?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
server: fix wording and punctuation in comment block for ctdb_reply_dmaster().
[Samba.git] / ctdb / server / ctdb_call.c
bloba4cab3fb252f2fcdf162a4f2d428f842fcffa7de
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, ctdb->pnn);
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
700 sort_keys:
701 for (i = 1; i < MAX_HOT_KEYS; i++) {
702 if (ctdb_db->statistics.hot_keys[i].count == 0) {
703 continue;
704 }
705 if (ctdb_db->statistics.hot_keys[i].count < ctdb_db->statistics.hot_keys[0].count) {
706 hopcount = ctdb_db->statistics.hot_keys[i].count;
707 ctdb_db->statistics.hot_keys[i].count = ctdb_db->statistics.hot_keys[0].count;
708 ctdb_db->statistics.hot_keys[0].count = hopcount;
709
710 key = ctdb_db->statistics.hot_keys[i].key;
711 ctdb_db->statistics.hot_keys[i].key = ctdb_db->statistics.hot_keys[0].key;
712 ctdb_db->statistics.hot_keys[0].key = key;
713 }
714 }
715 }
716
717 /*
718 called when a CTDB_REQ_CALL packet comes in
719 */
720 void ctdb_request_call(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
721 {
722 struct ctdb_req_call *c = (struct ctdb_req_call *)hdr;
723 TDB_DATA data;
724 struct ctdb_reply_call *r;
725 int ret, len;
726 struct ctdb_ltdb_header header;
727 struct ctdb_call *call;
728 struct ctdb_db_context *ctdb_db;
729 int tmp_count, bucket;
730
731 if (ctdb->methods == NULL) {
732 DEBUG(DEBUG_INFO,(__location__ " Failed ctdb_request_call. Transport is DOWN\n"));
733 return;
734 }
735
736
737 ctdb_db = find_ctdb_db(ctdb, c->db_id);
738 if (!ctdb_db) {
739 ctdb_send_error(ctdb, hdr, -1,
740 "Unknown database in request. db_id==0x%08x",
741 c->db_id);
742 return;
743 }
744
745 call = talloc(hdr, struct ctdb_call);
746 CTDB_NO_MEMORY_FATAL(ctdb, call);
747
748 call->call_id = c->callid;
749 call->key.dptr = c->data;
750 call->key.dsize = c->keylen;
751 call->call_data.dptr = c->data + c->keylen;
752 call->call_data.dsize = c->calldatalen;
753 call->reply_data.dptr = NULL;
754 call->reply_data.dsize = 0;
755
756
757 /* If this record is pinned down we should defer the
758 request until the pindown times out
759 */
760 if (ctdb_db->sticky) {
761 if (ctdb_defer_pinned_down_request(ctdb, ctdb_db, call->key, hdr) == 0) {
762 DEBUG(DEBUG_WARNING,("Defer request for pinned down record in %s\n", ctdb_db->db_name));
763 return;
764 }
765 }
766
767
768 /* determine if we are the dmaster for this key. This also
769 fetches the record data (if any), thus avoiding a 2nd fetch of the data
770 if the call will be answered locally */
771
772 ret = ctdb_ltdb_lock_fetch_requeue(ctdb_db, call->key, &header, hdr, &data,
773 ctdb_call_input_pkt, ctdb, false);
774 if (ret == -1) {
775 ctdb_send_error(ctdb, hdr, ret, "ltdb fetch failed in ctdb_request_call");
776 return;
777 }
778 if (ret == -2) {
779 DEBUG(DEBUG_INFO,(__location__ " deferred ctdb_request_call\n"));
780 return;
781 }
782
783 /* Dont do READONLY if we dont have a tracking database */
784 if ((c->flags & CTDB_WANT_READONLY) && !ctdb_db->readonly) {
785 c->flags &= ~CTDB_WANT_READONLY;
786 }
787
788 if (header.flags & CTDB_REC_RO_REVOKE_COMPLETE) {
789 header.flags &= ~CTDB_REC_RO_FLAGS;
790 CTDB_INCREMENT_STAT(ctdb, total_ro_revokes);
791 CTDB_INCREMENT_DB_STAT(ctdb_db, db_ro_revokes);
792 if (ctdb_ltdb_store(ctdb_db, call->key, &header, data) != 0) {
793 ctdb_fatal(ctdb, "Failed to write header with cleared REVOKE flag");
794 }
795 /* and clear out the tracking data */
796 if (tdb_delete(ctdb_db->rottdb, call->key) != 0) {
797 DEBUG(DEBUG_ERR,(__location__ " Failed to clear out trackingdb record\n"));
798 }
799 }
800
801 /* if we are revoking, we must defer all other calls until the revoke
802 * had completed.
803 */
804 if (header.flags & CTDB_REC_RO_REVOKING_READONLY) {
805 talloc_free(data.dptr);
806 ret = ctdb_ltdb_unlock(ctdb_db, call->key);
807
808 if (ctdb_add_revoke_deferred_call(ctdb, ctdb_db, call->key, hdr, ctdb_call_input_pkt, ctdb) != 0) {
809 ctdb_fatal(ctdb, "Failed to add deferred call for revoke child");
810 }
811 talloc_free(call);
812 return;
813 }
814
815 /*
816 * If we are not the dmaster and are not hosting any delegations,
817 * then we redirect the request to the node than can answer it
818 * (the lmaster or the dmaster).
