ctdb-daemon: Fix IP address comparisons for IPv6 addresses
[obnox/samba/samba-obnox.git] / ctdb / common / ctdb_util.c
index 5584c1732e548e2d5c516444dc13793565d53b58..137e0a8a095d2146fd383f45729b0770aa2afdc0 100644 (file)
 */
 
 #include "includes.h"
-#include "lib/tdb/include/tdb.h"
+#include "tdb.h"
 #include "system/network.h"
 #include "system/filesys.h"
 #include "system/wait.h"
-#include "system/shmem.h"
 #include "../include/ctdb_private.h"
 
-int LogLevel = DEBUG_NOTICE;
-int this_log_level = 0;
-
 /*
   return error string for last error
 */
@@ -59,6 +55,43 @@ void ctdb_fatal(struct ctdb_context *ctdb, const char *msg)
        abort();
 }
 
+/*
+  like ctdb_fatal() but a core/backtrace would not be useful
+*/
+void ctdb_die(struct ctdb_context *ctdb, const char *msg)
+{
+       DEBUG(DEBUG_ALERT,("ctdb exiting with error: %s\n", msg));
+       exit(1);
+}
+
+/* Invoke an external program to do some sort of tracing on the CTDB
+ * process.  This might block for a little while.  The external
+ * program is specified by the environment variable
+ * CTDB_EXTERNAL_TRACE.  This program should take one argument: the
+ * pid of the process to trace.  Commonly, the program would be a
+ * wrapper script around gcore.
+ */
+void ctdb_external_trace(void)
+{
+       int ret;
+       const char * t = getenv("CTDB_EXTERNAL_TRACE");
+       char * cmd;
+
+       if (t == NULL) {
+               return;
+       }
+
+       cmd = talloc_asprintf(NULL, "%s %lu", t, (unsigned long) getpid());
+       DEBUG(DEBUG_WARNING,("begin external trace: %s\n", cmd));
+       ret = system(cmd);
+       if (ret == -1) {
+               DEBUG(DEBUG_ERR,
+                     ("external trace command \"%s\" failed\n", cmd));
+       }
+       DEBUG(DEBUG_WARNING,("end external trace: %s\n", cmd));
+       talloc_free(cmd);
+}
+
 /*
   parse a IP:port pair
 */
@@ -67,12 +100,20 @@ int ctdb_parse_address(struct ctdb_context *ctdb,
                       struct ctdb_address *address)
 {
        struct servent *se;
+       ctdb_sock_addr addr;
 
        setservent(0);
        se = getservbyname("ctdb", "tcp");
        endservent();
-       
-       address->address = talloc_strdup(mem_ctx, str);
+
+       /* Parse IP address and re-convert to string.  This ensure correct
+        * string form for IPv6 addresses.
+        */
+       if (! parse_ip(str, NULL, 0, &addr)) {
+               return -1;
+       }
+
+       address->address = talloc_strdup(mem_ctx, ctdb_addr_to_str(&addr));
        CTDB_NO_MEMORY(ctdb, address->address);
 
        if (se == NULL) {
@@ -150,44 +191,67 @@ void ctdb_reqid_remove(struct ctdb_context *ctdb, uint32_t reqid)
 }
 
 
+static uint32_t ctdb_marshall_record_size(TDB_DATA key,
+                                         struct ctdb_ltdb_header *header,
+                                         TDB_DATA data)
+{
+       return offsetof(struct ctdb_rec_data, data) + key.dsize +
+              data.dsize + (header ? sizeof(*header) : 0);
+}
+
+static void ctdb_marshall_record_copy(struct ctdb_rec_data *rec,
+                                     uint32_t reqid,
+                                     TDB_DATA key,
+                                     struct ctdb_ltdb_header *header,
+                                     TDB_DATA data,
+                                     uint32_t length)
+{
+       uint32_t offset;
+
+       rec->length = length;
+       rec->reqid = reqid;
+       rec->keylen = key.dsize;
+       memcpy(&rec->data[0], key.dptr, key.dsize);
+       offset = key.dsize;
+
+       if (header) {
+               rec->datalen = data.dsize + sizeof(*header);
+               memcpy(&rec->data[offset], header, sizeof(*header));
+               offset += sizeof(*header);
+       } else {
+               rec->datalen = data.dsize;
+       }
+       memcpy(&rec->data[offset], data.dptr, data.dsize);
+}
+
 /*
   form a ctdb_rec_data record from a key/data pair
   
   note that header may be NULL. If not NULL then it is included in the data portion
   of the record
  */
-struct ctdb_rec_data *ctdb_marshall_record(TALLOC_CTX *mem_ctx, uint32_t reqid,        
-                                          TDB_DATA key, 
+struct ctdb_rec_data *ctdb_marshall_record(TALLOC_CTX *mem_ctx, uint32_t reqid,
+                                          TDB_DATA key,
                                           struct ctdb_ltdb_header *header,
                                           TDB_DATA data)
 {
        size_t length;
        struct ctdb_rec_data *d;
 
