source: branches/samba-3.2.x/source/winbindd/winbindd_dual.c

/* 
   Unix SMB/CIFS implementation.

   Winbind child daemons

   Copyright (C) Andrew Tridgell 2002
   Copyright (C) Volker Lendecke 2004,2005
   
   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/>.
*/

/*
 * We fork a child per domain to be able to act non-blocking in the main
 * winbind daemon. A domain controller thousands of miles away being being
 * slow replying with a 10.000 user list should not hold up netlogon calls
 * that can be handled locally.
 */

#include "includes.h"
#include "winbindd.h"

#undef DBGC_CLASS
#define DBGC_CLASS DBGC_WINBIND

extern bool override_logfile;
extern struct winbindd_methods cache_methods;

/* Read some data from a client connection */

static void child_read_request(struct winbindd_cli_state *state)
{
        NTSTATUS status;

        /* Read data */

        status = read_data(state->sock, (char *)&state->request,
                           sizeof(state->request));

        if (!NT_STATUS_IS_OK(status)) {
                DEBUG(3, ("child_read_request: read_data failed: %s\n",
                          nt_errstr(status)));
                state->finished = True;
                return;
        }

        if (state->request.extra_len == 0) {
                state->request.extra_data.data = NULL;
                return;
        }

        DEBUG(10, ("Need to read %d extra bytes\n", (int)state->request.extra_len));

        state->request.extra_data.data =
                SMB_MALLOC_ARRAY(char, state->request.extra_len + 1);

        if (state->request.extra_data.data == NULL) {
                DEBUG(0, ("malloc failed\n"));
                state->finished = True;
                return;
        }

        /* Ensure null termination */
        state->request.extra_data.data[state->request.extra_len] = '\0';

        status= read_data(state->sock, state->request.extra_data.data,
                          state->request.extra_len);

        if (!NT_STATUS_IS_OK(status)) {
                DEBUG(0, ("Could not read extra data: %s\n",
                          nt_errstr(status)));
                state->finished = True;
                return;
        }
}

/*
 * Machinery for async requests sent to children. You set up a
 * winbindd_request, select a child to query, and issue a async_request
 * call. When the request is completed, the callback function you specified is
 * called back with the private pointer you gave to async_request.
 */

struct winbindd_async_request {
        struct winbindd_async_request *next, *prev;
        TALLOC_CTX *mem_ctx;
        struct winbindd_child *child;
        struct winbindd_request *request;
        struct winbindd_response *response;
        void (*continuation)(void *private_data, bool success);
        struct timed_event *reply_timeout_event;
        pid_t child_pid; /* pid of the child we're waiting on. Used to detect
                            a restart of the child (child->pid != child_pid). */
        void *private_data;
};

static void async_request_fail(struct winbindd_async_request *state);
static void async_main_request_sent(void *private_data, bool success);
static void async_request_sent(void *private_data, bool success);
static void async_reply_recv(void *private_data, bool success);
static void schedule_async_request(struct winbindd_child *child);

void async_request(TALLOC_CTX *mem_ctx, struct winbindd_child *child,
                   struct winbindd_request *request,
                   struct winbindd_response *response,
                   void (*continuation)(void *private_data, bool success),
                   void *private_data)
{
        struct winbindd_async_request *state;

        SMB_ASSERT(continuation != NULL);

        DEBUG(10, ("Sending request to child pid %d (domain=%s)\n",
                (int)child->pid,
                (child->domain != NULL) ? child->domain->name : "''"));

        state = TALLOC_P(mem_ctx, struct winbindd_async_request);

        if (state == NULL) {
                DEBUG(0, ("talloc failed\n"));
                continuation(private_data, False);
                return;
        }

        state->mem_ctx = mem_ctx;
        state->child = child;
        state->reply_timeout_event = NULL;
        state->request = request;
        state->response = response;
        state->continuation = continuation;
        state->private_data = private_data;

        DLIST_ADD_END(child->requests, state, struct winbindd_async_request *);

        schedule_async_request(child);

        return;
}

static void async_main_request_sent(void *private_data, bool success)
{
        struct winbindd_async_request *state =
                talloc_get_type_abort(private_data, struct winbindd_async_request);

        if (!success) {
                DEBUG(5, ("Could not send async request\n"));
                async_request_fail(state);
                return;
        }

        if (state->request->extra_len == 0) {
                async_request_sent(private_data, True);
                return;
        }

        setup_async_write(&state->child->event, state->request->extra_data.data,
                          state->request->extra_len,
                          async_request_sent, state);
}