819 */
820 if ((header.dmaster != ctdb->pnn)
821 && (!(header.flags & CTDB_REC_RO_HAVE_DELEGATIONS)) ) {
822 talloc_free(data.dptr);
823 ctdb_call_send_redirect(ctdb, ctdb_db, call->key, c, &header);
824
825 ret = ctdb_ltdb_unlock(ctdb_db, call->key);
826 if (ret != 0) {
827 DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
828 }
829 return;
830 }
831
832 if ( (!(c->flags & CTDB_WANT_READONLY))
833 && (header.flags & (CTDB_REC_RO_HAVE_DELEGATIONS|CTDB_REC_RO_HAVE_READONLY)) ) {
834 header.flags |= CTDB_REC_RO_REVOKING_READONLY;
835 if (ctdb_ltdb_store(ctdb_db, call->key, &header, data) != 0) {
836 ctdb_fatal(ctdb, "Failed to store record with HAVE_DELEGATIONS set");
837 }
838 ret = ctdb_ltdb_unlock(ctdb_db, call->key);
839
840 if (ctdb_start_revoke_ro_record(ctdb, ctdb_db, call->key, &header, data) != 0) {
841 ctdb_fatal(ctdb, "Failed to start record revoke");
842 }
843 talloc_free(data.dptr);
844
845 if (ctdb_add_revoke_deferred_call(ctdb, ctdb_db, call->key, hdr, ctdb_call_input_pkt, ctdb) != 0) {
846 ctdb_fatal(ctdb, "Failed to add deferred call for revoke child");
847 }
848 talloc_free(call);
849
850 return;
851 }
852
853 /* If this is the first request for delegation. bump rsn and set
854 * the delegations flag
855 */
856 if ((c->flags & CTDB_WANT_READONLY)
857 && (c->callid == CTDB_FETCH_WITH_HEADER_FUNC)
858 && (!(header.flags & CTDB_REC_RO_HAVE_DELEGATIONS))) {
859 header.rsn += 3;
860 header.flags |= CTDB_REC_RO_HAVE_DELEGATIONS;
861 if (ctdb_ltdb_store(ctdb_db, call->key, &header, data) != 0) {
862 ctdb_fatal(ctdb, "Failed to store record with HAVE_DELEGATIONS set");
863 }
864 }
865 if ((c->flags & CTDB_WANT_READONLY)
866 && (call->call_id == CTDB_FETCH_WITH_HEADER_FUNC)) {
867 TDB_DATA tdata;
868
869 tdata = tdb_fetch(ctdb_db->rottdb, call->key);
870 if (ctdb_trackingdb_add_pnn(ctdb, &tdata, c->hdr.srcnode) != 0) {
871 ctdb_fatal(ctdb, "Failed to add node to trackingdb");
872 }
873 if (tdb_store(ctdb_db->rottdb, call->key, tdata, TDB_REPLACE) != 0) {
874 ctdb_fatal(ctdb, "Failed to store trackingdb data");
875 }
876 free(tdata.dptr);
877
878 ret = ctdb_ltdb_unlock(ctdb_db, call->key);
879 if (ret != 0) {
880 DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
881 }
882
883 len = offsetof(struct ctdb_reply_call, data) + data.dsize + sizeof(struct ctdb_ltdb_header);
884 r = ctdb_transport_allocate(ctdb, ctdb, CTDB_REPLY_CALL, len,
885 struct ctdb_reply_call);
886 CTDB_NO_MEMORY_FATAL(ctdb, r);
887 r->hdr.destnode = c->hdr.srcnode;
888 r->hdr.reqid = c->hdr.reqid;
889 r->status = 0;
890 r->datalen = data.dsize + sizeof(struct ctdb_ltdb_header);
891 header.rsn -= 2;
892 header.flags |= CTDB_REC_RO_HAVE_READONLY;
893 header.flags &= ~CTDB_REC_RO_HAVE_DELEGATIONS;
894 memcpy(&r->data[0], &header, sizeof(struct ctdb_ltdb_header));
895
896 if (data.dsize) {
897 memcpy(&r->data[sizeof(struct ctdb_ltdb_header)], data.dptr, data.dsize);
898 }
899
900 ctdb_queue_packet(ctdb, &r->hdr);
901 CTDB_INCREMENT_STAT(ctdb, total_ro_delegations);
902 CTDB_INCREMENT_DB_STAT(ctdb_db, db_ro_delegations);
903
904 talloc_free(r);
905 return;
906 }
907
908 CTDB_UPDATE_STAT(ctdb, max_hop_count, c->hopcount);
909 tmp_count = c->hopcount;
910 bucket = 0;
911 while (tmp_count) {
912 tmp_count >>= 2;
913 bucket++;
914 }
915 if (bucket >= MAX_COUNT_BUCKETS) {
916 bucket = MAX_COUNT_BUCKETS - 1;
917 }
918 CTDB_INCREMENT_STAT(ctdb, hop_count_bucket[bucket]);
919 CTDB_INCREMENT_DB_STAT(ctdb_db, hop_count_bucket[bucket]);
920 ctdb_update_db_stat_hot_keys(ctdb_db, call->key, c->hopcount);
921
922 /* If this database supports sticky records, then check if the
923 hopcount is big. If it is it means the record is hot and we
924 should make it sticky.
