(call->ctdb, call->hdr);
 437         talloc_free(call);
 438 }
 439
 440 static int dmaster_defer_queue_destructor(struct dmaster_defer_queue *ddq)
 441 {
 442         /* Ignore requests, if database recovery happens in-between. */
 443         if (ddq->generation != ddq->ctdb_db->generation) {
 444                 return 0;
 445         }
 446
 447         while (ddq->deferred_calls != NULL) {
 448                 struct dmaster_defer_call *call = ddq->deferred_calls;
 449
 450                 DLIST_REMOVE(ddq->deferred_calls, call);
 451
 452                 talloc_steal(call->ctdb, call);
 453                 tevent_add_timer(call->ctdb->ev, call, timeval_zero(),
 454                                  dmaster_defer_reprocess, call);
 455         }
 456         return 0;
 457 }
 458
 459 static void *insert_ddq_callback(void *parm, void *data)
 460 {
 461         if (data) {
 462                 talloc_free(data);
 463         }
 464         return parm;
 465 }
 466
 467 /**
 468  * This function is used to reigster a key in database that needs to be updated.
 469  * Any requests for that key should get deferred till this is completed.
 470  */
 471 static int dmaster_defer_setup(struct ctdb_db_context *ctdb_db,
 472                                struct ctdb_req_header *hdr,
 473                                TDB_DATA key)
 474 {
 475         uint32_t *k;
 476         struct dmaster_defer_queue *ddq;
 477
 478         k = ctdb_key_to_idkey(hdr, key);
 479         if (k == NULL) {
 480                 DEBUG(DEBUG_ERR, ("Failed to allocate key for dmaster defer setup\n"));
 481                 return -1;
 482         }
 483
 484         /* Already exists */
 485         ddq = trbt_lookuparray32(ctdb_db->defer_dmaster, k[0], k);
 486         if (ddq != NULL) {
 487                 if (ddq->generation == ctdb_db->generation) {
 488                         talloc_free(k);
 489                         return 0;
 490                 }
 491
 492                 /* Recovery ocurred - get rid of old queue. All the deferred
 493                  * requests will be resent anyway from ctdb_call_resend_db.
 494                  */
 495                 talloc_free(ddq);
 496         }
 497
 498         ddq = talloc(hdr, struct dmaster_defer_queue);
 499         if (ddq == NULL) {
 500                 DEBUG(DEBUG_ERR, ("Failed to allocate dmaster defer queue\n"));
 501                 talloc_free(k);
 502                 return -1;
 503         }
 504         ddq->ctdb_db = ctdb_db;
 505         ddq->generation = hdr->generation;
 506         ddq->deferred_calls = NULL;
 507
 508         trbt_insertarray32_callback(ctdb_db->defer_dmaster, k[0], k,
 509                                     insert_ddq_callback, ddq);
 510         talloc_set_destructor(ddq, dmaster_defer_queue_destructor);
 511
 512         talloc_free(k);
 513         return 0;
 514 }
 515
 516 static int dmaster_defer_add(struct ctdb_db_context *ctdb_db,
 517                              struct ctdb_req_header *hdr,
 518                              TDB_DATA key)
 519 {
 520         struct dmaster_defer_queue *ddq;
 521         struct dmaster_defer_call *call;
 522         uint32_t *k;
 523
 524         k = ctdb_key_to_idkey(hdr, key);
 525         if (k == NULL) {
 526                 DEBUG(DEBUG_ERR, ("Failed to allocate key for dmaster defer add\n"));
 527                 return -1;
 528         }
 529
 530         ddq = trbt_lookuparray32(ctdb_db->defer_dmaster, k[0], k);
 531         if (ddq == NULL) {
 532                 talloc_free(k);
 533                 return -1;
 534         }
 535
 536         talloc_free(k);
 537
 538         if (ddq->generation != hdr->generation) {
 539                 talloc_set_destructor(ddq, NULL);
 540                 talloc_free(ddq);
 541                 return -1;
 542         }
 543
 544         call = talloc(ddq, struct dmaster_defer_call);
 545         if (call == NULL) {
 546                 DEBUG(DEBUG_ERR, ("Failed to allocate dmaster defer call\n"));
 547                 return -1;
 548         }
 549
 550         call->ctdb = ctdb_db->ctdb;
 551         call->hdr = talloc_steal(call, hdr);
 552
 553         DLIST_ADD_END(ddq->deferred_calls, call, NULL);
 554
 555         return 0;
 556 }
 557
 558 /*
 559   called when a CTDB_REQ_DMASTER packet comes in
 560
 561   this comes into the lmaster for a record when the current dmaster
 562   wants to give up the dmaster role and give it to someone else
 563 */
 564 void ctdb_request_dmaster(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
 565 {
 566         struct ctdb_req_dmaster *c = (struct ctdb_req_dmaster *)hdr;
 567         TDB_DATA key, data, data2;
 568         struct ctdb_ltdb_header header;
 569         struct ctdb_db_context *ctdb_db;
 570         uint32_t record_flags = 0;
 571         size_t len;
 572         int ret;
 573
 574         ctdb_db = find_ctdb_db(ctdb, c->db_id);
 575         if (!ctdb_db) {
 576                 ctdb_send_error(ctdb, hdr, -1,
 577                                 "Unknown database in request. db_id==0x%08x",
 578                                 c->db_id);
 579                 return;
 580         }
 581
 582         if (hdr->generation != ctdb_db->generation) {
 583                 DEBUG(DEBUG_DEBUG,
 584                       ("ctdb operation %u request %u from node %u to %u had an"
 585                        " invalid generation:%u while our generation is:%u\n",
 586                        hdr->operation, hdr->reqid, hdr->srcnode, hdr->destnode,
 587                        hdr->generation, ctdb_db->generation));
 588                 return;
 589         }
 590
 591         key.dptr = c->data;
 592         key.dsize = c->keylen;
 593         data.dptr = c->data + c->keylen;
 594         data.dsize = c->datalen;
 595         len = offsetof(struct ctdb_req_dmaster, data) + key.dsize + data.dsize
 596                         + sizeof(uint32_t);
 597         if (len <= c->hdr.length) {
 598                 memcpy(&record_flags, &c->data[c->keylen + c->datalen],
 599                        sizeof(record_flags));
 600         }
 601
 602         dmaster_defer_setup(ctdb_db, hdr, key);
 603
 604         /* fetch the current record */
 605         ret = ctdb_ltdb_lock_fetch_requeue(ctdb_db, key, &header, hdr, &data2,
 606                                            ctdb_call_input_pkt, ctdb, false);
 607         if (ret == -1) {
 608                 ctdb_fatal(ctdb, "ctdb_req_dmaster failed to fetch record");
 609                 return;
 610         }
 611         if (ret == -2) {
 612                 DEBUG(DEBUG_INFO,(__location__ " deferring ctdb_request_dmaster\n"));
 613                 return;
 614         }
 615
 616         if (ctdb_lmaster(ctdb, &key) != ctdb->pnn) {
 617                 DEBUG(DEBUG_ALERT,("pnn %u dmaster request to non-lmaster lmaster=%u gen=%u curgen=%u\n",
 618                          ctdb->pnn, ctdb_lmaster(ctdb, &key),
 619                          hdr->generation, ctdb->vnn_map->generation));
 620                 ctdb_fatal(ctdb, "ctdb_req_dmaster to non-lmaster");
 621         }
 622
 623         DEBUG(DEBUG_DEBUG,("pnn %u dmaster request on %08x for %u from %u\n",
 624                  ctdb->pnn, ctdb_hash(&key), c->dmaster, c->hdr.srcnode));
 625
 626         /* its a protocol error if the sending node is not the current dmaster */
 627         if (header.dmaster != hdr->srcnode) {
 628                 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",
