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/tevent/tevent.h"
22 #include "lib/tdb/include/tdb.h"
23 #include "system/network.h"
24 #include "system/filesys.h"
25 #include "system/wait.h"
26 #include "system/shmem.h"
27 #include "../include/ctdb_private.h"
29 int LogLevel = DEBUG_NOTICE;
30 int this_log_level = 0;
33 return error string for last error
35 const char *ctdb_errstr(struct ctdb_context *ctdb)
42 remember an error message
44 void ctdb_set_error(struct ctdb_context *ctdb, const char *fmt, ...)
47 talloc_free(ctdb->err_msg);
49 ctdb->err_msg = talloc_vasprintf(ctdb, fmt, ap);
50 DEBUG(DEBUG_ERR,("ctdb error: %s\n", ctdb->err_msg));
55 a fatal internal error occurred - no hope for recovery
57 void ctdb_fatal(struct ctdb_context *ctdb, const char *msg)
59 DEBUG(DEBUG_ALERT,("ctdb fatal error: %s\n", msg));
66 int ctdb_parse_address(struct ctdb_context *ctdb,
67 TALLOC_CTX *mem_ctx, const char *str,
68 struct ctdb_address *address)
73 se = getservbyname("ctdb", "tcp");
76 address->address = talloc_strdup(mem_ctx, str);
77 CTDB_NO_MEMORY(ctdb, address->address);
80 address->port = CTDB_PORT;
82 address->port = ntohs(se->s_port);
89 check if two addresses are the same
91 bool ctdb_same_address(struct ctdb_address *a1, struct ctdb_address *a2)
93 return strcmp(a1->address, a2->address) == 0 && a1->port == a2->port;
98 hash function for mapping data to a VNN - taken from tdb
100 uint32_t ctdb_hash(const TDB_DATA *key)
102 return tdb_jenkins_hash(discard_const(key));
106 a type checking varient of idr_find
108 static void *_idr_find_type(struct idr_context *idp, int id, const char *type, const char *location)
110 void *p = idr_find(idp, id);
111 if (p && talloc_check_name(p, type) == NULL) {
112 DEBUG(DEBUG_ERR,("%s idr_find_type expected type %s but got %s\n",
113 location, type, talloc_get_name(p)));
119 uint32_t ctdb_reqid_new(struct ctdb_context *ctdb, void *state)
121 int id = idr_get_new_above(ctdb->idr, state, ctdb->lastid+1, INT_MAX);
123 DEBUG(DEBUG_DEBUG, ("Reqid wrap!\n"));
124 id = idr_get_new(ctdb->idr, state, INT_MAX);
130 void *_ctdb_reqid_find(struct ctdb_context *ctdb, uint32_t reqid, const char *type, const char *location)
134 p = _idr_find_type(ctdb->idr, reqid, type, location);
136 DEBUG(DEBUG_WARNING, ("Could not find idr:%u\n",reqid));
143 void ctdb_reqid_remove(struct ctdb_context *ctdb, uint32_t reqid)
147 ret = idr_remove(ctdb->idr, reqid);
149 DEBUG(DEBUG_ERR, ("Removing idr that does not exist\n"));
155 form a ctdb_rec_data record from a key/data pair
157 note that header may be NULL. If not NULL then it is included in the data portion
160 struct ctdb_rec_data *ctdb_marshall_record(TALLOC_CTX *mem_ctx, uint32_t reqid,
162 struct ctdb_ltdb_header *header,
166 struct ctdb_rec_data *d;
168 length = offsetof(struct ctdb_rec_data, data) + key.dsize +
169 data.dsize + (header?sizeof(*header):0);
170 d = (struct ctdb_rec_data *)talloc_size(mem_ctx, length);
176 d->keylen = key.dsize;
177 memcpy(&d->data[0], key.dptr, key.dsize);