-       length = offsetof(struct ctdb_rec_data, data) + key.dsize + 
-               data.dsize + (header?sizeof(*header):0);
+       length = ctdb_marshall_record_size(key, header, data);
+
        d = (struct ctdb_rec_data *)talloc_size(mem_ctx, length);
        if (d == NULL) {
                return NULL;
        }
-       d->length = length;
-       d->reqid = reqid;
-       d->keylen = key.dsize;
-       memcpy(&d->data[0], key.dptr, key.dsize);
-       if (header) {
-               d->datalen = data.dsize + sizeof(*header);
-               memcpy(&d->data[key.dsize], header, sizeof(*header));
-               memcpy(&d->data[key.dsize+sizeof(*header)], data.dptr, data.dsize);
-       } else {
-               d->datalen = data.dsize;
-               memcpy(&d->data[key.dsize], data.dptr, data.dsize);
-       }
+
+       ctdb_marshall_record_copy(d, reqid, key, header, data, length);
        return d;
 }
 
 
 /* helper function for marshalling multiple records */
-struct ctdb_marshall_buffer *ctdb_marshall_add(TALLOC_CTX *mem_ctx, 
+struct ctdb_marshall_buffer *ctdb_marshall_add(TALLOC_CTX *mem_ctx,
                                               struct ctdb_marshall_buffer *m,
                                               uint64_t db_id,
                                               uint32_t reqid,
@@ -196,36 +260,29 @@ struct ctdb_marshall_buffer *ctdb_marshall_add(TALLOC_CTX *mem_ctx,
                                               TDB_DATA data)
 {
        struct ctdb_rec_data *r;
-       size_t m_size, r_size;
        struct ctdb_marshall_buffer *m2;
+       uint32_t length, offset;
 
-       r = ctdb_marshall_record(mem_ctx, reqid, key, header, data);
-       if (r == NULL) {
-               talloc_free(m);
-               return NULL;
-       }
+       length = ctdb_marshall_record_size(key, header, data);
 
        if (m == NULL) {
-               m = talloc_zero_size(mem_ctx, offsetof(struct ctdb_marshall_buffer, data));
-               if (m == NULL) {
-                       return NULL;
-               }
-               m->db_id = db_id;
+               offset = offsetof(struct ctdb_marshall_buffer, data);
+               m2 = talloc_zero_size(mem_ctx, offset + length);
+       } else {
+               offset = talloc_get_size(m);
+               m2 = talloc_realloc_size(mem_ctx, m, offset + length);
        }
-
-       m_size = talloc_get_size(m);
-       r_size = talloc_get_size(r);
-
-       m2 = talloc_realloc_size(mem_ctx, m,  m_size + r_size);
        if (m2 == NULL) {
-               talloc_free(m);
+               TALLOC_FREE(m);
                return NULL;
        }
 
-       memcpy(m_size + (uint8_t *)m2, r, r_size);
-
-       talloc_free(r);
+       if (m == NULL) {
+               m2->db_id = db_id;
+       }
 
+       r = (struct ctdb_rec_data *)((uint8_t *)m2 + offset);
+       ctdb_marshall_record_copy(r, reqid, key, header, data, length);
        m2->count++;
 
        return m2;
@@ -278,233 +335,12 @@ struct ctdb_rec_data *ctdb_marshall_loop_next(struct ctdb_marshall_buffer *m, st
                if (r->datalen < sizeof(*header)) {
                        return NULL;
                }
-               *header = *(struct ctdb_ltdb_header *)&r->data[r->keylen];
+               memcpy(header, &r->data[r->keylen], sizeof(*header));
        }
 
