corosync  3.0.2
totempg.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2003-2005 MontaVista Software, Inc.
3  * Copyright (c) 2005 OSDL.
4  * Copyright (c) 2006-2012 Red Hat, Inc.
5  *
6  * All rights reserved.
7  *
8  * Author: Steven Dake (sdake@redhat.com)
9  * Author: Mark Haverkamp (markh@osdl.org)
10  *
11  * This software licensed under BSD license, the text of which follows:
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions are met:
15  *
16  * - Redistributions of source code must retain the above copyright notice,
17  * this list of conditions and the following disclaimer.
18  * - Redistributions in binary form must reproduce the above copyright notice,
19  * this list of conditions and the following disclaimer in the documentation
20  * and/or other materials provided with the distribution.
21  * - Neither the name of the MontaVista Software, Inc. nor the names of its
22  * contributors may be used to endorse or promote products derived from this
23  * software without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
26  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
29  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
35  * THE POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 /*
39  * FRAGMENTATION AND PACKING ALGORITHM:
40  *
41  * Assemble the entire message into one buffer
42  * if full fragment
43  * store fragment into lengths list
44  * for each full fragment
45  * multicast fragment
46  * set length and fragment fields of pg mesage
47  * store remaining multicast into head of fragmentation data and set lens field
48  *
49  * If a message exceeds the maximum packet size allowed by the totem
50  * single ring protocol, the protocol could lose forward progress.
51  * Statically calculating the allowed data amount doesn't work because
52  * the amount of data allowed depends on the number of fragments in
53  * each message. In this implementation, the maximum fragment size
54  * is dynamically calculated for each fragment added to the message.
55 
56  * It is possible for a message to be two bytes short of the maximum
57  * packet size. This occurs when a message or collection of
58  * messages + the mcast header + the lens are two bytes short of the
59  * end of the packet. Since another len field consumes two bytes, the
60  * len field would consume the rest of the packet without room for data.
61  *
62  * One optimization would be to forgo the final len field and determine
63  * it from the size of the udp datagram. Then this condition would no
64  * longer occur.
65  */
66 
67 /*
68  * ASSEMBLY AND UNPACKING ALGORITHM:
69  *
70  * copy incoming packet into assembly data buffer indexed by current
71  * location of end of fragment
72  *
73  * if not fragmented
74  * deliver all messages in assembly data buffer
75  * else
76  * if msg_count > 1 and fragmented
77  * deliver all messages except last message in assembly data buffer
78  * copy last fragmented section to start of assembly data buffer
79  * else
80  * if msg_count = 1 and fragmented
81  * do nothing
82  *
83  */
84 
85 #include <config.h>
86 
87 #ifdef HAVE_ALLOCA_H
88 #include <alloca.h>
89 #endif
90 #include <sys/types.h>
91 #include <sys/socket.h>
92 #include <netinet/in.h>
93 #include <arpa/inet.h>
94 #include <sys/uio.h>
95 #include <stdio.h>
96 #include <stdlib.h>
97 #include <string.h>
98 #include <assert.h>
99 #include <pthread.h>
100 #include <errno.h>
101 #include <limits.h>
102 
103 #include <corosync/swab.h>
104 #include <qb/qblist.h>
105 #include <qb/qbloop.h>
106 #include <qb/qbipcs.h>
107 #include <corosync/totem/totempg.h>
108 #define LOGSYS_UTILS_ONLY 1
109 #include <corosync/logsys.h>
110 
111 #include "util.h"
112 #include "totemsrp.h"
113 
115  short version;
116  short type;
117 };
118 
119 #if !(defined(__i386__) || defined(__x86_64__))
120 /*
121  * Need align on architectures different then i386 or x86_64
122  */
123 #define TOTEMPG_NEED_ALIGN 1
124 #endif
125 
126 /*
127  * totempg_mcast structure
128  *
129  * header: Identify the mcast.
130  * fragmented: Set if this message continues into next message
131  * continuation: Set if this message is a continuation from last message
132  * msg_count Indicates how many packed messages are contained
133  * in the mcast.
134  * Also, the size of each packed message and the messages themselves are
135  * appended to the end of this structure when sent.
136  */
139  unsigned char fragmented;
140  unsigned char continuation;
141  unsigned short msg_count;
142  /*
143  * short msg_len[msg_count];
144  */
145  /*
146  * data for messages
147  */
148 };
149 
150 /*
151  * Maximum packet size for totem pg messages
152  */
153 #define TOTEMPG_PACKET_SIZE (totempg_totem_config->net_mtu - \
154  sizeof (struct totempg_mcast))
155 
156 /*
157  * Local variables used for packing small messages
158  */
159 static unsigned short mcast_packed_msg_lens[FRAME_SIZE_MAX];
160 
161 static int mcast_packed_msg_count = 0;
162 
163 static int totempg_reserved = 1;
164 
165 static unsigned int totempg_size_limit;
166 
167 static totem_queue_level_changed_fn totem_queue_level_changed = NULL;
168 
169 static uint32_t totempg_threaded_mode = 0;
170 
171 static void *totemsrp_context;
172 
173 /*
174  * Function and data used to log messages
175  */
176 static int totempg_log_level_security;
177 static int totempg_log_level_error;
178 static int totempg_log_level_warning;
179 static int totempg_log_level_notice;
180 static int totempg_log_level_debug;
181 static int totempg_subsys_id;
182 static void (*totempg_log_printf) (
183  int level,
184  int subsys,
185  const char *function,
186  const char *file,
187  int line,
188  const char *format, ...) __attribute__((format(printf, 6, 7)));
189 
191 
192 static totempg_stats_t totempg_stats;
193 
197 };
198 
199 struct assembly {
200  unsigned int nodeid;
201  unsigned char data[MESSAGE_SIZE_MAX+KNET_MAX_PACKET_SIZE];
202  int index;
203  unsigned char last_frag_num;
205  struct qb_list_head list;
206 };
207 
208 static void assembly_deref (struct assembly *assembly);
209 
210 static int callback_token_received_fn (enum totem_callback_token_type type,
211  const void *data);
212 
213 QB_LIST_DECLARE(assembly_list_inuse);
214 
215 /*
216  * Free list is used both for transitional and operational assemblies
217  */
218 QB_LIST_DECLARE(assembly_list_free);
219 
220 QB_LIST_DECLARE(assembly_list_inuse_trans);
221 
222 QB_LIST_DECLARE(totempg_groups_list);
223 
224 /*
225  * Staging buffer for packed messages. Messages are staged in this buffer
226  * before sending. Multiple messages may fit which cuts down on the
227  * number of mcasts sent. If a message doesn't completely fit, then
228  * the mcast header has a fragment bit set that says that there are more
229  * data to follow. fragment_size is an index into the buffer. It indicates
230  * the size of message data and where to place new message data.
231  * fragment_contuation indicates whether the first packed message in
232  * the buffer is a continuation of a previously packed fragment.
