diff options
Diffstat (limited to 'source3/winbindd/winbindd.c')
-rw-r--r-- | source3/winbindd/winbindd.c | 1742 |
1 files changed, 1742 insertions, 0 deletions
diff --git a/source3/winbindd/winbindd.c b/source3/winbindd/winbindd.c new file mode 100644 index 0000000..29a24a9 --- /dev/null +++ b/source3/winbindd/winbindd.c @@ -0,0 +1,1742 @@ +/* + Unix SMB/CIFS implementation. + + Winbind daemon for ntdom nss module + + Copyright (C) by Tim Potter 2000-2002 + Copyright (C) Andrew Tridgell 2002 + Copyright (C) Jelmer Vernooij 2003 + Copyright (C) Volker Lendecke 2004 + + 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 3 of the License, or + (at your option) any later version. + + This program 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 program. If not, see <http://www.gnu.org/licenses/>. +*/ + +#include "includes.h" +#include "lib/cmdline/cmdline.h" +#include "winbindd.h" +#include "nsswitch/winbind_client.h" +#include "nsswitch/wb_reqtrans.h" +#include "ntdomain.h" +#include "librpc/rpc/dcesrv_core.h" +#include "librpc/gen_ndr/ndr_lsa_scompat.h" +#include "librpc/gen_ndr/ndr_samr_scompat.h" +#include "librpc/gen_ndr/ndr_winbind_scompat.h" +#include "secrets.h" +#include "rpc_client/cli_netlogon.h" +#include "idmap.h" +#include "lib/addrchange.h" +#include "auth.h" +#include "messages.h" +#include "../lib/util/pidfile.h" +#include "util_cluster.h" +#include "source4/lib/messaging/irpc.h" +#include "source4/lib/messaging/messaging.h" +#include "lib/param/param.h" +#include "lib/async_req/async_sock.h" +#include "libsmb/samlogon_cache.h" +#include "libcli/auth/netlogon_creds_cli.h" +#include "passdb.h" +#include "lib/util/tevent_req_profile.h" +#include "lib/gencache.h" +#include "rpc_server/rpc_config.h" +#include "lib/global_contexts.h" +#include "source3/lib/substitute.h" +#include "winbindd_traceid.h" +#include "lib/util/util_process.h" + +#undef DBGC_CLASS +#define DBGC_CLASS DBGC_WINBIND + +#define SCRUB_CLIENTS_INTERVAL 5 + +static bool client_is_idle(struct winbindd_cli_state *state); +static void remove_client(struct winbindd_cli_state *state); + +static bool interactive = False; + +/* Reload configuration */ + + + +static void winbindd_status(void) +{ + struct winbindd_cli_state *tmp; + + DEBUG(0, ("winbindd status:\n")); + + /* Print client state information */ + + DEBUG(0, ("\t%d clients currently active\n", winbindd_num_clients())); + + if (DEBUGLEVEL >= 2 && winbindd_num_clients()) { + DEBUG(2, ("\tclient list:\n")); + for(tmp = winbindd_client_list(); tmp; tmp = tmp->next) { + DEBUGADD(2, ("\t\tpid %lu, sock %d (%s)\n", + (unsigned long)tmp->pid, tmp->sock, + client_is_idle(tmp) ? "idle" : "active")); + } + } +} + +/* + handle stdin becoming readable when we are in --foreground mode + */ +static void winbindd_stdin_handler(struct tevent_context *ev, + struct tevent_fd *fde, + uint16_t flags, + void *private_data) +{ + char c; + if (read(0, &c, 1) != 1) { + bool *is_parent = talloc_get_type_abort(private_data, bool); + + /* we have reached EOF on stdin, which means the + parent has exited. Shutdown the server */ + DEBUG(0,("EOF on stdin (is_parent=%d)\n", + (int)*is_parent)); + winbindd_terminate(*is_parent); + } +} + +bool winbindd_setup_stdin_handler(bool parent, bool foreground) +{ + bool *is_parent; + + if (foreground) { + struct stat st; + + is_parent = talloc(global_event_context(), bool); + if (!is_parent) { + return false; + } + + *is_parent = parent; + + /* if we are running in the foreground then look for + EOF on stdin, and exit if it happens. This allows + us to die if the parent process dies + Only do this on a pipe or socket, no other device. + */ + if (fstat(0, &st) != 0) { + return false; + } + if (S_ISFIFO(st.st_mode) || S_ISSOCK(st.st_mode)) { + tevent_add_fd(global_event_context(), + is_parent, + 0, + TEVENT_FD_READ, + winbindd_stdin_handler, + is_parent); + } + } + + return true; +} + +static void winbindd_sig_chld_handler(struct tevent_context *ev, + struct tevent_signal *se, + int signum, + int count, + void *siginfo, + void *private_data) +{ + pid_t pid; + + while ((pid = waitpid(-1, NULL, WNOHANG)) > 0) { + winbind_child_died(pid); + } +} + +static bool winbindd_setup_sig_chld_handler(void) +{ + struct tevent_signal *se; + + se = tevent_add_signal(global_event_context(), + global_event_context(), + SIGCHLD, 0, + winbindd_sig_chld_handler, + NULL); + if (!se) { + return false; + } + + return true; +} + +static void winbindd_sig_usr2_handler(struct tevent_context *ev, + struct tevent_signal *se, + int signum, + int count, + void *siginfo, + void *private_data) +{ + winbindd_status(); +} + +static bool winbindd_setup_sig_usr2_handler(void) +{ + struct tevent_signal *se; + + se = tevent_add_signal(global_event_context(), + global_event_context(), + SIGUSR2, 0, + winbindd_sig_usr2_handler, + NULL); + if (!se) { + return false; + } + + return true; +} + +/* React on 'smbcontrol winbindd shutdown' in the same way as on SIGTERM*/ +static void msg_shutdown(struct messaging_context *msg, + void *private_data, + uint32_t msg_type, + struct server_id server_id, + DATA_BLOB *data) +{ + /* only the parent waits for this message */ + DEBUG(0,("Got shutdown message\n")); + winbindd_terminate(true); +} + + +static void winbind_msg_validate_cache(struct messaging_context *msg_ctx, + void *private_data, + uint32_t msg_type, + struct server_id server_id, + DATA_BLOB *data) +{ + uint8_t ret; + pid_t child_pid; + NTSTATUS status; + + DEBUG(10, ("winbindd_msg_validate_cache: got validate-cache " + "message.