diff --git a/src/config.h b/src/config.h index 1d93d1c..35566ff 100644 --- a/src/config.h +++ b/src/config.h @@ -1,58 +1,58 @@ /* * Copyright (C) 2011 Jiaju Zhang * Copyright (C) 2013 Philipp Marek * * This program is free software; you can redistribute it and/or * modify it under the terms of the GNU General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * * This software is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * General Public License for more details. * * You should have received a copy of the GNU General Public * License along with this library; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ #ifndef _CONFIG_H #define _CONFIG_H #include #include "booth.h" #include "transport.h" #define MAX_NODES 16 #define TICKET_ALLOC 16 struct ticket_config { int weight[MAX_NODES]; int expiry; boothc_ticket name; }; struct booth_config { int node_count; int ticket_count; transport_layer_t proto; uint16_t port; struct booth_node node[MAX_NODES]; struct ticket_config ticket[0]; }; struct booth_config *booth_conf; int read_config(const char *path); int check_config(int type); int find_site_in_config(unsigned char *site, struct booth_node **node); -static inline struct booth_transport *transport(void) { +static inline struct booth_transport const *transport(void) { return booth_transport + booth_conf->proto; } #endif /* _CONFIG_H */ diff --git a/src/main.c b/src/main.c index 3ce786a..25cfb3b 100644 --- a/src/main.c +++ b/src/main.c @@ -1,1083 +1,1057 @@ /* * Copyright (C) 2011 Jiaju Zhang * Copyright (C) 2013 Philipp Marek * * This program is free software; you can redistribute it and/or * modify it under the terms of the GNU General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * * This software is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * General Public License for more details. * * You should have received a copy of the GNU General Public * License along with this library; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "log.h" #include "booth.h" #include "config.h" #include "transport.h" #include "timer.h" #include "pacemaker.h" #include "ticket.h" #define RELEASE_VERSION "1.0" #define CLIENT_NALLOC 32 int log_logfile_priority = LOG_INFO; int log_syslog_priority = LOG_ERR; int log_stderr_priority = LOG_ERR; int daemonize = 0; static int client_maxi; static int client_size = 0; struct client *client = NULL; struct pollfd *pollfd = NULL; typedef enum { BOOTHD_STARTED=0, BOOTHD_STARTING } BOOTH_DAEMON_STATE; int poll_timeout = -1; typedef enum { ACT_ARBITRATOR = 1, ACT_SITE, ACT_CLIENT, } booth_role_t; typedef enum { OP_LIST = 1, OP_GRANT, OP_REVOKE, } operation_t; struct command_line { int type; /* ACT_ */ int op; /* OP_ */ char configfile[BOOTH_PATH_LEN]; char lockfile[BOOTH_PATH_LEN]; struct boothc_site_ticket_msg msg; }; static struct command_line cl; int do_read(int fd, void *buf, size_t count) { int rv, off = 0; while (off < count) { rv = read(fd, (char *)buf + off, count - off); if (rv == 0) return -1; if (rv == -1 && errno == EINTR) continue; if (rv == -1) return -1; off += rv; } return 0; } int do_write(int fd, void *buf, size_t count) { int rv, off = 0; retry: rv = write(fd, (char *)buf + off, count); if (rv == -1 && errno == EINTR) goto retry; if (rv < 0) { log_error("write failed: %s (%d)", strerror(errno), errno); return rv; } if (rv != count) { count -= rv; off += rv; goto retry; } return 0; } -static int do_connect(void) + +static int do_local_connect_and_write(void *data, int len, struct booth_node **ret) { - int rv, fd; struct booth_node *node; - struct sockaddr_in sin4; - struct sockaddr_in6 sin6; + int rv; + + + if (ret) + *ret = NULL; - rv = -1; /* Use locally reachable address, ie. in same cluster. */ - if (find_myself(&node, 1)) { + if (!find_myself(&node, 1)) { + log_error("Cannot find local cluster."); + return ENOENT; + } - /* Always use TCP within cluster. */ - fd = socket(PF_INET, SOCK_STREAM, 0); - if (fd < 0) { - log_error("failed to create socket: %s (%d)", strerror(errno), errno); - goto out; - } + if (ret) + *ret = node; - switch(node->family) { - case AF_INET: - sin4.sin_family = node->family; - sin4.sin_port = htons(booth_conf->port); - memcpy(&sin4.sin_addr, &node->in4, node->addrlen); - rv = connect(fd, (struct sockaddr *) &sin4, sizeof(sin4)); - break; - case AF_INET6: - sin6.sin6_family = node->family; - sin6.sin6_flowinfo = 0; - sin6.sin6_port = htons(booth_conf->port); - memcpy(&sin6.sin6_addr, &node->in6, node->addrlen); - rv = connect(fd, (struct sockaddr *) &sin6, sizeof(sin6)); - break; - default: - log_error("unknown family %d", node->family); - goto