 629                          ctdb->pnn, c->dmaster, hdr->srcnode, header.dmaster, ctdb_hash(&key),
 630                          ctdb_db->db_id, hdr->generation, ctdb_db->generation,
 631                          (unsigned long long)c->rsn, (unsigned long long)header.rsn, c->hdr.reqid,
 632                          (key.dsize >= 4)?(*(uint32_t *)key.dptr):0));
 633                 if (header.rsn != 0 || header.dmaster != ctdb->pnn) {
 634                         DEBUG(DEBUG_ERR,("ctdb_req_dmaster from non-master. Force a recovery.\n"));
 635
 636                         ctdb->recovery_mode = CTDB_RECOVERY_ACTIVE;
 637                         ctdb_ltdb_unlock(ctdb_db, key);
 638                         return;
 639                 }
 640         }
 641
 642         if (header.rsn > c->rsn) {
 643                 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",
 644                          ctdb->pnn, c->dmaster, hdr->srcnode, header.dmaster, ctdb_hash(&key),
 645                          ctdb_db->db_id, hdr->generation, ctdb_db->generation,
 646                          (unsigned long long)c->rsn, (unsigned long long)header.rsn, c->hdr.reqid));
 647         }
 648
 649         /* use the rsn from the sending node */
 650         header.rsn = c->rsn;
 651
 652         /* store the record flags from the sending node */
 653         header.flags = record_flags;
 654
 655         /* check if the new dmaster is the lmaster, in which case we
 656            skip the dmaster reply */
 657         if (c->dmaster == ctdb->pnn) {
 658                 ctdb_become_dmaster(ctdb_db, hdr, key, data, c->rsn, record_flags);
 659         } else {
 660                 ctdb_send_dmaster_reply(ctdb_db, &header, key, data, c->dmaster, hdr->reqid);
 661
 662                 ret = ctdb_ltdb_unlock(ctdb_db, key);
 663                 if (ret != 0) {
 664                         DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 665                 }
 666         }
 667 }
 668
 669 static void ctdb_sticky_record_timeout(struct event_context *ev, struct timed_event *te,
 670                                        struct timeval t, void *private_data)
 671 {
 672         struct ctdb_sticky_record *sr = talloc_get_type(private_data,
 673                                                        struct ctdb_sticky_record);
 674         talloc_free(sr);
 675 }
 676
 677 static void *ctdb_make_sticky_record_callback(void *parm, void *data)
 678 {
 679         if (data) {
 680                 DEBUG(DEBUG_ERR,("Already have sticky record registered. Free old %p and create new %p\n", data, parm));
 681                 talloc_free(data);
 682         }
 683         return parm;
 684 }
 685
 686 static int
 687 ctdb_make_record_sticky(struct ctdb_context *ctdb, struct ctdb_db_context *ctdb_db, TDB_DATA key)
 688 {
 689         TALLOC_CTX *tmp_ctx = talloc_new(NULL);
 690         uint32_t *k;
 691         struct ctdb_sticky_record *sr;
 692
 693         k = ctdb_key_to_idkey(tmp_ctx, key);
 694         if (k == NULL) {
 695                 DEBUG(DEBUG_ERR,("Failed to allocate key for sticky record\n"));
 696                 talloc_free(tmp_ctx);
 697                 return -1;
 698         }
 699
 700         sr = trbt_lookuparray32(ctdb_db->sticky_records, k[0], &k[0]);
 701         if (sr != NULL) {
 702                 talloc_free(tmp_ctx);
 703                 return 0;
 704         }
 705
 706         sr = talloc(ctdb_db->sticky_records, struct ctdb_sticky_record);
 707         if (sr == NULL) {
 708                 talloc_free(tmp_ctx);
 709                 DEBUG(DEBUG_ERR,("Failed to allocate sticky record structure\n"));
 710                 return -1;
 711         }
 712
 713         sr->ctdb    = ctdb;
 714         sr->ctdb_db = ctdb_db;
 715         sr->pindown = NULL;
 716
 717         DEBUG(DEBUG_ERR,("Make record sticky for %d seconds in db %s key:0x%08x.\n",
 718                          ctdb->tunable.sticky_duration,
 719                          ctdb_db->db_name, ctdb_hash(&key)));
 720
 721         trbt_insertarray32_callback(ctdb_db->sticky_records, k[0], &k[0], ctdb_make_sticky_record_callback, sr);
 722
 723         event_add_timed(ctdb->ev, sr, timeval_current_ofs(ctdb->tunable.sticky_duration, 0), ctdb_sticky_record_timeout, sr);
 724
 725         talloc_free(tmp_ctx);
 726         return 0;
 727 }
 728
 729 struct pinned_down_requeue_handle {
 730         struct ctdb_context *ctdb;
 731         struct ctdb_req_header *hdr;
 732 };
 733
 734 struct pinned_down_deferred_call {
 735         struct ctdb_context *ctdb;
 736         struct ctdb_req_header *hdr;
 737 };
 738
 739 static void pinned_down_requeue(struct event_context *ev, struct timed_event *te,
 740                        struct timeval t, void *private_data)
 741 {
 742         struct pinned_down_requeue_handle *handle = talloc_get_type(private_data, struct pinned_down_requeue_handle);
 743         struct ctdb_context *ctdb = handle->ctdb;
 744
 745         talloc_steal(ctdb, handle->hdr);
 746         ctdb_call_input_pkt(ctdb, handle->hdr);
 747
 748         talloc_free(handle);
 749 }
 750
 751 static int pinned_down_destructor(struct pinned_down_deferred_call *pinned_down)
 752 {
 753         struct ctdb_context *ctdb = pinned_down->ctdb;
 754         struct pinned_down_requeue_handle *handle = talloc(ctdb, struct pinned_down_requeue_handle);
 755
 756         handle->ctdb = pinned_down->ctdb;
 757         handle->hdr  = pinned_down->hdr;
 758         talloc_steal(handle, handle->hdr);
 759
 760         event_add_timed(ctdb->ev, handle, timeval_zero(), pinned_down_requeue, handle);
 761
 762         return 0;
 763 }
 764
 765 static int
 766 ctdb_defer_pinned_down_request(struct ctdb_context *ctdb, struct ctdb_db_context *ctdb_db, TDB_DATA key, struct ctdb_req_header *hdr)
 767 {
 768         TALLOC_CTX *tmp_ctx = talloc_new(NULL);
 769         uint32_t *k;
 770         struct ctdb_sticky_record *sr;
 771         struct pinned_down_deferred_call *pinned_down;
 772
 773         k = ctdb_key_to_idkey(tmp_ctx, key);
 774         if (k == NULL) {
 775                 DEBUG(DEBUG_ERR,("Failed to allocate key for sticky record\n"));
 776                 talloc_free(tmp_ctx);
 777                 return -1;
 778         }
 779
 780         sr = trbt_lookuparray32(ctdb_db->sticky_records, k[0], &k[0]);
 781         if (sr == NULL) {
 782                 talloc_free(tmp_ctx);
 783                 return -1;
 784         }
 785
 786         talloc_free(tmp_ctx);
 787
 788         if (sr->pindown == NULL) {
 789                 return -1;
 790         }
 791
 792         pinned_down = talloc(sr->pindown, struct pinned_down_deferred_call);
 793         if (pinned_down == NULL) {
 794                 DEBUG(DEBUG_ERR,("Failed to allocate structure for deferred pinned down request\n"));
 795                 return -1;
 796         }
 797
 798         pinned_down->ctdb = ctdb;
 799         pinned_down->hdr  = hdr;
 800
 801         talloc_set_destructor(pinned_down, pinned_down_destructor);
 802         talloc_steal(pinned_down, hdr);
 803
 804         return 0;
 805 }
 806
 807 static void
 808 ctdb_update_db_stat_hot_keys(struct ctdb_db_context *ctdb_db, TDB_DATA key, int hopcount)
 809 {
 810         int i, id;
 811
 812         /* smallest value is always at index 0 */
 813         if (hopcount <= ctdb_db->statistics.hot_keys[0].count) {
 814                 return;
 815         }
 816
 817         /* see if we already know this key */
 818         for (i = 0; i < MAX_HOT_KEYS; i++) {