925 */
926 if (ctdb_db->sticky && c->hopcount >= ctdb->tunable.hopcount_make_sticky) {
927 ctdb_make_record_sticky(ctdb, ctdb_db, call->key);
928 }
929
930
931 /* if this nodes has done enough consecutive calls on the same record
932 then give them the record
933 or if the node requested an immediate migration
934 */
935 if ( c->hdr.srcnode != ctdb->pnn &&
936 ((header.laccessor == c->hdr.srcnode
937 && header.lacount >= ctdb->tunable.max_lacount
938 && ctdb->tunable.max_lacount != 0)
939 || (c->flags & CTDB_IMMEDIATE_MIGRATION)) ) {
940 if (ctdb_db->transaction_active) {
941 DEBUG(DEBUG_INFO, (__location__ " refusing migration"
942 " of key %s while transaction is active\n",
943 (char *)call->key.dptr));
944 } else {
945 DEBUG(DEBUG_DEBUG,("pnn %u starting migration of %08x to %u\n",
946 ctdb->pnn, ctdb_hash(&(call->key)), c->hdr.srcnode));
947 ctdb_call_send_dmaster(ctdb_db, c, &header, &(call->key), &data);
948 talloc_free(data.dptr);
949
950 ret = ctdb_ltdb_unlock(ctdb_db, call->key);
951 if (ret != 0) {
952 DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
953 }
954 return;
955 }
956 }
957
958 ret = ctdb_call_local(ctdb_db, call, &header, hdr, &data, true, c->hdr.srcnode);
959 if (ret != 0) {
960 DEBUG(DEBUG_ERR,(__location__ " ctdb_call_local failed\n"));
961 call->status = -1;
962 }
963
964 ret = ctdb_ltdb_unlock(ctdb_db, call->key);
965 if (ret != 0) {
966 DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
967 }
968
969 len = offsetof(struct ctdb_reply_call, data) + call->reply_data.dsize;
970 r = ctdb_transport_allocate(ctdb, ctdb, CTDB_REPLY_CALL, len,
971 struct ctdb_reply_call);
972 CTDB_NO_MEMORY_FATAL(ctdb, r);
973 r->hdr.destnode = hdr->srcnode;
974 r->hdr.reqid = hdr->reqid;
975 r->status = call->status;
976 r->datalen = call->reply_data.dsize;
977 if (call->reply_data.dsize) {
978 memcpy(&r->data[0], call->reply_data.dptr, call->reply_data.dsize);
979 }
980
981 ctdb_queue_packet(ctdb, &r->hdr);
982
983 talloc_free(r);
984 }
985
986 /*
987 called when a CTDB_REPLY_CALL packet comes in
988
989 This packet comes in response to a CTDB_REQ_CALL request packet. It
990 contains any reply data from the call
991 */
992 void ctdb_reply_call(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
993 {
994 struct ctdb_reply_call *c = (struct ctdb_reply_call *)hdr;
995 struct ctdb_call_state *state;
996
997 state = ctdb_reqid_find(ctdb, hdr->reqid, struct ctdb_call_state);
998 if (state == NULL) {
999 DEBUG(DEBUG_ERR, (__location__ " reqid %u not found\n", hdr->reqid));
1000 return;
1001 }
1002
1003 if (hdr->reqid != state->reqid) {
1004 /* we found a record but it was the wrong one */
1005 DEBUG(DEBUG_ERR, ("Dropped orphaned call reply with reqid:%u\n",hdr->reqid));
1006 return;
1007 }
1008
1009
1010 /* read only delegation processing */
1011 /* If we got a FETCH_WITH_HEADER we should check if this is a ro
1012 * delegation since we may need to update the record header
1013 */
1014 if (state->c->callid == CTDB_FETCH_WITH_HEADER_FUNC) {
1015 struct ctdb_db_context *ctdb_db = state->ctdb_db;
1016 struct ctdb_ltdb_header *header = (struct ctdb_ltdb_header *)&c->data[0];
1017 struct ctdb_ltdb_header oldheader;
1018 TDB_DATA key, data, olddata;
1019 int ret;
1020
1021 if (!(header->flags & CTDB_REC_RO_HAVE_READONLY)) {
1022 goto finished_ro;
1023 return;
1024 }
1025
1026 key.dsize = state->c->keylen;
1027 key.dptr = state->c->data;
1028 ret = ctdb_ltdb_lock_requeue(ctdb_db, key, hdr,
1029 ctdb_call_input_pkt, ctdb, false);
1030 if (ret == -2) {
1031 return;
1032 }
1033 if (ret != 0) {
1034 DEBUG(DEBUG_ERR,(__location__ " Failed to get lock in ctdb_reply_call\n"));
1035 return;
1036 }
1037
1038 ret = ctdb_ltdb_fetch(ctdb_db, key, &oldheader, state, &olddata);
1039 if (ret != 0) {
1040 DEBUG(DEBUG_ERR, ("Failed to fetch old record in ctdb_reply_call\n"));
1041 ctdb_ltdb_unlock(ctdb_db, key);
1042 goto finished_ro;
1043 }
1044
1045 if (header->rsn <= oldheader.rsn) {
1046 ctdb_ltdb_unlock(ctdb_db, key);
1047 goto finished_ro;
1048 }
1049
1050 if (c->datalen < sizeof(struct ctdb_ltdb_header)) {
1051 DEBUG(DEBUG_ERR,(__location__ " Got FETCH_WITH_HEADER reply with too little data: %d bytes\n", c->datalen));
1052 ctdb_ltdb_unlock(ctdb_db, key);
1053 goto finished_ro;
1054 }
1055
1056 data.dsize = c->datalen - sizeof(struct ctdb_ltdb_header);
1057 data.dptr = &c->data[sizeof(struct ctdb_ltdb_header)];
1058 ret = ctdb_ltdb_store(ctdb_db, key, header, data);
1059 if (ret != 0) {
1060 DEBUG(DEBUG_ERR, ("Failed to store new record in ctdb_reply_call\n"));
1061 ctdb_ltdb_unlock(ctdb_db, key);
1062 goto finished_ro;
1063 }
1064
1065 ctdb_ltdb_unlock(ctdb_db, key);
1066 }
1067 finished_ro:
1068
1069 state->call->reply_data.dptr = c->data;
1070 state->call->reply_data.dsize = c->datalen;
1071 state->call->status = c->status;
1072
1073 talloc_steal(state, c);
1074
1075 state->state = CTDB_CALL_DONE;
1076 if (state->async.fn) {
1077 state->async.fn(state);
1078 }
1079 }
1080
1081
1082 /*
1083 called when a CTDB_REPLY_DMASTER packet comes in
1084
1085 This packet comes in from the lmaster in response to a CTDB_REQ_CALL
1086 request packet. It means that the current dmaster wants to give us
1087 the dmaster role.