        return r;
 }
 
-
-#if HAVE_SCHED_H
-#include <sched.h>
-#endif
-
-/*
-  if possible, make this task real time
- */
-void ctdb_set_scheduler(struct ctdb_context *ctdb)
-{
-#if HAVE_SCHED_SETSCHEDULER    
-       struct sched_param p;
-       if (ctdb->saved_scheduler_param == NULL) {
-               ctdb->saved_scheduler_param = talloc_size(ctdb, sizeof(p));
-       }
-       
-       if (sched_getparam(0, (struct sched_param *)ctdb->saved_scheduler_param) == -1) {
-               DEBUG(DEBUG_ERR,("Unable to get old scheduler params\n"));
-               return;
-       }
-
-       p = *(struct sched_param *)ctdb->saved_scheduler_param;
-       p.sched_priority = 1;
-
-       if (sched_setscheduler(0, SCHED_FIFO, &p) == -1) {
-               DEBUG(DEBUG_CRIT,("Unable to set scheduler to SCHED_FIFO (%s)\n", 
-                        strerror(errno)));
-       } else {
-               DEBUG(DEBUG_NOTICE,("Set scheduler to SCHED_FIFO\n"));
-       }
-#endif
-}
-
-/*
-  restore previous scheduler parameters
- */
-void ctdb_restore_scheduler(struct ctdb_context *ctdb)
-{
-#if HAVE_SCHED_SETSCHEDULER    
-       if (ctdb->saved_scheduler_param == NULL) {
-               ctdb_fatal(ctdb, "No saved scheduler parameters\n");
-       }
-       if (sched_setscheduler(0, SCHED_OTHER, (struct sched_param *)ctdb->saved_scheduler_param) == -1) {
-               ctdb_fatal(ctdb, "Unable to restore old scheduler parameters\n");
-       }
-#endif
-}
-
-/*
- * This function forks a child process and drops the realtime 
- * scheduler for the child process.
- */
-pid_t ctdb_fork(struct ctdb_context *ctdb)
-{
-       pid_t pid;
-
-       pid = fork();
-       if (pid == 0) {
-               if (ctdb->do_setsched) {
-                       ctdb_restore_scheduler(ctdb);
-               }
-               ctdb->can_send_controls = false;
-       }
-       return pid;
-}
-
-void set_nonblocking(int fd)
-{
-       unsigned v;
-       v = fcntl(fd, F_GETFL, 0);
-        fcntl(fd, F_SETFL, v | O_NONBLOCK);
-}
-
-void set_close_on_exec(int fd)
-{
-       unsigned v;
-       v = fcntl(fd, F_GETFD, 0);
-        fcntl(fd, F_SETFD, v | FD_CLOEXEC);
-}
-
-
-bool parse_ipv4(const char *s, unsigned port, struct sockaddr_in *sin)
-{
-       sin->sin_family = AF_INET;
-       sin->sin_port   = htons(port);
-
-       if (inet_pton(AF_INET, s, &sin->sin_addr) != 1) {
-               DEBUG(DEBUG_ERR, (__location__ " Failed to translate %s into sin_addr\n", s));
-               return false;
-       }
-
-       return true;
-}
-
-static bool parse_ipv6(const char *s, const char *ifaces, unsigned port, ctdb_sock_addr *saddr)
-{
-       saddr->ip6.sin6_family   = AF_INET6;
-       saddr->ip6.sin6_port     = htons(port);
-       saddr->ip6.sin6_flowinfo = 0;
-       saddr->ip6.sin6_scope_id = 0;
-
-       if (inet_pton(AF_INET6, s, &saddr->ip6.sin6_addr) != 1) {
-               DEBUG(DEBUG_ERR, (__location__ " Failed to translate %s into sin6_addr\n", s));
-               return false;
-       }
-
-       if (ifaces && IN6_IS_ADDR_LINKLOCAL(&saddr->ip6.sin6_addr)) {
-               if (strchr(ifaces, ',')) {
-                       DEBUG(DEBUG_ERR, (__location__ " Link local address %s "
-                                         "is specified for multiple ifaces %s\n",