/****************************************************************
 Handler triggered if the child winbindd doesn't respond within
 a given timeout.
****************************************************************/

static void async_request_timeout_handler(struct event_context *ctx,
                                        struct timed_event *te,
                                        const struct timeval *now,
                                        void *private_data)
{
        struct winbindd_async_request *state =
                talloc_get_type_abort(private_data, struct winbindd_async_request);

        DEBUG(0,("async_request_timeout_handler: child pid %u is not responding. "
                "Closing connection to it.\n",
                state->child_pid ));

        /* Deal with the reply - set to error. */
        async_reply_recv(private_data, False);
}

/**************************************************************
 Common function called on both async send and recv fail.
 Cleans up the child and schedules the next request.
**************************************************************/

static void async_request_fail(struct winbindd_async_request *state)
{
        DLIST_REMOVE(state->child->requests, state);

        TALLOC_FREE(state->reply_timeout_event);

        /* If child exists and is not already reaped,
           send kill signal to child. */

        if ((state->child->pid != (pid_t)0) &&
                        (state->child->pid != (pid_t)-1) &&
                        (state->child->pid == state->child_pid)) {
                kill(state->child_pid, SIGTERM);

                /* 
                 * Close the socket to the child.
                 */
                winbind_child_died(state->child_pid);
        }

        state->response->length = sizeof(struct winbindd_response);
        state->response->result = WINBINDD_ERROR;
        state->continuation(state->private_data, False);
}

static void async_request_sent(void *private_data_data, bool success)
{
        struct winbindd_async_request *state =
                talloc_get_type_abort(private_data_data, struct winbindd_async_request);

        if (!success) {
                DEBUG(5, ("Could not send async request to child pid %u\n",
                        (unsigned int)state->child_pid ));
                async_request_fail(state);
                return;
        }

        /* Request successfully sent to the child, setup the wait for reply */

        setup_async_read(&state->child->event,
                         &state->response->result,
                         sizeof(state->response->result),
                         async_reply_recv, state);

        /* 
         * Set up a timeout of 300 seconds for the response.
         * If we don't get it close the child socket and
         * report failure.
         */

        state->reply_timeout_event = event_add_timed(winbind_event_context(),
                                                        NULL,
                                                        timeval_current_ofs(300,0),
                                                        "async_request_timeout",
                                                        async_request_timeout_handler,
                                                        state);
        if (!state->reply_timeout_event) {
                smb_panic("async_request_sent: failed to add timeout handler.\n");
        }
}

static void async_reply_recv(void *private_data, bool success)
{
        struct winbindd_async_request *state =
                talloc_get_type_abort(private_data, struct winbindd_async_request);
        struct winbindd_child *child = state->child;

        TALLOC_FREE(state->reply_timeout_event);

        state->response->length = sizeof(struct winbindd_response);

        if (!success) {
                DEBUG(5, ("Could not receive async reply from child pid %u\n",
                        (unsigned int)state->child_pid ));

                cache_cleanup_response(state->child_pid);
                async_request_fail(state);
                return;
        }

        SMB_ASSERT(cache_retrieve_response(state->child_pid,
                                           state->response));

        cache_cleanup_response(state->child_pid);
        
        DLIST_REMOVE(child->requests, state);

        schedule_async_request(child);

        state->continuation(state->private_data, True);
}

static bool fork_domain_child(struct winbindd_child *child);

static void schedule_async_request(struct winbindd_child *child)
{
        struct winbindd_async_request *request = child->requests;

        if (request == NULL) {
                return;
        }

        if (child->event.flags != 0) {
                return;         /* Busy */
        }

        /*
         * This may be a reschedule, so we might
         * have an existing timeout event pending on
         * the first entry in the child->requests list
         * (we only send one request at a time).
         * Ensure we free it before we reschedule.
         * Bug #5814, from hargagan <shargagan@novell.com>.
         * JRA.
         */

        TALLOC_FREE(request->reply_timeout_event);

        if ((child->pid == 0) && (!fork_domain_child(child))) {
                /* fork_domain_child failed.
                   Cancel all outstanding requests */

                while (request != NULL) {
                        /* request might be free'd in the continuation */
                        struct winbindd_async_request *next = request->next;