 819                 if (key.dsize != ctdb_db->statistics.hot_keys[i].key.dsize) {
 820                         continue;
 821                 }
 822                 if (memcmp(key.dptr, ctdb_db->statistics.hot_keys[i].key.dptr, key.dsize)) {
 823                         continue;
 824                 }
 825                 /* found an entry for this key */
 826                 if (hopcount <= ctdb_db->statistics.hot_keys[i].count) {
 827                         return;
 828                 }
 829                 ctdb_db->statistics.hot_keys[i].count = hopcount;
 830                 goto sort_keys;
 831         }
 832
 833         if (ctdb_db->statistics.num_hot_keys < MAX_HOT_KEYS) {
 834                 id = ctdb_db->statistics.num_hot_keys;
 835                 ctdb_db->statistics.num_hot_keys++;
 836         } else {
 837                 id = 0;
 838         }
 839
 840         if (ctdb_db->statistics.hot_keys[id].key.dptr != NULL) {
 841                 talloc_free(ctdb_db->statistics.hot_keys[id].key.dptr);
 842         }
 843         ctdb_db->statistics.hot_keys[id].key.dsize = key.dsize;
 844         ctdb_db->statistics.hot_keys[id].key.dptr  = talloc_memdup(ctdb_db, key.dptr, key.dsize);
 845         ctdb_db->statistics.hot_keys[id].count = hopcount;
 846         DEBUG(DEBUG_NOTICE,("Updated hot key database=%s key=0x%08x id=%d hop_count=%d\n",
 847                             ctdb_db->db_name, ctdb_hash(&key), id, hopcount));
 848
 849 sort_keys:
 850         for (i = 1; i < MAX_HOT_KEYS; i++) {
 851                 if (ctdb_db->statistics.hot_keys[i].count == 0) {
 852                         continue;
 853                 }
 854                 if (ctdb_db->statistics.hot_keys[i].count < ctdb_db->statistics.hot_keys[0].count) {
 855                         hopcount = ctdb_db->statistics.hot_keys[i].count;
 856                         ctdb_db->statistics.hot_keys[i].count = ctdb_db->statistics.hot_keys[0].count;
 857                         ctdb_db->statistics.hot_keys[0].count = hopcount;
 858
 859                         key = ctdb_db->statistics.hot_keys[i].key;
 860                         ctdb_db->statistics.hot_keys[i].key = ctdb_db->statistics.hot_keys[0].key;
 861                         ctdb_db->statistics.hot_keys[0].key = key;
 862                 }
 863         }
 864 }
 865
 866 /*
 867   called when a CTDB_REQ_CALL packet comes in
 868 */
 869 void ctdb_request_call(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
 870 {
 871         struct ctdb_req_call *c = (struct ctdb_req_call *)hdr;
 872         TDB_DATA data;
 873         struct ctdb_reply_call *r;
 874         int ret, len;
 875         struct ctdb_ltdb_header header;
 876         struct ctdb_call *call;
 877         struct ctdb_db_context *ctdb_db;
 878         int tmp_count, bucket;
 879
 880         if (ctdb->methods == NULL) {
 881                 DEBUG(DEBUG_INFO,(__location__ " Failed ctdb_request_call. Transport is DOWN\n"));
 882                 return;
 883         }
 884
 885         ctdb_db = find_ctdb_db(ctdb, c->db_id);
 886         if (!ctdb_db) {
 887                 ctdb_send_error(ctdb, hdr, -1,
 888                                 "Unknown database in request. db_id==0x%08x",
 889                                 c->db_id);
 890                 return;
 891         }
 892
 893         if (hdr->generation != ctdb_db->generation) {
 894                 DEBUG(DEBUG_DEBUG,
 895                       ("ctdb operation %u request %u from node %u to %u had an"
 896                        " invalid generation:%u while our generation is:%u\n",
 897                        hdr->operation, hdr->reqid, hdr->srcnode, hdr->destnode,
 898                        hdr->generation, ctdb_db->generation));
 899                 return;
 900         }
 901
 902         call = talloc(hdr, struct ctdb_call);
 903         CTDB_NO_MEMORY_FATAL(ctdb, call);
 904
 905         call->call_id  = c->callid;
 906         call->key.dptr = c->data;
 907         call->key.dsize = c->keylen;
 908         call->call_data.dptr = c->data + c->keylen;
 909         call->call_data.dsize = c->calldatalen;
 910         call->reply_data.dptr  = NULL;
 911         call->reply_data.dsize = 0;
 912
 913
 914         /* If this record is pinned down we should defer the
 915            request until the pindown times out
 916         */
 917         if (ctdb_db->sticky) {
 918                 if (ctdb_defer_pinned_down_request(ctdb, ctdb_db, call->key, hdr) == 0) {
 919                         DEBUG(DEBUG_WARNING,
 920                               ("Defer request for pinned down record in %s\n", ctdb_db->db_name));
 921                         talloc_free(call);
 922                         return;
 923                 }
 924         }
 925
 926         if (dmaster_defer_add(ctdb_db, hdr, call->key) == 0) {
 927                 talloc_free(call);
 928                 return;
 929         }
 930
 931         /* determine if we are the dmaster for this key. This also
 932            fetches the record data (if any), thus avoiding a 2nd fetch of the data
 933            if the call will be answered locally */
 934
 935         ret = ctdb_ltdb_lock_fetch_requeue(ctdb_db, call->key, &header, hdr, &data,
 936                                            ctdb_call_input_pkt, ctdb, false);
 937         if (ret == -1) {
 938                 ctdb_send_error(ctdb, hdr, ret, "ltdb fetch failed in ctdb_request_call");
 939                 talloc_free(call);
 940                 return;
 941         }
 942         if (ret == -2) {
 943                 DEBUG(DEBUG_INFO,(__location__ " deferred ctdb_request_call\n"));
 944                 talloc_free(call);
 945                 return;
 946         }
 947
 948         /* Dont do READONLY if we dont have a tracking database */
 949         if ((c->flags & CTDB_WANT_READONLY) && !ctdb_db->readonly) {
 950                 c->flags &= ~CTDB_WANT_READONLY;
 951         }
 952
 953         if (header.flags & CTDB_REC_RO_REVOKE_COMPLETE) {
 954                 header.flags &= ~CTDB_REC_RO_FLAGS;
 955                 CTDB_INCREMENT_STAT(ctdb, total_ro_revokes);
 956                 CTDB_INCREMENT_DB_STAT(ctdb_db, db_ro_revokes);
 957                 if (ctdb_ltdb_store(ctdb_db, call->key, &header, data) != 0) {
 958                         ctdb_fatal(ctdb, "Failed to write header with cleared REVOKE flag");
 959                 }
 960                 /* and clear out the tracking data */
 961                 if (tdb_delete(ctdb_db->rottdb, call->key) != 0) {
 962                         DEBUG(DEBUG_ERR,(__location__ " Failed to clear out trackingdb record\n"));
 963                 }
 964         }
 965
 966         /* if we are revoking, we must defer all other calls until the revoke
 967          * had completed.
 968          */
 969         if (header.flags & CTDB_REC_RO_REVOKING_READONLY) {
 970                 talloc_free(data.dptr);
 971                 ret = ctdb_ltdb_unlock(ctdb_db, call->key);
 972
 973                 if (ctdb_add_revoke_deferred_call(ctdb, ctdb_db, call->key, hdr, ctdb_call_input_pkt, ctdb) != 0) {
 974                         ctdb_fatal(ctdb, "Failed to add deferred call for revoke child");
 975                 }
 976                 talloc_free(call);
 977                 return;
 978         }
 979
 980         /*
 981          * If we are not the dmaster and are not hosting any delegations,
 982          * then we redirect the request to the node than can answer it
 983          * (the lmaster or the dmaster).