out; - } - } + /* Always use TCP within cluster. */ + rv = booth_tcp_open(node); + if (rv < 0) + goto out; + + rv = booth_tcp_send(node, data, len); - if (rv < 0) { - if (errno == ECONNREFUSED) - log_error("Connection to boothd was refused; " - "please ensure that you are on a " - "machine which has boothd running."); - else - log_error("failed to connect: %s (%d)", strerror(errno), errno); - close(fd); - fd = rv; - } out: - return fd; + return rv; } + static void init_header(struct boothc_header *h, int cmd, int result, int data_len) { memset(h, 0, sizeof(struct boothc_header)); h->magic = BOOTHC_MAGIC; h->version = BOOTHC_VERSION; h->len = data_len; h->cmd = cmd; h->result = result; } static void client_alloc(void) { int i; if (!client) { client = malloc(CLIENT_NALLOC * sizeof(struct client)); pollfd = malloc(CLIENT_NALLOC * sizeof(struct pollfd)); } else { client = realloc(client, (client_size + CLIENT_NALLOC) * sizeof(struct client)); pollfd = realloc(pollfd, (client_size + CLIENT_NALLOC) * sizeof(struct pollfd)); if (!pollfd) log_error("can't alloc for pollfd"); } if (!client || !pollfd) log_error("can't alloc for client array"); for (i = client_size; i < client_size + CLIENT_NALLOC; i++) { client[i].workfn = NULL; client[i].deadfn = NULL; client[i].fd = -1; pollfd[i].fd = -1; pollfd[i].revents = 0; } client_size += CLIENT_NALLOC; } static void client_dead(int ci) { close(client[ci].fd); client[ci].workfn = NULL; client[ci].fd = -1; pollfd[ci].fd = -1; } int client_add(int fd, void (*workfn)(int ci), void (*deadfn)(int ci)) { int i; if (!client) client_alloc(); again: for (i = 0; i < client_size; i++) { if (client[i].fd == -1) { client[i].workfn = workfn; if (deadfn) client[i].deadfn = deadfn; else client[i].deadfn = client_dead; client[i].fd = fd; pollfd[i].fd = fd; pollfd[i].events = POLLIN; if (i > client_maxi) client_maxi = i; return i; } } client_alloc(); goto again; } void process_connection(int ci) { struct boothc_site_ticket_msg msg; char *data = NULL; int ticket_owner; int local, rv; void (*deadfn) (int ci); rv = do_read(client[ci].fd, &msg.header, sizeof(msg.header)); if (rv < 0) { if (errno == ECONNRESET) log_debug("client %d connection reset for fd %d", ci, client[ci].fd); deadfn = client[ci].deadfn; if(deadfn) { deadfn(ci); } return; } if (msg.header.magic != BOOTHC_MAGIC) { log_error("connection %d magic error %x", ci, msg.header.magic); return; } if (msg.header.version != BOOTHC_VERSION) { log_error("connection %d version error %x", ci, msg.header.version); return; } if (msg.header.len) { if (msg.header.len != sizeof(msg) - sizeof(msg.header)) { log_error("got wrong length %u", msg.header.len); return; } rv = do_read(client[ci].fd, msg.header.data, msg.header.len); if (rv < 0) { log_error("connection %d read data error %d", ci, rv); goto out; } } switch (msg.header.cmd) { case BOOTHC_CMD_LIST: assert(!data); msg.header.result = list_ticket(&data, &msg.header.len); break; case BOOTHC_CMD_GRANT: msg.header.len = 0; if (!check_ticket(msg.ticket)) { msg.header.result = BOOTHC_RLT_INVALID_ARG; goto reply; } if (get_ticket_info(msg.ticket, &ticket_owner, NULL) == 0) { if (ticket_owner > -1) { log_error("client want to get an granted " "ticket %s", msg.ticket); msg.header.result = BOOTHC_RLT_OVERGRANT; goto reply; } } else { log_error("can not get ticket %s's info", msg.ticket); msg.header.result = BOOTHC_RLT_INVALID_ARG; goto reply; } if (!check_site(msg.site, &local)) { msg.header.result = BOOTHC_RLT_INVALID_ARG; goto reply; } if (local) msg.header.result = grant_ticket(msg.ticket); else msg.header.result = BOOTHC_RLT_REMOTE_OP; break; case BOOTHC_CMD_REVOKE: msg.header.len = 0; if (!check_ticket(msg.ticket)) { msg.header.result = BOOTHC_RLT_INVALID_ARG; goto reply; } if (!check_site(msg.site, &local)) { msg.header.result = BOOTHC_RLT_INVALID_ARG; goto reply; } if (local) msg.header.result = revoke_ticket(msg.ticket); else msg.header.result = BOOTHC_RLT_REMOTE_OP; break; case BOOTHC_CMD_CATCHUP: msg.header.result = catchup_ticket(&data, msg.header.len); break; default: log_error("connection %d cmd %x unknown", ci, msg.header.cmd); break; } reply: rv = do_write(client[ci].fd, &msg.header, sizeof(msg.header)); if (rv < 0) log_error("connection %d write error %d", ci, rv); if (msg.header.len) { rv = do_write(client[ci].fd, data, msg.header.len); if (rv < 0) log_error("connection %d write error %d", ci, rv); } out: free(data); } static void process_listener(int ci) { int fd, i; fd = accept(client[ci].fd, NULL, NULL); if (fd < 0) { log_error("process_listener: accept error for fd %d: %s (%d)", client[ci].fd, strerror(errno), errno); if (client[ci].deadfn) client[ci].deadfn(ci); return; } i = client_add(fd, process_connection, NULL); log_debug("add client connection %d fd %d", i, fd); } static int setup_config(int type) { int rv; rv = read_config(cl.configfile); if (rv < 0) goto out; rv = check_config(type); if (rv < 0) goto out; out: return rv; } static int setup_transport(void) { int rv; rv = transport()->init(ticket_recv); if (rv < 0) { log_error("failed to init booth_transport %s", transport()->name); goto out; } rv = booth_transport[TCP].init(NULL); if (rv < 0) { log_error("failed to init booth_transport[TCP]"); goto out; } out: return rv; } static int setup_timer(void) { return timerlist_init(); } static int write_daemon_state(int fd, int state) { char buf[16]; int rv=0; memset(buf, 0, sizeof(buf)); snprintf(buf, sizeof(buf), "%d %d", getpid(), state); rv = lseek(fd, 0, SEEK_SET); if (rv < 0) { log_error("lseek set fd(%d) offset to 0 error, return(%d), message(%s)", fd, rv, strerror(errno)); rv = -1; return rv; } rv = write(fd, buf, strlen(buf)); if (rv <= 0) { log_error("write to fd(%d) error, return(%d), message(%s)", fd, rv, strerror(errno)); rv = -1; return rv; } rv = 0; return rv; } static int loop(int fd) { void (*workfn) (int ci); void (*deadfn) (int ci); int rv, i; rv = setup_timer(); if (rv < 0) goto fail; rv = setup_transport(); if (rv < 0) goto fail; rv = setup_ticket(); if (rv < 0) goto fail; client_add(rv, process_listener, NULL); rv = write_daemon_state(fd, BOOTHD_STARTED); if (rv != 0) { log_error("write daemon state %d to lockfile error %s: %s", BOOTHD_STARTED, cl.lockfile, strerror(errno)); goto fail; } if (cl.type == ARBITRATOR) log_info("BOOTH arbitrator daemon started"); else if (cl.type == SITE) log_info("BOOTH cluster site daemon started"); while (1) { rv = poll(pollfd, client_maxi + 1, poll_timeout); if (rv == -1 && errno == EINTR) continue; if (rv < 0) { log_error("poll failed: %s (%d)", strerror(errno), errno); goto fail; } for (i = 0; i <= client_maxi; i++) { if (client[i].fd < 0) continue; if (pollfd[i].revents & POLLIN) { workfn = client[i].workfn; if (workfn) workfn(i); } if (pollfd[i].revents & (POLLERR | POLLHUP | POLLNVAL)) { deadfn = client[i].deadfn; if (deadfn) deadfn(i); } } process_timerlist(); } return 0; fail: return -1; } + static int do_list(void) { + struct booth_node *node; struct boothc_header h, *rh; char *reply = NULL, *data; int data_len; - int fd, rv; + int rv; init_header(&h, BOOTHC_CMD_LIST, 0, 0); - fd = do_connect(); - if (fd < 0) { - rv = fd; - goto out; - } - - rv = do_write(fd, &h, sizeof(h)); + rv = do_local_connect_and_write(&h, sizeof(h), &node); if (rv < 0) - goto out_close; + goto out; reply = malloc(sizeof(struct boothc_header)); if (!reply) { rv = -ENOMEM; goto out_close; } - rv = do_read(fd, reply, sizeof(struct boothc_header)); + rv = local_transport->recv(node, reply, sizeof(struct boothc_header)); if (rv < 0) goto out_free; rh = (struct boothc_header *)reply; data_len = rh->len; reply = realloc(reply, sizeof(struct boothc_header) + data_len); if (!reply) { rv = -ENOMEM; goto out_free; } data = reply + sizeof(struct boothc_header); - rv = do_read(fd, data, data_len); + rv = local_transport->recv(node, data, data_len); if (rv < 0) goto out_free; do_write(STDOUT_FILENO, data, data_len); rv = 0; out_free: free(reply); out_close: - close(fd); + local_transport->close(node); out: return rv; } static int do_command(cmd_request_t cmd) { + struct booth_node *node, *to; struct boothc_header reply; - int fd, rv; + int rv; + + node = NULL; + to = NULL; init_header(&cl.msg.header, cmd, 0, sizeof(cl.msg) - sizeof(cl.msg.header)); - fd = do_connect(); - if (fd < 0) { - rv = fd; - goto out; - } - - rv = do_write(fd, &cl.msg, sizeof(cl.msg)); + rv = do_local_connect_and_write(&cl.msg, sizeof(cl.msg), &node); if (rv < 0) goto out_close; - rv = do_read(fd, &reply, sizeof(reply)); + rv = local_transport->recv(node, &reply, sizeof(reply)); if (rv < 0) goto out_close; if (reply.result == BOOTHC_RLT_INVALID_ARG) { log_info("invalid argument!"); rv = -1; goto out_close; } if (reply.result == BOOTHC_RLT_OVERGRANT) { log_info("You're granting a granted ticket " "If you wanted to migrate a ticket," "use revoke first, then use grant"); rv = -1; goto out_close; } if (reply.result == BOOTHC_RLT_REMOTE_OP) { - struct booth_node *to; if (!find_site_in_config(cl.msg.site, &to)) { log_error("Redirected to unknown site %s.", cl.msg.site); rv = -1; goto out_close; } rv = booth_transport[TCP].open(to); if (rv < 0) { goto out_close; } rv = booth_transport[TCP].send(to, &cl.msg, sizeof(cl.msg)); if (rv < 