233  */
234 static unsigned char *fragmentation_data;
235 
236 static int fragment_size = 0;
237 
238 static int fragment_continuation = 0;
239 
240 static int totempg_waiting_transack = 0;
241 
243  void (*deliver_fn) (
244  unsigned int nodeid,
245  const void *msg,
246  unsigned int msg_len,
247  int endian_conversion_required);
248 
249  void (*confchg_fn) (
250  enum totem_configuration_type configuration_type,
251  const unsigned int *member_list, size_t member_list_entries,
252  const unsigned int *left_list, size_t left_list_entries,
253  const unsigned int *joined_list, size_t joined_list_entries,
254  const struct memb_ring_id *ring_id);
255 
257 
259  int32_t q_level;
260 
261  struct qb_list_head list;
262 };
263 
264 static unsigned char next_fragment = 1;
265 
266 static pthread_mutex_t totempg_mutex = PTHREAD_MUTEX_INITIALIZER;
267 
268 static pthread_mutex_t callback_token_mutex = PTHREAD_MUTEX_INITIALIZER;
269 
270 static pthread_mutex_t mcast_msg_mutex = PTHREAD_MUTEX_INITIALIZER;
271 
272 #define log_printf(level, format, args...) \
273 do { \
274  totempg_log_printf(level, \
275  totempg_subsys_id, \
276  __FUNCTION__, __FILE__, __LINE__, \
277  format, ##args); \
278 } while (0);
279 
280 static int msg_count_send_ok (int msg_count);
281 
282 static int byte_count_send_ok (int byte_count);
283 
284 static void totempg_waiting_trans_ack_cb (int waiting_trans_ack)
285 {
286  log_printf(LOG_DEBUG, "waiting_trans_ack changed to %u", waiting_trans_ack);
287  totempg_waiting_transack = waiting_trans_ack;
288 }
289 
290 static struct assembly *assembly_ref (unsigned int nodeid)
291 {
292  struct assembly *assembly;
293  struct qb_list_head *list;
294  struct qb_list_head *active_assembly_list_inuse;
295 
296  if (totempg_waiting_transack) {
297  active_assembly_list_inuse = &assembly_list_inuse_trans;
298  } else {
299  active_assembly_list_inuse = &assembly_list_inuse;
300  }
301 
302  /*
303  * Search inuse list for node id and return assembly buffer if found
304  */
305  qb_list_for_each(list, active_assembly_list_inuse) {
306  assembly = qb_list_entry (list, struct assembly, list);
307 
308  if (nodeid == assembly->nodeid) {
309  return (assembly);
310  }
311  }
312 
313  /*
314  * Nothing found in inuse list get one from free list if available
315  */
316  if (qb_list_empty (&assembly_list_free) == 0) {
317  assembly = qb_list_first_entry (&assembly_list_free, struct assembly, list);
318  qb_list_del (&assembly->list);
319  qb_list_add (&assembly->list, active_assembly_list_inuse);
320  assembly->nodeid = nodeid;
321  assembly->index = 0;
322  assembly->last_frag_num = 0;
324  return (assembly);
325  }
326 
327  /*
328  * Nothing available in inuse or free list, so allocate a new one
329  */
330  assembly = malloc (sizeof (struct assembly));
331  /*
332  * TODO handle memory allocation failure here
333  */
334  assert (assembly);
335  assembly->nodeid = nodeid;
336  assembly->data[0] = 0;
337  assembly->index = 0;
338  assembly->last_frag_num = 0;
340  qb_list_init (&assembly->list);
341  qb_list_add (&assembly->list, active_assembly_list_inuse);
342 
343  return (assembly);
344 }
345 
346 static void assembly_deref (struct assembly *assembly)
347 {
348  qb_list_del (&assembly->list);
349  qb_list_add (&assembly->list, &assembly_list_free);
350 }
351 
352 static void assembly_deref_from_normal_and_trans (int nodeid)
353 {
354  int j;
355  struct qb_list_head *list, *tmp_iter;
356  struct qb_list_head *active_assembly_list_inuse;
357  struct assembly *assembly;
358 
359  for (j = 0; j < 2; j++) {
360  if (j == 0) {
361  active_assembly_list_inuse = &assembly_list_inuse;
362  } else {
363  active_assembly_list_inuse = &assembly_list_inuse_trans;
364  }
365 
366  qb_list_for_each_safe(list, tmp_iter, active_assembly_list_inuse) {
367  assembly = qb_list_entry (list, struct assembly, list);
368 
369  if (nodeid == assembly->nodeid) {
370  qb_list_del (&assembly->list);
371  qb_list_add (&assembly->list, &assembly_list_free);
372  }
373  }
374  }
375 
376 }
377 
378 static inline void app_confchg_fn (
379  enum totem_configuration_type configuration_type,
380  const unsigned int *member_list, size_t member_list_entries,
381  const unsigned int *left_list, size_t left_list_entries,
382  const unsigned int *joined_list, size_t joined_list_entries,
383  const struct memb_ring_id *ring_id)
384 {
385  int i;
386  struct totempg_group_instance *instance;
387  struct qb_list_head *list;
388 
389  /*
390  * For every leaving processor, add to free list
391  * This also has the side effect of clearing out the dataset
392  * In the leaving processor's assembly buffer.
393  */
394  for (i = 0; i < left_list_entries; i++) {
395  assembly_deref_from_normal_and_trans (left_list[i]);
396  }
397 
398  qb_list_for_each(list, &totempg_groups_list) {
399  instance = qb_list_entry (list, struct totempg_group_instance, list);
400 
401  if (instance->confchg_fn) {
402  instance->confchg_fn (
403  configuration_type,
404  member_list,
405  member_list_entries,
406  left_list,
407  left_list_entries,
408  joined_list,
409  joined_list_entries,
410  ring_id);
411  }
412  }
413 }
414 
415 static inline void group_endian_convert (
416  void *msg,
417  int msg_len)
418 {
419  unsigned short *group_len;
420  int i;
421  char *aligned_msg;
422 
423 #ifdef TOTEMPG_NEED_ALIGN
424  /*
425  * Align data structure for not i386 or x86_64
426  */
427  if ((size_t)msg % 4 != 0) {
428  aligned_msg = alloca(msg_len);
429  memcpy(aligned_msg, msg, msg_len);
430  } else {
431  aligned_msg = msg;
432  }
433 #else
434  aligned_msg = msg;
435 #endif
436 
437  group_len = (unsigned short *)aligned_msg;
438  group_len[0] = swab16(group_len[0]);
439  for (i = 1; i < group_len[0] + 1; i++) {
440  group_len[i] = swab16(group_len[i]);
441  }
442 
443  if (aligned_msg != msg) {
444  memcpy(msg, aligned_msg, msg_len);
445  }
446 }
447 
448 static inline int group_matches (
449  struct iovec *iovec,
450  unsigned int iov_len,
451  struct totempg_group *groups_b,
452  unsigned int group_b_cnt,
453  unsigned int *adjust_iovec)
454 {
455  unsigned short *group_len;
456  char *group_name;
457  int i;
458  int j;
459 #ifdef TOTEMPG_NEED_ALIGN
460  struct iovec iovec_aligned = { NULL, 0 };
461 #endif
462 
463  assert (iov_len == 1);
464 
465 #ifdef TOTEMPG_NEED_ALIGN
466  /*
467  * Align data structure for not i386 or x86_64
468  */
469  if ((size_t)iovec->iov_base % 4 != 0) {
470  iovec_aligned.iov_base = alloca(iovec->iov_len);
471  memcpy(iovec_aligned.iov_base, iovec->iov_base, iovec->iov_len);
472  iovec_aligned.iov_len = iovec->iov_len;
473  iovec = &iovec_aligned;
474  }
475 #endif
476 
477  group_len = (unsigned short *)iovec->iov_base;
478  group_name = ((char *)iovec->iov_base) +
479  sizeof (unsigned short) * (group_len[0] + 1);
480 
481 
482  /*
483  * Calculate amount to adjust the iovec by before delivering to app
484  */
485  *adjust_iovec = sizeof (unsigned short) * (group_len[0] + 1);
486  for (i = 1; i < group_len[0] + 1; i++) {
487  *adjust_iovec += group_len[i];
488  }
489 
490  /*
491  * Determine if this message should be delivered to this instance
492  */
493  for (i = 1; i < group_len[0] + 1; i++) {
494  for (j = 0; j < group_b_cnt; j++) {
495  if ((group_len[i] == groups_b[j].group_len) &&
496  (memcmp (groups_b[j].group, group_name, group_len[i]) == 0)) {
497  return (1);
498  }
499  }
500  group_name += group_len[i];
501  }
502  return (0);
503 }
504 
505 
506 static inline void app_deliver_fn (
507  unsigned int nodeid,
508  void *msg,
509  unsigned int msg_len,
510  int endian_conversion_required)
511 {
512  struct totempg_group_instance *instance;
513  struct iovec stripped_iovec;
514  unsigned int adjust_iovec;
515  struct iovec *iovec;
