4 Copyright (C) Andrew Tridgell 2006
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.
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.
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/>.
21 #include "lib/tdb/include/tdb.h"
22 #include "system/network.h"
23 #include "system/filesys.h"
24 #include "system/wait.h"
25 #include "system/shmem.h"
26 #include "../include/ctdb_private.h"
28 int LogLevel = DEBUG_NOTICE;
29 int this_log_level = 0;
32 return error string for last error
34 const char *ctdb_errstr(struct ctdb_context *ctdb)
41 remember an error message
43 void ctdb_set_error(struct ctdb_context *ctdb, const char *fmt, ...)
46 talloc_free(ctdb->err_msg);
48 ctdb->err_msg = talloc_vasprintf(ctdb, fmt, ap);
49 DEBUG(DEBUG_ERR,("ctdb error: %s\n", ctdb->err_msg));
54 a fatal internal error occurred - no hope for recovery
56 void ctdb_fatal(struct ctdb_context *ctdb, const char *msg)
58 DEBUG(DEBUG_ALERT,("ctdb fatal error: %s\n", msg));
62 /* Invoke an external program to do some sort of tracing on the CTDB
63 * process. This might block for a little while. The external
64 * program is specified by the environment variable
65 * CTDB_EXTERNAL_TRACE. This program should take one argument: the
66 * pid of the process to trace. Commonly, the program would be a
67 * wrapper script around gcore.
69 void ctdb_external_trace(void)
72 const char * t = getenv("CTDB_EXTERNAL_TRACE");
79 cmd = talloc_asprintf(NULL, "%s %lu", t, (unsigned long) getpid());
80 DEBUG(DEBUG_WARNING,("begin external trace: %s\n", cmd));
82 DEBUG(DEBUG_WARNING,("end external trace: %s\n", cmd));
89 int ctdb_parse_address(struct ctdb_context *ctdb,
90 TALLOC_CTX *mem_ctx, const char *str,
91 struct ctdb_address *address)
96 se = getservbyname("ctdb", "tcp");
99 address->address = talloc_strdup(mem_ctx, str);
100 CTDB_NO_MEMORY(ctdb, address->address);
103 address->port = CTDB_PORT;
105 address->port = ntohs(se->s_port);
112 check if two addresses are the same
114 bool ctdb_same_address(struct ctdb_address *a1, struct ctdb_address *a2)
116 return strcmp(a1->address, a2->address) == 0 && a1->port == a2->port;
121 hash function for mapping data to a VNN - taken from tdb
123 uint32_t ctdb_hash(const TDB_DATA *key)
125 return tdb_jenkins_hash(discard_const(key));
129 a type checking varient of idr_find
131 static void *_idr_find_type(struct idr_context *idp, int id, const char *type, const char *location)
133 void *p = idr_find(idp, id);
134 if (p && talloc_check_name(p, type) == NULL) {
135 DEBUG(DEBUG_ERR,("%s idr_find_type expected type %s but got %s\n",
136 location, type, talloc_get_name(p)));
142 uint32_t ctdb_reqid_new(struct ctdb_context *ctdb, void *state)
144 int id = idr_get_new_above(ctdb->idr, state, ctdb->lastid+1, INT_MAX);
146 DEBUG(DEBUG_DEBUG, ("Reqid wrap!\n"));
147 id = idr_get_new(ctdb->idr, state, INT_MAX);
153 void *_ctdb_reqid_find(struct ctdb_context *ctdb, uint32_t reqid, const char *type, const char *location)
157 p = _idr_find_type(ctdb->idr, reqid, type, location);
159 DEBUG(DEBUG_WARNING, ("Could not find idr:%u\n",reqid));
166 void ctdb_reqid_remove(struct ctdb_context *ctdb, uint32_t reqid)
170 ret = idr_remove(ctdb->idr, reqid);
172 DEBUG(DEBUG_ERR, ("Removing idr that does not exist\n"));
178 form a ctdb_rec_data record from a key/data pair
180 note that header may be NULL. If not NULL then it is included in the data portion
183 struct ctdb_rec_data *ctdb_marshall_record(TALLOC_CTX *mem_ctx, uint32_t reqid,
185 struct ctdb_ltdb_header *header,