 984          */
 985         if ((header.dmaster != ctdb->pnn)
 986             && (!(header.flags & CTDB_REC_RO_HAVE_DELEGATIONS)) ) {
 987                 talloc_free(data.dptr);
 988                 ctdb_call_send_redirect(ctdb, ctdb_db, call->key, c, &header);
 989
 990                 ret = ctdb_ltdb_unlock(ctdb_db, call->key);
 991                 if (ret != 0) {
 992                         DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
 993                 }
 994                 talloc_free(call);
 995                 return;
 996         }
 997
 998         if ( (!(c->flags & CTDB_WANT_READONLY))
 999         && (header.flags & (CTDB_REC_RO_HAVE_DELEGATIONS|CTDB_REC_RO_HAVE_READONLY)) ) {
1000                 header.flags   |= CTDB_REC_RO_REVOKING_READONLY;
1001                 if (ctdb_ltdb_store(ctdb_db, call->key, &header, data) != 0) {
1002                         ctdb_fatal(ctdb, "Failed to store record with HAVE_DELEGATIONS set");
1003                 }
1004                 ret = ctdb_ltdb_unlock(ctdb_db, call->key);
1005
1006                 if (ctdb_start_revoke_ro_record(ctdb, ctdb_db, call->key, &header, data) != 0) {
1007                         ctdb_fatal(ctdb, "Failed to start record revoke");
1008                 }
1009                 talloc_free(data.dptr);
1010
1011                 if (ctdb_add_revoke_deferred_call(ctdb, ctdb_db, call->key, hdr, ctdb_call_input_pkt, ctdb) != 0) {
1012                         ctdb_fatal(ctdb, "Failed to add deferred call for revoke child");
1013                 }
1014                 talloc_free(call);
1015
1016                 return;
1017         }
1018
1019         /* If this is the first request for delegation. bump rsn and set
1020          * the delegations flag
1021          */
1022         if ((c->flags & CTDB_WANT_READONLY)
1023         &&  (c->callid == CTDB_FETCH_WITH_HEADER_FUNC)
1024         &&  (!(header.flags & CTDB_REC_RO_HAVE_DELEGATIONS))) {
1025                 header.rsn     += 3;
1026                 header.flags   |= CTDB_REC_RO_HAVE_DELEGATIONS;
1027                 if (ctdb_ltdb_store(ctdb_db, call->key, &header, data) != 0) {
1028                         ctdb_fatal(ctdb, "Failed to store record with HAVE_DELEGATIONS set");
1029                 }
1030         }
1031         if ((c->flags & CTDB_WANT_READONLY)
1032         &&  (call->call_id == CTDB_FETCH_WITH_HEADER_FUNC)) {
1033                 TDB_DATA tdata;
1034
1035                 tdata = tdb_fetch(ctdb_db->rottdb, call->key);
1036                 if (ctdb_trackingdb_add_pnn(ctdb, &tdata, c->hdr.srcnode) != 0) {
1037                         ctdb_fatal(ctdb, "Failed to add node to trackingdb");
1038                 }
1039                 if (tdb_store(ctdb_db->rottdb, call->key, tdata, TDB_REPLACE) != 0) {
1040                         ctdb_fatal(ctdb, "Failed to store trackingdb data");
1041                 }
1042                 free(tdata.dptr);
1043
1044                 ret = ctdb_ltdb_unlock(ctdb_db, call->key);
1045                 if (ret != 0) {
1046                         DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
1047                 }
1048
1049                 len = offsetof(struct ctdb_reply_call, data) + data.dsize + sizeof(struct ctdb_ltdb_header);
1050                 r = ctdb_transport_allocate(ctdb, ctdb, CTDB_REPLY_CALL, len,
1051                                             struct ctdb_reply_call);
1052                 CTDB_NO_MEMORY_FATAL(ctdb, r);
1053                 r->hdr.destnode  = c->hdr.srcnode;
1054                 r->hdr.reqid     = c->hdr.reqid;
1055                 r->hdr.generation = ctdb_db->generation;
1056                 r->status        = 0;
1057                 r->datalen       = data.dsize + sizeof(struct ctdb_ltdb_header);
1058                 header.rsn      -= 2;
1059                 header.flags   |= CTDB_REC_RO_HAVE_READONLY;
1060                 header.flags   &= ~CTDB_REC_RO_HAVE_DELEGATIONS;
1061                 memcpy(&r->data[0], &header, sizeof(struct ctdb_ltdb_header));
1062
1063                 if (data.dsize) {
1064                         memcpy(&r->data[sizeof(struct ctdb_ltdb_header)], data.dptr, data.dsize);
1065                 }
1066
1067                 ctdb_queue_packet(ctdb, &r->hdr);
1068                 CTDB_INCREMENT_STAT(ctdb, total_ro_delegations);
1069                 CTDB_INCREMENT_DB_STAT(ctdb_db, db_ro_delegations);
1070
1071                 talloc_free(r);
1072                 talloc_free(call);
1073                 return;
1074         }
1075
1076         CTDB_UPDATE_STAT(ctdb, max_hop_count, c->hopcount);
1077         tmp_count = c->hopcount;
1078         bucket = 0;
1079         while (tmp_count) {
1080                 tmp_count >>= 2;
1081                 bucket++;
1082         }
1083         if (bucket >= MAX_COUNT_BUCKETS) {
1084                 bucket = MAX_COUNT_BUCKETS - 1;
1085         }
1086         CTDB_INCREMENT_STAT(ctdb, hop_count_bucket[bucket]);
1087         CTDB_INCREMENT_DB_STAT(ctdb_db, hop_count_bucket[bucket]);
1088         ctdb_update_db_stat_hot_keys(ctdb_db, call->key, c->hopcount);
1089
1090         /* If this database supports sticky records, then check if the
1091            hopcount is big. If it is it means the record is hot and we
1092            should make it sticky.
1093         */
1094         if (ctdb_db->sticky && c->hopcount >= ctdb->tunable.hopcount_make_sticky) {
1095                 ctdb_make_record_sticky(ctdb, ctdb_db, call->key);
1096         }
1097
1098
1099         /* Try if possible to migrate the record off to the caller node.
1100          * From the clients perspective a fetch of the data is just as
1101          * expensive as a migration.