516  struct qb_list_head *list;
517 
518  struct iovec aligned_iovec = { NULL, 0 };
519 
520  if (endian_conversion_required) {
521  group_endian_convert (msg, msg_len);
522  }
523 
524  /*
525  * TODO: segmentation/assembly need to be redesigned to provide aligned access
526  * in all cases to avoid memory copies on non386 archs. Probably broke backwars
527  * compatibility
528  */
529 
530 #ifdef TOTEMPG_NEED_ALIGN
531  /*
532  * Align data structure for not i386 or x86_64
533  */
534  aligned_iovec.iov_base = alloca(msg_len);
535  aligned_iovec.iov_len = msg_len;
536  memcpy(aligned_iovec.iov_base, msg, msg_len);
537 #else
538  aligned_iovec.iov_base = msg;
539  aligned_iovec.iov_len = msg_len;
540 #endif
541 
542  iovec = &aligned_iovec;
543 
544  qb_list_for_each(list, &totempg_groups_list) {
545  instance = qb_list_entry (list, struct totempg_group_instance, list);
546  if (group_matches (iovec, 1, instance->groups, instance->groups_cnt, &adjust_iovec)) {
547  stripped_iovec.iov_len = iovec->iov_len - adjust_iovec;
548  stripped_iovec.iov_base = (char *)iovec->iov_base + adjust_iovec;
549 
550 #ifdef TOTEMPG_NEED_ALIGN
551  /*
552  * Align data structure for not i386 or x86_64
553  */
554  if ((char *)iovec->iov_base + adjust_iovec % 4 != 0) {
555  /*
556  * Deal with misalignment
557  */
558  stripped_iovec.iov_base =
559  alloca (stripped_iovec.iov_len);
560  memcpy (stripped_iovec.iov_base,
561  (char *)iovec->iov_base + adjust_iovec,
562  stripped_iovec.iov_len);
563  }
564 #endif
565  instance->deliver_fn (
566  nodeid,
567  stripped_iovec.iov_base,
568  stripped_iovec.iov_len,
569  endian_conversion_required);
570  }
571  }
572 }
573 
574 static void totempg_confchg_fn (
575  enum totem_configuration_type configuration_type,
576  const unsigned int *member_list, size_t member_list_entries,
577  const unsigned int *left_list, size_t left_list_entries,
578  const unsigned int *joined_list, size_t joined_list_entries,
579  const struct memb_ring_id *ring_id)
580 {
581 // TODO optimize this
582  app_confchg_fn (configuration_type,
583  member_list, member_list_entries,
584  left_list, left_list_entries,
585  joined_list, joined_list_entries,
586  ring_id);
587 }
588 
589 static void totempg_deliver_fn (
590  unsigned int nodeid,
591  const void *msg,
592  unsigned int msg_len,
593  int endian_conversion_required)
594 {
595  struct totempg_mcast *mcast;
596  unsigned short *msg_lens;
597  int i;
598  struct assembly *assembly;
599  char header[FRAME_SIZE_MAX];
600  int msg_count;
601  int continuation;
602  int start;
603  const char *data;
604  int datasize;
605  struct iovec iov_delv;
606  size_t expected_msg_len;
607 
608  assembly = assembly_ref (nodeid);
609  assert (assembly);
610 
611  if (msg_len < sizeof(struct totempg_mcast)) {
612  log_printf(LOG_WARNING,
613  "Message (totempg_mcast) received from node " CS_PRI_NODE_ID " is too short... Ignoring.", nodeid);
614 
615  return ;
616  }
617 
618  /*
619  * Assemble the header into one block of data and
620  * assemble the packet contents into one block of data to simplify delivery
621  */
622 
623  mcast = (struct totempg_mcast *)msg;
624  if (endian_conversion_required) {
625  mcast->msg_count = swab16 (mcast->msg_count);
626  }
627 
628  msg_count = mcast->msg_count;
629  datasize = sizeof (struct totempg_mcast) +
630  msg_count * sizeof (unsigned short);
631 
632  if (msg_len < datasize) {
633  log_printf(LOG_WARNING,
634  "Message (totempg_mcast datasize) received from node " CS_PRI_NODE_ID
635  " is too short... Ignoring.", nodeid);
636 
637  return ;
638  }
639 
640  memcpy (header, msg, datasize);
641  data = msg;
642 
643  msg_lens = (unsigned short *) (header + sizeof (struct totempg_mcast));
644  expected_msg_len = datasize;
645  for (i = 0; i < mcast->msg_count; i++) {
646  if (endian_conversion_required) {
647  msg_lens[i] = swab16 (msg_lens[i]);
648  }
649 
650  expected_msg_len += msg_lens[i];
651  }
652 
653  if (msg_len != expected_msg_len) {
654  log_printf(LOG_WARNING,
655  "Message (totempg_mcast) received from node " CS_PRI_NODE_ID
656  " doesn't have expected length of %zu (has %u) bytes... Ignoring.",
657  nodeid, expected_msg_len, msg_len);
658 
659  return ;
660  }
661 
662  assert((assembly->index+msg_len) < sizeof(assembly->data));
663  memcpy (&assembly->data[assembly->index], &data[datasize],
664  msg_len - datasize);
665 
666  /*
667  * If the last message in the buffer is a fragment, then we
668  * can't deliver it. We'll first deliver the full messages
669  * then adjust the assembly buffer so we can add the rest of the
670  * fragment when it arrives.
671  */
672  msg_count = mcast->fragmented ? mcast->msg_count - 1 : mcast->msg_count;
673  continuation = mcast->continuation;
674  iov_delv.iov_base = (void *)&assembly->data[0];
675  iov_delv.iov_len = assembly->index + msg_lens[0];
676 
677  /*
678  * Make sure that if this message is a continuation, that it
679  * matches the sequence number of the previous fragment.
680  * Also, if the first packed message is a continuation
681  * of a previous message, but the assembly buffer
682  * is empty, then we need to discard it since we can't
683  * assemble a complete message. Likewise, if this message isn't a
684  * continuation and the assembly buffer is empty, we have to discard
685  * the continued message.
686  */
687  start = 0;
688 
689  if (assembly->throw_away_mode == THROW_AWAY_ACTIVE) {
690  /* Throw away the first msg block */
691  if (mcast->fragmented == 0 || mcast->fragmented == 1) {
693 
694  assembly->index += msg_lens[0];
695  iov_delv.iov_base = (void *)&assembly->data[assembly->index];
696  iov_delv.iov_len = msg_lens[1];
697  start = 1;
698  }
699  } else
700  if (assembly->throw_away_mode == THROW_AWAY_INACTIVE) {
701  if (continuation == assembly->last_frag_num) {
702  assembly->last_frag_num = mcast->fragmented;
703  for (i = start; i < msg_count; i++) {
704  app_deliver_fn(nodeid, iov_delv.iov_base, iov_delv.iov_len,
705  endian_conversion_required);
706  assembly->index += msg_lens[i];
707  iov_delv.iov_base = (void *)&assembly->data[assembly->index];
708  if (i < (msg_count - 1)) {
709  iov_delv.iov_len = msg_lens[i + 1];
710  }
711  }
712  } else {
713  log_printf (LOG_DEBUG, "fragmented continuation %u is not equal to assembly last_frag_num %u",
714  continuation, assembly->last_frag_num);
716  }
717  }
718 
719  if (mcast->fragmented == 0) {
720  /*
721  * End of messages, dereference assembly struct
722  */
723  assembly->last_frag_num = 0;
724  assembly->index = 0;
725  assembly_deref (assembly);
726  } else {
727  /*
728  * Message is fragmented, keep around assembly list
729  */
730  if (mcast->msg_count > 1) {
731  memmove (&assembly->data[0],
732  &assembly->data[assembly->index],
733  msg_lens[msg_count]);
734 
735  assembly->index = 0;
736  }
737  assembly->index += msg_lens[msg_count];
738  }
739 }
740 
741 /*
742  * Totem Process Group Abstraction
743  * depends on poll abstraction, POSIX, IPV4
744  */
745 
747 
748 int callback_token_received_fn (enum totem_callback_token_type type,
749  const void *data)
750 {
751  struct totempg_mcast mcast;
752  struct iovec iovecs[3];
753 
754  if (totempg_threaded_mode == 1) {
755  pthread_mutex_lock (&mcast_msg_mutex);
756  }
757  if (mcast_packed_msg_count == 0) {
758  if (totempg_threaded_mode == 1) {
759  pthread_mutex_unlock (&mcast_msg_mutex);
760  }
761  return (0);
762  }
763  if (totemsrp_avail(totemsrp_context) == 0) {
764  if (totempg_threaded_mode == 1) {
765  pthread_mutex_unlock (&mcast_msg_mutex);
766  }
767  return (0);
768  }
769  mcast.header.version = 0;
770  mcast.header.type = 0;
771  mcast.fragmented = 0;
772 
773  /*
774  * Was the first message in this buffer a continuation of a
775  * fragmented message?