                        async_request_fail(request);
                        request = next;
                }
                return;
        }

        /* Now we know who we're sending to - remember the pid. */
        request->child_pid = child->pid;

        setup_async_write(&child->event, request->request,
                          sizeof(*request->request),
                          async_main_request_sent, request);

        return;
}

struct domain_request_state {
        TALLOC_CTX *mem_ctx;
        struct winbindd_domain *domain;
        struct winbindd_request *request;
        struct winbindd_response *response;
        void (*continuation)(void *private_data_data, bool success);
        void *private_data_data;
};

static void domain_init_recv(void *private_data_data, bool success);

void async_domain_request(TALLOC_CTX *mem_ctx,
                          struct winbindd_domain *domain,
                          struct winbindd_request *request,
                          struct winbindd_response *response,
                          void (*continuation)(void *private_data_data, bool success),
                          void *private_data_data)
{
        struct domain_request_state *state;

        if (domain->initialized) {
                async_request(mem_ctx, &domain->child, request, response,
                              continuation, private_data_data);
                return;
        }

        state = TALLOC_P(mem_ctx, struct domain_request_state);
        if (state == NULL) {
                DEBUG(0, ("talloc failed\n"));
                continuation(private_data_data, False);
                return;
        }

        state->mem_ctx = mem_ctx;
        state->domain = domain;
        state->request = request;
        state->response = response;
        state->continuation = continuation;
        state->private_data_data = private_data_data;

        init_child_connection(domain, domain_init_recv, state);
}

static void domain_init_recv(void *private_data_data, bool success)
{
        struct domain_request_state *state =
                talloc_get_type_abort(private_data_data, struct domain_request_state);

        if (!success) {
                DEBUG(5, ("Domain init returned an error\n"));
                state->continuation(state->private_data_data, False);
                return;
        }

        async_request(state->mem_ctx, &state->domain->child,
                      state->request, state->response,
                      state->continuation, state->private_data_data);
}

static void recvfrom_child(void *private_data_data, bool success)
{
        struct winbindd_cli_state *state =
                talloc_get_type_abort(private_data_data, struct winbindd_cli_state);
        enum winbindd_result result = state->response.result;

        /* This is an optimization: The child has written directly to the
         * response buffer. The request itself is still in pending state,
         * state that in the result code. */

        state->response.result = WINBINDD_PENDING;

        if ((!success) || (result != WINBINDD_OK)) {
                request_error(state);
                return;
        }

        request_ok(state);
}

void sendto_child(struct winbindd_cli_state *state,
                  struct winbindd_child *child)
{
        async_request(state->mem_ctx, child, &state->request,
                      &state->response, recvfrom_child, state);
}

void sendto_domain(struct winbindd_cli_state *state,
                   struct winbindd_domain *domain)
{
        async_domain_request(state->mem_ctx, domain,
                             &state->request, &state->response,
                             recvfrom_child, state);
}

static void child_process_request(struct winbindd_child *child,
                                  struct winbindd_cli_state *state)
{
        struct winbindd_domain *domain = child->domain;
        const struct winbindd_child_dispatch_table *table = child->table;

        /* Free response data - we may be interrupted and receive another
           command before being able to send this data off. */

        state->response.result = WINBINDD_ERROR;
        state->response.length = sizeof(struct winbindd_response);

        /* as all requests in the child are sync, we can use talloc_tos() */
        state->mem_ctx = talloc_tos();

        /* Process command */

        for (; table->name; table++) {
                if (state->request.cmd == table->struct_cmd) {
                        DEBUG(10,("child_process_request: request fn %s\n",
                                  table->name));
                        state->response.result = table->struct_fn(domain, state);
                        return;
                }
        }

        DEBUG(1 ,("child_process_request: unknown request fn number %d\n",
                  (int)state->request.cmd));
        state->response.result = WINBINDD_ERROR;
}

void setup_child(struct winbindd_child *child,
                 const struct winbindd_child_dispatch_table *table,
                 const char *logprefix,
                 const char *logname)
{
        if (logprefix && logname) {
                if (asprintf(&child->logfilename, "%s/%s-%s",
                             get_dyn_LOGFILEBASE(), logprefix, logname) < 0) {
                        smb_panic("Internal error: asprintf failed");
                }
        } else {
                smb_panic("Internal error: logprefix == NULL && "
                          "logname == NULL");
        }

        child->domain = NULL;
        child->table = table;
}

struct winbindd_child *children = NULL;

void winbind_child_died(pid_t pid)
{
        struct winbindd_child *child;

        for (child = children; child != NULL; child = child->next) {
                if (child->pid == pid) {
                        break;
                }
        }

        if (child == NULL) {
                DEBUG(5, ("Already reaped child %u died\n", (unsigned int)pid));
                return;
        }