1102          */
1103         if (c->hdr.srcnode != ctdb->pnn) {
1104                 if (ctdb_db->persistent_state) {
1105                         DEBUG(DEBUG_INFO, (__location__ " refusing migration"
1106                               " of key %s while transaction is active\n",
1107                               (char *)call->key.dptr));
1108                 } else {
1109                         DEBUG(DEBUG_DEBUG,("pnn %u starting migration of %08x to %u\n",
1110                                  ctdb->pnn, ctdb_hash(&(call->key)), c->hdr.srcnode));
1111                         ctdb_call_send_dmaster(ctdb_db, c, &header, &(call->key), &data);
1112                         talloc_free(data.dptr);
1113
1114                         ret = ctdb_ltdb_unlock(ctdb_db, call->key);
1115                         if (ret != 0) {
1116                                 DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
1117                         }
1118                 }
1119                 talloc_free(call);
1120                 return;
1121         }
1122
1123         ret = ctdb_call_local(ctdb_db, call, &header, hdr, &data, true);
1124         if (ret != 0) {
1125                 DEBUG(DEBUG_ERR,(__location__ " ctdb_call_local failed\n"));
1126                 call->status = -1;
1127         }
1128
1129         ret = ctdb_ltdb_unlock(ctdb_db, call->key);
1130         if (ret != 0) {
1131                 DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
1132         }
1133
1134         len = offsetof(struct ctdb_reply_call, data) + call->reply_data.dsize;
1135         r = ctdb_transport_allocate(ctdb, ctdb, CTDB_REPLY_CALL, len,
1136                                     struct ctdb_reply_call);
1137         CTDB_NO_MEMORY_FATAL(ctdb, r);
1138         r->hdr.destnode  = hdr->srcnode;
1139         r->hdr.reqid     = hdr->reqid;
1140         r->hdr.generation = ctdb_db->generation;
1141         r->status        = call->status;
1142         r->datalen       = call->reply_data.dsize;
1143         if (call->reply_data.dsize) {
1144                 memcpy(&r->data[0], call->reply_data.dptr, call->reply_data.dsize);
1145         }
1146
1147         ctdb_queue_packet(ctdb, &r->hdr);
1148
1149         talloc_free(r);
1150         talloc_free(call);
1151 }
1152
1153 /**
1154  * called when a CTDB_REPLY_CALL packet comes in
1155  *
1156  * This packet comes in response to a CTDB_REQ_CALL request packet. It
1157  * contains any reply data from the call
1158  */
1159 void ctdb_reply_call(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
1160 {
1161         struct ctdb_reply_call *c = (struct ctdb_reply_call *)hdr;
1162         struct ctdb_call_state *state;
1163
1164         state = reqid_find(ctdb->idr, hdr->reqid, struct ctdb_call_state);
1165         if (state == NULL) {
1166                 DEBUG(DEBUG_ERR, (__location__ " reqid %u not found\n", hdr->reqid));
1167                 return;
1168         }
1169
1170         if (hdr->reqid != state->reqid) {
1171                 /* we found a record  but it was the wrong one */
1172                 DEBUG(DEBUG_ERR, ("Dropped orphaned call reply with reqid:%u\n",hdr->reqid));
1173                 return;
1174         }
1175
1176         if (hdr->generation != state->generation) {
1177                 DEBUG(DEBUG_DEBUG,
1178                       ("ctdb operation %u request %u from node %u to %u had an"
1179                        " invalid generation:%u while our generation is:%u\n",
1180                        hdr->operation, hdr->reqid, hdr->srcnode, hdr->destnode,
1181                        hdr->generation, state->generation));
1182                 return;
1183         }
1184
1185
1186         /* read only delegation processing */
1187         /* If we got a FETCH_WITH_HEADER we should check if this is a ro
1188          * delegation since we may need to update the record header
1189          */
1190         if (state->c->callid == CTDB_FETCH_WITH_HEADER_FUNC) {
1191                 struct ctdb_db_context *ctdb_db = state->ctdb_db;
1192                 struct ctdb_ltdb_header *header = (struct ctdb_ltdb_header *)&c->data[0];
1193                 struct ctdb_ltdb_header oldheader;
1194                 TDB_DATA key, data, olddata;
1195                 int ret;
1196
1197                 if (!(header->flags & CTDB_REC_RO_HAVE_READONLY)) {
1198                         goto finished_ro;
1199                         return;
1200                 }
1201
1202                 key.dsize = state->c->keylen;
1203                 key.dptr  = state->c->data;
1204                 ret = ctdb_ltdb_lock_requeue(ctdb_db, key, hdr,
1205                                      ctdb_call_input_pkt, ctdb, false);
1206                 if (ret == -2) {
1207                         return;
1208                 }
1209                 if (ret != 0) {
1210                         DEBUG(DEBUG_ERR,(__location__ " Failed to get lock in ctdb_reply_call\n"));
1211                         return;
1212                 }
1213
1214                 ret = ctdb_ltdb_fetch(ctdb_db, key, &oldheader, state, &olddata);
1215                 if (ret != 0) {
1216                         DEBUG(DEBUG_ERR, ("Failed to fetch old record in ctdb_reply_call\n"));
1217                         ctdb_ltdb_unlock(ctdb_db, key);
1218                         goto finished_ro;
1219                 }
1220
1221                 if (header->rsn <= oldheader.rsn) {
1222                         ctdb_ltdb_unlock(ctdb_db, key);
1223                         goto finished_ro;
1224                 }
1225
1226                 if (c->datalen < sizeof(struct ctdb_ltdb_header)) {
1227                         DEBUG(DEBUG_ERR,(__location__ " Got FETCH_WITH_HEADER reply with too little data: %d bytes\n", c->datalen));
1228                         ctdb_ltdb_unlock(ctdb_db, key);
1229                         goto finished_ro;
1230                 }
1231
1232                 data.dsize = c->datalen - sizeof(struct ctdb_ltdb_header);
1233                 data.dptr  = &c->data[sizeof(struct ctdb_ltdb_header)];
1234                 ret = ctdb_ltdb_store(ctdb_db, key, header, data);
1235                 if (ret != 0) {
1236                         DEBUG(DEBUG_ERR, ("Failed to store new record in ctdb_reply_call\n"));
1237                         ctdb_ltdb_unlock(ctdb_db, key);
1238                         goto finished_ro;
1239                 }
1240
1241                 ctdb_ltdb_unlock(ctdb_db, key);
1242         }
1243 finished_ro:
1244
1245         state->call->reply_data.dptr = c->data;
1246         state->call->reply_data.dsize = c->datalen;
1247         state->call->status = c->status;
1248
1249         talloc_steal(state, c);
1250
1251         state->state = CTDB_CALL_DONE;
1252         if (state->async.fn) {
1253                 state->async.fn(state);
1254         }
1255 }
1256
1257
1258 /**
1259  * called when a CTDB_REPLY_DMASTER packet comes in
1260  *
1261  * This packet comes in from the lmaster in response to a CTDB_REQ_CALL
1262  * request packet. It means that the current dmaster wants to give us
1263  * the dmaster role.
1264  */
1265 void ctdb_reply_dmaster(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
1266 {
1267         struct ctdb_reply_dmaster *c = (struct ctdb_reply_dmaster *)hdr;
1268         struct ctdb_db_context *ctdb_db;
1269         TDB_DATA key, data;
1270         uint32_t record_flags = 0;
1271         size_t len;
1272         int ret;
1273
1274         ctdb_db = find_ctdb_db(ctdb, c->db_id);
1275         if (ctdb_db == NULL) {
1276                 DEBUG(DEBUG_ERR,("Unknown db_id 0x%x in ctdb_reply_dmaster\n", c->db_id));
1277                 return;
1278         }
1279
1280         if (hdr->generation != ctdb_db->generation) {
1281                 DEBUG(DEBUG_DEBUG,
1282                       ("ctdb operation %u request %u from node %u to %u had an"
1283                        " invalid generation:%u while our generation is:%u\n",
1284                        hdr->operation, hdr->reqid, hdr->srcnode, hdr->destnode,
1285                        hdr->generation, ctdb_db->generation));
1286                 return;
1287         }
1288
1289         key.dptr = c->data;
1290         key.dsize = c->keylen;
1291         data.dptr = &c->data[key.dsize];
1292         data.dsize = c->datalen;
1293         len = offsetof(struct ctdb_reply_dmaster, data) + key.dsize + data.dsize
1294                 + sizeof(uint32_t);
1295         if (len <= c->hdr.length) {
1296                 memcpy(&record_flags, &c->data[c->keylen + c->datalen],