776  */
777  mcast.continuation = fragment_continuation;
778  fragment_continuation = 0;
779 
780  mcast.msg_count = mcast_packed_msg_count;
781 
782  iovecs[0].iov_base = (void *)&mcast;
783  iovecs[0].iov_len = sizeof (struct totempg_mcast);
784  iovecs[1].iov_base = (void *)mcast_packed_msg_lens;
785  iovecs[1].iov_len = mcast_packed_msg_count * sizeof (unsigned short);
786  iovecs[2].iov_base = (void *)&fragmentation_data[0];
787  iovecs[2].iov_len = fragment_size;
788  (void)totemsrp_mcast (totemsrp_context, iovecs, 3, 0);
789 
790  mcast_packed_msg_count = 0;
791  fragment_size = 0;
792 
793  if (totempg_threaded_mode == 1) {
794  pthread_mutex_unlock (&mcast_msg_mutex);
795  }
796  return (0);
797 }
798 
799 /*
800  * Initialize the totem process group abstraction
801  */
803  qb_loop_t *poll_handle,
804  struct totem_config *totem_config)
805 {
806  int res;
807 
808  totempg_totem_config = totem_config;
809  totempg_log_level_security = totem_config->totem_logging_configuration.log_level_security;
810  totempg_log_level_error = totem_config->totem_logging_configuration.log_level_error;
811  totempg_log_level_warning = totem_config->totem_logging_configuration.log_level_warning;
812  totempg_log_level_notice = totem_config->totem_logging_configuration.log_level_notice;
813  totempg_log_level_debug = totem_config->totem_logging_configuration.log_level_debug;
814  totempg_log_printf = totem_config->totem_logging_configuration.log_printf;
815  totempg_subsys_id = totem_config->totem_logging_configuration.log_subsys_id;
816 
817  fragmentation_data = malloc (TOTEMPG_PACKET_SIZE);
818  if (fragmentation_data == 0) {
819  return (-1);
820  }
821 
822  totemsrp_net_mtu_adjust (totem_config);
823 
824  res = totemsrp_initialize (
825  poll_handle,
826  &totemsrp_context,
827  totem_config,
828  &totempg_stats,
829  totempg_deliver_fn,
830  totempg_confchg_fn,
831  totempg_waiting_trans_ack_cb);
832 
833  if (res == -1) {
834  goto error_exit;
835  }
836 
838  totemsrp_context,
839  &callback_token_received_handle,
841  0,
842  callback_token_received_fn,
843  0);
844 
845  totempg_size_limit = (totemsrp_avail(totemsrp_context) - 1) *
846  (totempg_totem_config->net_mtu -
847  sizeof (struct totempg_mcast) - 16);
848 
849  qb_list_init (&totempg_groups_list);
850 
851 error_exit:
852  return (res);
853 }
854 
855 void totempg_finalize (void)
856 {
857  if (totempg_threaded_mode == 1) {
858  pthread_mutex_lock (&totempg_mutex);
859  }
860  totemsrp_finalize (totemsrp_context);
861  if (totempg_threaded_mode == 1) {
862  pthread_mutex_unlock (&totempg_mutex);
863  }
864 }
865 
866 /*
867  * Multicast a message
868  */
869 static int mcast_msg (
870  struct iovec *iovec_in,
871  unsigned int iov_len,
872  int guarantee)
873 {
874  int res = 0;
875  struct totempg_mcast mcast;
876  struct iovec iovecs[3];
877  struct iovec iovec[64];
878  int i;
879  int dest, src;
880  int max_packet_size = 0;
881  int copy_len = 0;
882  int copy_base = 0;
883  int total_size = 0;
884 
885  if (totempg_threaded_mode == 1) {
886  pthread_mutex_lock (&mcast_msg_mutex);
887  }
888  totemsrp_event_signal (totemsrp_context, TOTEM_EVENT_NEW_MSG, 1);
889 
890  /*
891  * Remove zero length iovectors from the list
892  */
893  assert (iov_len < 64);
894  for (dest = 0, src = 0; src < iov_len; src++) {
895  if (iovec_in[src].iov_len) {
896  memcpy (&iovec[dest++], &iovec_in[src],
897  sizeof (struct iovec));
898  }
899  }
900  iov_len = dest;
901 
902  max_packet_size = TOTEMPG_PACKET_SIZE -
903  (sizeof (unsigned short) * (mcast_packed_msg_count + 1));
904 
905  mcast_packed_msg_lens[mcast_packed_msg_count] = 0;
906 
907  /*
908  * Check if we would overwrite new message queue
909  */
910  for (i = 0; i < iov_len; i++) {
911  total_size += iovec[i].iov_len;
912  }
913 
914  if (byte_count_send_ok (total_size + sizeof(unsigned short) *
915  (mcast_packed_msg_count)) == 0) {
916 
917  if (totempg_threaded_mode == 1) {
918  pthread_mutex_unlock (&mcast_msg_mutex);
919  }
920  return(-1);
921  }
922 
923  mcast.header.version = 0;
924  for (i = 0; i < iov_len; ) {
925  mcast.fragmented = 0;
926  mcast.continuation = fragment_continuation;
927  copy_len = iovec[i].iov_len - copy_base;
928 
929  /*
930  * If it all fits with room left over, copy it in.
931  * We need to leave at least sizeof(short) + 1 bytes in the
932  * fragment_buffer on exit so that max_packet_size + fragment_size
933  * doesn't exceed the size of the fragment_buffer on the next call.
934  */
935  if ((iovec[i].iov_len + fragment_size) <
936  (max_packet_size - sizeof (unsigned short))) {
937 
938  memcpy (&fragmentation_data[fragment_size],
939  (char *)iovec[i].iov_base + copy_base, copy_len);
940  fragment_size += copy_len;
941  mcast_packed_msg_lens[mcast_packed_msg_count] += copy_len;
942  next_fragment = 1;
943  copy_len = 0;
944  copy_base = 0;
945  i++;
946  continue;
947 
948  /*
949  * If it just fits or is too big, then send out what fits.
950  */
951  } else {
952  unsigned char *data_ptr;
953 
954  copy_len = min(copy_len, max_packet_size - fragment_size);
955  if( copy_len == max_packet_size )
956  data_ptr = (unsigned char *)iovec[i].iov_base + copy_base;
957  else {
958  data_ptr = fragmentation_data;
959  }
960 
961  memcpy (&fragmentation_data[fragment_size],
962  (unsigned char *)iovec[i].iov_base + copy_base, copy_len);
963  mcast_packed_msg_lens[mcast_packed_msg_count] += copy_len;
964 
965  /*
966  * if we're not on the last iovec or the iovec is too large to
967  * fit, then indicate a fragment. This also means that the next
968  * message will have the continuation of this one.