189 struct ctdb_rec_data *d;
191 length = offsetof(struct ctdb_rec_data, data) + key.dsize +
192 data.dsize + (header?sizeof(*header):0);
193 d = (struct ctdb_rec_data *)talloc_size(mem_ctx, length);
199 d->keylen = key.dsize;
200 memcpy(&d->data[0], key.dptr, key.dsize);
202 d->datalen = data.dsize + sizeof(*header);
203 memcpy(&d->data[key.dsize], header, sizeof(*header));
204 memcpy(&d->data[key.dsize+sizeof(*header)], data.dptr, data.dsize);
206 d->datalen = data.dsize;
207 memcpy(&d->data[key.dsize], data.dptr, data.dsize);
213 /* helper function for marshalling multiple records */
214 struct ctdb_marshall_buffer *ctdb_marshall_add(TALLOC_CTX *mem_ctx,
215 struct ctdb_marshall_buffer *m,
219 struct ctdb_ltdb_header *header,
222 struct ctdb_rec_data *r;
223 size_t m_size, r_size;
224 struct ctdb_marshall_buffer *m2;
226 r = ctdb_marshall_record(mem_ctx, reqid, key, header, data);
233 m = talloc_zero_size(mem_ctx, offsetof(struct ctdb_marshall_buffer, data));
240 m_size = talloc_get_size(m);
241 r_size = talloc_get_size(r);
243 m2 = talloc_realloc_size(mem_ctx, m, m_size + r_size);
249 memcpy(m_size + (uint8_t *)m2, r, r_size);
258 /* we've finished marshalling, return a data blob with the marshalled records */
259 TDB_DATA ctdb_marshall_finish(struct ctdb_marshall_buffer *m)
262 data.dptr = (uint8_t *)m;
263 data.dsize = talloc_get_size(m);
268 loop over a marshalling buffer
270 - pass r==NULL to start
271 - loop the number of times indicated by m->count
273 struct ctdb_rec_data *ctdb_marshall_loop_next(struct ctdb_marshall_buffer *m, struct ctdb_rec_data *r,
275 struct ctdb_ltdb_header *header,
276 TDB_DATA *key, TDB_DATA *data)
279 r = (struct ctdb_rec_data *)&m->data[0];
281 r = (struct ctdb_rec_data *)(r->length + (uint8_t *)r);
289 key->dptr = &r->data[0];
290 key->dsize = r->keylen;
293 data->dptr = &r->data[r->keylen];
294 data->dsize = r->datalen;
295 if (header != NULL) {
296 data->dptr += sizeof(*header);
297 data->dsize -= sizeof(*header);
301 if (header != NULL) {
302 if (r->datalen < sizeof(*header)) {
305 *header = *(struct ctdb_ltdb_header *)&r->data[r->keylen];
317 if possible, make this task real time
319 void ctdb_set_scheduler(struct ctdb_context *ctdb)
321 #if HAVE_SCHED_SETSCHEDULER
322 struct sched_param p;
323 if (ctdb->saved_scheduler_param == NULL) {
324 ctdb->saved_scheduler_param = talloc_size(ctdb, sizeof(p));
327 if (sched_getparam(0, (struct sched_param *)ctdb->saved_scheduler_param) == -1) {
328 DEBUG(DEBUG_ERR,("Unable to get old scheduler params\n"));
332 p = *(struct sched_param *)ctdb->saved_scheduler_param;
333 p.sched_priority = 1;
335 if (sched_setscheduler(0, SCHED_FIFO, &p) == -1) {
336 DEBUG(DEBUG_CRIT,("Unable to set scheduler to SCHED_FIFO (%s)\n",
339 DEBUG(DEBUG_NOTICE,("Set scheduler to SCHED_FIFO\n"));
345 restore previous scheduler parameters
347 void ctdb_restore_scheduler(struct ctdb_context *ctdb)
349 #if HAVE_SCHED_SETSCHEDULER
350 if (ctdb->saved_scheduler_param == NULL) {
351 ctdb_fatal(ctdb, "No saved scheduler parameters\n");
353 if (sched_setscheduler(0, SCHED_OTHER, (struct sched_param *)ctdb->saved_scheduler_param) == -1) {
354 ctdb_fatal(ctdb, "Unable to restore old scheduler parameters\n");
359 void set_nonblocking(int fd)
362 v = fcntl(fd, F_GETFL, 0);
363 fcntl(fd, F_SETFL, v | O_NONBLOCK);
366 void set_close_on_exec(int fd)
369 v = fcntl(fd, F_GETFD, 0);
370 fcntl(fd, F_SETFD, v | FD_CLOEXEC);
374 bool parse_ipv4(const char *s, unsigned port, struct sockaddr_in *sin)
376 sin->sin_family = AF_INET;