1297                        sizeof(record_flags));
1298         }
1299
1300         dmaster_defer_setup(ctdb_db, hdr, key);
1301
1302         ret = ctdb_ltdb_lock_requeue(ctdb_db, key, hdr,
1303                                      ctdb_call_input_pkt, ctdb, false);
1304         if (ret == -2) {
1305                 return;
1306         }
1307         if (ret != 0) {
1308                 DEBUG(DEBUG_ERR,(__location__ " Failed to get lock in ctdb_reply_dmaster\n"));
1309                 return;
1310         }
1311
1312         ctdb_become_dmaster(ctdb_db, hdr, key, data, c->rsn, record_flags);
1313 }
1314
1315
1316 /*
1317   called when a CTDB_REPLY_ERROR packet comes in
1318 */
1319 void ctdb_reply_error(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
1320 {
1321         struct ctdb_reply_error *c = (struct ctdb_reply_error *)hdr;
1322         struct ctdb_call_state *state;
1323
1324         state = reqid_find(ctdb->idr, hdr->reqid, struct ctdb_call_state);
1325         if (state == NULL) {
1326                 DEBUG(DEBUG_ERR,("pnn %u Invalid reqid %u in ctdb_reply_error\n",
1327                          ctdb->pnn, hdr->reqid));
1328                 return;
1329         }
1330
1331         if (hdr->reqid != state->reqid) {
1332                 /* we found a record  but it was the wrong one */
1333                 DEBUG(DEBUG_ERR, ("Dropped orphaned error reply with reqid:%u\n",hdr->reqid));
1334                 return;
1335         }
1336
1337         talloc_steal(state, c);
1338
1339         state->state  = CTDB_CALL_ERROR;
1340         state->errmsg = (char *)c->msg;
1341         if (state->async.fn) {
1342                 state->async.fn(state);
1343         }
1344 }
1345
1346
1347 /*
1348   destroy a ctdb_call
1349 */
1350 static int ctdb_call_destructor(struct ctdb_call_state *state)
1351 {
1352         DLIST_REMOVE(state->ctdb_db->pending_calls, state);
1353         reqid_remove(state->ctdb_db->ctdb->idr, state->reqid);
1354         return 0;
1355 }
1356
1357
1358 /*
1359   called when a ctdb_call needs to be resent after a reconfigure event
1360 */
1361 static void ctdb_call_resend(struct ctdb_call_state *state)
1362 {
1363         struct ctdb_context *ctdb = state->ctdb_db->ctdb;
1364
1365         state->generation = state->ctdb_db->generation;
1366
1367         /* use a new reqid, in case the old reply does eventually come in */
1368         reqid_remove(ctdb->idr, state->reqid);
1369         state->reqid = reqid_new(ctdb->idr, state);
1370         state->c->hdr.reqid = state->reqid;
1371
1372         /* update the generation count for this request, so its valid with the new vnn_map */
1373         state->c->hdr.generation = state->generation;
1374
1375         /* send the packet to ourselves, it will be redirected appropriately */
1376         state->c->hdr.destnode = ctdb->pnn;
1377
1378         ctdb_queue_packet(ctdb, &state->c->hdr);
1379         DEBUG(DEBUG_NOTICE,("resent ctdb_call for db %s reqid %u generation %u\n",
1380                             state->ctdb_db->db_name, state->reqid, state->generation));
1381 }
1382
1383 /*
1384   resend all pending calls on recovery
1385  */
1386 void ctdb_call_resend_db(struct ctdb_db_context *ctdb_db)
1387 {
1388         struct ctdb_call_state *state, *next;
1389
1390         for (state = ctdb_db->pending_calls; state; state = next) {
1391                 next = state->next;
1392                 ctdb_call_resend(state);
1393         }
1394 }
1395
1396 void ctdb_call_resend_all(struct ctdb_context *ctdb)
1397 {
1398         struct ctdb_db_context *ctdb_db;
1399
1400         for (ctdb_db = ctdb->db_list; ctdb_db; ctdb_db = ctdb_db->next) {
1401                 ctdb_call_resend_db(ctdb_db);
1402         }
1403 }
1404
1405 /*
1406   this allows the caller to setup a async.fn
1407 */
1408 static void call_local_trigger(struct event_context *ev, struct timed_event *te,
1409                        struct timeval t, void *private_data)
1410 {
1411         struct ctdb_call_state *state = talloc_get_type(private_data, struct ctdb_call_state);
1412         if (state->async.fn) {
1413                 state->async.fn(state);
1414         }
1415 }
1416
1417
1418 /*
1419   construct an event driven local ctdb_call
1420
1421   this is used so that locally processed ctdb_call requests are processed
1422   in an event driven manner
1423 */
1424 struct ctdb_call_state *ctdb_call_local_send(struct ctdb_db_context *ctdb_db,
1425                                              struct ctdb_call *call,
1426                                              struct ctdb_ltdb_header *header,
1427                                              TDB_DATA *data)
1428 {
1429         struct ctdb_call_state *state;
1430         struct ctdb_context *ctdb = ctdb_db->ctdb;
1431         int ret;
1432
1433         state = talloc_zero(ctdb_db, struct ctdb_call_state);
1434         CTDB_NO_MEMORY_NULL(ctdb, state);
1435
1436         talloc_steal(state, data->dptr);
1437
1438         state->state = CTDB_CALL_DONE;
1439         state->call  = talloc(state, struct ctdb_call);
1440         CTDB_NO_MEMORY_NULL(ctdb, state->call);
1441         *(state->call) = *call;
1442         state->ctdb_db = ctdb_db;
1443
1444         ret = ctdb_call_local(ctdb_db, state->call, header, state, data, true);
1445         if (ret != 0) {
1446                 DEBUG(DEBUG_DEBUG,("ctdb_call_local() failed, ignoring return code %d\n", ret));
1447         }
1448
1449         event_add_timed(ctdb->ev, state, timeval_zero(), call_local_trigger, state);
1450
1451         return state;
1452 }
1453
1454
1455 /*
1456   make a remote ctdb call - async send. Called in daemon context.
1457
1458   This constructs a ctdb_call request and queues it for processing.
1459   This call never blocks.
1460 */
1461 struct ctdb_call_state *ctdb_daemon_call_send_remote(struct ctdb_db_context *ctdb_db,
1462                                                      struct ctdb_call *call,
1463                                                      struct ctdb_ltdb_header *header)
1464 {
1465         uint32_t len;
1466         struct ctdb_call_state *state;
1467         struct ctdb_context *ctdb = ctdb_db->ctdb;
1468
1469         if (ctdb->methods == NULL) {
1470                 DEBUG(DEBUG_INFO,(__location__ " Failed send packet. Transport is down\n"));
1471                 return NULL;
1472         }
1473
1474         state = talloc_zero(ctdb_db, struct ctdb_call_state);
1475         CTDB_NO_MEMORY_NULL(ctdb, state);
1476         state->call = talloc(state, struct ctdb_call);
1477         CTDB_NO_MEMORY_NULL(ctdb, state->call);
1478
1479         state->reqid = reqid_new(ctdb->idr, state);
1480         state->ctdb_db = ctdb_db;
1481         talloc_set_destructor(state, ctdb_call_destructor);
1482
1483         len = offsetof(struct ctdb_req_call, data) + call->key.dsize + call->call_data.dsize;
1484         state->c = ctdb_transport_allocate(ctdb, state, CTDB_REQ_CALL, len,
1485                                            struct ctdb_req_call);
1486         CTDB_NO_MEMORY_NULL(ctdb, state->c);
1487         state->c->hdr.destnode  = header->dmaster;
1488
1489         /* this limits us to 16k outstanding messages - not unreasonable */
1490         state->c->hdr.reqid     = state->reqid;
1491         state->c->hdr.generation = ctdb_db->generation;
1492         state->c->flags         = call->flags;
1493         state->c->db_id         = ctdb_db->db_id;
1494         state->c->callid        = call->call_id;
1495         state->c->hopcount      = 0;
1496         state->c->keylen        = call->key.dsize;
1497         state->c->calldatalen   = call->call_data.dsize;
1498         memcpy(&state->c->data[0], call->key.dptr, call->key.dsize);
1499         memcpy(&state->c->data[call->key.dsize],
1500                call->call_data.dptr, call->call_data.dsize);
1501         *(state->call)              = *call;
1502         state->call->call_data.dptr = &state->c->data[call->key.dsize];
1503         state->call->key.dptr       = &state->c->data[0];
1504
1505         state->state  = CTDB_CALL_WAIT;
1506         state->generation = ctdb_db->generation;
1507
1508         DLIST_ADD(ctdb_db->pending_calls, state);
1509
1510         ctdb_queue_packet(ctdb, &state->c->hdr);
1511
1512         return state;
1513 }
1514
1515 /*
1516   make a remote ctdb call - async recv - called in daemon context
1517
1518   This is called when the program wants to wait for a ctdb_call to complete and get the
1519   results. This call will block unless the call has already completed.