1088 */
1089 void ctdb_reply_dmaster(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
1090 {
1091 struct ctdb_reply_dmaster *c = (struct ctdb_reply_dmaster *)hdr;
1092 struct ctdb_db_context *ctdb_db;
1093 TDB_DATA key, data;
1094 uint32_t record_flags = 0;
1095 size_t len;
1096 int ret;
1097
1098 ctdb_db = find_ctdb_db(ctdb, c->db_id);
1099 if (ctdb_db == NULL) {
1100 DEBUG(DEBUG_ERR,("Unknown db_id 0x%x in ctdb_reply_dmaster\n", c->db_id));
1101 return;
1102 }
1103
1104 key.dptr = c->data;
1105 key.dsize = c->keylen;
1106 data.dptr = &c->data[key.dsize];
1107 data.dsize = c->datalen;
1108 len = offsetof(struct ctdb_reply_dmaster, data) + key.dsize + data.dsize
1109 + sizeof(uint32_t);
1110 if (len <= c->hdr.length) {
1111 record_flags = *(uint32_t *)&c->data[c->keylen + c->datalen];
1112 }
1113
1114 ret = ctdb_ltdb_lock_requeue(ctdb_db, key, hdr,
1115 ctdb_call_input_pkt, ctdb, false);
1116 if (ret == -2) {
1117 return;
1118 }
1119 if (ret != 0) {
1120 DEBUG(DEBUG_ERR,(__location__ " Failed to get lock in ctdb_reply_dmaster\n"));
1121 return;
1122 }
1123
1124 ctdb_become_dmaster(ctdb_db, hdr, key, data, c->rsn, record_flags);
1125 }
1126
1127
1128 /*
1129 called when a CTDB_REPLY_ERROR packet comes in
1130 */
1131 void ctdb_reply_error(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
1132 {
1133 struct ctdb_reply_error *c = (struct ctdb_reply_error *)hdr;
1134 struct ctdb_call_state *state;
1135
1136 state = ctdb_reqid_find(ctdb, hdr->reqid, struct ctdb_call_state);
1137 if (state == NULL) {
1138 DEBUG(DEBUG_ERR,("pnn %u Invalid reqid %u in ctdb_reply_error\n",
1139 ctdb->pnn, hdr->reqid));
1140 return;
1141 }
1142
1143 if (hdr->reqid != state->reqid) {
1144 /* we found a record but it was the wrong one */
1145 DEBUG(DEBUG_ERR, ("Dropped orphaned error reply with reqid:%u\n",hdr->reqid));
1146 return;
1147 }
1148
1149 talloc_steal(state, c);
1150
1151 state->state = CTDB_CALL_ERROR;
1152 state->errmsg = (char *)c->msg;
1153 if (state->async.fn) {
1154 state->async.fn(state);
1155 }
1156 }
1157
1158
1159 /*
1160 destroy a ctdb_call
1161 */
1162 static int ctdb_call_destructor(struct ctdb_call_state *state)
1163 {
1164 DLIST_REMOVE(state->ctdb_db->ctdb->pending_calls, state);
1165 ctdb_reqid_remove(state->ctdb_db->ctdb, state->reqid);
1166 return 0;
1167 }
1168
1169
1170 /*
1171 called when a ctdb_call needs to be resent after a reconfigure event
1172 */
1173 static void ctdb_call_resend(struct ctdb_call_state *state)
1174 {
1175 struct ctdb_context *ctdb = state->ctdb_db->ctdb;
1176
1177 state->generation = ctdb->vnn_map->generation;
1178
1179 /* use a new reqid, in case the old reply does eventually come in */
1180 ctdb_reqid_remove(ctdb, state->reqid);
1181 state->reqid = ctdb_reqid_new(ctdb, state);
1182 state->c->hdr.reqid = state->reqid;
1183
1184 /* update the generation count for this request, so its valid with the new vnn_map */
1185 state->c->hdr.generation = state->generation;
1186
1187 /* send the packet to ourselves, it will be redirected appropriately */
1188 state->c->hdr.destnode = ctdb->pnn;
1189
1190 ctdb_queue_packet(ctdb, &state->c->hdr);
1191 DEBUG(DEBUG_NOTICE,("resent ctdb_call\n"));
1192 }
1193
1194 /*
1195 resend all pending calls on recovery
1196 */
1197 void ctdb_call_resend_all(struct ctdb_context *ctdb)
1198 {
1199 struct ctdb_call_state *state, *next;
1200 for (state=ctdb->pending_calls;state;state=next) {
1201 next = state->next;
1202 ctdb_call_resend(state);
1203 }
1204 }
1205
1206 /*
1207 this allows the caller to setup a async.fn
1208 */
1209 static void call_local_trigger(struct event_context *ev, struct timed_event *te,
1210 struct timeval t, void *private_data)
1211 {