\n")); + + /* + * call the validation code from a child: + * so we don't block the main winbindd and the validation + * code can safely use fork/waitpid... + */ + child_pid = fork(); + + if (child_pid == -1) { + DEBUG(1, ("winbind_msg_validate_cache: Could not fork: %s\n", + strerror(errno))); + return; + } + + if (child_pid != 0) { + /* parent */ + DEBUG(5, ("winbind_msg_validate_cache: child created with " + "pid %d.\n", (int)child_pid)); + return; + } + + /* child */ + + status = winbindd_reinit_after_fork(NULL, NULL); + if (!NT_STATUS_IS_OK(status)) { + DEBUG(1, ("winbindd_reinit_after_fork failed: %s\n", + nt_errstr(status))); + _exit(0); + } + + /* install default SIGCHLD handler: validation code uses fork/waitpid */ + CatchSignal(SIGCHLD, SIG_DFL); + + process_set_title("wb: check cache", "validate cache child"); + + ret = (uint8_t)winbindd_validate_cache_nobackup(); + DEBUG(10, ("winbindd_msg_validata_cache: got return value %d\n", ret)); + messaging_send_buf(msg_ctx, server_id, MSG_WINBIND_VALIDATE_CACHE, &ret, + (size_t)1); + _exit(0); +} + +static struct winbindd_bool_dispatch_table { + enum winbindd_cmd cmd; + bool (*fn)(struct winbindd_cli_state *state); + const char *cmd_name; +} bool_dispatch_table[] = { + { WINBINDD_INTERFACE_VERSION, + winbindd_interface_version, + "INTERFACE_VERSION" }, + { WINBINDD_INFO, + winbindd_info, + "INFO" }, + { WINBINDD_PING, + winbindd_ping, + "PING" }, + { WINBINDD_DOMAIN_NAME, + winbindd_domain_name, + "DOMAIN_NAME" }, + { WINBINDD_NETBIOS_NAME, + winbindd_netbios_name, + "NETBIOS_NAME" }, + { WINBINDD_DC_INFO, + winbindd_dc_info, + "DC_INFO" }, + { WINBINDD_CCACHE_NTLMAUTH, + winbindd_ccache_ntlm_auth, + "NTLMAUTH" }, + { WINBINDD_CCACHE_SAVE, + winbindd_ccache_save, + "CCACHE_SAVE" }, + { WINBINDD_PRIV_PIPE_DIR, + winbindd_priv_pipe_dir, + "WINBINDD_PRIV_PIPE_DIR" }, + { WINBINDD_LIST_TRUSTDOM, + winbindd_list_trusted_domains, + "LIST_TRUSTDOM" }, +}; + +struct winbindd_async_dispatch_table { + enum winbindd_cmd cmd; + const char *cmd_name; + struct tevent_req *(*send_req)(TALLOC_CTX *mem_ctx, + struct tevent_context *ev, + struct winbindd_cli_state *cli, + struct winbindd_request *request); + NTSTATUS (*recv_req)(struct tevent_req *req, + struct winbindd_response *presp); +}; + +static struct winbindd_async_dispatch_table async_nonpriv_table[] = { + { WINBINDD_LOOKUPSID, "LOOKUPSID", + winbindd_lookupsid_send, winbindd_lookupsid_recv }, + { WINBINDD_LOOKUPSIDS, "LOOKUPSIDS", + winbindd_lookupsids_send, winbindd_lookupsids_recv }, + { WINBINDD_LOOKUPNAME, "LOOKUPNAME", + winbindd_lookupname_send, winbindd_lookupname_recv }, + { WINBINDD_SIDS_TO_XIDS, "SIDS_TO_XIDS", + winbindd_sids_to_xids_send, winbindd_sids_to_xids_recv }, + { WINBINDD_XIDS_TO_SIDS, "XIDS_TO_SIDS", + winbindd_xids_to_sids_send, winbindd_xids_to_sids_recv }, + { WINBINDD_GETPWSID, "GETPWSID", + winbindd_getpwsid_send, winbindd_getpwsid_recv }, + { WINBINDD_GETPWNAM, "GETPWNAM", + winbindd_getpwnam_send, winbindd_getpwnam_recv }, + { WINBINDD_GETPWUID, "GETPWUID", + winbindd_getpwuid_send, winbindd_getpwuid_recv }, + { WINBINDD_GETSIDALIASES, "GETSIDALIASES", + winbindd_getsidaliases_send, winbindd_getsidaliases_recv }, + { WINBINDD_GETUSERDOMGROUPS, "GETUSERDOMGROUPS", + winbindd_getuserdomgroups_send, winbindd_getuserdomgroups_recv }, + { WINBINDD_GETGROUPS, "GETGROUPS", + winbindd_getgroups_send, winbindd_getgroups_recv }, + { WINBINDD_SHOW_SEQUENCE, "SHOW_SEQUENCE", + winbindd_show_sequence_send, winbindd_show_sequence_recv }, + { WINBINDD_GETGRGID, "GETGRGID", + winbindd_getgrgid_send, winbindd_getgrgid_recv }, + { WINBINDD_GETGRNAM, "GETGRNAM", + winbindd_getgrnam_send, winbindd_getgrnam_recv }, + { WINBINDD_GETUSERSIDS, "GETUSERSIDS", + winbindd_getusersids_send, winbindd_getusersids_recv }, + { WINBINDD_LOOKUPRIDS, "LOOKUPRIDS", + winbindd_lookuprids_send, winbindd_lookuprids_recv }, + { WINBINDD_SETPWENT, "SETPWENT", + winbindd_setpwent_send, winbindd_setpwent_recv }, + { WINBINDD_GETPWENT, "GETPWENT", + winbindd_getpwent_send, winbindd_getpwent_recv }, + { WINBINDD_ENDPWENT, "ENDPWENT", + winbindd_endpwent_send, winbindd_endpwent_recv }, + { WINBINDD_DSGETDCNAME, "DSGETDCNAME", + winbindd_dsgetdcname_send, winbindd_dsgetdcname_recv }, + { WINBINDD_GETDCNAME, "GETDCNAME", + winbindd_getdcname_send, winbindd_getdcname_recv }, + { WINBINDD_SETGRENT, "SETGRENT", + winbindd_setgrent_send, winbindd_setgrent_recv }, + { WINBINDD_GETGRENT, "GETGRENT", + winbindd_getgrent_send, winbindd_getgrent_recv }, + { WINBINDD_ENDGRENT, "ENDGRENT", + winbindd_endgrent_send, winbindd_endgrent_recv }, + { WINBINDD_LIST_USERS, "LIST_USERS", + winbindd_list_users_send, winbindd_list_users_recv }, + { WINBINDD_LIST_GROUPS, "LIST_GROUPS", + winbindd_list_groups_send, winbindd_list_groups_recv }, + { WINBINDD_CHECK_MACHACC, "CHECK_MACHACC", + winbindd_check_machine_acct_send, winbindd_check_machine_acct_recv }, + { WINBINDD_PING_DC, "PING_DC", + winbindd_ping_dc_send, winbindd_ping_dc_recv }, + { WINBINDD_PAM_AUTH, "PAM_AUTH", + winbindd_pam_auth_send, winbindd_pam_auth_recv }, + { WINBINDD_PAM_LOGOFF, "PAM_LOGOFF", + winbindd_pam_logoff_send, winbindd_pam_logoff_recv }, + { WINBINDD_PAM_CHAUTHTOK, "PAM_CHAUTHTOK", + winbindd_pam_chauthtok_send, winbindd_pam_chauthtok_recv }, + { WINBINDD_PAM_CHNG_PSWD_AUTH_CRAP, "PAM_CHNG_PSWD_AUTH_CRAP", + winbindd_pam_chng_pswd_auth_crap_send, + winbindd_pam_chng_pswd_auth_crap_recv }, + { WINBINDD_WINS_BYIP, "WINS_BYIP", + winbindd_wins_byip_send, winbindd_wins_byip_recv }, + { WINBINDD_WINS_BYNAME, "WINS_BYNAME", + winbindd_wins_byname_send, winbindd_wins_byname_recv }, + { WINBINDD_DOMAIN_INFO, "DOMAIN_INFO", + winbindd_domain_info_send, winbindd_domain_info_recv }, + + { 0, NULL, NULL, NULL } +}; + +static struct winbindd_async_dispatch_table async_priv_table[] = { + { WINBINDD_ALLOCATE_UID, "ALLOCATE_UID", + winbindd_allocate_uid_send, winbindd_allocate_uid_recv }, + { WINBINDD_ALLOCATE_GID, "ALLOCATE_GID", + winbindd_allocate_gid_send, winbindd_allocate_gid_recv }, + { WINBINDD_CHANGE_MACHACC, "CHANGE_MACHACC", + winbindd_change_machine_acct_send, winbindd_change_machine_acct_recv }, + { WINBINDD_PAM_AUTH_CRAP, "PAM_AUTH_CRAP", + winbindd_pam_auth_crap_send, winbindd_pam_auth_crap_recv }, + + { 0, NULL, NULL, NULL } +}; + +struct process_request_state { + struct winbindd_cli_state *cli_state; + struct tevent_context *ev; +}; + +static void process_request_done(struct tevent_req *subreq); +static void process_request_written(struct tevent_req *subreq); + +static struct tevent_req *process_request_send( + TALLOC_CTX *mem_ctx, + struct tevent_context *ev, + struct winbindd_cli_state *cli_state) +{ + struct tevent_req *req, *subreq; + struct process_request_state *state; + struct winbindd_async_dispatch_table *atable; + enum winbindd_cmd cmd = cli_state->request->cmd; + size_t i; + bool ok; + static uint64_t request_index = 1; + + /* + * debug traceid values: + * 0 .. inactive + * 1 .. not processing a winbind request, but in other code (timers) + * >=2 .. winbind request processing + */ + if (debug_traceid_get() != 0) { + request_index = ++request_index == 0 ? 2 : request_index; + debug_traceid_set(request_index); + } + req = tevent_req_create(mem_ctx, &state, + struct process_request_state); + if (req == NULL) { + return NULL; + } + state->cli_state = cli_state; + state->ev = ev; + + ok = tevent_req_set_profile(req); + if (!ok) { + return tevent_req_post(req, ev); + } + + SMB_ASSERT(cli_state->mem_ctx == NULL); + cli_state->mem_ctx = talloc_named(cli_state, 0, "winbind request"); + if (tevent_req_nomem(cli_state->mem_ctx, req)) { + return tevent_req_post(req, ev); + } + + cli_state->response = talloc_zero( + cli_state->mem_ctx, + struct winbindd_response); + if (tevent_req_nomem(cli_state->response, req)) { + return tevent_req_post(req, ev); + } + cli_state->response->result = WINBINDD_PENDING; + cli_state->response->length = sizeof(struct winbindd_response); + + /* Remember who asked us. */ + cli_state->pid = cli_state->request->pid; + memcpy(cli_state->client_name, + cli_state->request->client_name, + sizeof(cli_state->client_name)); + + cli_state->cmd_name = "unknown request"; + cli_state->recv_fn = NULL; + + /* client is newest */ + winbindd_promote_client(cli_state); + + for (atable = async_nonpriv_table; atable->send_req; atable += 1) { + if (cmd == atable->cmd) { + break; + } + } + + if ((atable->send_req == NULL) && cli_state->privileged) { + for (atable = async_priv_table; atable->send_req; + atable += 1) { + if (cmd == atable->cmd) { + break; + } + } + } + + if (atable->send_req != NULL) { + cli_state->cmd_name = atable->cmd_name; + cli_state->recv_fn = atable->recv_req; + + DBG_NOTICE("[%s (%d)] Handling async request: %s\n", + cli_state->client_name, + (int)cli_state->pid, + cli_state->cmd_name); + + subreq = atable->send_req( + state, + state->ev, + cli_state, + cli_state->request); + if (tevent_req_nomem(subreq, req)) { + return tevent_req_post(req, ev); + } + tevent_req_set_callback(subreq, process_request_done, req); + return req; + } + + for (i=0; i<ARRAY_SIZE(bool_dispatch_table); i++) { + if (cmd == bool_dispatch_table[i].cmd) { + break; + } + } + + ok = false; + + if (i < ARRAY_SIZE(bool_dispatch_table)) { + cli_state->cmd_name = bool_dispatch_table[i].cmd_name; + + DBG_DEBUG("process_request: request fn %s\n", + bool_dispatch_table[i].cmd_name); + ok = bool_dispatch_table[i].fn(cli_state); + } + + cli_state->response->result = ok ? WINBINDD_OK : WINBINDD_ERROR; + + TALLOC_FREE(cli_state->io_req); + TALLOC_FREE(cli_state->request); + + subreq = wb_resp_write_send( + state, + state->ev, + cli_state->out_queue, + cli_state->sock, + cli_state->response); + if (tevent_req_nomem(subreq, req)) { + return tevent_req_post(req, ev); + } + tevent_req_set_callback(subreq, process_request_written, req); + + cli_state->io_req = subreq; + + return req; +} + +static void process_request_done(struct tevent_req *subreq) +{ + struct tevent_req *req = tevent_req_callback_data( + subreq, struct tevent_req); + struct process_request_state *state = tevent_req_data( + req, struct process_request_state); + struct winbindd_cli_state *cli_state = state->cli_state; + NTSTATUS status; + bool ok; + + status = cli_state->recv_fn(subreq, cli_state->response); + TALLOC_FREE(subreq); + + DBG_NOTICE("[%s(%d):%s]: %s\n", + cli_state->client_name, + (int)cli_state->pid, + cli_state->cmd_name, + nt_errstr(status)); + + ok = NT_STATUS_IS_OK(status); + cli_state->response->result = ok ? WINBINDD_OK : WINBINDD_ERROR; + + TALLOC_FREE(cli_state->io_req); + TALLOC_FREE(cli_state->request); + + subreq = wb_resp_write_send( + state, + state->ev, + cli_state->out_queue, + cli_state->sock, + cli_state->response); + if (tevent_req_nomem(subreq, req)) { + return; + } + tevent_req_set_callback(subreq, process_request_written, req); + + cli_state->io_req = subreq; +} + +static void process_request_written(struct tevent_req *subreq) +{ + struct tevent_req *req = tevent_req_callback_data( + subreq, struct tevent_req); + struct process_request_state *state = tevent_req_data( + req, struct process_request_state); + struct winbindd_cli_state *cli_state = state->cli_state; + ssize_t ret; + int err; + + cli_state->io_req = NULL; + + ret = wb_resp_write_recv(subreq, &err); + TALLOC_FREE(subreq); + if (ret == -1) { + tevent_req_nterror(req, map_nt_error_from_unix(err)); + return; + } + + DBG_DEBUG("[%s(%d):%s]: delivered response to client\n", + cli_state->client_name, + (int)cli_state->pid, + cli_state->cmd_name); + + TALLOC_FREE(cli_state->mem_ctx); + cli_state->response = NULL; + cli_state->cmd_name = "no request"; + cli_state->recv_fn = NULL; + + tevent_req_done(req); +} + +static NTSTATUS process_request_recv( + struct tevent_req *req, + TALLOC_CTX *mem_ctx, + struct tevent_req_profile **profile) +{ + NTSTATUS status; + + if (tevent_req_is_nterror(req, &status)) { + tevent_req_received(req); + return status; + } + + *profile = tevent_req_move_profile(req, mem_ctx); + tevent_req_received(req); + return NT_STATUS_OK; +} + +/* + * This is the main event loop of winbind requests. It goes through a + * state-machine of 3 read/write requests, 4 if you have extra data to send. + * + * An idle winbind client has a read request of 4 bytes outstanding, + * finalizing function is request_len_recv, checking the length. request_recv + * then processes the packet. The processing function then at some point has + * to call request_finished which schedules sending the response. + */ + +static void winbind_client_request_read(struct tevent_req *req); +static void winbind_client_activity(struct tevent_req *req); +static void winbind_client_processed(struct tevent_req *req); + +/* Process a new connection by adding it to the client connection list */ + +static void new_connection(int listen_sock, bool privileged) +{ + struct sockaddr_un sunaddr; + struct winbindd_cli_state *state; + struct tevent_req *req; + socklen_t len; + int sock; + + /* Accept connection */ + + len = sizeof(sunaddr); + + sock = accept(listen_sock, (struct sockaddr *)(void *)&sunaddr, &len); + + if (sock == -1) { + if (errno != EINTR) { + D_ERR("Failed to accept socket: %s\n", strerror(errno)); + } + return; + } + smb_set_close_on_exec(sock); + + D_INFO("Accepted client socket %d\n", sock); + + /* Create new connection structure */ + + if ((state = talloc_zero(NULL, struct winbindd_cli_state)) == NULL) { + close(sock); + return; + } + + state->sock = sock; + + state->out_queue = tevent_queue_create(state, "winbind client reply"); + if (state->out_queue == NULL) { + close(sock); + TALLOC_FREE(state); + return; + } + + state->privileged = privileged; + + req = wb_req_read_send(state, global_event_context(), state->sock, + WINBINDD_MAX_EXTRA_DATA); + if (req == NULL) { + TALLOC_FREE(state); + close(sock); + return; + } + tevent_req_set_callback(req, winbind_client_request_read, state); + state->io_req = req; + + /* Add to connection list */ + + winbindd_add_client(state); +} + +static void winbind_client_request_read(struct tevent_req *req) +{ + struct winbindd_cli_state *state = tevent_req_callback_data( + req, struct winbindd_cli_state); + ssize_t ret; + int err; + + state->io_req = NULL; + + ret = wb_req_read_recv(req, state, &state->request, &err); + TALLOC_FREE(req); + if (ret == -1) { + if (err == EPIPE) { + DEBUG(6, ("closing socket %d, client exited\n", + state->sock)); + } else { + DEBUG(2, ("Could not read client request from fd %d: " + "%s\n", state->sock, strerror(err))); + } + close(state->sock); + state->sock = -1; + remove_client(state); + return; + } + + req = wait_for_read_send(state, global_event_context(), state->sock, + true); + if (req == NULL) { + DEBUG(0, ("winbind_client_request_read[%d:%s]:" + " wait_for_read_send failed - removing client\n", + (int)state->pid, state->cmd_name)); + remove_client(state); + return; + } + tevent_req_set_callback(req, winbind_client_activity, state); + state->io_req = req; + + req = process_request_send(state, global_event_context(), state); + if (req == NULL) { + DBG_ERR("process_request_send failed\n"); + remove_client(state); + return; + } + tevent_req_set_callback(req, winbind_client_processed, state); +} + +static void winbind_client_activity(struct tevent_req *req) +{ + struct winbindd_cli_state *state = + tevent_req_callback_data(req, struct winbindd_cli_state); + int err; + bool has_data; + + has_data = wait_for_read_recv(req, &err); + + if (has_data) { + DEBUG(0, ("winbind_client_activity[%d:%s]:" + "unexpected data from client - removing client\n", + (int)state->pid, state->cmd_name)); + } else { + if (err == EPIPE) { + DEBUG(6, ("winbind_client_activity[%d:%s]: " + "client has closed connection - removing " + "client\n", + (int)state->pid, state->cmd_name)); + } else { + DEBUG(2, ("winbind_client_activity[%d:%s]: " + "client socket error (%s) - removing " + "client\n", + (int)state->pid, state->cmd_name, + strerror(err))); + } + } + + remove_client(state); +} + +static void winbind_client_processed(struct tevent_req *req) +{ + struct winbindd_cli_state *cli_state = tevent_req_callback_data( + req, struct winbindd_cli_state); + struct tevent_req_profile *profile = NULL; + struct timeval start, stop, diff; + int threshold; + NTSTATUS status; + + status = process_request_recv(req, cli_state, &profile); + TALLOC_FREE(req); + if (!NT_STATUS_IS_OK(status)) { + DBG_DEBUG("process_request failed: %s\n", nt_errstr(status)); + remove_client(cli_state); + return; + } + + tevent_req_profile_get_start(profile, NULL, &start); + tevent_req_profile_get_stop(profile, NULL, &stop); + diff = tevent_timeval_until(&start, &stop); + + threshold = lp_parm_int(-1, "winbind", "request profile threshold", 60); + + if (diff.tv_sec >= threshold) { + int depth; + char *str; + + depth = lp_parm_int( + -1, + "winbind", + "request profile depth", + INT_MAX); + + DBG_ERR("request took %u.