179 d->datalen = data.dsize + sizeof(*header);
180 memcpy(&d->data[key.dsize], header, sizeof(*header));
181 memcpy(&d->data[key.dsize+sizeof(*header)], data.dptr, data.dsize);
183 d->datalen = data.dsize;
184 memcpy(&d->data[key.dsize], data.dptr, data.dsize);
190 /* helper function for marshalling multiple records */
191 struct ctdb_marshall_buffer *ctdb_marshall_add(TALLOC_CTX *mem_ctx,
192 struct ctdb_marshall_buffer *m,
196 struct ctdb_ltdb_header *header,
199 struct ctdb_rec_data *r;
200 size_t m_size, r_size;
201 struct ctdb_marshall_buffer *m2;
203 r = ctdb_marshall_record(mem_ctx, reqid, key, header, data);
210 m = talloc_zero_size(mem_ctx, offsetof(struct ctdb_marshall_buffer, data));
217 m_size = talloc_get_size(m);
218 r_size = talloc_get_size(r);
220 m2 = talloc_realloc_size(mem_ctx, m, m_size + r_size);
226 memcpy(m_size + (uint8_t *)m2, r, r_size);
235 /* we've finished marshalling, return a data blob with the marshalled records */
236 TDB_DATA ctdb_marshall_finish(struct ctdb_marshall_buffer *m)
239 data.dptr = (uint8_t *)m;
240 data.dsize = talloc_get_size(m);
245 loop over a marshalling buffer
247 - pass r==NULL to start
248 - loop the number of times indicated by m->count
250 struct ctdb_rec_data *ctdb_marshall_loop_next(struct ctdb_marshall_buffer *m, struct ctdb_rec_data *r,
252 struct ctdb_ltdb_header *header,
253 TDB_DATA *key, TDB_DATA *data)
256 r = (struct ctdb_rec_data *)&m->data[0];
258 r = (struct ctdb_rec_data *)(r->length + (uint8_t *)r);
266 key->dptr = &r->data[0];
267 key->dsize = r->keylen;
270 data->dptr = &r->data[r->keylen];
271 data->dsize = r->datalen;
272 if (header != NULL) {
273 data->dptr += sizeof(*header);
274 data->dsize -= sizeof(*header);
278 if (header != NULL) {
279 if (r->datalen < sizeof(*header)) {
282 *header = *(struct ctdb_ltdb_header *)&r->data[r->keylen];
294 if possible, make this task real time
296 void ctdb_set_scheduler(struct ctdb_context *ctdb)
298 #if HAVE_SCHED_SETSCHEDULER
299 struct sched_param p;
300 if (ctdb->saved_scheduler_param == NULL) {
301 ctdb->saved_scheduler_param = talloc_size(ctdb, sizeof(p));
304 if (sched_getparam(0, (struct sched_param *)ctdb->saved_scheduler_param) == -1) {
305 DEBUG(DEBUG_ERR,("Unable to get old scheduler params\n"));
309 p = *(struct sched_param *)ctdb->saved_scheduler_param;
310 p.sched_priority = 1;
312 if (sched_setscheduler(0, SCHED_FIFO, &p) == -1) {
313 DEBUG(DEBUG_CRIT,("Unable to set scheduler to SCHED_FIFO (%s)\n",
316 DEBUG(DEBUG_NOTICE,("Set scheduler to SCHED_FIFO\n"));
322 restore previous scheduler parameters
324 void ctdb_restore_scheduler(struct ctdb_context *ctdb)
326 #if HAVE_SCHED_SETSCHEDULER
327 if (ctdb->saved_scheduler_param == NULL) {
328 ctdb_fatal(ctdb, "No saved scheduler parameters\n");
330 if (sched_setscheduler(0, SCHED_OTHER, (struct sched_param *)ctdb->saved_scheduler_param) == -1) {
331 ctdb_fatal(ctdb, "Unable to restore old scheduler parameters\n");
337 * This function forks a child process and drops the realtime
338 * scheduler for the child process.