1212 struct ctdb_call_state *state = talloc_get_type(private_data, struct ctdb_call_state);
1213 if (state->async.fn) {
1214 state->async.fn(state);
1215 }
1216 }
1217
1218
1219 /*
1220 construct an event driven local ctdb_call
1221
1222 this is used so that locally processed ctdb_call requests are processed
1223 in an event driven manner
1224 */
1225 struct ctdb_call_state *ctdb_call_local_send(struct ctdb_db_context *ctdb_db,
1226 struct ctdb_call *call,
1227 struct ctdb_ltdb_header *header,
1228 TDB_DATA *data)
1229 {
1230 struct ctdb_call_state *state;
1231 struct ctdb_context *ctdb = ctdb_db->ctdb;
1232 int ret;
1233
1234 state = talloc_zero(ctdb_db, struct ctdb_call_state);
1235 CTDB_NO_MEMORY_NULL(ctdb, state);
1236
1237 talloc_steal(state, data->dptr);
1238
1239 state->state = CTDB_CALL_DONE;
1240 state->call = talloc(state, struct ctdb_call);
1241 CTDB_NO_MEMORY_NULL(ctdb, state->call);
1242 *(state->call) = *call;
1243 state->ctdb_db = ctdb_db;
1244
1245 ret = ctdb_call_local(ctdb_db, state->call, header, state, data, true, ctdb->pnn);
1246 if (ret != 0) {
1247 DEBUG(DEBUG_DEBUG,("ctdb_call_local() failed, ignoring return code %d\n", ret));
1248 }
1249
1250 event_add_timed(ctdb->ev, state, timeval_zero(), call_local_trigger, state);
1251
1252 return state;
1253 }
1254
1255
1256 /*
1257 make a remote ctdb call - async send. Called in daemon context.
1258
1259 This constructs a ctdb_call request and queues it for processing.
1260 This call never blocks.
1261 */
1262 struct ctdb_call_state *ctdb_daemon_call_send_remote(struct ctdb_db_context *ctdb_db,
1263 struct ctdb_call *call,
1264 struct ctdb_ltdb_header *header)
1265 {
1266 uint32_t len;
1267 struct ctdb_call_state *state;
1268 struct ctdb_context *ctdb = ctdb_db->ctdb;
1269
1270 if (ctdb->methods == NULL) {
1271 DEBUG(DEBUG_INFO,(__location__ " Failed send packet. Transport is down\n"));
1272 return NULL;
1273 }
1274
1275 state = talloc_zero(ctdb_db, struct ctdb_call_state);
1276 CTDB_NO_MEMORY_NULL(ctdb, state);
1277 state->call = talloc(state, struct ctdb_call);
1278 CTDB_NO_MEMORY_NULL(ctdb, state->call);
1279
1280 state->reqid = ctdb_reqid_new(ctdb, state);
1281 state->ctdb_db = ctdb_db;
1282 talloc_set_destructor(state, ctdb_call_destructor);
1283
1284 len = offsetof(struct ctdb_req_call, data) + call->key.dsize + call->call_data.dsize;
1285 state->c = ctdb_transport_allocate(ctdb, state, CTDB_REQ_CALL, len,
1286 struct ctdb_req_call);
1287 CTDB_NO_MEMORY_NULL(ctdb, state->c);
1288 state->c->hdr.destnode = header->dmaster;
1289
1290 /* this limits us to 16k outstanding messages - not unreasonable */
1291 state->c->hdr.reqid = state->reqid;
1292 state->c->flags = call->flags;
1293 state->c->db_id = ctdb_db->db_id;
1294 state->c->callid = call->call_id;
1295 state->c->hopcount = 0;
1296 state->c->keylen = call->key.dsize;
1297 state->c->calldatalen = call->call_data.dsize;
1298 memcpy(&state->c->data[0], call->key.dptr, call->key.dsize);
1299 memcpy(&state->c->data[call->key.dsize],
1300 call->call_data.dptr, call->call_data.dsize);
1301 *(state->call) = *call;
1302 state->call->call_data.dptr = &state->c->data[call->key.dsize];
1303 state->call->key.dptr = &state->c->data[0];
1304
1305 state->state = CTDB_CALL_WAIT;
1306 state->generation = ctdb->vnn_map->generation;
1307
1308 DLIST_ADD(ctdb->pending_calls, state);
1309
1310 ctdb_queue_packet(ctdb, &state->c->hdr);
1311
1312 return state;
1313 }
1314
1315 /*
1316 make a remote ctdb call - async recv - called in daemon context
1317
1318 This is called when the program wants to wait for a ctdb_call to complete and get the
1319 results. This call will block unless the call has already completed.