        /* This will be re-added in fork_domain_child() */

        DLIST_REMOVE(children, child);
        
        remove_fd_event(&child->event);
        close(child->event.fd);
        child->event.fd = 0;
        child->event.flags = 0;
        child->pid = 0;

        if (child->requests) {
                /*
                 * schedule_async_request() will also
                 * clear this event but the call is
                 * idempotent so it doesn't hurt to
                 * cover all possible future code
                 * paths. JRA.
                 */
                TALLOC_FREE(child->requests->reply_timeout_event);
        }

        schedule_async_request(child);
}

/* Ensure any negative cache entries with the netbios or realm names are removed. */

void winbindd_flush_negative_conn_cache(struct winbindd_domain *domain)
{
        flush_negative_conn_cache_for_domain(domain->name);
        if (*domain->alt_name) {
                flush_negative_conn_cache_for_domain(domain->alt_name);
        }
}

/* 
 * Parent winbindd process sets its own debug level first and then
 * sends a message to all the winbindd children to adjust their debug
 * level to that of parents.
 */

void winbind_msg_debug(struct messaging_context *msg_ctx,
                         void *private_data,
                         uint32_t msg_type,
                         struct server_id server_id,
                         DATA_BLOB *data)
{
        struct winbindd_child *child;

        DEBUG(10,("winbind_msg_debug: got debug message.\n"));
        
        debug_message(msg_ctx, private_data, MSG_DEBUG, server_id, data);

        for (child = children; child != NULL; child = child->next) {

                DEBUG(10,("winbind_msg_debug: sending message to pid %u.\n",
                        (unsigned int)child->pid));

                messaging_send_buf(msg_ctx, pid_to_procid(child->pid),
                           MSG_DEBUG,
                           data->data,
                           strlen((char *) data->data) + 1);
        }
}

/* Set our domains as offline and forward the offline message to our children. */

void winbind_msg_offline(struct messaging_context *msg_ctx,
                         void *private_data,
                         uint32_t msg_type,
                         struct server_id server_id,
                         DATA_BLOB *data)
{
        struct winbindd_child *child;
        struct winbindd_domain *domain;

        DEBUG(10,("winbind_msg_offline: got offline message.\n"));

        if (!lp_winbind_offline_logon()) {
                DEBUG(10,("winbind_msg_offline: rejecting offline message.\n"));
                return;
        }

        /* Set our global state as offline. */
        if (!set_global_winbindd_state_offline()) {
                DEBUG(10,("winbind_msg_offline: offline request failed.\n"));
                return;
        }

        /* Set all our domains as offline. */
        for (domain = domain_list(); domain; domain = domain->next) {
                if (domain->internal) {
                        continue;
                }
                DEBUG(5,("winbind_msg_offline: marking %s offline.\n", domain->name));
                set_domain_offline(domain);
        }

        for (child = children; child != NULL; child = child->next) {
                /* Don't send message to internal childs.  We've already
                   done so above. */
                if (!child->domain || winbindd_internal_child(child)) {
                        continue;
                }

                /* Or internal domains (this should not be possible....) */
                if (child->domain->internal) {
                        continue;
                }

                /* Each winbindd child should only process requests for one domain - make sure
                   we only set it online / offline for that domain. */

                DEBUG(10,("winbind_msg_offline: sending message to pid %u for domain %s.\n",
                        (unsigned int)child->pid, domain->name ));

                messaging_send_buf(msg_ctx, pid_to_procid(child->pid),
                                   MSG_WINBIND_OFFLINE,
                                   (uint8 *)child->domain->name,
                                   strlen(child->domain->name)+1);
        }
}

/* Set our domains as online and forward the online message to our children. */

void winbind_msg_online(struct messaging_context *msg_ctx,
                        void *private_data,
                        uint32_t msg_type,
                        struct server_id server_id,
                        DATA_BLOB *data)
{
        struct winbindd_child *child;
        struct winbindd_domain *domain;