377 sin->sin_port = htons(port);
379 if (inet_pton(AF_INET, s, &sin->sin_addr) != 1) {
380 DEBUG(DEBUG_ERR, (__location__ " Failed to translate %s into sin_addr\n", s));
387 static bool parse_ipv6(const char *s, const char *ifaces, unsigned port, ctdb_sock_addr *saddr)
389 saddr->ip6.sin6_family = AF_INET6;
390 saddr->ip6.sin6_port = htons(port);
391 saddr->ip6.sin6_flowinfo = 0;
392 saddr->ip6.sin6_scope_id = 0;
394 if (inet_pton(AF_INET6, s, &saddr->ip6.sin6_addr) != 1) {
395 DEBUG(DEBUG_ERR, (__location__ " Failed to translate %s into sin6_addr\n", s));
399 if (ifaces && IN6_IS_ADDR_LINKLOCAL(&saddr->ip6.sin6_addr)) {
400 if (strchr(ifaces, ',')) {
401 DEBUG(DEBUG_ERR, (__location__ " Link local address %s "
402 "is specified for multiple ifaces %s\n",
406 saddr->ip6.sin6_scope_id = if_nametoindex(ifaces);
414 bool parse_ip_port(const char *addr, ctdb_sock_addr *saddr)
416 TALLOC_CTX *tmp_ctx = talloc_new(NULL);
422 s = talloc_strdup(tmp_ctx, addr);
424 DEBUG(DEBUG_ERR, (__location__ " Failed strdup()\n"));
425 talloc_free(tmp_ctx);
431 DEBUG(DEBUG_ERR, (__location__ " This addr: %s does not contain a port number\n", s));
432 talloc_free(tmp_ctx);
436 port = strtoul(p+1, &endp, 10);
437 if (endp == NULL || *endp != 0) {
438 /* trailing garbage */
439 DEBUG(DEBUG_ERR, (__location__ " Trailing garbage after the port in %s\n", s));
440 talloc_free(tmp_ctx);
446 /* now is this a ipv4 or ipv6 address ?*/
447 ret = parse_ip(s, NULL, port, saddr);
449 talloc_free(tmp_ctx);
456 bool parse_ip(const char *addr, const char *ifaces, unsigned port, ctdb_sock_addr *saddr)
461 /* now is this a ipv4 or ipv6 address ?*/
462 p = index(addr, ':');
464 ret = parse_ipv4(addr, port, &saddr->ip);
466 ret = parse_ipv6(addr, ifaces, port, saddr);
475 bool parse_ip_mask(const char *str, const char *ifaces, ctdb_sock_addr *addr, unsigned *mask)
477 TALLOC_CTX *tmp_ctx = talloc_new(NULL);
483 s = talloc_strdup(tmp_ctx, str);
485 DEBUG(DEBUG_ERR, (__location__ " Failed strdup()\n"));
486 talloc_free(tmp_ctx);
492 DEBUG(DEBUG_ERR, (__location__ " This addr: %s does not contain a mask\n", s));
493 talloc_free(tmp_ctx);
497 *mask = strtoul(p+1, &endp, 10);
498 if (endp == NULL || *endp != 0) {
499 /* trailing garbage */
500 DEBUG(DEBUG_ERR, (__location__ " Trailing garbage after the mask in %s\n", s));
501 talloc_free(tmp_ctx);
507 /* now is this a ipv4 or ipv6 address ?*/
508 ret = parse_ip(s, ifaces, 0, addr);
510 talloc_free(tmp_ctx);
515 This is used to canonicalize a ctdb_sock_addr structure.
517 void ctdb_canonicalize_ip(const ctdb_sock_addr *ip, ctdb_sock_addr *cip)
519 char prefix[12] = { 0,0,0,0,0,0,0,0,0,0,0xff,0xff };
521 memcpy(cip, ip, sizeof (*cip));
523 if ( (ip->sa.sa_family == AF_INET6)
524 && !memcmp(&ip->ip6.sin6_addr, prefix, 12)) {
525 memset(cip, 0, sizeof(*cip));
526 #ifdef HAVE_SOCK_SIN_LEN
527 cip->ip.sin_len = sizeof(*cip);
529 cip->ip.sin_family = AF_INET;
530 cip->ip.sin_port = ip->ip6.sin6_port;
531 memcpy(&cip->ip.sin_addr, &ip->ip6.sin6_addr.s6_addr[12], 4);
535 bool ctdb_same_ip(const ctdb_sock_addr *tip1, const ctdb_sock_addr *tip2)
537 ctdb_sock_addr ip1, ip2;
539 ctdb_canonicalize_ip(tip1, &ip1);
540 ctdb_canonicalize_ip(tip2, &ip2);
542 if (ip1.sa.sa_family != ip2.sa.sa_family) {
546 switch (ip1.sa.sa_family) {
548 return ip1.ip.sin_addr.s_addr == ip2.ip.sin_addr.s_addr;
550 return !memcmp(&ip1.ip6.sin6_addr.s6_addr[0],
551 &ip2.ip6.sin6_addr.s6_addr[0],
554 DEBUG(DEBUG_ERR, (__location__ " CRITICAL Can not compare sockaddr structures of type %u\n", ip1.sa.sa_family));