%.6u seconds\n", + (unsigned)diff.tv_sec, (unsigned)diff.tv_usec); + + str = tevent_req_profile_string( + talloc_tos(), profile, 0, depth); + if (str != NULL) { + /* No "\n", already contained in "str" */ + DEBUGADD(0, ("%s", str)); + } + TALLOC_FREE(str); + } + + TALLOC_FREE(profile); + + req = wb_req_read_send( + cli_state, + global_event_context(), + cli_state->sock, + WINBINDD_MAX_EXTRA_DATA); + if (req == NULL) { + remove_client(cli_state); + return; + } + tevent_req_set_callback(req, winbind_client_request_read, cli_state); + cli_state->io_req = req; +} + +/* Remove a client connection from client connection list */ + +static void remove_client(struct winbindd_cli_state *state) +{ + /* It's a dead client - hold a funeral */ + + if (state == NULL) { + return; + } + + /* + * We need to remove a pending wb_req_read_* + * or wb_resp_write_* request before closing the + * socket. + * + * This is important as they might have used tevent_add_fd() and we + * use the epoll * backend on linux. So we must remove the tevent_fd + * before closing the fd. + * + * Otherwise we might hit a race with close_conns_after_fork() (via + * winbindd_reinit_after_fork()) where a file descriptor + * is still open in a child, which means it's still active in + * the parents epoll queue, but the related tevent_fd is already + * already gone in the parent. + * + * See bug #11141. + */ + TALLOC_FREE(state->io_req); + + if (state->sock != -1) { + char c = 0; + int nwritten; + + /* tell client, we are closing ... */ + nwritten = write(state->sock, &c, sizeof(c)); + if (nwritten == -1) { + DEBUG(2, ("final write to client failed: %s\n", + strerror(errno))); + } + + /* Close socket */ + + close(state->sock); + state->sock = -1; + } + + TALLOC_FREE(state->mem_ctx); + + /* Remove from list and free */ + + winbindd_remove_client(state); + TALLOC_FREE(state); +} + +/* Is a client idle? */ + +static bool client_is_idle(struct winbindd_cli_state *state) { + return (state->request == NULL && + state->response == NULL && + !state->pwent_state && !state->grent_state); +} + +/* Shutdown client connection which has been idle for the longest time */ + +static bool remove_idle_client(void) +{ + struct winbindd_cli_state *state, *remove_state = NULL; + int nidle = 0; + + for (state = winbindd_client_list(); state; state = state->next) { + if (client_is_idle(state)) { + nidle++; + /* list is sorted by access time */ + remove_state = state; + } + } + + if (remove_state) { + DEBUG(5,("Found %d idle client connections, shutting down sock %d, pid %u\n", + nidle, remove_state->sock, (unsigned int)remove_state->pid)); + remove_client(remove_state); + return True; + } + + return False; +} + +/* + * Terminate all clients whose requests have taken longer than + * "winbind request timeout" seconds to process, or have been + * idle for more than "winbind request timeout" seconds. + */ + +static void remove_timed_out_clients(void) +{ + struct winbindd_cli_state *state, *prev = NULL; + time_t curr_time = time(NULL); + int timeout_val = lp_winbind_request_timeout(); + + for (state = winbindd_client_list_tail(); state; state = prev) { + time_t expiry_time; + + prev = winbindd_client_list_prev(state); + expiry_time = state->last_access + timeout_val; + + if (curr_time <= expiry_time) { + /* list is sorted, previous clients in + list are newer */ + break; + } + + if (client_is_idle(state)) { + DEBUG(5,("Idle client timed out, " + "shutting down sock %d, pid %u\n", + state->sock, + (unsigned int)state->pid)); + } else { + DEBUG(5,("Client request timed out, " + "shutting down sock %d, pid %u\n", + state->sock, + (unsigned int)state->pid)); + } + + remove_client(state); + } +} + +static void winbindd_scrub_clients_handler(struct tevent_context *ev, + struct tevent_timer *te, + struct timeval current_time, + void *private_data) +{ + remove_timed_out_clients(); + if (tevent_add_timer(ev, ev, + timeval_current_ofs(SCRUB_CLIENTS_INTERVAL, 0), + winbindd_scrub_clients_handler, NULL) == NULL) { + DEBUG(0, ("winbindd: failed to reschedule client scrubber\n")); + exit(1); + } +} + +struct winbindd_listen_state { + bool privileged; + int fd; +}; + +static void winbindd_listen_fde_handler(struct tevent_context *ev, + struct tevent_fd *fde, + uint16_t flags, + void *private_data) +{ + struct winbindd_listen_state *s = talloc_get_type_abort(private_data, + struct winbindd_listen_state); + + while (winbindd_num_clients() > lp_winbind_max_clients() - 1) { + DEBUG(5,("winbindd: Exceeding %d client " + "connections, removing idle " + "connection.\n", lp_winbind_max_clients())); + if (!remove_idle_client()) { + DEBUG(0,("winbindd: Exceeding %d " + "client connections, no idle " + "connection found\n", + lp_winbind_max_clients())); + break; + } + } + remove_timed_out_clients(); + new_connection(s->fd, s->privileged); +} + +/* + * Winbindd socket accessor functions + */ + +static bool winbindd_setup_listeners(void) +{ + struct winbindd_listen_state *pub_state = NULL; + struct winbindd_listen_state *priv_state = NULL; + struct tevent_fd *fde; + int rc; + char *socket_path; + + pub_state = talloc(global_event_context(), + struct winbindd_listen_state); + if (!pub_state) { + goto failed; + } + + pub_state->privileged = false; + pub_state->fd = create_pipe_sock( + lp_winbindd_socket_directory(), WINBINDD_SOCKET_NAME, 0755); + if (pub_state->fd == -1) { + goto failed; + } + rc = listen(pub_state->fd, 5); + if (rc < 0) { + goto failed; + } + + fde = tevent_add_fd(global_event_context(), pub_state, pub_state->fd, + TEVENT_FD_READ, winbindd_listen_fde_handler, + pub_state); + if (fde == NULL) { + close(pub_state->fd); + goto failed; + } + tevent_fd_set_auto_close(fde); + + priv_state = talloc(global_event_context(), + struct winbindd_listen_state); + if (!priv_state) { + goto failed; + } + + socket_path = get_winbind_priv_pipe_dir(); + if (socket_path == NULL) { + goto failed; + } + + priv_state->privileged = true; + priv_state->fd = create_pipe_sock( + socket_path, WINBINDD_SOCKET_NAME, 0750); + TALLOC_FREE(socket_path); + if (priv_state->fd == -1) { + goto failed; + } + rc = listen(priv_state->fd, 5); + if (rc < 0) { + goto failed; + } + + fde = tevent_add_fd(global_event_context(), priv_state, + priv_state->fd, TEVENT_FD_READ, + winbindd_listen_fde_handler, priv_state); + if (fde == NULL) { + close(priv_state->fd); + goto failed; + } + tevent_fd_set_auto_close(fde); + + winbindd_scrub_clients_handler(global_event_context(), NULL, + timeval_current(), NULL); + return true; +failed: + TALLOC_FREE(pub_state); + TALLOC_FREE(priv_state); + return false; +} + +static void winbindd_register_handlers(struct messaging_context *msg_ctx, + bool foreground) +{ + bool scan_trusts = true; + NTSTATUS status; + struct tevent_timer *te = NULL; + + /* Setup signal handlers */ + + if (!winbindd_setup_sig_term_handler(true)) + exit(1); + if (!winbindd_setup_stdin_handler(true, foreground)) + exit(1); + if (!winbindd_setup_sig_hup_handler(NULL)) + exit(1); + if (!winbindd_setup_sig_chld_handler()) + exit(1); + if (!winbindd_setup_sig_usr2_handler()) + exit(1); + + CatchSignal(SIGPIPE, SIG_IGN); /* Ignore sigpipe */ + + /* + * Ensure all cache and idmap caches are consistent + * and initialized before we startup. + */ + if (!winbindd_cache_validate_and_initialize()) { + exit(1); + } + + /* React on 'smbcontrol winbindd reload-config' in the same way + as to SIGHUP signal */ + messaging_register(msg_ctx, NULL, + MSG_SMB_CONF_UPDATED, + winbindd_msg_reload_services_parent); + messaging_register(msg_ctx, NULL, + MSG_SHUTDOWN, msg_shutdown); + + /* Handle online/offline messages. */ + messaging_register(msg_ctx, NULL, + MSG_WINBIND_OFFLINE, winbind_msg_offline); + messaging_register(msg_ctx, NULL, + MSG_WINBIND_ONLINE, winbind_msg_online); + messaging_register(msg_ctx, NULL, + MSG_WINBIND_ONLINESTATUS, winbind_msg_onlinestatus); + + /* Handle domain online/offline messages for domains */ + messaging_register(global_messaging_context(), NULL, + MSG_WINBIND_DOMAIN_OFFLINE, winbind_msg_domain_offline); + messaging_register(global_messaging_context(), NULL, + MSG_WINBIND_DOMAIN_ONLINE, winbind_msg_domain_online); + + messaging_register(msg_ctx, NULL, + MSG_WINBIND_VALIDATE_CACHE, + winbind_msg_validate_cache); + + messaging_register(msg_ctx, NULL, + MSG_WINBIND_DUMP_DOMAIN_LIST, + winbind_msg_dump_domain_list); + + messaging_register(msg_ctx, NULL, + MSG_WINBIND_IP_DROPPED, + winbind_msg_ip_dropped_parent); + + /* Register handler for MSG_DEBUG. */ + messaging_register(msg_ctx, NULL, + MSG_DEBUG, + winbind_msg_debug); + + messaging_register(msg_ctx, NULL, + MSG_WINBIND_DISCONNECT_DC, + winbind_disconnect_dc_parent); + + netsamlogon_cache_init(); /* Non-critical */ + + /* clear the cached list of trusted domains */ + + wcache_tdc_clear(); + + if (!init_domain_list()) { + DEBUG(0,("unable to initialize domain list\n")); + exit(1); + } + + status = init_idmap_child(global_event_context()); + if (NT_STATUS_IS_ERR(status)) { + DBG_ERR("Unable to start idmap child: %s\n", nt_errstr(status)); + exit(1); + } + + status = init_locator_child(global_event_context()); + if (NT_STATUS_IS_ERR(status)) { + DBG_ERR("Unable to start locator child: %s\n", nt_errstr(status)); + exit(1); + } + + smb_nscd_flush_user_cache(); + smb_nscd_flush_group_cache(); + + if (!lp_winbind_scan_trusted_domains()) { + scan_trusts = false; + } + + if (!lp_allow_trusted_domains()) { + scan_trusts = false; + } + + if (IS_DC) { + scan_trusts = false; + } + + if (scan_trusts) { + if (tevent_add_timer(global_event_context(), NULL, timeval_zero(), + rescan_trusted_domains, NULL) == NULL) { + DEBUG(0, ("Could not trigger rescan_trusted_domains()\n")); + exit(1); + } + } + + te = tevent_add_timer(global_event_context(), + NULL, + timeval_zero(), + winbindd_ping_offline_domains, + NULL); + if (te == NULL) { + DBG_ERR("Failed to schedule winbindd_ping_offline_domains()\n"); + exit(1); + } + + status = wb_irpc_register(); + + if (!NT_STATUS_IS_OK(status)) { + DEBUG(0, ("Could not register IRPC handlers\n")); + exit(1); + } +} + +struct winbindd_addrchanged_state { + struct addrchange_context *ctx; + struct tevent_context *ev; + struct messaging_context *msg_ctx; +}; + +static void winbindd_addr_changed(struct tevent_req *req); + +static void winbindd_init_addrchange(TALLOC_CTX *mem_ctx, + struct tevent_context *ev, + struct messaging_context *msg_ctx) +{ + struct winbindd_addrchanged_state *state; + struct tevent_req *req; + NTSTATUS status; + + state = talloc(mem_ctx, struct winbindd_addrchanged_state); + if (state == NULL) { + DEBUG(10, ("talloc failed\n")); + return; + } + state->ev = ev; + state->msg_ctx = msg_ctx; + + status = addrchange_context_create(state, &state->ctx); + if (!NT_STATUS_IS_OK(status)) { + DEBUG(10, ("addrchange_context_create failed: %s\n", + nt_errstr(status))); + TALLOC_FREE(state); + return; + } + req = addrchange_send(state, ev, state->ctx); + if (req == NULL) { + DEBUG(0, ("addrchange_send failed\n")); + TALLOC_FREE(state); + return; + } + tevent_req_set_callback(req, winbindd_addr_changed, state); +} + +static void winbindd_addr_changed(struct tevent_req *req) +{ + struct winbindd_addrchanged_state *state = tevent_req_callback_data( + req, struct winbindd_addrchanged_state); + enum addrchange_type type; + struct sockaddr_storage addr; + NTSTATUS status; + + status = addrchange_recv(req, &type, &addr); + TALLOC_FREE(req); + if (!NT_STATUS_IS_OK(status)) { + DEBUG(10, ("addrchange_recv failed: %s, stop listening\n", + nt_errstr(status))); + TALLOC_FREE(state); + return; + } + if (type == ADDRCHANGE_DEL) { + char addrstr[INET6_ADDRSTRLEN]; + DATA_BLOB blob; + + print_sockaddr(addrstr, sizeof(addrstr), &addr); + + DEBUG(3, ("winbindd: kernel (AF_NETLINK) dropped ip %s\n", + addrstr)); + + blob = data_blob_const(addrstr, strlen(addrstr)+1); + + status = messaging_send(state->msg_ctx, + messaging_server_id(state->msg_ctx), + MSG_WINBIND_IP_DROPPED, &blob); + if (!NT_STATUS_IS_OK(status)) { + DEBUG(10, ("messaging_send failed: %s - ignoring\n", + nt_errstr(status))); + } + } + req = addrchange_send(state, state->ev, state->ctx); + if (req == NULL) { + DEBUG(0, ("addrchange_send failed\n")); + TALLOC_FREE(state); + return; + } + tevent_req_set_callback(req, winbindd_addr_changed, state); +} + +/* Main function */ + +int main(int argc, const char **argv) +{ + static bool log_stdout = False; + struct samba_cmdline_daemon_cfg *cmdline_daemon_cfg = NULL; + struct poptOption long_options[] = { + POPT_AUTOHELP + { + .longName = "no-caching", + .shortName = 'n', + .argInfo = POPT_ARG_NONE, + .arg = NULL, + .val = 'n', + .descrip = "Disable caching", + }, + POPT_COMMON_SAMBA + POPT_COMMON_DAEMON + POPT_COMMON_VERSION + POPT_TABLEEND + }; + const struct loadparm_substitution *lp_sub = + loadparm_s3_global_substitution(); + poptContext pc; + int opt; + TALLOC_CTX *frame; + NTSTATUS status; + bool ok; + const struct dcesrv_endpoint_server *ep_server = NULL; + struct dcesrv_context *dce_ctx = NULL; + size_t winbindd_socket_dir_len = 0; + char *winbindd_priv_socket_dir = NULL; + size_t winbindd_priv_socket_dir_len = 0; + + setproctitle_init(argc, discard_const(argv), environ); + + /* + * Do this before any other talloc operation + */ + talloc_enable_null_tracking(); + frame = talloc_stackframe(); + + /* + * We want total control over the permissions on created files, + * so set our umask to 0. + */ + umask(0); + + smb_init_locale(); + + /* glibc (?) likes to print "User defined signal 1" and exit if a + SIGUSR[12] is received before a handler is installed */ + + CatchSignal(SIGUSR1, SIG_IGN); + CatchSignal(SIGUSR2, SIG_IGN); + + ok = samba_cmdline_init(frame, + SAMBA_CMDLINE_CONFIG_SERVER, + true /* require_smbconf */); + if (!ok) { + DBG_ERR("Failed to setup cmdline parser\n"); + TALLOC_FREE(frame); + exit(1); + } + + cmdline_daemon_cfg = samba_cmdline_get_daemon_cfg(); + + pc = samba_popt_get_context(getprogname(), argc, argv, long_options, 0); + if (pc == NULL) { + DBG_ERR("Failed to setup popt parser!\n"); + TALLOC_FREE(frame); + exit(1); + } + + while ((opt = poptGetNextOpt(pc)) != -1) { + switch (opt) { + case 'n': + winbindd_set_use_cache(false); + break; + default: + d_fprintf(stderr, "\nInvalid option %s: %s\n\n", + poptBadOption(pc, 0), poptStrerror(opt)); + poptPrintUsage(pc, stderr, 0); + exit(1); + } + } + + /* Set environment variable so we don't recursively call ourselves. + This may also be useful interactively. */ + if ( !winbind_off() ) { + DEBUG(0,("Failed to disable recursive winbindd calls. Exiting.\n")); + exit(1); + } + + /* Initialise for running in non-root mode */ + sec_init(); + + set_remote_machine_name("winbindd", False); + + dump_core_setup("winbindd", lp_logfile(talloc_tos(), lp_sub)); + if (cmdline_daemon_cfg->daemon && cmdline_daemon_cfg->interactive) { + d_fprintf(stderr,"\nERROR: " + "Option -i|--interactive is not allowed together with -D|--daemon\n\n"); + poptPrintUsage(pc, stderr, 0); + exit(1); + } + + log_stdout = (debug_get_log_type() == DEBUG_STDOUT); + if (cmdline_daemon_cfg->interactive) { + /* + * libcmdline POPT_DAEMON callback sets "fork" to false if "-i" + * for interactive is passed on the commandline. Set it back to + * true. TODO: check if this is correct, smbd and nmbd don't do + * this. + */ + cmdline_daemon_cfg->fork = true; + log_stdout = true; + } + + if (log_stdout && cmdline_daemon_cfg->fork) { + d_fprintf(stderr, "\nERROR: " + "Can't log to stdout (-S) unless daemon is in " + "foreground (-F) or interactive (-i)\n\n"); + poptPrintUsage(pc, stderr, 0); + exit(1); + } + + poptFreeContext(pc); + + reopen_logs(); + + DBG_STARTUP_NOTICE("winbindd version %s started.\n%s\n", + samba_version_string(), + samba_copyright_string()); + + /* After parsing the configuration file we setup the core path one more time + * as the log file might have been set in the configuration and cores's + * path is by default basename(lp_logfile()). + */ + dump_core_setup("winbindd", lp_logfile(talloc_tos(), lp_sub)); + + if (lp_server_role() == ROLE_ACTIVE_DIRECTORY_DC) { + if (!lp_parm_bool(-1, "server role check", "inhibit", false)) { + DBG_ERR("server role = 'active directory domain controller' not compatible with running the winbindd binary. \n"); + DEBUGADD(0, ("You should start 'samba' instead, and it will control starting the internal AD DC winbindd implementation, which is not the same as this one\n")); + exit(1); + } + /* Main 'samba' daemon will notify */ + daemon_sd_notifications(false); + } + + if (lp_security() == SEC_ADS) { + const char *realm = lp_realm(); + const char *workgroup = lp_workgroup(); + + if (workgroup == NULL || strlen(workgroup) == 0) { + DBG_ERR("For 'secuirity = ADS' mode, the 'workgroup' " + "parameter is required to be set!\n"); + exit(1); + } + + if (realm == NULL || strlen(realm) == 0) { + DBG_ERR("For 'secuirity = ADS' mode, the 'realm' " + "parameter is required to be set!