-                                         s, ifaces));
-                       return false;
-               }
-               saddr->ip6.sin6_scope_id = if_nametoindex(ifaces);
-       }
-
-       return true;
-}
-/*
-  parse a ip:port pair
- */
-bool parse_ip_port(const char *addr, ctdb_sock_addr *saddr)
-{
-       TALLOC_CTX *tmp_ctx = talloc_new(NULL);
-       char *s, *p;
-       unsigned port;
-       char *endp = NULL;
-       bool ret;
-
-       s = talloc_strdup(tmp_ctx, addr);
-       if (s == NULL) {
-               DEBUG(DEBUG_ERR, (__location__ " Failed strdup()\n"));
-               talloc_free(tmp_ctx);
-               return false;
-       }
-
-       p = rindex(s, ':');
-       if (p == NULL) {
-               DEBUG(DEBUG_ERR, (__location__ " This addr: %s does not contain a port number\n", s));
-               talloc_free(tmp_ctx);
-               return false;
-       }
-
-       port = strtoul(p+1, &endp, 10);
-       if (endp == NULL || *endp != 0) {
-               /* trailing garbage */
-               DEBUG(DEBUG_ERR, (__location__ " Trailing garbage after the port in %s\n", s));
-               talloc_free(tmp_ctx);
-               return false;
-       }
-       *p = 0;
-
-
-       /* now is this a ipv4 or ipv6 address ?*/
-       ret = parse_ip(s, NULL, port, saddr);
-
-       talloc_free(tmp_ctx);
-       return ret;
-}
-
-/*
-  parse an ip
- */
-bool parse_ip(const char *addr, const char *ifaces, unsigned port, ctdb_sock_addr *saddr)
-{
-       char *p;
-       bool ret;
-
-       /* now is this a ipv4 or ipv6 address ?*/
-       p = index(addr, ':');
-       if (p == NULL) {
-               ret = parse_ipv4(addr, port, &saddr->ip);
-       } else {
-               ret = parse_ipv6(addr, ifaces, port, saddr);
-       }
-
-       return ret;
-}
-
-/*
-  parse a ip/mask pair
- */
-bool parse_ip_mask(const char *str, const char *ifaces, ctdb_sock_addr *addr, unsigned *mask)
-{
-       TALLOC_CTX *tmp_ctx = talloc_new(NULL);
-       char *s, *p;
-       char *endp = NULL;
-       bool ret;
-
-       ZERO_STRUCT(*addr);
-       s = talloc_strdup(tmp_ctx, str);
-       if (s == NULL) {
-               DEBUG(DEBUG_ERR, (__location__ " Failed strdup()\n"));
-               talloc_free(tmp_ctx);
-               return false;
-       }
-
-       p = rindex(s, '/');
-       if (p == NULL) {
-               DEBUG(DEBUG_ERR, (__location__ " This addr: %s does not contain a mask\n", s));
-               talloc_free(tmp_ctx);
-               return false;
-       }
-
-       *mask = strtoul(p+1, &endp, 10);
-       if (endp == NULL || *endp != 0) {
-               /* trailing garbage */
-               DEBUG(DEBUG_ERR, (__location__ " Trailing garbage after the mask in %s\n", s));
-               talloc_free(tmp_ctx);
-               return false;
-       }
-       *p = 0;
-
-
-       /* now is this a ipv4 or ipv6 address ?*/
-       ret = parse_ip(s, ifaces, 0, addr);
-
-       talloc_free(tmp_ctx);
-       return ret;
-}
-
 /*
    This is used to canonicalize a ctdb_sock_addr structure.
 */
@@ -573,6 +409,7 @@ char *ctdb_addr_to_str(ctdb_sock_addr *addr)
                break;
        default:
                DEBUG(DEBUG_ERR, (__location__ " ERROR, unknown family %u\n", addr->sa.sa_family));
+               ctdb_external_trace();
        }
 