1520 */
1521 int ctdb_daemon_call_recv(struct ctdb_call_state *state, struct ctdb_call *call)
1522 {
1523         while (state->state < CTDB_CALL_DONE) {
1524                 event_loop_once(state->ctdb_db->ctdb->ev);
1525         }
1526         if (state->state != CTDB_CALL_DONE) {
1527                 ctdb_set_error(state->ctdb_db->ctdb, "%s", state->errmsg);
1528                 talloc_free(state);
1529                 return -1;
1530         }
1531
1532         if (state->call->reply_data.dsize) {
1533                 call->reply_data.dptr = talloc_memdup(call,
1534                                                       state->call->reply_data.dptr,
1535                                                       state->call->reply_data.dsize);
1536                 call->reply_data.dsize = state->call->reply_data.dsize;
1537         } else {
1538                 call->reply_data.dptr = NULL;
1539                 call->reply_data.dsize = 0;
1540         }
1541         call->status = state->call->status;
1542         talloc_free(state);
1543         return 0;
1544 }
1545
1546
1547 /*
1548    send a keepalive packet to the other node
1549 */
1550 void ctdb_send_keepalive(struct ctdb_context *ctdb, uint32_t destnode)
1551 {
1552         struct ctdb_req_keepalive *r;
1553
1554         if (ctdb->methods == NULL) {
1555                 DEBUG(DEBUG_INFO,(__location__ " Failed to send keepalive. Transport is DOWN\n"));
1556                 return;
1557         }
1558
1559         r = ctdb_transport_allocate(ctdb, ctdb, CTDB_REQ_KEEPALIVE,
1560                                     sizeof(struct ctdb_req_keepalive),
1561                                     struct ctdb_req_keepalive);
1562         CTDB_NO_MEMORY_FATAL(ctdb, r);
1563         r->hdr.destnode  = destnode;
1564         r->hdr.reqid     = 0;
1565
1566         CTDB_INCREMENT_STAT(ctdb, keepalive_packets_sent);
1567
1568         ctdb_queue_packet(ctdb, &r->hdr);
1569
1570         talloc_free(r);
1571 }
1572
1573
1574
1575 struct revokechild_deferred_call {
1576         struct ctdb_context *ctdb;
1577         struct ctdb_req_header *hdr;
1578         deferred_requeue_fn fn;
1579         void *ctx;
1580 };
1581
1582 struct revokechild_handle {
1583         struct revokechild_handle *next, *prev;
1584         struct ctdb_context *ctdb;
1585         struct ctdb_db_context *ctdb_db;
1586         struct fd_event *fde;
1587         int status;
1588         int fd[2];
1589         pid_t child;
1590         TDB_DATA key;
1591 };
1592
1593 struct revokechild_requeue_handle {
1594         struct ctdb_context *ctdb;
1595         struct ctdb_req_header *hdr;
1596         deferred_requeue_fn fn;
1597         void *ctx;
1598 };
1599
1600 static void deferred_call_requeue(struct event_context *ev, struct timed_event *te,
1601                        struct timeval t, void *private_data)
1602 {
1603         struct revokechild_requeue_handle *requeue_handle = talloc_get_type(private_data, struct revokechild_requeue_handle);
1604
1605         requeue_handle->fn(requeue_handle->ctx, requeue_handle->hdr);
1606         talloc_free(requeue_handle);
1607 }
1608
1609 static int deferred_call_destructor(struct revokechild_deferred_call *deferred_call)
1610 {
1611         struct ctdb_context *ctdb = deferred_call->ctdb;
1612         struct revokechild_requeue_handle *requeue_handle = talloc(ctdb, struct revokechild_requeue_handle);
1613         struct ctdb_req_call *c = (struct ctdb_req_call *)deferred_call->hdr;
1614
1615         requeue_handle->ctdb = ctdb;
1616         requeue_handle->hdr  = deferred_call->hdr;
1617         requeue_handle->fn   = deferred_call->fn;
1618         requeue_handle->ctx  = deferred_call->ctx;
1619         talloc_steal(requeue_handle, requeue_handle->hdr);
1620
1621         /* when revoking, any READONLY requests have 1 second grace to let read/write finish first */
1622         event_add_timed(ctdb->ev, requeue_handle, timeval_current_ofs(c->flags & CTDB_WANT_READONLY ? 1 : 0, 0), deferred_call_requeue, requeue_handle);
1623
1624         return 0;
1625 }
1626
1627
1628 static int revokechild_destructor(struct revokechild_handle *rc)
1629 {
1630         if (rc->fde != NULL) {
1631                 talloc_free(rc->fde);
1632         }
1633
1634         if (rc->fd[0] != -1) {
1635                 close(rc->fd[0]);
1636         }
1637         if (rc->fd[1] != -1) {
1638                 close(rc->fd[1]);
1639         }
1640         ctdb_kill(rc->ctdb, rc->child, SIGKILL);
1641
1642         DLIST_REMOVE(rc->ctdb_db->revokechild_active, rc);
1643         return 0;
1644 }
1645
1646 static void revokechild_handler(struct event_context *ev, struct fd_event *fde,
1647                              uint16_t flags, void *private_data)
1648 {
1649         struct revokechild_handle *rc = talloc_get_type(private_data,
1650                                                      struct revokechild_handle);
1651         int ret;
1652         char c;
1653
1654         ret = sys_read(rc->fd[0], &c, 1);
1655         if (ret != 1) {
1656                 DEBUG(DEBUG_ERR,("Failed to read status from revokechild. errno:%d\n", errno));
1657                 rc->status = -1;
1658                 talloc_free(rc);
1659                 return;
1660         }
1661         if (c != 0) {
1662                 DEBUG(DEBUG_ERR,("revokechild returned failure. status:%d\n", c));
1663                 rc->status = -1;
1664                 talloc_free(rc);
1665                 return;
1666         }
1667
1668         talloc_free(rc);
1669 }
1670
1671 struct ctdb_revoke_state {
1672         struct ctdb_db_context *ctdb_db;
1673         TDB_DATA key;
1674         struct ctdb_ltdb_header *header;
1675         TDB_DATA data;
1676         int count;
1677         int status;
1678         int finished;
1679 };
1680
1681 static void update_record_cb(struct ctdb_client_control_state *state)
1682 {
1683         struct ctdb_revoke_state *revoke_state;
1684         int ret;
1685         int32_t res;
1686
1687         if (state == NULL) {
1688                 return;
1689         }
1690         revoke_state = state->async.private_data;
1691
1692         state->async.fn = NULL;
1693         ret = ctdb_control_recv(state->ctdb, state, state, NULL, &res, NULL);
1694         if ((ret != 0) || (res != 0)) {
1695                 DEBUG(DEBUG_ERR,("Recv for revoke update record failed ret:%d res:%d\n", ret, res));
1696                 revoke_state->status = -1;
1697         }
1698
1699         revoke_state->count--;
1700         if (revoke_state->count <= 0) {
1701                 revoke_state->finished = 1;
1702         }
1703 }
1704
1705 static void revoke_send_cb(struct ctdb_context *ctdb, uint32_t pnn, void *private_data)
1706 {
1707         struct ctdb_revoke_state *revoke_state = private_data;
1708         struct ctdb_client_control_state *state;
1709
1710         state = ctdb_ctrl_updaterecord_send(ctdb, revoke_state, timeval_current_ofs(ctdb->tunable.control_timeout,0), pnn, revoke_state->ctdb_db, revoke_state->key, revoke_state->header, revoke_state->data);
1711         if (state == NULL) {
1712                 DEBUG(DEBUG_ERR,("Failure to send update record to revoke readonly delegation\n"));
1713                 revoke_state->status = -1;
1714                 return;
1715         }
1716         state->async.fn           = update_record_cb;
1717         state->async.private_data = revoke_state;
1718
1719         revoke_state->count++;
1720
1721 }
1722
1723 static void ctdb_revoke_timeout_handler(struct event_context *ev, struct timed_event *te,
1724                               struct timeval yt, void *private_data)
1725 {
1726         struct ctdb_revoke_state *state = private_data;
1727
1728         DEBUG(DEBUG_ERR,("Timed out waiting for revoke to finish\n"));
1729         state->finished = 1;
1730         state->status   = -1;
1731 }
1732
1733 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)
1734 {
1735         struct ctdb_revoke_state *state = talloc_zero(ctdb, struct ctdb_revoke_state);
1736         struct ctdb_ltdb_header new_header;
1737         TDB_DATA new_data;
1738
1739         state->ctdb_db = ctdb_db;
1740         state->key     = key;
1741         state->header  = header;
1742         state->data    = data;
1743
1744         ctdb_trackingdb_traverse(ctdb, tdata, revoke_send_cb, state);
1745
1746         event_add_timed(ctdb->ev, state, timeval_current_ofs(ctdb->tunable.control_timeout, 0), ctdb_revoke_timeout_handler, state);
1747
1748         while (state->finished == 0) {
1749                 event_loop_once(ctdb->ev);
1750         }
1751
1752         if (ctdb_ltdb_lock(ctdb_db, key) != 0) {
1753                 DEBUG(DEBUG_ERR,("Failed to chainlock the database in revokechild\n"));
1754                 talloc_free(state);
1755                 return -1;
1756         }
1757         if (ctdb_ltdb_fetch(ctdb_db, key, &new_header, state, &new_data) != 0) {
1758                 ctdb_ltdb_unlock(ctdb_db, key);
1759                 DEBUG(DEBUG_ERR,("Failed for fetch tdb record in revokechild\n"));
1760                 talloc_free(state);
1761                 return -1;
1762         }
1763         header->rsn++;
1764         if (new_header.rsn > header->rsn) {
1765                 ctdb_ltdb_unlock(ctdb_db, key);
1766                 DEBUG(DEBUG_ERR,("RSN too high in tdb record in revokechild\n"));
1767                 talloc_free(state);
1768                 return -1;
1769         }
1770         if ( (new_header.flags & (CTDB_REC_RO_REVOKING_READONLY|CTDB_REC_RO_HAVE_DELEGATIONS)) != (CTDB_REC_RO_REVOKING_READONLY|CTDB_REC_RO_HAVE_DELEGATIONS) ) {
1771                 ctdb_ltdb_unlock(ctdb_db, key);
1772                 DEBUG(DEBUG_ERR,("Flags are wrong in tdb record in revokechild\n"));
1773                 talloc_free(state);
1774                 return -1;
1775         }
1776
1777         /*
1778          * If revoke on all nodes succeed, revoke is complete.  Otherwise,
1779          * remove CTDB_REC_RO_REVOKING_READONLY flag and retry.