969  */
970  if ((i < (iov_len - 1)) ||
971  ((copy_base + copy_len) < iovec[i].iov_len)) {
972  if (!next_fragment) {
973  next_fragment++;
974  }
975  fragment_continuation = next_fragment;
976  mcast.fragmented = next_fragment++;
977  assert(fragment_continuation != 0);
978  assert(mcast.fragmented != 0);
979  } else {
980  fragment_continuation = 0;
981  }
982 
983  /*
984  * assemble the message and send it
985  */
986  mcast.msg_count = ++mcast_packed_msg_count;
987  iovecs[0].iov_base = (void *)&mcast;
988  iovecs[0].iov_len = sizeof(struct totempg_mcast);
989  iovecs[1].iov_base = (void *)mcast_packed_msg_lens;
990  iovecs[1].iov_len = mcast_packed_msg_count *
991  sizeof(unsigned short);
992  iovecs[2].iov_base = (void *)data_ptr;
993  iovecs[2].iov_len = fragment_size + copy_len;
994  assert (totemsrp_avail(totemsrp_context) > 0);
995  res = totemsrp_mcast (totemsrp_context, iovecs, 3, guarantee);
996  if (res == -1) {
997  goto error_exit;
998  }
999 
1000  /*
1001  * Recalculate counts and indexes for the next.
1002  */
1003  mcast_packed_msg_lens[0] = 0;
1004  mcast_packed_msg_count = 0;
1005  fragment_size = 0;
1006  max_packet_size = TOTEMPG_PACKET_SIZE - (sizeof(unsigned short));
1007 
1008  /*
1009  * If the iovec all fit, go to the next iovec
1010  */
1011  if ((copy_base + copy_len) == iovec[i].iov_len) {
1012  copy_len = 0;
1013  copy_base = 0;
1014  i++;
1015 
1016  /*
1017  * Continue with the rest of the current iovec.
1018  */
1019  } else {
1020  copy_base += copy_len;
1021  }
1022  }
1023  }
1024 
1025  /*
1026  * Bump only if we added message data. This may be zero if
1027  * the last buffer just fit into the fragmentation_data buffer
1028  * and we were at the last iovec.
1029  */
1030  if (mcast_packed_msg_lens[mcast_packed_msg_count]) {
1031  mcast_packed_msg_count++;
1032  }
1033 
1034 error_exit:
1035  if (totempg_threaded_mode == 1) {
1036  pthread_mutex_unlock (&mcast_msg_mutex);
1037  }
1038  return (res);
1039 }
1040 
1041 /*
1042  * Determine if a message of msg_size could be queued
1043  */
1044 static int msg_count_send_ok (
1045  int msg_count)
1046 {
1047  int avail = 0;
1048 
1049  avail = totemsrp_avail (totemsrp_context);
1050  totempg_stats.msg_queue_avail = avail;
1051 
1052  return ((avail - totempg_reserved) > msg_count);
1053 }
1054 
1055 static int byte_count_send_ok (
1056  int byte_count)
1057 {
1058  unsigned int msg_count = 0;
1059  int avail = 0;
1060 
1061  avail = totemsrp_avail (totemsrp_context);
1062 
1063  msg_count = (byte_count / (totempg_totem_config->net_mtu - sizeof (struct totempg_mcast) - 16)) + 1;
1064 
1065  return (avail >= msg_count);
1066 }
1067 
1068 static int send_reserve (
1069  int msg_size)
1070 {
1071  unsigned int msg_count = 0;
1072 
1073  msg_count = (msg_size / (totempg_totem_config->net_mtu - sizeof (struct totempg_mcast) - 16)) + 1;
1074  totempg_reserved += msg_count;
1075  totempg_stats.msg_reserved = totempg_reserved;
1076 
1077  return (msg_count);
1078 }
1079 
1080 static void send_release (
1081  int msg_count)
1082 {
1083  totempg_reserved -= msg_count;
1084  totempg_stats.msg_reserved = totempg_reserved;
1085 }
1086 
1087 #ifndef HAVE_SMALL_MEMORY_FOOTPRINT
1088 #undef MESSAGE_QUEUE_MAX
1089 #define MESSAGE_QUEUE_MAX ((4 * MESSAGE_SIZE_MAX) / totempg_totem_config->net_mtu)
1090 #endif /* HAVE_SMALL_MEMORY_FOOTPRINT */
1091 
1092 static uint32_t q_level_precent_used(void)
1093 {
1094  return (100 - (((totemsrp_avail(totemsrp_context) - totempg_reserved) * 100) / MESSAGE_QUEUE_MAX));
1095 }
1096 
1098  void **handle_out,
1100  int delete,
1101  int (*callback_fn) (enum totem_callback_token_type type, const void *),
1102  const void *data)
1103 {
1104  unsigned int res;
1105  if (totempg_threaded_mode == 1) {
1106  pthread_mutex_lock (&callback_token_mutex);
1107  }
1108  res = totemsrp_callback_token_create (totemsrp_context, handle_out, type, delete,
1109  callback_fn, data);
1110  if (totempg_threaded_mode == 1) {
1111  pthread_mutex_unlock (&callback_token_mutex);
1112  }
1113  return (res);
1114 }
1115 
1117  void *handle_out)
1118 {
1119  if (totempg_threaded_mode == 1) {
1120  pthread_mutex_lock (&callback_token_mutex);
1121  }
1122  totemsrp_callback_token_destroy (totemsrp_context, handle_out);
1123  if (totempg_threaded_mode == 1) {
1124  pthread_mutex_unlock (&callback_token_mutex);
1125  }
1126 }
1127 
1128 /*
1129  * vi: set autoindent tabstop=4 shiftwidth=4 :
1130  */
1131 
1133  void **totempg_groups_instance,
1134 
1135  void (*deliver_fn) (
1136  unsigned int nodeid,
1137  const void *msg,
1138  unsigned int msg_len,
1139  int endian_conversion_required),
1140 
1141  void (*confchg_fn) (
1142  enum totem_configuration_type configuration_type,
1143  const unsigned int *member_list, size_t member_list_entries,
1144  const unsigned int *left_list, size_t left_list_entries,
1145  const unsigned int *joined_list, size_t joined_list_entries,
1146  const struct memb_ring_id *ring_id))
1147 {
1148  struct totempg_group_instance *instance;
1149 
1150  if (totempg_threaded_mode == 1) {
1151  pthread_mutex_lock (&totempg_mutex);
1152  }
1153 
1154  instance = malloc (sizeof (struct totempg_group_instance));
1155  if (instance == NULL) {
1156  goto error_exit;
1157  }
1158 
1159  instance->deliver_fn = deliver_fn;
1160  instance->confchg_fn = confchg_fn;
1161  instance->groups = 0;
1162  instance->groups_cnt = 0;
1163  instance->q_level = QB_LOOP_MED;
1164  qb_list_init (&instance->list);
1165  qb_list_add (&instance->list, &totempg_groups_list);
1166 
1167  if (totempg_threaded_mode == 1) {
1168  pthread_mutex_unlock (&totempg_mutex);
1169  }
1170  *totempg_groups_instance = instance;
1171  return (0);
1172 
1173 error_exit:
1174  if (totempg_threaded_mode == 1) {
1175  pthread_mutex_unlock (&totempg_mutex);
1176  }
1177  return (-1);
1178 }
1179 
1181  void *totempg_groups_instance,
1182  const struct totempg_group *groups,
1183  size_t group_cnt)
1184 {
1185  struct totempg_group_instance *instance = (struct totempg_group_instance *)totempg_groups_instance;
1186  struct totempg_group *new_groups;
1187  int res = 0;
1188 
1189  if (totempg_threaded_mode == 1) {
1190  pthread_mutex_lock (&totempg_mutex);
1191  }
1192 
1193  new_groups = realloc (instance->groups,
1194  sizeof (struct totempg_group) *
1195  (instance->groups_cnt + group_cnt));
1196  if (new_groups == 0) {
1197  res = -1;
1198  goto error_exit;
1199  }