0) { booth_transport[TCP].close(to); goto out_close; } rv = booth_transport[TCP].recv(to, &reply, sizeof(struct boothc_header)); if (rv < 0) { booth_transport[TCP].close(to); goto out_close; } booth_transport[TCP].close(to); } if (reply.result == BOOTHC_RLT_ASYNC) { if (cmd == BOOTHC_CMD_GRANT) log_info("grant command sent, result will be returned " "asynchronously, you can get the result from " "the log files"); else if (cmd == BOOTHC_CMD_REVOKE) log_info("revoke command sent, result will be returned " "asynchronously, you can get the result from " "the log files."); else log_error("internal error reading reply result!"); rv = 0; } else if (reply.result == BOOTHC_RLT_SYNC_SUCC) { if (cmd == BOOTHC_CMD_GRANT) log_info("grant succeeded!"); else if (cmd == BOOTHC_CMD_REVOKE) log_info("revoke succeeded!"); rv = 0; } else if (reply.result == BOOTHC_RLT_SYNC_FAIL) { if (cmd == BOOTHC_CMD_GRANT) log_info("grant failed!"); else if (cmd == BOOTHC_CMD_REVOKE) log_info("revoke failed!"); rv = -1; } else { log_error("internal error!"); rv = -1; } out_close: - close(fd); -out: + if (node) + local_transport->close(node); + if (to) + booth_transport[TCP].close(to); return rv; } static int do_grant(void) { return do_command(BOOTHC_CMD_GRANT); } static int do_revoke(void) { return do_command(BOOTHC_CMD_REVOKE); } static int lockfile(void) { struct flock lock; int fd, rv; fd = open(cl.lockfile, O_CREAT|O_WRONLY, 0666); if (fd < 0) { log_error("lockfile open error %s: %s", cl.lockfile, strerror(errno)); return -1; } lock.l_type = F_WRLCK; lock.l_start = 0; lock.l_whence = SEEK_SET; lock.l_len = 0; rv = fcntl(fd, F_SETLK, &lock); if (rv < 0) { log_error("lockfile setlk error %s: %s", cl.lockfile, strerror(errno)); goto fail; } rv = ftruncate(fd, 0); if (rv < 0) { log_error("lockfile truncate error %s: %s", cl.lockfile, strerror(errno)); goto fail; } rv = write_daemon_state(fd, BOOTHD_STARTING); if (rv != 0) { log_error("write daemon state %d to lockfile error %s: %s", BOOTHD_STARTING, cl.lockfile, strerror(errno)); goto fail; } return fd; fail: close(fd); return -1; } static void unlink_lockfile(int fd) { unlink(cl.lockfile); close(fd); } static void print_usage(void) { printf("Usage:\n"); printf("booth [options]\n"); printf("\n"); printf("Types:\n"); printf(" arbitrator: daemon running on arbitrator\n"); printf(" site: daemon running on cluster site\n"); printf(" client: command running from client\n"); printf("\n"); printf("Operations:\n"); printf("Please note that operations are valid iff type is client!\n"); printf(" list: List all the tickets\n"); printf(" grant: Grant ticket T(-t T) to site S(-s S)\n"); printf(" revoke: Revoke ticket T(-t T) from site S(-s S)\n"); printf("\n"); printf("Options:\n"); printf(" -c FILE Specify config file [default " BOOTH_DEFAULT_CONF "]\n"); printf(" -l LOCKFILE Specify lock file [default " BOOTH_DEFAULT_LOCKFILE "]\n"); printf(" -D Enable debugging to stderr and don't fork\n"); printf(" -t ticket name\n"); printf(" -s site name\n"); printf(" -h Print this help, then exit\n"); } #define OPTION_STRING "c:Dl:t:s:h" static char *logging_entity = NULL; void safe_copy(char *dest, char *value, size_t buflen, const char *description) { int content_len = buflen - 1; if (strlen(value) >= content_len) { fprintf(stderr, "'%s' exceeds maximum %s length of %d\n", value, description, content_len); exit(EXIT_FAILURE); } strncpy(dest, value, content_len); dest[content_len] = 0; } static int host_convert(char *hostname, char *ip_str, size_t ip_size) { struct addrinfo *result = NULL, hints = {0}; int re = -1; memset(&hints, 0, sizeof(hints)); hints.ai_family = BOOTH_PROTO_FAMILY; hints.ai_socktype = SOCK_DGRAM; re = getaddrinfo(hostname, NULL, &hints, &result); if (re == 0) { struct in_addr addr = ((struct sockaddr_in *)result->ai_addr)->sin_addr; const char *re_ntop = inet_ntop(BOOTH_PROTO_FAMILY, &addr, ip_str, ip_size); if (re_ntop == NULL) { re = -1; } } freeaddrinfo(result); return re; } static int read_arguments(int argc, char **argv) { int optchar; char *arg1 = argv[1]; char *op = NULL; char site_arg[INET_ADDRSTRLEN] = {0}; if (argc < 2 || !strcmp(arg1, "help") || !strcmp(arg1, "--help") || !strcmp(arg1, "-h")) { print_usage(); exit(EXIT_SUCCESS); } if (!strcmp(arg1, "version") || !strcmp(arg1, "--version") || !strcmp(arg1, "-V")) { printf("%s %s (built %s %s)\n", argv[0], RELEASE_VERSION, __DATE__, __TIME__); exit(EXIT_SUCCESS); } if (!strcmp(arg1, "arbitrator")) { cl.type = ACT_ARBITRATOR; logging_entity = (char *) DAEMON_NAME "-arbitrator"; optind = 2; } else if (!strcmp(arg1, "site")) { cl.type = ACT_SITE; logging_entity = (char *) DAEMON_NAME "-site"; optind = 2; } else if (!strcmp(arg1, "client")) { cl.type = ACT_CLIENT; if (argc < 3) { print_usage(); exit(EXIT_FAILURE); } op = argv[2]; optind = 3; } else { cl.type = ACT_CLIENT; op = argv[1]; optind = 2; } switch (cl.type) { case ACT_ARBITRATOR: break; case ACT_SITE: break; case ACT_CLIENT: if (!strcmp(op, "list")) cl.op = OP_LIST; else if (!strcmp(op, "grant")) cl.op = OP_GRANT; else if (!strcmp(op, "revoke")) cl.op = OP_REVOKE; else { fprintf(stderr, "client operation \"%s\" is unknown\n", op); exit(EXIT_FAILURE); } break; } while (optind < argc) { optchar = getopt(argc, argv, OPTION_STRING); switch (optchar) { case 'c': safe_copy(cl.configfile, optarg, sizeof(cl.configfile), "config file"); break; case 'D': daemonize = 1; debug_level = 1; log_logfile_priority = LOG_DEBUG; log_syslog_priority = LOG_DEBUG; break; case 'l': safe_copy(cl.lockfile, optarg, sizeof(cl.lockfile), "lock file"); break; case 't': if (cl.op == OP_GRANT || cl.op == OP_REVOKE) { safe_copy(cl.msg.ticket, optarg, sizeof(cl.msg.ticket), "ticket name"); } else { print_usage(); exit(EXIT_FAILURE); } break; case 's': if (cl.op == OP_GRANT || cl.op == OP_REVOKE) { int re = host_convert(optarg, site_arg, INET_ADDRSTRLEN); if (re == 0) { safe_copy(cl.msg.site, site_arg, sizeof(cl.msg.ticket), "site name"); } else { safe_copy(cl.msg.site, optarg, sizeof(cl.msg.ticket), "site name"); } } else { print_usage(); exit(EXIT_FAILURE); } break; case 'h': print_usage(); exit(EXIT_SUCCESS); break; case ':': case '?': fprintf(stderr, "Please use '-h' for usage.\n"); exit(EXIT_FAILURE); break; default: fprintf(stderr, "unknown option: %s\n", argv[optind]); exit(EXIT_FAILURE); break; }; } return 0; } static void set_scheduler(void) { struct sched_param sched_param; struct rlimit rlimit; int rv; rlimit.rlim_cur = RLIM_INFINITY; rlimit.rlim_max = RLIM_INFINITY; setrlimit(RLIMIT_MEMLOCK, &rlimit); rv = mlockall(MCL_CURRENT | MCL_FUTURE); if (rv < 0) { log_error("mlockall failed"); } rv = sched_get_priority_max(SCHED_RR); if (rv != -1) { sched_param.sched_priority = rv; rv = sched_setscheduler(0, SCHED_RR, &sched_param); if (rv == -1) log_error("could not set SCHED_RR priority %d: %s (%d)", sched_param.sched_priority, strerror(errno), errno); } else { log_error("could not get maximum scheduler priority err %d", errno); } } static void set_oom_adj(int val) { FILE *fp; fp = fopen("/proc/self/oom_adj", "w"); if (!fp) return; fprintf(fp, "%i", val); fclose(fp); } static int do_server(int type) { int fd = -1; int rv = -1; rv = setup_config(type); if (rv < 0) goto out; if (!daemonize) { if (daemon(0, 0) < 0) { perror("daemon error"); exit(EXIT_FAILURE); } } /* The lock cannot be obtained before the call to daemon(), otherwise the lockfile would contain the pid of the parent, not the daemon. */ fd = lockfile(); if (fd < 0) return fd; if (type == ARBITRATOR) log_info("BOOTH arbitrator daemon is starting."); else if (type == SITE) log_info("BOOTH cluster site daemon is starting."); set_scheduler(); set_oom_adj(-16); rv = loop(fd); out: if (fd >= 0) unlink_lockfile(fd); return rv; } static int do_client(void) { int rv = -1; rv = setup_config(CLIENT); if (rv < 0) { log_error("cannot read config"); goto out; } switch (cl.op) { case OP_LIST: rv = do_list(); break; case OP_GRANT: rv = do_grant(); break; case OP_REVOKE: rv = do_revoke(); break; } out: return rv; } int main(int argc, char *argv[]) { int rv; memset(&cl, 0, sizeof(cl)); strncpy(cl.configfile, BOOTH_DEFAULT_CONF, BOOTH_PATH_LEN - 1); strncpy(cl.lockfile, BOOTH_DEFAULT_LOCKFILE, BOOTH_PATH_LEN - 1); rv = read_arguments(argc, argv); if (rv < 0) goto out; if (cl.type == ACT_CLIENT) { cl_log_enable_stderr(TRUE); cl_log_set_facility(0); } else { cl_log_set_entity(logging_entity); cl_log_enable_stderr(debug_level ? TRUE : FALSE); cl_log_set_facility(HA_LOG_FACILITY); } cl_inherit_logging_environment(0); switch (cl.type) { case ACT_ARBITRATOR: rv = do_server(ARBITRATOR); break; case ACT_SITE: rv = do_server(SITE); break; case ACT_CLIENT: rv = do_client(); break; } out: return rv ? EXIT_FAILURE : EXIT_SUCCESS; } diff --git a/src/transport.c b/src/transport.c index d82bfdd..7e9e35e 100644 --- a/src/transport.c +++ b/src/transport.c @@ -1,634 +1,637 @@ /* * Copyright (C) 2011 Jiaju Zhang * Copyright (C) 2013 Philipp Marek * * This program is free software; you can redistribute it and/or * modify it under the terms of the GNU General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * * This software is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * General Public License for more details. * * You