        DEBUG(10,("winbind_msg_online: got online message.\n"));

        if (!lp_winbind_offline_logon()) {
                DEBUG(10,("winbind_msg_online: rejecting online message.\n"));
                return;
        }

        /* Set our global state as online. */
        set_global_winbindd_state_online();

        smb_nscd_flush_user_cache();
        smb_nscd_flush_group_cache();

        /* Set all our domains as online. */
        for (domain = domain_list(); domain; domain = domain->next) {
                if (domain->internal) {
                        continue;
                }
                DEBUG(5,("winbind_msg_online: requesting %s to go online.\n", domain->name));

                winbindd_flush_negative_conn_cache(domain);
                set_domain_online_request(domain);

                /* Send an online message to the idmap child when our
                   primary domain comes back online */

                if ( domain->primary ) {
                        struct winbindd_child *idmap = idmap_child();
                        
                        if ( idmap->pid != 0 ) {
                                messaging_send_buf(msg_ctx,
                                                   pid_to_procid(idmap->pid), 
                                                   MSG_WINBIND_ONLINE,
                                                   (uint8 *)domain->name,
                                                   strlen(domain->name)+1);
                        }
                        
                }
        }

        for (child = children; child != NULL; child = child->next) {
                /* Don't send message to internal childs. */
                if (!child->domain || winbindd_internal_child(child)) {
                        continue;
                }

                /* Or internal domains (this should not be possible....) */
                if (child->domain->internal) {
                        continue;
                }

                /* Each winbindd child should only process requests for one domain - make sure
                   we only set it online / offline for that domain. */

                DEBUG(10,("winbind_msg_online: sending message to pid %u for domain %s.\n",
                        (unsigned int)child->pid, child->domain->name ));

                messaging_send_buf(msg_ctx, pid_to_procid(child->pid),
                                   MSG_WINBIND_ONLINE,
                                   (uint8 *)child->domain->name,
                                   strlen(child->domain->name)+1);
        }
}

/* Forward the online/offline messages to our children. */
void winbind_msg_onlinestatus(struct messaging_context *msg_ctx,
                              void *private_data,
                              uint32_t msg_type,
                              struct server_id server_id,
                              DATA_BLOB *data)
{
        struct winbindd_child *child;

        DEBUG(10,("winbind_msg_onlinestatus: got onlinestatus message.\n"));

        for (child = children; child != NULL; child = child->next) {
                if (child->domain && child->domain->primary) {
                        DEBUG(10,("winbind_msg_onlinestatus: "
                                  "sending message to pid %u of primary domain.\n",
                                  (unsigned int)child->pid));
                        messaging_send_buf(msg_ctx, pid_to_procid(child->pid), 
                                           MSG_WINBIND_ONLINESTATUS,
                                           (uint8 *)data->data,
                                           data->length);
                        break;
                }
        }
}

void winbind_msg_dump_event_list(struct messaging_context *msg_ctx,
                                 void *private_data,
                                 uint32_t msg_type,
                                 struct server_id server_id,
                                 DATA_BLOB *data)
{
        struct winbindd_child *child;

        DEBUG(10,("winbind_msg_dump_event_list received\n"));

        dump_event_list(winbind_event_context());

        for (child = children; child != NULL; child = child->next) {

                DEBUG(10,("winbind_msg_dump_event_list: sending message to pid %u\n",
                        (unsigned int)child->pid));

                messaging_send_buf(msg_ctx, pid_to_procid(child->pid),
                                   MSG_DUMP_EVENT_LIST,
                                   NULL, 0);
        }

}

void winbind_msg_dump_domain_list(struct messaging_context *msg_ctx,
                                  void *private_data,
                                  uint32_t msg_type,
                                  struct server_id server_id,
                                  DATA_BLOB *data)
{
        TALLOC_CTX *mem_ctx;
        const char *message = NULL;
        struct server_id *sender = NULL;
        const char *domain = NULL;
        char *s = NULL;
        NTSTATUS status;
        struct winbindd_domain *dom = NULL;