1780          */
1781         if (state->status == 0) {
1782                 new_header.rsn++;
1783                 new_header.flags |= CTDB_REC_RO_REVOKE_COMPLETE;
1784         } else {
1785                 DEBUG(DEBUG_NOTICE, ("Revoke all delegations failed, retrying.\n"));
1786                 new_header.flags &= ~CTDB_REC_RO_REVOKING_READONLY;
1787         }
1788         if (ctdb_ltdb_store(ctdb_db, key, &new_header, new_data) != 0) {
1789                 ctdb_ltdb_unlock(ctdb_db, key);
1790                 DEBUG(DEBUG_ERR,("Failed to write new record in revokechild\n"));
1791                 talloc_free(state);
1792                 return -1;
1793         }
1794         ctdb_ltdb_unlock(ctdb_db, key);
1795
1796         talloc_free(state);
1797         return 0;
1798 }
1799
1800
1801 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)
1802 {
1803         TDB_DATA tdata;
1804         struct revokechild_handle *rc;
1805         pid_t parent = getpid();
1806         int ret;
1807
1808         header->flags &= ~(CTDB_REC_RO_REVOKING_READONLY|CTDB_REC_RO_HAVE_DELEGATIONS|CTDB_REC_RO_HAVE_READONLY);
1809         header->flags |= CTDB_REC_FLAG_MIGRATED_WITH_DATA;
1810         header->rsn   -= 1;
1811
1812         if ((rc = talloc_zero(ctdb_db, struct revokechild_handle)) == NULL) {
1813                 DEBUG(DEBUG_ERR,("Failed to allocate revokechild_handle\n"));
1814                 return -1;
1815         }
1816
1817         tdata = tdb_fetch(ctdb_db->rottdb, key);
1818         if (tdata.dsize > 0) {
1819                 uint8_t *tmp;
1820
1821                 tmp = tdata.dptr;
1822                 tdata.dptr = talloc_memdup(rc, tdata.dptr, tdata.dsize);
1823                 free(tmp);
1824         }
1825
1826         rc->status    = 0;
1827         rc->ctdb      = ctdb;
1828         rc->ctdb_db   = ctdb_db;
1829         rc->fd[0]     = -1;
1830         rc->fd[1]     = -1;
1831
1832         talloc_set_destructor(rc, revokechild_destructor);
1833
1834         rc->key.dsize = key.dsize;
1835         rc->key.dptr  = talloc_memdup(rc, key.dptr, key.dsize);
1836         if (rc->key.dptr == NULL) {
1837                 DEBUG(DEBUG_ERR,("Failed to allocate key for revokechild_handle\n"));
1838                 talloc_free(rc);
1839                 return -1;
1840         }
1841
1842         ret = pipe(rc->fd);
1843         if (ret != 0) {
1844                 DEBUG(DEBUG_ERR,("Failed to allocate key for revokechild_handle\n"));
1845                 talloc_free(rc);
1846                 return -1;
1847         }
1848
1849
1850         rc->child = ctdb_fork(ctdb);
1851         if (rc->child == (pid_t)-1) {
1852                 DEBUG(DEBUG_ERR,("Failed to fork child for revokechild\n"));
1853                 talloc_free(rc);
1854                 return -1;
1855         }
1856
1857         if (rc->child == 0) {
1858                 char c = 0;
1859                 close(rc->fd[0]);
1860                 debug_extra = talloc_asprintf(NULL, "revokechild-%s:", ctdb_db->db_name);
1861
1862                 ctdb_set_process_name("ctdb_revokechild");
1863                 if (switch_from_server_to_client(ctdb, "revokechild-%s", ctdb_db->db_name) != 0) {
1864                         DEBUG(DEBUG_ERR,("Failed to switch from server to client for revokechild process\n"));
1865                         c = 1;
1866                         goto child_finished;
1867                 }
1868
1869                 c = ctdb_revoke_all_delegations(ctdb, ctdb_db, tdata, key, header, data);
1870
1871 child_finished:
1872                 sys_write(rc->fd[1], &c, 1);
1873                 /* make sure we die when our parent dies */
1874                 while (ctdb_kill(ctdb, parent, 0) == 0 || errno != ESRCH) {
1875                         sleep(5);
1876                 }
1877                 _exit(0);
1878         }
1879
1880         close(rc->fd[1]);
1881         rc->fd[1] = -1;
1882         set_close_on_exec(rc->fd[0]);
1883
1884         /* This is an active revokechild child process */
1885         DLIST_ADD_END(ctdb_db->revokechild_active, rc, NULL);
1886
1887         rc->fde = event_add_fd(ctdb->ev, rc, rc->fd[0],
1888                                    EVENT_FD_READ, revokechild_handler,
1889                                    (void *)rc);
1890         if (rc->fde == NULL) {
1891                 DEBUG(DEBUG_ERR,("Failed to set up fd event for revokechild process\n"));
1892                 talloc_free(rc);
1893         }
1894         tevent_fd_set_auto_close(rc->fde);
1895
1896         return 0;
1897 }
1898
1899 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)
1900 {
1901         struct revokechild_handle *rc;
1902         struct revokechild_deferred_call *deferred_call;
1903
1904         for (rc = ctdb_db->revokechild_active; rc; rc = rc->next) {
1905                 if (rc->key.dsize == 0) {
1906                         continue;
1907                 }
1908                 if (rc->key.dsize != key.dsize) {
1909                         continue;
1910                 }
1911                 if (!memcmp(rc->key.dptr, key.dptr, key.dsize)) {
1912                         break;
1913                 }
1914         }
1915
1916         if (rc == NULL) {
1917                 DEBUG(DEBUG_ERR,("Failed to add deferred call to revoke list. revoke structure not found\n"));
1918                 return -1;
1919         }
1920
1921         deferred_call = talloc(rc, struct revokechild_deferred_call);
1922         if (deferred_call == NULL) {
1923                 DEBUG(DEBUG_ERR,("Failed to allocate deferred call structure for revoking record\n"));
1924                 return -1;
1925         }
1926
1927         deferred_call->ctdb = ctdb;
1928         deferred_call->hdr  = hdr;
1929         deferred_call->fn   = fn;
1930         deferred_call->ctx  = call_context;
1931
1932         talloc_set_destructor(deferred_call, deferred_call_destructor);
1933         talloc_steal(deferred_call, hdr);
1934
1935         return 0;
1936 }