340 pid_t ctdb_fork(struct ctdb_context *ctdb)
346 if (ctdb->do_setsched) {
347 ctdb_restore_scheduler(ctdb);
353 void set_nonblocking(int fd)
356 v = fcntl(fd, F_GETFL, 0);
357 fcntl(fd, F_SETFL, v | O_NONBLOCK);
360 void set_close_on_exec(int fd)
363 v = fcntl(fd, F_GETFD, 0);
364 fcntl(fd, F_SETFD, v | FD_CLOEXEC);
368 bool parse_ipv4(const char *s, unsigned port, struct sockaddr_in *sin)
370 sin->sin_family = AF_INET;
371 sin->sin_port = htons(port);
373 if (inet_pton(AF_INET, s, &sin->sin_addr) != 1) {
374 DEBUG(DEBUG_ERR, (__location__ " Failed to translate %s into sin_addr\n", s));
381 static bool parse_ipv6(const char *s, const char *ifaces, unsigned port, ctdb_sock_addr *saddr)
383 saddr->ip6.sin6_family = AF_INET6;
384 saddr->ip6.sin6_port = htons(port);
385 saddr->ip6.sin6_flowinfo = 0;
386 saddr->ip6.sin6_scope_id = 0;
388 if (inet_pton(AF_INET6, s, &saddr->ip6.sin6_addr) != 1) {
389 DEBUG(DEBUG_ERR, (__location__ " Failed to translate %s into sin6_addr\n", s));
393 if (ifaces && IN6_IS_ADDR_LINKLOCAL(&saddr->ip6.sin6_addr)) {
394 if (strchr(ifaces, ',')) {
395 DEBUG(DEBUG_ERR, (__location__ " Link local address %s "
396 "is specified for multiple ifaces %s\n",
400 saddr->ip6.sin6_scope_id = if_nametoindex(ifaces);
408 bool parse_ip_port(const char *addr, ctdb_sock_addr *saddr)
410 TALLOC_CTX *tmp_ctx = talloc_new(NULL);
416 s = talloc_strdup(tmp_ctx, addr);
418 DEBUG(DEBUG_ERR, (__location__ " Failed strdup()\n"));
419 talloc_free(tmp_ctx);
425 DEBUG(DEBUG_ERR, (__location__ " This addr: %s does not contain a port number\n", s));
426 talloc_free(tmp_ctx);
430 port = strtoul(p+1, &endp, 10);
431 if (endp == NULL || *endp != 0) {
432 /* trailing garbage */
433 DEBUG(DEBUG_ERR, (__location__ " Trailing garbage after the port in %s\n", s));
434 talloc_free(tmp_ctx);
440 /* now is this a ipv4 or ipv6 address ?*/
441 ret = parse_ip(s, NULL, port, saddr);
443 talloc_free(tmp_ctx);
450 bool parse_ip(const char *addr, const char *ifaces, unsigned port, ctdb_sock_addr *saddr)
455 /* now is this a ipv4 or ipv6 address ?*/
456 p = index(addr, ':');
458 ret = parse_ipv4(addr, port, &saddr->ip);
460 ret = parse_ipv6(addr, ifaces, port, saddr);
469 bool parse_ip_mask(const char *str, const char *ifaces, ctdb_sock_addr *addr, unsigned *mask)
471 TALLOC_CTX *tmp_ctx = talloc_new(NULL);
477 s = talloc_strdup(tmp_ctx, str);
479 DEBUG(DEBUG_ERR, (__location__ " Failed strdup()\n"));
480 talloc_free(tmp_ctx);
486 DEBUG(DEBUG_ERR, (__location__ " This addr: %s does not contain a mask\n", s));
487 talloc_free(tmp_ctx);
491 *mask = strtoul(p+1, &endp, 10);
492 if (endp == NULL || *endp != 0) {
493 /* trailing garbage */
494 DEBUG(DEBUG_ERR, (__location__ " Trailing garbage after the mask in %s\n", s));
495 talloc_free(tmp_ctx);
501 /* now is this a ipv4 or ipv6 address ?*/
502 ret = parse_ip(s, ifaces, 0, addr);
504 talloc_free(tmp_ctx);