        DEBUG(5,("winbind_msg_dump_domain_list received.\n"));

        if (!data || !data->data) {
                return;
        }

        if (data->length < sizeof(struct server_id)) {
                return;
        }

        mem_ctx = talloc_init("winbind_msg_dump_domain_list");
        if (!mem_ctx) {
                return;
        }

        sender = (struct server_id *)data->data;
        if (data->length > sizeof(struct server_id)) {
                domain = (const char *)data->data+sizeof(struct server_id);
        }

        if (domain) {

                DEBUG(5,("winbind_msg_dump_domain_list for domain: %s\n",
                        domain));

                message = NDR_PRINT_STRUCT_STRING(mem_ctx, winbindd_domain,
                                                  find_domain_from_name_noinit(domain));
                if (!message) {
                        talloc_destroy(mem_ctx);
                        return;
                }

                messaging_send_buf(msg_ctx, *sender,
                                   MSG_WINBIND_DUMP_DOMAIN_LIST,
                                   (uint8_t *)message, strlen(message) + 1);

                talloc_destroy(mem_ctx);

                return;
        }

        DEBUG(5,("winbind_msg_dump_domain_list all domains\n"));

        for (dom = domain_list(); dom; dom=dom->next) {
                message = NDR_PRINT_STRUCT_STRING(mem_ctx, winbindd_domain, dom);
                if (!message) {
                        talloc_destroy(mem_ctx);
                        return;
                }

                s = talloc_asprintf_append(s, "%s\n", message);
                if (!s) {
                        talloc_destroy(mem_ctx);
                        return;
                }
        }

        status = messaging_send_buf(msg_ctx, *sender,
                                    MSG_WINBIND_DUMP_DOMAIN_LIST,
                                    (uint8_t *)s, strlen(s) + 1);
        if (!NT_STATUS_IS_OK(status)) {
                DEBUG(0,("failed to send message: %s\n",
                nt_errstr(status)));
        }

        talloc_destroy(mem_ctx);
}

static void account_lockout_policy_handler(struct event_context *ctx,
                                           struct timed_event *te,
                                           const struct timeval *now,
                                           void *private_data)
{
        struct winbindd_child *child =
                (struct winbindd_child *)private_data;
        TALLOC_CTX *mem_ctx = NULL;
        struct winbindd_methods *methods;
        struct samr_DomInfo12 lockout_policy;
        NTSTATUS result;

        DEBUG(10,("account_lockout_policy_handler called\n"));

        TALLOC_FREE(child->lockout_policy_event);

        if ( !winbindd_can_contact_domain( child->domain ) ) {
                DEBUG(10,("account_lockout_policy_handler: Removing myself since I "
                          "do not have an incoming trust to domain %s\n", 
                          child->domain->name));

                return;         
        }

        methods = child->domain->methods;

        mem_ctx = talloc_init("account_lockout_policy_handler ctx");
        if (!mem_ctx) {
                result = NT_STATUS_NO_MEMORY;
        } else {
                result = methods->lockout_policy(child->domain, mem_ctx, &lockout_policy);
        }
        TALLOC_FREE(mem_ctx);

        if (!NT_STATUS_IS_OK(result)) {
                DEBUG(10,("account_lockout_policy_handler: lockout_policy failed error %s\n",
                         nt_errstr(result)));
        }

        child->lockout_policy_event = event_add_timed(winbind_event_context(), NULL,
                                                      timeval_current_ofs(3600, 0),
                                                      "account_lockout_policy_handler",
                                                      account_lockout_policy_handler,
                                                      child);
}

/* Deal with a request to go offline. */

static void child_msg_offline(struct messaging_context *msg,
                              void *private_data,
                              uint32_t msg_type,
                              struct server_id server_id,
                              DATA_BLOB *data)
{
        struct winbindd_domain *domain;
        struct winbindd_domain *primary_domain = NULL;
        const char *domainname = (const char *)data->data;

        if (data->data == NULL || data->length == 0) {
                return;
        }

        DEBUG(5,("child_msg_offline received for domain %s.\n", domainname));

        if (!lp_winbind_offline_logon()) {
                DEBUG(10,("child_msg_offline: rejecting offline message.\n"));
                return;
        }

        primary_domain = find_our_domain();

        /* Mark the requested domain offline. */

        for (domain = domain_list(); domain; domain = domain->next) {
                if (domain->internal) {
                        continue;
                }
                if (strequal(domain->name, domainname)) {
                        DEBUG(5,("child_msg_offline: marking %s offline.\n", domain->name));
                        set_domain_offline(domain);
                        /* we are in the trusted domain, set the primary domain 
                         * offline too */
                        if (domain != primary_domain) {
                                set_domain_offline(primary_domain);
                        }
                }
        }
}