        return cip;
@@ -594,94 +431,88 @@ unsigned ctdb_addr_to_port(ctdb_sock_addr *addr)
        return 0;
 }
 
-void ctdb_block_signal(int signum)
-{
-       sigset_t set;
-       sigemptyset(&set);
-       sigaddset(&set,signum);
-       sigprocmask(SIG_BLOCK,&set,NULL);
-}
 
-void ctdb_unblock_signal(int signum)
-{
-       sigset_t set;
-       sigemptyset(&set);
-       sigaddset(&set,signum);
-       sigprocmask(SIG_UNBLOCK,&set,NULL);
-}
+const char *ctdb_eventscript_call_names[] = {
+       "init",
+       "setup",
+       "startup",
+       "startrecovery",
+       "recovered",
+       "takeip",
+       "releaseip",
+       "stopped",
+       "monitor",
+       "status",
+       "shutdown",
+       "reload",
+       "updateip",
+       "ipreallocated"
+};
 
-struct debug_levels debug_levels[] = {
-       {DEBUG_EMERG,   "EMERG"},
-       {DEBUG_ALERT,   "ALERT"},
-       {DEBUG_CRIT,    "CRIT"},
-       {DEBUG_ERR,     "ERR"},
-       {DEBUG_WARNING, "WARNING"},
-       {DEBUG_NOTICE,  "NOTICE"},
-       {DEBUG_INFO,    "INFO"},
-       {DEBUG_DEBUG,   "DEBUG"},
-       {0, NULL}
+/* Runstate handling */
+static struct {
+       enum ctdb_runstate runstate;
+       const char * label;
+} runstate_map[] = {
+       { CTDB_RUNSTATE_UNKNOWN, "UNKNOWN" },
+       { CTDB_RUNSTATE_INIT, "INIT" },
+       { CTDB_RUNSTATE_SETUP, "SETUP" },
+       { CTDB_RUNSTATE_FIRST_RECOVERY, "FIRST_RECOVERY" },
+       { CTDB_RUNSTATE_STARTUP, "STARTUP" },
+       { CTDB_RUNSTATE_RUNNING, "RUNNING" },
+       { CTDB_RUNSTATE_SHUTDOWN, "SHUTDOWN" },
+       { -1, NULL },
 };
 
-const char *get_debug_by_level(int32_t level)
+const char *runstate_to_string(enum ctdb_runstate runstate)
 {
        int i;
-
-       for (i=0; debug_levels[i].description != NULL; i++) {
-               if (debug_levels[i].level == level) {
-                       return debug_levels[i].description;
+       for (i=0; runstate_map[i].label != NULL ; i++) {
+               if (runstate_map[i].runstate == runstate) {
+                       return runstate_map[i].label;
                }
        }
-       return "Unknown";
+
+       return runstate_map[0].label;
 }
 
-int32_t get_debug_by_desc(const char *desc)
+enum ctdb_runstate runstate_from_string(const char *label)
 {
        int i;
-
-       for (i=0; debug_levels[i].description != NULL; i++) {
-               if (!strcmp(debug_levels[i].description, desc)) {
-                       return debug_levels[i].level;
+       for (i=0; runstate_map[i].label != NULL; i++) {
+               if (strcasecmp(runstate_map[i].label, label) == 0) {
+                       return runstate_map[i].runstate;
                }
        }
 
-       return DEBUG_ERR;
+       return CTDB_RUNSTATE_UNKNOWN;
 }
 
-/* we don't lock future pages here; it would increase the chance that
- * we'd fail to mmap later on. */
-void ctdb_lockdown_memory(struct ctdb_context *ctdb)
+void ctdb_set_runstate(struct ctdb_context *ctdb, enum ctdb_runstate runstate)
 {
-#ifdef HAVE_MLOCKALL
-       /* Extra stack, please! */
-       char dummy[10000];
-       memset(dummy, 0, sizeof(dummy));
-
-       if (ctdb->valgrinding) {
-               return;
+       if (runstate <= ctdb->runstate) {
+               ctdb_fatal(ctdb, "runstate must always increase");
        }
 
-       /* Avoid compiler optimizing out dummy. */
-       mlock(dummy, sizeof(dummy));
-       if (mlockall(MCL_CURRENT) != 0) {
-               DEBUG(DEBUG_WARNING,("Failed to lock memory: %s'\n",
-                                    strerror(errno)));
-       }
-#endif
+       DEBUG(DEBUG_NOTICE,("Set runstate to %s (%d)\n",
+                           runstate_to_string(runstate), runstate));
+       ctdb->runstate = runstate;
 }
 
-const char *ctdb_eventscript_call_names[] = {
-       "init",
-       "setup",
-       "startup",
-       "startrecovery",
-       "recovered",
-       "takeip",
-       "releaseip",
-       "stopped",
-       "monitor",
-       "status",
-       "shutdown",
-       "reload",
-       "updateip",
-       "ipreallocated"
-};
+/* Convert arbitrary data to 4-byte boundary padded uint32 array */
+uint32_t *ctdb_key_to_idkey(TALLOC_CTX *mem_ctx, TDB_DATA key)
+{
+       uint32_t idkey_size, *k;
+
+       idkey_size = 1 + (key.dsize + sizeof(uint32_t)-1) / sizeof(uint32_t);
+
+       k = talloc_zero_array(mem_ctx, uint32_t, idkey_size);
+       if (k == NULL) {
+               return NULL;
+       }
+
+       k[0] = idkey_size;
+       memcpy(&k[1], key.dptr, key.dsize);
+
+       return k;
+}