509 This is used to canonicalize a ctdb_sock_addr structure.
511 void ctdb_canonicalize_ip(const ctdb_sock_addr *ip, ctdb_sock_addr *cip)
513 char prefix[12] = { 0,0,0,0,0,0,0,0,0,0,0xff,0xff };
515 memcpy(cip, ip, sizeof (*cip));
517 if ( (ip->sa.sa_family == AF_INET6)
518 && !memcmp(&ip->ip6.sin6_addr, prefix, 12)) {
519 memset(cip, 0, sizeof(*cip));
520 #ifdef HAVE_SOCK_SIN_LEN
521 cip->ip.sin_len = sizeof(*cip);
523 cip->ip.sin_family = AF_INET;
524 cip->ip.sin_port = ip->ip6.sin6_port;
525 memcpy(&cip->ip.sin_addr, &ip->ip6.sin6_addr.s6_addr32[3], 4);
529 bool ctdb_same_ip(const ctdb_sock_addr *tip1, const ctdb_sock_addr *tip2)
531 ctdb_sock_addr ip1, ip2;
533 ctdb_canonicalize_ip(tip1, &ip1);
534 ctdb_canonicalize_ip(tip2, &ip2);
536 if (ip1.sa.sa_family != ip2.sa.sa_family) {
540 switch (ip1.sa.sa_family) {
542 return ip1.ip.sin_addr.s_addr == ip2.ip.sin_addr.s_addr;
544 return !memcmp(&ip1.ip6.sin6_addr.s6_addr[0],
545 &ip2.ip6.sin6_addr.s6_addr[0],
548 DEBUG(DEBUG_ERR, (__location__ " CRITICAL Can not compare sockaddr structures of type %u\n", ip1.sa.sa_family));
556 compare two ctdb_sock_addr structures
558 bool ctdb_same_sockaddr(const ctdb_sock_addr *ip1, const ctdb_sock_addr *ip2)
560 return ctdb_same_ip(ip1, ip2) && ip1->ip.sin_port == ip2->ip.sin_port;
563 char *ctdb_addr_to_str(ctdb_sock_addr *addr)
565 static char cip[128] = "";
567 switch (addr->sa.sa_family) {
569 inet_ntop(addr->ip.sin_family, &addr->ip.sin_addr, cip, sizeof(cip));
572 inet_ntop(addr->ip6.sin6_family, &addr->ip6.sin6_addr, cip, sizeof(cip));
575 DEBUG(DEBUG_ERR, (__location__ " ERROR, unknown family %u\n", addr->sa.sa_family));
581 unsigned ctdb_addr_to_port(ctdb_sock_addr *addr)
583 switch (addr->sa.sa_family) {
585 return ntohs(addr->ip.sin_port);
588 return ntohs(addr->ip6.sin6_port);
591 DEBUG(DEBUG_ERR, (__location__ " ERROR, unknown family %u\n", addr->sa.sa_family));
597 void ctdb_block_signal(int signum)
601 sigaddset(&set,signum);
602 sigprocmask(SIG_BLOCK,&set,NULL);
605 void ctdb_unblock_signal(int signum)
609 sigaddset(&set,signum);
610 sigprocmask(SIG_UNBLOCK,&set,NULL);
613 struct debug_levels debug_levels[] = {
614 {DEBUG_EMERG, "EMERG"},
615 {DEBUG_ALERT, "ALERT"},
616 {DEBUG_CRIT, "CRIT"},
618 {DEBUG_WARNING, "WARNING"},
619 {DEBUG_NOTICE, "NOTICE"},
620 {DEBUG_INFO, "INFO"},
621 {DEBUG_DEBUG, "DEBUG"},
625 const char *get_debug_by_level(int32_t level)
629 for (i=0; debug_levels[i].description != NULL; i++) {
630 if (debug_levels[i].level == level) {
631 return debug_levels[i].description;
637 int32_t get_debug_by_desc(const char *desc)
641 for (i=0; debug_levels[i].description != NULL; i++) {
642 if (!strcmp(debug_levels[i].description, desc)) {
643 return debug_levels[i].level;
650 /* we don't lock future pages here; it would increase the chance that
651 * we'd fail to mmap later on. */
652 void ctdb_lockdown_memory(struct ctdb_context *ctdb)
655 /* Extra stack, please! */
657 memset(dummy, 0, sizeof(dummy));
659 if (ctdb->valgrinding) {
663 /* Avoid compiler optimizing out dummy. */
664 mlock(dummy, sizeof(dummy));
665 if (mlockall(MCL_CURRENT) != 0) {
666 DEBUG(DEBUG_WARNING,("Failed to lock memory: %s'\n",
672 const char *ctdb_eventscript_call_names[] = {