/* Deal with a request to go online. */

static void child_msg_online(struct messaging_context *msg,
                             void *private_data,
                             uint32_t msg_type,
                             struct server_id server_id,
                             DATA_BLOB *data)
{
        struct winbindd_domain *domain;
        struct winbindd_domain *primary_domain = NULL;
        const char *domainname = (const char *)data->data;

        if (data->data == NULL || data->length == 0) {
                return;
        }

        DEBUG(5,("child_msg_online received for domain %s.\n", domainname));

        if (!lp_winbind_offline_logon()) {
                DEBUG(10,("child_msg_online: rejecting online message.\n"));
                return;
        }

        primary_domain = find_our_domain();

        /* Set our global state as online. */
        set_global_winbindd_state_online();

        /* Try and mark everything online - delete any negative cache entries
           to force a reconnect now. */

        for (domain = domain_list(); domain; domain = domain->next) {
                if (domain->internal) {
                        continue;
                }
                if (strequal(domain->name, domainname)) {
                        DEBUG(5,("child_msg_online: requesting %s to go online.\n", domain->name));
                        winbindd_flush_negative_conn_cache(domain);
                        set_domain_online_request(domain);

                        /* we can be in trusted domain, which will contact primary domain
                         * we have to bring primary domain online in trusted domain process
                         * see, winbindd_dual_pam_auth() --> winbindd_dual_pam_auth_samlogon()
                         * --> contact_domain = find_our_domain()
                         * */
                        if (domain != primary_domain) {
                                winbindd_flush_negative_conn_cache(primary_domain);
                                set_domain_online_request(primary_domain);
                        }
                }
        }
}

static const char *collect_onlinestatus(TALLOC_CTX *mem_ctx)
{
        struct winbindd_domain *domain;
        char *buf = NULL;

        if ((buf = talloc_asprintf(mem_ctx, "global:%s ", 
                                   get_global_winbindd_state_offline() ? 
                                   "Offline":"Online")) == NULL) {
                return NULL;
        }

        for (domain = domain_list(); domain; domain = domain->next) {
                if ((buf = talloc_asprintf_append_buffer(buf, "%s:%s ", 
                                                  domain->name, 
                                                  domain->online ?
                                                  "Online":"Offline")) == NULL) {
                        return NULL;
                }
        }

        buf = talloc_asprintf_append_buffer(buf, "\n");

        DEBUG(5,("collect_onlinestatus: %s", buf));

        return buf;
}

static void child_msg_onlinestatus(struct messaging_context *msg_ctx,
                                   void *private_data,
                                   uint32_t msg_type,
                                   struct server_id server_id,
                                   DATA_BLOB *data)
{
        TALLOC_CTX *mem_ctx;
        const char *message;
        struct server_id *sender;
        
        DEBUG(5,("winbind_msg_onlinestatus received.\n"));

        if (!data->data) {
                return;
        }

        sender = (struct server_id *)data->data;

        mem_ctx = talloc_init("winbind_msg_onlinestatus");
        if (mem_ctx == NULL) {
                return;
        }
        
        message = collect_onlinestatus(mem_ctx);
        if (message == NULL) {
                talloc_destroy(mem_ctx);
                return;
        }

        messaging_send_buf(msg_ctx, *sender, MSG_WINBIND_ONLINESTATUS, 
                           (uint8 *)message, strlen(message) + 1);

        talloc_destroy(mem_ctx);
}

static void child_msg_dump_event_list(struct messaging_context *msg,
                                      void *private_data,
                                      uint32_t msg_type,
                                      struct server_id server_id,
                                      DATA_BLOB *data)
{
        DEBUG(5,("child_msg_dump_event_list received\n"));

        dump_event_list(winbind_event_context());
}

void ccache_remove_all_after_fork(void);

bool winbindd_reinit_after_fork(const char *logfilename)
{
        struct winbindd_domain *domain;
        struct winbindd_child *cl;

        if (!reinit_after_fork(winbind_messaging_context(),
                               winbind_event_context(), true)) {
                DEBUG(0,("reinit_after_fork() failed\n"));
                return false;
        }

        close_conns_after_fork();

        if (!override_logfile && logfilename) {
                lp_set_logfile(logfilename);
                reopen_logs();
        }