1200  memcpy (&new_groups[instance->groups_cnt],
1201  groups, group_cnt * sizeof (struct totempg_group));
1202  instance->groups = new_groups;
1203  instance->groups_cnt += group_cnt;
1204 
1205 error_exit:
1206  if (totempg_threaded_mode == 1) {
1207  pthread_mutex_unlock (&totempg_mutex);
1208  }
1209  return (res);
1210 }
1211 
1213  void *totempg_groups_instance,
1214  const struct totempg_group *groups,
1215  size_t group_cnt)
1216 {
1217  if (totempg_threaded_mode == 1) {
1218  pthread_mutex_lock (&totempg_mutex);
1219  }
1220 
1221  if (totempg_threaded_mode == 1) {
1222  pthread_mutex_unlock (&totempg_mutex);
1223  }
1224  return (0);
1225 }
1226 
1227 #define MAX_IOVECS_FROM_APP 32
1228 #define MAX_GROUPS_PER_MSG 32
1229 
1231  void *totempg_groups_instance,
1232  const struct iovec *iovec,
1233  unsigned int iov_len,
1234  int guarantee)
1235 {
1236  struct totempg_group_instance *instance = (struct totempg_group_instance *)totempg_groups_instance;
1237  unsigned short group_len[MAX_GROUPS_PER_MSG + 1];
1238  struct iovec iovec_mcast[MAX_GROUPS_PER_MSG + 1 + MAX_IOVECS_FROM_APP];
1239  int i;
1240  unsigned int res;
1241 
1242  if (totempg_threaded_mode == 1) {
1243  pthread_mutex_lock (&totempg_mutex);
1244  }
1245 
1246  /*
1247  * Build group_len structure and the iovec_mcast structure
1248  */
1249  group_len[0] = instance->groups_cnt;
1250  for (i = 0; i < instance->groups_cnt; i++) {
1251  group_len[i + 1] = instance->groups[i].group_len;
1252  iovec_mcast[i + 1].iov_len = instance->groups[i].group_len;
1253  iovec_mcast[i + 1].iov_base = (void *) instance->groups[i].group;
1254  }
1255  iovec_mcast[0].iov_len = (instance->groups_cnt + 1) * sizeof (unsigned short);
1256  iovec_mcast[0].iov_base = group_len;
1257  for (i = 0; i < iov_len; i++) {
1258  iovec_mcast[i + instance->groups_cnt + 1].iov_len = iovec[i].iov_len;
1259  iovec_mcast[i + instance->groups_cnt + 1].iov_base = iovec[i].iov_base;
1260  }
1261 
1262  res = mcast_msg (iovec_mcast, iov_len + instance->groups_cnt + 1, guarantee);
1263 
1264  if (totempg_threaded_mode == 1) {
1265  pthread_mutex_unlock (&totempg_mutex);
1266  }
1267 
1268  return (res);
1269 }
1270 
1271 static void check_q_level(
1272  void *totempg_groups_instance)
1273 {
1274  struct totempg_group_instance *instance = (struct totempg_group_instance *)totempg_groups_instance;
1275  int32_t old_level = instance->q_level;
1276  int32_t percent_used = q_level_precent_used();
1277 
1278  if (percent_used >= 75 && instance->q_level != TOTEM_Q_LEVEL_CRITICAL) {
1279  instance->q_level = TOTEM_Q_LEVEL_CRITICAL;
1280  } else if (percent_used < 30 && instance->q_level != TOTEM_Q_LEVEL_LOW) {
1281  instance->q_level = TOTEM_Q_LEVEL_LOW;
1282  } else if (percent_used > 40 && percent_used < 50 && instance->q_level != TOTEM_Q_LEVEL_GOOD) {
1283  instance->q_level = TOTEM_Q_LEVEL_GOOD;
1284  } else if (percent_used > 60 && percent_used < 70 && instance->q_level != TOTEM_Q_LEVEL_HIGH) {
1285  instance->q_level = TOTEM_Q_LEVEL_HIGH;
1286  }
1287  if (totem_queue_level_changed && old_level != instance->q_level) {
1288  totem_queue_level_changed(instance->q_level);
1289  }
1290 }
1291 
1293  void *totempg_groups_instance)
1294 {
1295  struct totempg_group_instance *instance = (struct totempg_group_instance *)totempg_groups_instance;
1296 
1297  check_q_level(instance);
1298 }
1299 
1301  void *totempg_groups_instance,
1302  const struct iovec *iovec,
1303  unsigned int iov_len)
1304 {
1305  struct totempg_group_instance *instance = (struct totempg_group_instance *)totempg_groups_instance;
1306  unsigned int size = 0;
1307  unsigned int i;
1308  unsigned int reserved = 0;
1309 
1310  if (totempg_threaded_mode == 1) {
1311  pthread_mutex_lock (&totempg_mutex);
1312  pthread_mutex_lock (&mcast_msg_mutex);
1313  }
1314 
1315  for (i = 0; i < instance->groups_cnt; i++) {
1316  size += instance->groups[i].group_len;
1317  }
1318  for (i = 0; i < iov_len; i++) {
1319  size += iovec[i].iov_len;
1320  }
1321 
1322  if (size >= totempg_size_limit) {
1323  reserved = -1;
1324  goto error_exit;
1325  }
1326 
1327  if (byte_count_send_ok (size)) {
1328  reserved = send_reserve (size);
1329  } else {
1330  reserved = 0;
1331  }
1332 
1333 error_exit:
1334  check_q_level(instance);
1335 
1336  if (totempg_threaded_mode == 1) {
1337  pthread_mutex_unlock (&mcast_msg_mutex);
1338  pthread_mutex_unlock (&totempg_mutex);
1339  }
1340  return (reserved);
1341 }
1342 
1343 
1345 {
1346  if (totempg_threaded_mode == 1) {
1347  pthread_mutex_lock (&totempg_mutex);
1348  pthread_mutex_lock (&mcast_msg_mutex);
1349  }
1350  send_release (msg_count);
1351  if (totempg_threaded_mode == 1) {
1352  pthread_mutex_unlock (&mcast_msg_mutex);
1353  pthread_mutex_unlock (&totempg_mutex);
1354  }
1355  return 0;
1356 }
1357 
1359  void *totempg_groups_instance,
1360  int guarantee,
1361  const struct totempg_group *groups,
1362  size_t groups_cnt,
1363  const struct iovec *iovec,
1364  unsigned int iov_len)
1365 {
1366  unsigned short group_len[MAX_GROUPS_PER_MSG + 1];
1367  struct iovec iovec_mcast[MAX_GROUPS_PER_MSG + 1 + MAX_IOVECS_FROM_APP];
1368  int i;
1369  unsigned int res;
1370 
1371  if (totempg_threaded_mode == 1) {
1372  pthread_mutex_lock (&totempg_mutex);
1373  }
1374 
1375  /*
1376  * Build group_len structure and the iovec_mcast structure
1377  */
1378  group_len[0] = groups_cnt;
1379  for (i = 0; i < groups_cnt; i++) {
1380  group_len[i + 1] = groups[i].group_len;
1381  iovec_mcast[i + 1].iov_len = groups[i].group_len;
1382  iovec_mcast[i + 1].iov_base = (void *) groups[i].group;
1383  }
1384  iovec_mcast[0].iov_len = (groups_cnt + 1) * sizeof (unsigned short);
1385  iovec_mcast[0].iov_base = group_len;
1386  for (i = 0; i < iov_len; i++) {
1387  iovec_mcast[i + groups_cnt + 1].iov_len = iovec[i].iov_len;
1388  iovec_mcast[i + groups_cnt + 1].iov_base = iovec[i].iov_base;
1389  }
1390 
1391  res = mcast_msg (iovec_mcast, iov_len + groups_cnt + 1, guarantee);
1392 
1393  if (totempg_threaded_mode == 1) {
1394  pthread_mutex_unlock (&totempg_mutex);
1395  }
1396  return (res);
1397 }
1398 
1399 /*
1400  * Returns -1 if error, 0 if can't send, 1 if can send the message
1401  */
1403  void *totempg_groups_instance,
1404  const struct totempg_group *groups,
1405  size_t groups_cnt,
1406  const struct iovec *iovec,
1407  unsigned int iov_len)
1408 {
1409  unsigned int size = 0;
1410  unsigned int i;
1411  unsigned int res;