1320 */
1321 int ctdb_daemon_call_recv(struct ctdb_call_state *state, struct ctdb_call *call)
1322 {
1323 while (state->state < CTDB_CALL_DONE) {
1324 event_loop_once(state->ctdb_db->ctdb->ev);
1325 }
1326 if (state->state != CTDB_CALL_DONE) {
1327 ctdb_set_error(state->ctdb_db->ctdb, "%s", state->errmsg);
1328 talloc_free(state);
1329 return -1;
1330 }
1331
1332 if (state->call->reply_data.dsize) {
1333 call->reply_data.dptr = talloc_memdup(call,
1334 state->call->reply_data.dptr,
1335 state->call->reply_data.dsize);
1336 call->reply_data.dsize = state->call->reply_data.dsize;
1337 } else {
1338 call->reply_data.dptr = NULL;
1339 call->reply_data.dsize = 0;
1340 }
1341 call->status = state->call->status;
1342 talloc_free(state);
1343 return 0;
1344 }
1345
1346
1347 /*
1348 send a keepalive packet to the other node
1349 */
1350 void ctdb_send_keepalive(struct ctdb_context *ctdb, uint32_t destnode)
1351 {
1352 struct ctdb_req_keepalive *r;
1353
1354 if (ctdb->methods == NULL) {
1355 DEBUG(DEBUG_INFO,(__location__ " Failed to send keepalive. Transport is DOWN\n"));
1356 return;
1357 }
1358
1359 r = ctdb_transport_allocate(ctdb, ctdb, CTDB_REQ_KEEPALIVE,
1360 sizeof(struct ctdb_req_keepalive),
1361 struct ctdb_req_keepalive);
1362 CTDB_NO_MEMORY_FATAL(ctdb, r);
1363 r->hdr.destnode = destnode;
1364 r->hdr.reqid = 0;
1365
1366 CTDB_INCREMENT_STAT(ctdb, keepalive_packets_sent);
1367
1368 ctdb_queue_packet(ctdb, &r->hdr);
1369
1370 talloc_free(r);
1371 }
1372
1373
1374
1375 struct revokechild_deferred_call {
1376 struct ctdb_context *ctdb;
1377 struct ctdb_req_header *hdr;
1378 deferred_requeue_fn fn;
1379 void *ctx;
1380 };
1381
1382 struct revokechild_handle {
1383 struct revokechild_handle *next, *prev;
1384 struct ctdb_context *ctdb;
1385 struct ctdb_db_context *ctdb_db;
1386 struct fd_event *fde;
1387 int status;
1388 int fd[2];
1389 pid_t child;
1390 TDB_DATA key;
1391 };
1392
1393 struct revokechild_requeue_handle {
1394 struct ctdb_context *ctdb;
1395 struct ctdb_req_header *hdr;
1396 deferred_requeue_fn fn;
1397 void *ctx;
1398 };
1399
1400 static void deferred_call_requeue(struct event_context *ev, struct timed_event *te,
1401 struct timeval t, void *private_data)
1402 {
1403 struct revokechild_requeue_handle *requeue_handle = talloc_get_type(private_data, struct revokechild_requeue_handle);
1404
1405 requeue_handle->fn(requeue_handle->ctx, requeue_handle->hdr);
1406 talloc_free(requeue_handle);
1407 }
1408
1409 static int deferred_call_destructor(struct revokechild_deferred_call *deferred_call)
1410 {
1411 struct ctdb_context *ctdb = deferred_call->ctdb;
1412 struct revokechild_requeue_handle *requeue_handle = talloc(ctdb, struct revokechild_requeue_handle);
1413 struct ctdb_req_call *c = (struct ctdb_req_call *)deferred_call->hdr;
1414
1415 requeue_handle->ctdb = ctdb;
1416 requeue_handle->hdr = deferred_call->hdr;
1417 requeue_handle->fn = deferred_call->fn;
1418 requeue_handle->ctx = deferred_call->ctx;
1419 talloc_steal(requeue_handle, requeue_handle->hdr);
1420
1421 /* when revoking, any READONLY requests have 1 second grace to let read/write finish first */
1422 event_add_timed(ctdb->ev, requeue_handle, timeval_current_ofs(c->flags & CTDB_WANT_READONLY ? 1 : 0, 0), deferred_call_requeue, requeue_handle);
1423
1424 return 0;
1425 }
1426
1427
1428 static int revokechild_destructor(struct revokechild_handle *rc)
1429 {
1430 if (rc->fde != NULL) {
1431 talloc_free(rc->fde);
1432 }
1433
1434 if (rc->fd[0] != -1) {
1435 close(rc->fd[0]);
1436 }
1437 if (rc->fd[1] != -1) {
1438 close(rc->fd[1]);
1439 }
1440 ctdb_kill(rc->ctdb, rc->child, SIGKILL);
1441
1442 DLIST_REMOVE(rc->ctdb_db->revokechild_active, rc);
1443 return 0;
1444 }
1445
1446 static void revokechild_handler(struct event_context *ev, struct fd_event *fde,
1447 uint16_t flags, void *private_data)
1448 {
1449 struct revokechild_handle *rc = talloc_get_type(private_data,
1450 struct revokechild_handle);
1451 int ret;
1452 char c;
1453
1454 ret = read(rc->fd[0], &c, 1);
1455 if (ret != 1) {