should have received a copy of the GNU General Public * License along with this library; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ #include #include #include #include #include #include #include #include #include #include #include #include #include "list.h" #include "booth.h" #include "log.h" #include "config.h" #include "paxos_lease.h" #include "transport.h" #define BOOTH_IPADDR_LEN (sizeof(struct in6_addr)) #define NETLINK_BUFSIZE 16384 #define SOCKET_BUFFER_SIZE 160000 #define FRAME_SIZE_MAX 10000 extern struct client *client; extern struct pollfd *pollfd; static struct booth_node local; struct tcp_conn { int s; struct sockaddr to; struct list_head list; }; static LIST_HEAD(tcp); struct udp_context { int s; struct iovec iov_recv; char iov_buffer[FRAME_SIZE_MAX]; } udp; static int (*deliver_fn) (void *msg, int msglen); static void parse_rtattr(struct rtattr *tb[], int max, struct rtattr *rta, int len) { while (RTA_OK(rta, len)) { if (rta->rta_type <= max) tb[rta->rta_type] = rta; rta = RTA_NEXT(rta,len); } } static int find_address(unsigned char ipaddr[BOOTH_IPADDR_LEN], int family, int prefixlen, int fuzzy_allowed, struct booth_node **me) { int i; struct booth_node *node; int bytes, bits_left, mask; unsigned char node_bits, ip_bits; uint8_t *n_a; bytes = prefixlen / 8; bits_left = prefixlen % 8; /* One bit left to check means ignore 7 lowest bits. */ mask = ~( (1 << (8 - bits_left)) -1); for (i = 0; i < booth_conf->node_count; i++) { node = booth_conf->node + i; if (family != node->family) continue; n_a = node_to_addr_pointer(node); if (memcmp(ipaddr, n_a, node->addrlen) == 0) { found: *me = node; return 1; } if (!fuzzy_allowed) continue; // assert(bytes <= node->addrlen); //#include // printf("node->addr %s, fam %d, prefix %d; %llx vs %llx, bytes %d\n", node->addr, node->family, prefixlen, *((long long*)&node->in6), *((long long*)ipaddr), bytes); /* Check prefix, whole bytes */ if (memcmp(ipaddr, n_a, bytes) != 0) continue; //printf("bits %d\n", bits_left); if (!bits_left) goto found; node_bits = n_a[bytes]; ip_bits = ipaddr[bytes]; //printf("nodebits %x ip %x mask %x\n", node_bits, ip_bits, mask); if (((node_bits ^ ip_bits) & mask) == 0) goto found; } return 0; } int find_myself(struct booth_node **me, int fuzzy_allowed) { int fd; struct sockaddr_nl nladdr; unsigned char ipaddr[BOOTH_IPADDR_LEN]; static char rcvbuf[NETLINK_BUFSIZE]; struct { struct nlmsghdr nlh; struct rtgenmsg g; } req; *me = NULL; fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE); if (fd < 0) { log_error("failed to create netlink socket"); return 0; } setsockopt(fd, SOL_SOCKET, SO_RCVBUF, &rcvbuf, sizeof(rcvbuf)); memset(&nladdr, 0, sizeof(nladdr)); nladdr.nl_family = AF_NETLINK; memset(&req, 0, sizeof(req)); req.nlh.nlmsg_len = sizeof(req); req.nlh.nlmsg_type = RTM_GETADDR; req.nlh.nlmsg_flags = NLM_F_ROOT|NLM_F_MATCH|NLM_F_REQUEST; req.nlh.nlmsg_pid = 0; req.nlh.nlmsg_seq = 1; req.g.rtgen_family = AF_INET; if (sendto(fd, (void *)&req, sizeof(req), 0, (struct sockaddr*)&nladdr, sizeof(nladdr)) < 0) { close(fd); log_error("failed to send data to netlink socket"); return 0; } while (1) { int status; struct nlmsghdr *h; struct iovec iov = { rcvbuf, sizeof(rcvbuf) }; struct msghdr msg = { (void *)&nladdr, sizeof(nladdr), &iov, 1, NULL, 0, 0 }; status = recvmsg(fd, &msg, 0); if (!status) { close(fd); log_error("failed to recvmsg from netlink socket"); return 0; } h = (struct nlmsghdr *)rcvbuf; if (h->nlmsg_type == NLMSG_DONE) break; if (h->nlmsg_type == NLMSG_ERROR) { close(fd); log_error("netlink socket recvmsg error"); return 0; } while (NLMSG_OK(h, status)) { if (h->nlmsg_type == RTM_NEWADDR) { struct ifaddrmsg *ifa = NLMSG_DATA(h); struct rtattr *tb[IFA_MAX+1]; int len = h->nlmsg_len - NLMSG_LENGTH(sizeof(*ifa)); memset(tb, 0, sizeof(tb)); parse_rtattr(tb, IFA_MAX, IFA_RTA(ifa), len); memset(ipaddr, 0, BOOTH_IPADDR_LEN); memcpy(ipaddr, RTA_DATA(tb[IFA_ADDRESS]), BOOTH_IPADDR_LEN); if (find_address(ipaddr, ifa->ifa_family, ifa->ifa_prefixlen, fuzzy_allowed, me)) goto out; } h = NLMSG_NEXT(h, status); } } out: close(fd); return *me != NULL; } static int load_myid(void) { struct booth_node *me; if (find_myself(&me, 0)) { me->local = 1; if (!local.family) memcpy(&local, me, sizeof(struct booth_node)); return me->nodeid; } return -1; } static int booth_get_myid(void) { if (local.local) return local.nodeid; else return -1; } static void process_dead(int ci) { struct tcp_conn *conn, *safe; list_for_each_entry_safe(conn, safe, &tcp, list) { if (conn->s == client[ci].fd) { list_del(&conn->list); free(conn); break; } } close(client[ci].fd); client[ci].workfn = NULL; client[ci].fd = -1; pollfd[ci].fd = -1; } static void process_tcp_listener(int ci) { int fd, i, one = 1; socklen_t addrlen = sizeof(struct sockaddr); struct sockaddr addr; struct tcp_conn *conn; fd = accept(client[ci].fd, &addr, &addrlen); if (fd < 0) { log_error("process_tcp_listener: accept error %d %d", fd, errno); return; } setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, (char *)&one, sizeof(one)); conn = malloc(sizeof(struct tcp_conn)); if (!conn) { log_error("failed to alloc mem"); return; } memset(conn, 0, sizeof(struct tcp_conn)); conn->s = fd; memcpy(&conn->to, &addr, sizeof(struct sockaddr)); list_add_tail(&conn->list, &tcp); i = client_add(fd, process_connection, process_dead); log_debug("client connection %d fd %d", i, fd); } static int setup_tcp_listener(void) { int s, rv; s = socket(local.family, SOCK_STREAM, 0); if (s == -1) { log_error("failed to create tcp socket %s", strerror(errno)); return s; } rv = bind(s, &local.sa6, local.addrlen); if (rv == -1) { log_error("failed to bind socket %s", strerror(errno)); return rv; } rv = listen(s, 5); if (rv == -1) { log_error("failed to listen on socket %s", strerror(errno)); return rv; } return s; } static int booth_tcp_init(void * unused __attribute__((unused))) { int rv; if (!local.local) return -1; rv = setup_tcp_listener(); if (rv < 0) return rv; client_add(rv, process_tcp_listener, NULL); return 0; } static int connect_nonb(int sockfd, const struct sockaddr *saptr, socklen_t salen, int sec) { int flags, n, error; socklen_t len; fd_set rset, wset; struct timeval tval; flags = fcntl(sockfd, F_GETFL, 0); fcntl(sockfd, F_SETFL, flags | O_NONBLOCK); error = 0; if ( (n = connect(sockfd, saptr, salen)) < 0) if (errno != EINPROGRESS) return -1; if (n == 0) goto done; /* connect completed immediately */ FD_ZERO(&rset); FD_SET(sockfd, &rset); wset = rset; tval.tv_sec = sec; tval.tv_usec = 0; if ((n = select(sockfd + 1, &rset, &wset, NULL, sec ? &tval : NULL)) == 0) { /* leave outside function to close */ /* timeout */ /* close(sockfd); */ errno = ETIMEDOUT; return -1; } if (FD_ISSET(sockfd, &rset) || FD_ISSET(sockfd, &wset)) { len = sizeof(error); if (getsockopt(sockfd, SOL_SOCKET, SO_ERROR, &error, &len) < 0) return -1; /* Solaris pending error */ } else { log_error("select error: sockfd not set"); return -1; } done: fcntl(sockfd, F_SETFL, flags); /* restore file status flags */ if (error) { /* leave outside function to close */ /* close(sockfd); */ errno = error; return -1; } return 0; } -static int booth_tcp_open(struct booth_node *to) +int booth_tcp_open(struct booth_node *to) { int s, rv; if (to->tcp_fd >= 0) goto found; - s = socket(BOOTH_PROTO_FAMILY, SOCK_STREAM, 0); - if (s == -1) + s = socket(to->family, SOCK_STREAM, 0); + if (s == -1) { + log_error("cannot create socket of family %d", to->family); return -1; +} rv = connect_nonb(s, (struct sockaddr *)&to->sa6, to->addrlen, 10); if (rv == -1) { if( errno == ETIMEDOUT) log_error("connection to %s timeout", to->addr_string); else log_error("connection to %s error %s", to->addr_string, strerror(errno)); goto error; } to->tcp_fd = s; found: return 1; error: if (s >= 0) close(s); return -1; } -static int booth_tcp_send(struct booth_node *to, void *buf, int len) +int booth_tcp_send(struct booth_node *to, void *buf, int len) { return do_write(to->tcp_fd, buf, len); } static int booth_tcp_recv(struct booth_node *from, void *buf, int len) { return do_read(from->tcp_fd, buf, len); } static int booth_tcp_close(struct booth_node *to) { if (to && to->tcp_fd >= 0) { close(to->tcp_fd); to->tcp_fd = -1; } return 0; } static int booth_tcp_exit(void) { return 0; } static int setup_udp_server(void) { int rv; unsigned int recvbuf_size; udp.s = socket(local.family, SOCK_DGRAM, 0); if (udp.s == -1) { log_error("failed to create udp socket %s", strerror(errno)); return -1; } rv = fcntl(udp.s, F_SETFL, O_NONBLOCK); if (rv == -1) { log_error("failed to set non-blocking operation " "on udp socket: %s", strerror(errno)); close(udp.s); return -1; } rv = bind(udp.s, (struct sockaddr *)&local.sa6, local.addrlen); if (rv == -1) { log_error("failed to bind socket %s", strerror(errno)); close(udp.s); return -1; } recvbuf_size = SOCKET_BUFFER_SIZE; rv = setsockopt(udp.s, SOL_SOCKET, SO_RCVBUF, &recvbuf_size, sizeof(recvbuf_size)); if (rv == -1) { log_error("failed to set recvbuf size"); close(udp.s); return -1; } return udp.s; } static void process_recv(int ci) { struct msghdr msg_recv; struct sockaddr_storage system_from; int received; unsigned char *msg_offset; msg_recv.msg_name = &system_from; msg_recv.msg_namelen = sizeof (struct sockaddr_storage); msg_recv.msg_iov = &udp.iov_recv; msg_recv.msg_iovlen = 1; msg_recv.msg_control = 0; msg_recv.msg_controllen = 0; msg_recv.msg_flags = 0; received = recvmsg(client[ci].fd, &msg_recv, MSG_NOSIGNAL | MSG_DONTWAIT); if (received == -1) return; msg_offset = udp.iov_recv.iov_base; deliver_fn(msg_offset, received); } static int booth_udp_init(void *f) { int myid = -1; memset(&local, 0, sizeof(struct booth_node)); myid = load_myid(); if (myid < 0) { log_error("can't find myself in config file"); return -1; } memset(&udp, 0, sizeof(struct udp_context)); udp.iov_recv.iov_base = udp.iov_buffer; udp.iov_recv.iov_len = FRAME_SIZE_MAX; udp.s = setup_udp_server(); if (udp.s == -1) return -1; deliver_fn = f; client_add(udp.s, process_recv, NULL); return 0; } static int booth_udp_send(struct booth_node *to, void *buf, int len) { struct msghdr msg; struct iovec iovec; unsigned int iov_len; int rv; iovec.iov_base = (void *)buf; iovec.iov_len = len; iov_len = 1; msg.msg_name = &to->sa6; msg.msg_namelen = to->addrlen; msg.msg_iov = (void *)&iovec; msg.msg_iovlen = iov_len; msg.msg_control = 0; msg.msg_controllen = 0; msg.msg_flags = 0; rv = sendmsg(udp.s, &msg, MSG_NOSIGNAL); if (rv < 0) return rv; return 0; } static int booth_udp_broadcast(void *buf, int len) { int i; if (!booth_conf || !booth_conf->node_count) return -1; for (i = 0; i < booth_conf->node_count; i++) booth_udp_send(booth_conf->node+i, buf, len); return 0; } static int booth_udp_exit(void) { return 0; } /* SCTP transport layer has not been developed yet */ static int booth_sctp_init(void *f __attribute__((unused))) { return 0; } static int booth_sctp_send(struct booth_node * to __attribute__((unused)), void *buf __attribute__((unused)), int len __attribute__((unused))) { return 0; } static int booth_sctp_broadcast(void *buf __attribute__((unused)), int len __attribute__((unused))) { return 0; } static int booth_sctp_exit(void) { return 0; } -struct booth_transport booth_transport[] = { - { +const struct booth_transport booth_transport[TRANSPORT_ENTRIES] = { + [TCP] = { .name = "TCP", .init = booth_tcp_init, .get_myid = booth_get_myid, .open = booth_tcp_open, .send = booth_tcp_send, .recv = booth_tcp_recv, .close = booth_tcp_close, .exit = booth_tcp_exit }, - { + [UDP] = { .name = "UDP", .init = booth_udp_init, .get_myid = booth_get_myid, .send = booth_udp_send, .broadcast = booth_udp_broadcast, .exit = booth_udp_exit }, - { + [SCTP] = { .name = "SCTP", .init = booth_sctp_init, .get_myid = booth_get_myid, .send = booth_sctp_send, .broadcast = booth_sctp_broadcast, .exit = booth_sctp_exit } }; +const struct booth_transport *local_transport = booth_transport+TCP; diff --git a/src/transport.h b/src/transport.h index 8601984..94d2c77 100644 --- a/src/transport.h +++ b/src/transport.h @@ -1,77 +1,82 @@ /* * Copyright (C) 2011 Jiaju Zhang * * This program is free software; you can redistribute it and/or * modify it under the terms of the GNU General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * * This software is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * General Public License for more details. * * You should have received a copy of the GNU General Public * License along with this library; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ #ifndef _TRANSPORT_H #define _TRANSPORT_H #include "booth.h" struct booth_node { int nodeid; int type; int local; char addr_string[BOOTH_NAME_LEN]; int tcp_fd; unsigned short family; union { struct sockaddr_in sa4; struct sockaddr_in6 sa6; }; int addrlen; } __attribute__((packed)); typedef enum { TCP = 0, UDP = 1, SCTP = 2, TRANSPORT_ENTRIES = 3, } transport_layer_t; typedef enum { ARBITRATOR = 1, SITE, CLIENT, } node_type_t; struct booth_transport { const char *name; int (*init) (void *); int (*get_myid) (void); int (*open) (struct booth_node *); int (*send) (struct booth_node *, void *, int); int (*recv) (struct booth_node *, void *, int); int (*broadcast) (void *, int); int (*close) (struct booth_node *); int (*exit) (void); }; -struct booth_transport booth_transport[TRANSPORT_ENTRIES]; +const struct booth_transport booth_transport[TRANSPORT_ENTRIES]; int find_myself(struct booth_node **me, int fuzzy_allowed); +int booth_tcp_open(struct booth_node *to); +int booth_tcp_send(struct booth_node *to, void *buf, int len); + inline static void * node_to_addr_pointer(struct booth_node *node) { switch (node->family) { case AF_INET: return &node->sa4.sin_addr; case AF_INET6: return &node->sa6.sin6_addr; } return NULL; } +extern const struct booth_transport *local_transport; + #endif /* _TRANSPORT_H */