1412 
1413  if (totempg_threaded_mode == 1) {
1414  pthread_mutex_lock (&totempg_mutex);
1415  }
1416 
1417  for (i = 0; i < groups_cnt; i++) {
1418  size += groups[i].group_len;
1419  }
1420  for (i = 0; i < iov_len; i++) {
1421  size += iovec[i].iov_len;
1422  }
1423 
1424  res = msg_count_send_ok (size);
1425 
1426  if (totempg_threaded_mode == 1) {
1427  pthread_mutex_unlock (&totempg_mutex);
1428  }
1429  return (res);
1430 }
1431 
1433  struct totem_ip_address *interface_addr,
1434  unsigned short ip_port,
1435  unsigned int iface_no)
1436 {
1437  int res;
1438 
1439  res = totemsrp_iface_set (
1440  totemsrp_context,
1441  interface_addr,
1442  ip_port,
1443  iface_no);
1444 
1445  return (res);
1446 }
1447 
1449  unsigned int nodeid,
1450  unsigned int *interface_id,
1451  struct totem_ip_address *interfaces,
1452  unsigned int interfaces_size,
1453  char ***status,
1454  unsigned int *iface_count)
1455 {
1456  int res;
1457 
1458  res = totemsrp_ifaces_get (
1459  totemsrp_context,
1460  nodeid,
1461  interface_id,
1462  interfaces,
1463  interfaces_size,
1464  status,
1465  iface_count);
1466 
1467  return (res);
1468 }
1469 
1471 {
1472  totemsrp_event_signal (totemsrp_context, type, value);
1473 }
1474 
1475 void* totempg_get_stats (void)
1476 {
1477  return &totempg_stats;
1478 }
1479 
1481  const char *cipher_type,
1482  const char *hash_type)
1483 {
1484  int res;
1485 
1486  res = totemsrp_crypto_set (totemsrp_context, cipher_type, hash_type);
1487 
1488  return (res);
1489 }
1490 
1491 #define ONE_IFACE_LEN 63
1492 const char *totempg_ifaces_print (unsigned int nodeid)
1493 {
1494  static char iface_string[256 * INTERFACE_MAX];
1495  char one_iface[ONE_IFACE_LEN+1];
1496  struct totem_ip_address interfaces[INTERFACE_MAX];
1497  unsigned int iface_count;
1498  unsigned int iface_ids[INTERFACE_MAX];
1499  unsigned int i;
1500  int res;
1501 
1502  iface_string[0] = '\0';
1503 
1504  res = totempg_ifaces_get (nodeid, iface_ids, interfaces, INTERFACE_MAX, NULL, &iface_count);
1505  if (res == -1) {
1506  return ("no interface found for nodeid");
1507  }
1508 
1509  res = totempg_ifaces_get (nodeid, iface_ids, interfaces, INTERFACE_MAX, NULL, &iface_count);
1510 
1511  for (i = 0; i < iface_count; i++) {
1512  if (!interfaces[i].family) {
1513  continue;
1514  }
1515  snprintf (one_iface, ONE_IFACE_LEN,
1516  "r(%d) ip(%s) ",
1517  i, totemip_print (&interfaces[i]));
1518  strcat (iface_string, one_iface);
1519  }
1520  return (iface_string);
1521 }
1522 
1523 unsigned int totempg_my_nodeid_get (void)
1524 {
1525  return (totemsrp_my_nodeid_get(totemsrp_context));
1526 }
1527 
1529 {
1530  return (totemsrp_my_family_get(totemsrp_context));
1531 }
1533  void (*totem_service_ready) (void))
1534 {
1535  totemsrp_service_ready_register (totemsrp_context, totem_service_ready);
1536 }
1537 
1539 {
1540  totem_queue_level_changed = fn;
1541 }
1542 
1544  const struct totem_ip_address *member,
1545  int ring_no)
1546 {
1547  return totemsrp_member_add (totemsrp_context, member, ring_no);
1548 }
1549 
1551  const struct totem_ip_address *member,
1552  int ring_no)
1553 {
1554  return totemsrp_member_remove (totemsrp_context, member, ring_no);
1555 }
1556 
1557 extern int totempg_reconfigure (void)
1558 {
1559  return totemsrp_reconfigure (totemsrp_context, totempg_totem_config);
1560 }
1561 
1562 extern void totempg_stats_clear (int flags)
1563 {
1564  if (flags & TOTEMPG_STATS_CLEAR_TOTEM) {
1565  totempg_stats.msg_reserved = 0;
1566  totempg_stats.msg_queue_avail = 0;
1567  }
1568  return totemsrp_stats_clear (totemsrp_context, flags);
1569 }
1570 
1572 {
1573  totempg_threaded_mode = 1;
1574  totemsrp_threaded_mode_enable (totemsrp_context);
1575 }
1576 
1578 {
1579  totemsrp_trans_ack (totemsrp_context);
1580 }
1581 
1583 {
1584  totemsrp_force_gather(totemsrp_context);
1585 }
unsigned char last_frag_num
Definition: totempg.c:203
int totemsrp_reconfigure(void *context, struct totem_config *totem_config)
Definition: totemsrp.c:5149
int totempg_initialize(qb_loop_t *poll_handle, struct totem_config *totem_config)
Initialize the totem process groups abstraction.
Definition: totempg.c:802
int totemsrp_iface_set(void *context, const struct totem_ip_address *interface_addr, unsigned short ip_port, unsigned int iface_no)
Definition: totemsrp.c:5026
#define min(a, b)
Definition: exec/util.h:66
#define TOTEMPG_STATS_CLEAR_TOTEM
Definition: totemstats.h:115
Totem Single Ring Protocol.
#define TOTEMPG_NEED_ALIGN
Definition: totempg.c:123
uint32_t value
int totempg_groups_initialize(void **totempg_groups_instance, void(*deliver_fn)(unsigned int nodeid, const void *msg, unsigned int msg_len, int endian_conversion_required), void(*confchg_fn)(enum totem_configuration_type configuration_type, const unsigned int *member_list, size_t member_list_entries, const unsigned int *left_list, size_t left_list_entries, const unsigned int *joined_list, size_t joined_list_entries, const struct memb_ring_id *ring_id))
Initialize a groups instance.
Definition: totempg.c:1132
int totemsrp_my_family_get(void *srp_context)
Definition: totemsrp.c:1112
size_t group_len
Definition: totempg.h:57
void * totempg_get_stats(void)
Definition: totempg.c:1475
The totem_ip_address struct.
Definition: coroapi.h:111
const char * totemip_print(const struct totem_ip_address *addr)
Definition: totemip.c:264
Totem Single Ring Protocol.
void totempg_stats_clear(int flags)
Definition: totempg.c:1562
void(* confchg_fn)(enum totem_configuration_type configuration_type, const unsigned int *member_list, size_t member_list_entries, const unsigned int *left_list, size_t left_list_entries, const unsigned int *joined_list, size_t joined_list_entries, const struct memb_ring_id *ring_id)
Definition: totempg.c:249
int totempg_groups_join(void *totempg_groups_instance, const struct totempg_group *groups, size_t group_cnt)
Definition: totempg.c:1180
unsigned char data[MESSAGE_SIZE_MAX+KNET_MAX_PACKET_SIZE]
Definition: totempg.c:201
void totempg_queue_level_register_callback(totem_queue_level_changed_fn fn)
Definition: totempg.c:1538
#define TOTEMPG_PACKET_SIZE
Definition: totempg.c:153
#define CS_PRI_NODE_ID
Definition: corotypes.h:59
int totemsrp_member_add(void *context, const struct totem_ip_address *member, int iface_no)
Definition: totemsrp.c:5107
unsigned char fragmented
Definition: totempg.c:139
totem_configuration_type
The totem_configuration_type enum.