1456 DEBUG(DEBUG_ERR,("Failed to read status from revokechild. errno:%d\n", errno));
1457 rc->status = -1;
1458 talloc_free(rc);
1459 return;
1460 }
1461 if (c != 0) {
1462 DEBUG(DEBUG_ERR,("revokechild returned failure. status:%d\n", c));
1463 rc->status = -1;
1464 talloc_free(rc);
1465 return;
1466 }
1467
1468 talloc_free(rc);
1469 }
1470
1471 struct ctdb_revoke_state {
1472 struct ctdb_db_context *ctdb_db;
1473 TDB_DATA key;
1474 struct ctdb_ltdb_header *header;
1475 TDB_DATA data;
1476 int count;
1477 int status;
1478 int finished;
1479 };
1480
1481 static void update_record_cb(struct ctdb_client_control_state *state)
1482 {
1483 struct ctdb_revoke_state *revoke_state;
1484 int ret;
1485 int32_t res;
1486
1487 if (state == NULL) {
1488 return;
1489 }
1490 revoke_state = state->async.private_data;
1491
1492 state->async.fn = NULL;
1493 ret = ctdb_control_recv(state->ctdb, state, state, NULL, &res, NULL);
1494 if ((ret != 0) || (res != 0)) {
1495 DEBUG(DEBUG_ERR,("Recv for revoke update record failed ret:%d res:%d\n", ret, res));
1496 revoke_state->status = -1;
1497 }
1498
1499 revoke_state->count--;
1500 if (revoke_state->count <= 0) {
1501 revoke_state->finished = 1;
1502 }
1503 }
1504
1505 static void revoke_send_cb(struct ctdb_context *ctdb, uint32_t pnn, void *private_data)
1506 {
1507 struct ctdb_revoke_state *revoke_state = private_data;
1508 struct ctdb_client_control_state *state;
1509
1510 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);
1511 if (state == NULL) {
1512 DEBUG(DEBUG_ERR,("Failure to send update record to revoke readonly delegation\n"));
1513 revoke_state->status = -1;
1514 return;
1515 }
1516 state->async.fn = update_record_cb;
1517 state->async.private_data = revoke_state;
1518
1519 revoke_state->count++;
1520
1521 }
1522
1523 static void ctdb_revoke_timeout_handler(struct event_context *ev, struct timed_event *te,
1524 struct timeval yt, void *private_data)
1525 {
1526 struct ctdb_revoke_state *state = private_data;
1527
1528 DEBUG(DEBUG_ERR,("Timed out waiting for revoke to finish\n"));
1529 state->finished = 1;
1530 state->status = -1;
1531 }
1532
1533 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)
1534 {
1535 struct ctdb_revoke_state *state = talloc_zero(ctdb, struct ctdb_revoke_state);
1536 int status;
1537
1538 state->ctdb_db = ctdb_db;
1539 state->key = key;
1540 state->header = header;
1541 state->data = data;
1542
1543 ctdb_trackingdb_traverse(ctdb, tdata, revoke_send_cb, state);
1544
1545 event_add_timed(ctdb->ev, state, timeval_current_ofs(5, 0), ctdb_revoke_timeout_handler, state);
1546
1547 while (state->finished == 0) {
1548 event_loop_once(ctdb->ev);
1549 }
1550
1551 status = state->status;
1552
1553 if (status == 0) {
1554 struct ctdb_ltdb_header new_header;
1555 TDB_DATA new_data;
1556
1557 if (ctdb_ltdb_lock(ctdb_db, key) != 0) {
1558 DEBUG(DEBUG_ERR,("Failed to chainlock the database in revokechild\n"));
1559 talloc_free(state);
1560 return -1;
1561 }
1562 if (ctdb_ltdb_fetch(ctdb_db, key, &new_header, state, &new_data) != 0) {
1563 ctdb_ltdb_unlock(ctdb_db, key);
1564 DEBUG(DEBUG_ERR,("Failed for fetch tdb record in revokechild\n"));
1565 talloc_free(state);
1566 return -1;
1567 }
1568 header->rsn++;
1569 if (new_header.rsn > header->rsn) {
1570 ctdb_ltdb_unlock(ctdb_db, key);
1571 DEBUG(DEBUG_ERR,("RSN too high in tdb record in revokechild\n"));
1572 talloc_free(state);
1573 return -1;
1574 }
1575 if ( (new_header.flags & (CTDB_REC_RO_REVOKING_READONLY|CTDB_REC_RO_HAVE_DELEGATIONS)) != (CTDB_REC_RO_REVOKING_READONLY|CTDB_REC_RO_HAVE_DELEGATIONS) ) {
1576 ctdb_ltdb_unlock(ctdb_db, key);
1577 DEBUG(DEBUG_ERR,("Flags are wrong in tdb record in revokechild\n"));
1578 talloc_free(state);
1579 return -1;
1580 }
1581 new_header.rsn++;
1582 new_header.flags |= CTDB_REC_RO_REVOKE_COMPLETE;
1583 if (ctdb_ltdb_store(ctdb_db, key, &new_header, new_data) != 0) {
1584 ctdb_ltdb_unlock(ctdb_db, key);
1585 DEBUG(DEBUG_ERR,("Failed to write new record in revokechild\n"));
1586 talloc_free(state);