562 compare two ctdb_sock_addr structures
564 bool ctdb_same_sockaddr(const ctdb_sock_addr *ip1, const ctdb_sock_addr *ip2)
566 return ctdb_same_ip(ip1, ip2) && ip1->ip.sin_port == ip2->ip.sin_port;
569 char *ctdb_addr_to_str(ctdb_sock_addr *addr)
571 static char cip[128] = "";
573 switch (addr->sa.sa_family) {
575 inet_ntop(addr->ip.sin_family, &addr->ip.sin_addr, cip, sizeof(cip));
578 inet_ntop(addr->ip6.sin6_family, &addr->ip6.sin6_addr, cip, sizeof(cip));
581 DEBUG(DEBUG_ERR, (__location__ " ERROR, unknown family %u\n", addr->sa.sa_family));
582 ctdb_external_trace();
588 unsigned ctdb_addr_to_port(ctdb_sock_addr *addr)
590 switch (addr->sa.sa_family) {
592 return ntohs(addr->ip.sin_port);
595 return ntohs(addr->ip6.sin6_port);
598 DEBUG(DEBUG_ERR, (__location__ " ERROR, unknown family %u\n", addr->sa.sa_family));
604 void ctdb_block_signal(int signum)
608 sigaddset(&set,signum);
609 sigprocmask(SIG_BLOCK,&set,NULL);
612 void ctdb_unblock_signal(int signum)
616 sigaddset(&set,signum);
617 sigprocmask(SIG_UNBLOCK,&set,NULL);
620 struct debug_levels debug_levels[] = {
621 {DEBUG_EMERG, "EMERG"},
622 {DEBUG_ALERT, "ALERT"},
623 {DEBUG_CRIT, "CRIT"},
625 {DEBUG_WARNING, "WARNING"},
626 {DEBUG_NOTICE, "NOTICE"},
627 {DEBUG_INFO, "INFO"},
628 {DEBUG_DEBUG, "DEBUG"},
632 const char *get_debug_by_level(int32_t level)
636 for (i=0; debug_levels[i].description != NULL; i++) {
637 if (debug_levels[i].level == level) {
638 return debug_levels[i].description;
644 int32_t get_debug_by_desc(const char *desc)
648 for (i=0; debug_levels[i].description != NULL; i++) {
649 if (!strcmp(debug_levels[i].description, desc)) {
650 return debug_levels[i].level;
657 /* we don't lock future pages here; it would increase the chance that
658 * we'd fail to mmap later on. */
659 void ctdb_lockdown_memory(struct ctdb_context *ctdb)
662 /* Extra stack, please! */
664 memset(dummy, 0, sizeof(dummy));
666 if (ctdb->valgrinding) {
670 /* TODO: Add a command line option to disable memory lockdown.
671 * This can be a performance issue on AIX since fork() copies
672 * all locked memory pages.
675 /* Ignore when running in local daemons mode */
680 /* Avoid compiler optimizing out dummy. */
681 mlock(dummy, sizeof(dummy));
682 if (mlockall(MCL_CURRENT) != 0) {
683 DEBUG(DEBUG_WARNING,("Failed to lockdown memory: %s'\n",
689 const char *ctdb_eventscript_call_names[] = {
706 /* Runstate handling */
708 enum ctdb_runstate runstate;
711 { CTDB_RUNSTATE_UNKNOWN, "UNKNOWN" },
712 { CTDB_RUNSTATE_INIT, "INIT" },
713 { CTDB_RUNSTATE_SETUP, "SETUP" },
714 { CTDB_RUNSTATE_FIRST_RECOVERY, "FIRST_RECOVERY" },
715 { CTDB_RUNSTATE_STARTUP, "STARTUP" },
716 { CTDB_RUNSTATE_RUNNING, "RUNNING" },
717 { CTDB_RUNSTATE_SHUTDOWN, "SHUTDOWN" },
721 const char *runstate_to_string(enum ctdb_runstate runstate)
724 for (i=0; runstate_map[i].label != NULL ; i++) {
725 if (runstate_map[i].runstate == runstate) {
726 return runstate_map[i].label;
730 return runstate_map[0].label;
733 enum ctdb_runstate runstate_from_string(const char *label)
736 for (i=0; runstate_map[i].label != NULL; i++) {
737 if (strcasecmp(runstate_map[i].label, label) == 0) {
738 return runstate_map[i].runstate;
742 return CTDB_RUNSTATE_UNKNOWN;
745 void ctdb_set_runstate(struct ctdb_context *ctdb, enum ctdb_runstate runstate)
747 if (runstate <= ctdb->runstate) {
748 ctdb_fatal(ctdb, "runstate must always increase");
751 DEBUG(DEBUG_NOTICE,("Set runstate to %s (%d)\n",
752 runstate_to_string(runstate), runstate));
753 ctdb->runstate = runstate;