Definition: coroapi.h:132
unsigned int totemsrp_my_nodeid_get(void *srp_context)
Definition: totemsrp.c:1101
struct totempg_group * groups
Definition: totempg.c:256
void totempg_trans_ack(void)
Definition: totempg.c:1577
void totemsrp_net_mtu_adjust(struct totem_config *totem_config)
Definition: totemsrp.c:5094
int totemsrp_crypto_set(void *srp_context, const char *cipher_type, const char *hash_type)
Definition: totemsrp.c:1087
int totemsrp_member_remove(void *context, const struct totem_ip_address *member, int iface_no)
Definition: totemsrp.c:5120
int guarantee
Definition: totemsrp.c:266
#define log_printf(level, format, args...)
Definition: totempg.c:272
void totemsrp_callback_token_destroy(void *srp_context, void **handle_out)
Definition: totemsrp.c:3491
int totemsrp_callback_token_create(void *srp_context, void **handle_out, enum totem_callback_token_type type, int delete, int(*callback_fn)(enum totem_callback_token_type type, const void *), const void *data)
Definition: totemsrp.c:3456
int totempg_groups_send_ok_groups(void *totempg_groups_instance, const struct totempg_group *groups, size_t groups_cnt, const struct iovec *iovec, unsigned int iov_len)
Definition: totempg.c:1402
void totemsrp_service_ready_register(void *context, void(*totem_service_ready)(void))
Definition: totemsrp.c:5098
int totempg_groups_mcast_groups(void *totempg_groups_instance, int guarantee, const struct totempg_group *groups, size_t groups_cnt, const struct iovec *iovec, unsigned int iov_len)
Definition: totempg.c:1358
void(*) in log_level_security)
Definition: totem.h:106
#define INTERFACE_MAX
Definition: coroapi.h:88
int totemsrp_mcast(void *srp_context, struct iovec *iovec, unsigned int iov_len, int guarantee)
Multicast a message.
Definition: totemsrp.c:2458
#define MAX_GROUPS_PER_MSG
Definition: totempg.c:1228
const char * totempg_ifaces_print(unsigned int nodeid)
Definition: totempg.c:1492
void totempg_threaded_mode_enable(void)
Definition: totempg.c:1571
void(* totem_queue_level_changed_fn)(enum totem_q_level level)
Definition: totempg.h:186
int totempg_groups_leave(void *totempg_groups_instance, const struct totempg_group *groups, size_t group_cnt)
Definition: totempg.c:1212
int totempg_reconfigure(void)
Definition: totempg.c:1557
int totempg_my_family_get(void)
Definition: totempg.c:1528
void(* deliver_fn)(unsigned int nodeid, const void *msg, unsigned int msg_len, int endian_conversion_required)
Definition: totempg.c:243
unsigned char continuation
Definition: totempg.c:140
uint32_t flags
const void * group
Definition: totempg.h:56
void * callback_token_received_handle
Definition: totempg.c:746
void totempg_force_gather(void)
Definition: totempg.c:1582
int totemsrp_avail(void *srp_context)
Return number of available messages that can be queued.
Definition: totemsrp.c:2529
int totempg_groups_mcast_joined(void *totempg_groups_instance, const struct iovec *iovec, unsigned int iov_len, int guarantee)
Definition: totempg.c:1230
void totemsrp_stats_clear(void *context, int flags)
Definition: totemsrp.c:5158
void totempg_check_q_level(void *totempg_groups_instance)
Definition: totempg.c:1292
unsigned int totempg_my_nodeid_get(void)
Definition: totempg.c:1523
typedef __attribute__
uint32_t msg_queue_avail
Definition: totemstats.h:98
totem_event_type
Definition: totem.h:259
void(* log_printf)(int level, int subsys, const char *function_name, const char *file_name, int file_line, const char *format,...) __attribute__((format(printf
Definition: totem.h:99
int totempg_groups_joined_release(int msg_count)
Definition: totempg.c:1344
void totempg_event_signal(enum totem_event_type type, int value)
Definition: totempg.c:1470
struct totem_message_header header
Definition: totemsrp.c:260
unsigned short msg_count
Definition: totempg.c:141
int totempg_iface_set(struct totem_ip_address *interface_addr, unsigned short ip_port, unsigned int iface_no)
Definition: totempg.c:1432
int totempg_member_remove(const struct totem_ip_address *member, int ring_no)
Definition: totempg.c:1550
void totempg_callback_token_destroy(void *handle_out)
Definition: totempg.c:1116
void totemsrp_threaded_mode_enable(void *context)
Definition: totemsrp.c:5133
unsigned int net_mtu
Definition: totem.h:202
struct qb_list_head list
Definition: totempg.c:205
int totemsrp_ifaces_get(void *srp_context, unsigned int nodeid, unsigned int *interface_id, struct totem_ip_address *interfaces, unsigned int interfaces_size, char ***status, unsigned int *iface_count)
Definition: totemsrp.c:1049
int totemsrp_initialize(qb_loop_t *poll_handle, void **srp_context, struct totem_config *totem_config, totempg_stats_t *stats, void(*deliver_fn)(unsigned int nodeid, const void *msg, unsigned int msg_len, int endian_conversion_required), void(*confchg_fn)(enum totem_configuration_type configuration_type, const unsigned int *member_list, size_t member_list_entries, const unsigned int *left_list, size_t left_list_entries, const unsigned int *joined_list, size_t joined_list_entries, const struct memb_ring_id *ring_id), void(*waiting_trans_ack_cb_fn)(int waiting_trans_ack))
Create a protocol instance.
Definition: totemsrp.c:818
void totemsrp_event_signal(void *srp_context, enum totem_event_type type, int value)
Definition: totemsrp.c:2449
throw_away_mode
Definition: totempg.c:194
int totempg_groups_joined_reserve(void *totempg_groups_instance, const struct iovec *iovec, unsigned int iov_len)
Definition: totempg.c:1300
void totemsrp_force_gather(void *context)
Definition: totemsrp.c:5168
#define MESSAGE_QUEUE_MAX
Definition: totempg.c:1089
The memb_ring_id struct.
Definition: coroapi.h:122
#define swab16(x)
The swab16 macro.
Definition: swab.h:39
void totemsrp_finalize(void *srp_context)
Definition: totemsrp.c:1026
unsigned short family
Definition: coroapi.h:76
int totempg_ifaces_get(unsigned int nodeid, unsigned int *interface_id, struct totem_ip_address *interfaces, unsigned int interfaces_size, char ***status, unsigned int *iface_count)
Definition: totempg.c:1448
uint32_t msg_reserved
Definition: totemstats.h:97
void totempg_service_ready_register(void(*totem_service_ready)(void))
Definition: totempg.c:1532
unsigned int nodeid
Definition: totempg.c:200
#define FRAME_SIZE_MAX
Definition: totem.h:52
QB_LIST_DECLARE(assembly_list_inuse)
#define ONE_IFACE_LEN
Definition: totempg.c:1491
struct totem_logging_configuration totem_logging_configuration
Definition: totem.h:200
void totemsrp_trans_ack(void *context)
Definition: totemsrp.c:5140
enum throw_away_mode throw_away_mode
Definition: totempg.c:204
int totempg_crypto_set(const char *cipher_type, const char *hash_type)
Definition: totempg.c:1480
static void(*) struct totem_config totempg_totem_config)
Definition: totempg.c:190
#define MESSAGE_SIZE_MAX
Definition: coroapi.h:97
unsigned int nodeid
Definition: coroapi.h:75
void totempg_finalize(void)
Definition: totempg.c:855
struct memb_ring_id ring_id
Definition: totemsrp.c:264
int totempg_member_add(const struct totem_ip_address *member, int ring_no)
Definition: totempg.c:1543
int index
Definition: totempg.c:202
int totempg_callback_token_create(void **handle_out, enum totem_callback_token_type type, int delete, int(*callback_fn)(enum totem_callback_token_type type, const void *), const void *data)
Definition: totempg.c:1097
struct totempg_mcast_header header
Definition: totempg.c:138
totem_callback_token_type
The totem_callback_token_type enum.
Definition: coroapi.h:142
struct qb_list_head list
Definition: totempg.c:261