1587 return -1;
1588 }
1589 ctdb_ltdb_unlock(ctdb_db, key);
1590 }
1591
1592 talloc_free(state);
1593 return status;
1594 }
1595
1596
1597 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)
1598 {
1599 TDB_DATA tdata;
1600 struct revokechild_handle *rc;
1601 pid_t parent = getpid();
1602 int ret;
1603
1604 header->flags &= ~(CTDB_REC_RO_REVOKING_READONLY|CTDB_REC_RO_HAVE_DELEGATIONS|CTDB_REC_RO_HAVE_READONLY);
1605 header->flags |= CTDB_REC_FLAG_MIGRATED_WITH_DATA;
1606 header->rsn -= 1;
1607
1608 if ((rc = talloc_zero(ctdb_db, struct revokechild_handle)) == NULL) {
1609 DEBUG(DEBUG_ERR,("Failed to allocate revokechild_handle\n"));
1610 return -1;
1611 }
1612
1613 tdata = tdb_fetch(ctdb_db->rottdb, key);
1614 if (tdata.dsize > 0) {
1615 uint8_t *tmp;
1616
1617 tmp = tdata.dptr;
1618 tdata.dptr = talloc_memdup(rc, tdata.dptr, tdata.dsize);
1619 free(tmp);
1620 }
1621
1622 rc->status = 0;
1623 rc->ctdb = ctdb;
1624 rc->ctdb_db = ctdb_db;
1625 rc->fd[0] = -1;
1626 rc->fd[1] = -1;
1627
1628 talloc_set_destructor(rc, revokechild_destructor);
1629
1630 rc->key.dsize = key.dsize;
1631 rc->key.dptr = talloc_memdup(rc, key.dptr, key.dsize);
1632 if (rc->key.dptr == NULL) {
1633 DEBUG(DEBUG_ERR,("Failed to allocate key for revokechild_handle\n"));
1634 talloc_free(rc);
1635 return -1;
1636 }
1637
1638 ret = pipe(rc->fd);
1639 if (ret != 0) {
1640 DEBUG(DEBUG_ERR,("Failed to allocate key for revokechild_handle\n"));
1641 talloc_free(rc);
1642 return -1;
1643 }
1644
1645
1646 rc->child = ctdb_fork(ctdb);
1647 if (rc->child == (pid_t)-1) {
1648 DEBUG(DEBUG_ERR,("Failed to fork child for revokechild\n"));
1649 talloc_free(rc);
1650 return -1;
1651 }
1652
1653 if (rc->child == 0) {
1654 char c = 0;
1655 close(rc->fd[0]);
1656 debug_extra = talloc_asprintf(NULL, "revokechild-%s:", ctdb_db->db_name);
1657
1658 ctdb_set_process_name("ctdb_revokechild");
1659 if (switch_from_server_to_client(ctdb, "revokechild-%s", ctdb_db->db_name) != 0) {
1660 DEBUG(DEBUG_ERR,("Failed to switch from server to client for revokechild process\n"));
1661 c = 1;
1662 goto child_finished;
1663 }
1664
1665 c = ctdb_revoke_all_delegations(ctdb, ctdb_db, tdata, key, header, data);
1666
1667 child_finished:
1668 write(rc->fd[1], &c, 1);
1669 /* make sure we die when our parent dies */
1670 while (ctdb_kill(ctdb, parent, 0) == 0 || errno != ESRCH) {
1671 sleep(5);
1672 }
1673 _exit(0);
1674 }
1675
1676 close(rc->fd[1]);
1677 rc->fd[1] = -1;
1678 set_close_on_exec(rc->fd[0]);
1679
1680 /* This is an active revokechild child process */
1681 DLIST_ADD_END(ctdb_db->revokechild_active, rc, NULL);
1682
1683 rc->fde = event_add_fd(ctdb->ev, rc, rc->fd[0],
1684 EVENT_FD_READ, revokechild_handler,
1685 (void *)rc);
1686 if (rc->fde == NULL) {
1687 DEBUG(DEBUG_ERR,("Failed to set up fd event for revokechild process\n"));
1688 talloc_free(rc);
1689 }
1690 tevent_fd_set_auto_close(rc->fde);
1691
1692 return 0;
1693 }
1694
1695 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)
1696 {
1697 struct revokechild_handle *rc;
1698 struct revokechild_deferred_call *deferred_call;
1699
1700 for (rc = ctdb_db->revokechild_active; rc; rc = rc->next) {
1701 if (rc->key.dsize == 0) {
1702 continue;
1703 }
1704 if (rc->key.dsize != key.dsize) {
1705 continue;
1706 }
1707 if (!memcmp(rc->key.dptr, key.dptr, key.dsize)) {
1708 break;
1709 }
1710 }
1711
1712 if (rc == NULL) {
1713 DEBUG(DEBUG_ERR,("Failed to add deferred call to revoke list. revoke structure not found\n"));
1714 return -1;
1715 }
1716
1717 deferred_call = talloc(rc, struct revokechild_deferred_call);
1718 if (deferred_call == NULL) {
1719 DEBUG(DEBUG_ERR,("Failed to allocate deferred call structure for revoking record\n"));
1720 return -1;
1721 }
1722
1723 deferred_call->ctdb = ctdb;
1724 deferred_call->hdr = hdr;
1725 deferred_call->fn = fn;
1726 deferred_call->ctx = call_context;
1727
1728 talloc_set_destructor(deferred_call, deferred_call_destructor);
1729 talloc_steal(deferred_call, hdr);
1730
1731 return 0;
1732 }
Samba GIT mirror