\n"); + exit(1); + } + } + + winbindd_socket_dir_len = strlen(lp_winbindd_socket_directory()); + if (winbindd_socket_dir_len > 0) { + size_t winbindd_socket_len = + winbindd_socket_dir_len + 1 + + strlen(WINBINDD_SOCKET_NAME); + struct sockaddr_un un = { + .sun_family = AF_UNIX, + }; + size_t sun_path_len = sizeof(un.sun_path); + + if (winbindd_socket_len >= sun_path_len) { + DBG_ERR("The winbind socket path [%s/%s] is too long " + "(%zu >= %zu)\n", + lp_winbindd_socket_directory(), + WINBINDD_SOCKET_NAME, + winbindd_socket_len, + sun_path_len); + exit(1); + } + } else { + DBG_ERR("'winbindd_socket_directory' parameter is empty\n"); + exit(1); + } + + winbindd_priv_socket_dir = get_winbind_priv_pipe_dir(); + winbindd_priv_socket_dir_len = strlen(winbindd_priv_socket_dir); + if (winbindd_priv_socket_dir_len > 0) { + size_t winbindd_priv_socket_len = + winbindd_priv_socket_dir_len + 1 + + strlen(WINBINDD_SOCKET_NAME); + struct sockaddr_un un = { + .sun_family = AF_UNIX, + }; + size_t sun_path_len = sizeof(un.sun_path); + + if (winbindd_priv_socket_len >= sun_path_len) { + DBG_ERR("The winbind privileged socket path [%s/%s] is too long " + "(%zu >= %zu)\n", + winbindd_priv_socket_dir, + WINBINDD_SOCKET_NAME, + winbindd_priv_socket_len, + sun_path_len); + exit(1); + } + } else { + DBG_ERR("'winbindd_priv_socket_directory' parameter is empty\n"); + exit(1); + } + TALLOC_FREE(winbindd_priv_socket_dir); + + if (!cluster_probe_ok()) { + exit(1); + } + + /* Initialise messaging system */ + + if (global_messaging_context() == NULL) { + exit(1); + } + + if (!winbindd_reload_services_file(NULL)) { + DEBUG(0, ("error opening config file\n")); + exit(1); + } + + { + size_t i; + const char *idmap_backend; + const char *invalid_backends[] = { + "ad", "rfc2307", "rid", + }; + + idmap_backend = lp_idmap_default_backend(); + for (i = 0; i < ARRAY_SIZE(invalid_backends); i++) { + ok = strequal(idmap_backend, invalid_backends[i]); + if (ok) { + DBG_ERR("FATAL: Invalid idmap backend %s " + "configured as the default backend!\n", + idmap_backend); + exit(1); + } + } + } + + ok = directory_create_or_exist(lp_lock_directory(), 0755); + if (!ok) { + DEBUG(0, ("Failed to create directory %s for lock files - %s\n", + lp_lock_directory(), strerror(errno))); + exit(1); + } + + ok = directory_create_or_exist(lp_pid_directory(), 0755); + if (!ok) { + DEBUG(0, ("Failed to create directory %s for pid files - %s\n", + lp_pid_directory(), strerror(errno))); + exit(1); + } + + load_interfaces(); + + if (!secrets_init()) { + + DEBUG(0,("Could not initialize domain trust account secrets. Giving up\n")); + return False; + } + + status = rpccli_pre_open_netlogon_creds(); + if (!NT_STATUS_IS_OK(status)) { + DEBUG(0, ("rpccli_pre_open_netlogon_creds() - %s\n", + nt_errstr(status))); + exit(1); + } + + /* Unblock all signals we are interested in as they may have been + blocked by the parent process. */ + + BlockSignals(False, SIGINT); + BlockSignals(False, SIGQUIT); + BlockSignals(False, SIGTERM); + BlockSignals(False, SIGUSR1); + BlockSignals(False, SIGUSR2); + BlockSignals(False, SIGHUP); + BlockSignals(False, SIGCHLD); + + if (!interactive) { + become_daemon(cmdline_daemon_cfg->fork, + cmdline_daemon_cfg->no_process_group, + log_stdout); + } else { + daemon_status("winbindd", "Starting process ..."); + } + + pidfile_create(lp_pid_directory(), "winbindd"); + +#ifdef HAVE_SETPGID + /* + * If we're interactive we want to set our own process group for + * signal management. + */ + if (cmdline_daemon_cfg->interactive && + !cmdline_daemon_cfg->no_process_group) + { + setpgid( (pid_t)0, (pid_t)0); + } +#endif + + TimeInit(); + + /* Don't use winbindd_reinit_after_fork here as + * we're just starting up and haven't created any + * winbindd-specific resources we must free yet. JRA. + */ + + status = reinit_after_fork(global_messaging_context(), + global_event_context(), + false); + if (!NT_STATUS_IS_OK(status)) { + exit_daemon("Winbindd reinit_after_fork() failed", map_errno_from_nt_status(status)); + } + + if (lp_winbind_debug_traceid()) { + winbind_debug_traceid_setup(global_event_context()); + winbind_debug_call_depth_setup(debug_call_depth_addr()); + tevent_thread_call_depth_set_callback( + debuglevel_get() > 1 ? winbind_call_flow : NULL, + NULL); + } + ok = initialize_password_db(true, global_event_context()); + if (!ok) { + exit_daemon("Failed to initialize passdb backend! " + "Check the 'passdb backend' variable in your " + "smb.conf file.", EINVAL); + } + + /* + * Do not initialize the parent-child-pipe before becoming + * a daemon: this is used to detect a died parent in the child + * process. + */ + status = init_before_fork(); + if (!NT_STATUS_IS_OK(status)) { + exit_daemon(nt_errstr(status), map_errno_from_nt_status(status)); + } + + winbindd_register_handlers(global_messaging_context(), + !cmdline_daemon_cfg->fork); + + if (!messaging_parent_dgm_cleanup_init(global_messaging_context())) { + exit(1); + } + + status = init_system_session_info(NULL); + if (!NT_STATUS_IS_OK(status)) { + exit_daemon("Winbindd failed to setup system user info", map_errno_from_nt_status(status)); + } + + DBG_INFO("Registering DCE/RPC endpoint servers\n"); + + ep_server = winbind_get_ep_server(); + if (ep_server == NULL) { + DBG_ERR("Failed to get 'winbind' endpoint server\n"); + exit(1); + } + status = dcerpc_register_ep_server(ep_server); + if (!NT_STATUS_IS_OK(status)) { + DBG_ERR("Failed to register 'winbind' endpoint " + "server: %s\n", nt_errstr(status)); + exit(1); + } + + dce_ctx = global_dcesrv_context(); + + DBG_INFO("Initializing DCE/RPC registered endpoint servers\n"); + + /* Init all registered ep servers */ + status = dcesrv_init_registered_ep_servers(dce_ctx); + if (!NT_STATUS_IS_OK(status)) { + DBG_ERR("Failed to init DCE/RPC endpoint servers: %s\n", + nt_errstr(status)); + exit(1); + } + + winbindd_init_addrchange(NULL, global_event_context(), + global_messaging_context()); + + /* setup listen sockets */ + + if (!winbindd_setup_listeners()) { + exit_daemon("Winbindd failed to setup listeners", EPIPE); + } + + irpc_add_name(winbind_imessaging_context(), "winbind_server"); + + TALLOC_FREE(frame); + + if (!interactive) { + daemon_ready("winbindd"); + } + + gpupdate_init(); + + /* Loop waiting for requests */ + while (1) { + frame = talloc_stackframe(); + + if (tevent_loop_once(global_event_context()) == -1) { + DEBUG(1, ("tevent_loop_once() failed: %s\n", + strerror(errno))); + return 1; + } + + TALLOC_FREE(frame); + } + + return 0; +} |