        /* Don't handle the same messages as our parent. */
        messaging_deregister(winbind_messaging_context(),
                             MSG_SMB_CONF_UPDATED, NULL);
        messaging_deregister(winbind_messaging_context(),
                             MSG_SHUTDOWN, NULL);
        messaging_deregister(winbind_messaging_context(),
                             MSG_WINBIND_OFFLINE, NULL);
        messaging_deregister(winbind_messaging_context(),
                             MSG_WINBIND_ONLINE, NULL);
        messaging_deregister(winbind_messaging_context(),
                             MSG_WINBIND_ONLINESTATUS, NULL);
        messaging_deregister(winbind_messaging_context(),
                             MSG_DUMP_EVENT_LIST, NULL);
        messaging_deregister(winbind_messaging_context(),
                             MSG_WINBIND_DUMP_DOMAIN_LIST, NULL);
        messaging_deregister(winbind_messaging_context(),
                             MSG_DEBUG, NULL);

        /* We have destroyed all events in the winbindd_event_context
         * in reinit_after_fork(), so clean out all possible pending
         * event pointers. */

        /* Deal with check_online_events. */

        for (domain = domain_list(); domain; domain = domain->next) {
                TALLOC_FREE(domain->check_online_event);
        }

        /* Ensure we're not handling a credential cache event inherited
         * from our parent. */

        ccache_remove_all_after_fork();

        /* Destroy all possible events in child list. */
        for (cl = children; cl != NULL; cl = cl->next) {
                struct winbindd_async_request *request;

                for (request = cl->requests; request; request = request->next) {
                        TALLOC_FREE(request->reply_timeout_event);
                }
                TALLOC_FREE(cl->lockout_policy_event);

                /* Children should never be able to send
                 * each other messages, all messages must
                 * go through the parent.
                 */
                cl->pid = (pid_t)0;
        }
        /*
         * This is a little tricky, children must not
         * send an MSG_WINBIND_ONLINE message to idmap_child().
         * If we are in a child of our primary domain or
         * in the process created by fork_child_dc_connect(),
         * and the primary domain cannot go online,
         * fork_child_dc_connection() sends MSG_WINBIND_ONLINE
         * periodically to idmap_child().
         *
         * The sequence is, fork_child_dc_connect() ---> getdcs() --->
         * get_dc_name_via_netlogon() ---> cm_connect_netlogon()
         * ---> init_dc_connection() ---> cm_open_connection --->
         * set_domain_online(), sends MSG_WINBIND_ONLINE to
         * idmap_child(). Disallow children sending messages
         * to each other, all messages must go through the parent.
         */
        cl = idmap_child();
        cl->pid = (pid_t)0;

        return true;
}

static bool fork_domain_child(struct winbindd_child *child)
{
        int fdpair[2];
        struct winbindd_cli_state state;
        struct winbindd_domain *primary_domain = NULL;

        if (child->domain) {
                DEBUG(10, ("fork_domain_child called for domain '%s'\n",
                           child->domain->name));
        } else {
                DEBUG(10, ("fork_domain_child called without domain.\n"));
        }

        if (socketpair(AF_UNIX, SOCK_STREAM, 0, fdpair) != 0) {
                DEBUG(0, ("Could not open child pipe: %s\n",
                          strerror(errno)));
                return False;
        }

        ZERO_STRUCT(state);
        state.pid = sys_getpid();

        child->pid = sys_fork();

        if (child->pid == -1) {
                DEBUG(0, ("Could not fork: %s\n", strerror(errno)));
                return False;
        }

        if (child->pid != 0) {
                /* Parent */
                close(fdpair[0]);
                child->next = child->prev = NULL;
                DLIST_ADD(children, child);
                child->event.fd = fdpair[1];
                child->event.flags = 0;
                add_fd_event(&child->event);
                return True;
        }

        /* Child */

        /* Stop zombies in children */
        CatchChild();

        state.sock = fdpair[0];
        close(fdpair[1]);

        if (!winbindd_reinit_after_fork(child->logfilename)) {
                DEBUG(0, ("winbindd_reinit_after_fork failed.\n"));
                _exit(0);
        }
 
        /* Handle online/offline messages. */
        messaging_register(winbind_messaging_context(), NULL,
                           MSG_WINBIND_OFFLINE, child_msg_offline);
        messaging_register(winbind_messaging_context(), NULL,
                           MSG_WINBIND_ONLINE, child_msg_online);
        messaging_register(winbind_messaging_context(), NULL,
                           MSG_WINBIND_ONLINESTATUS, child_msg_onlinestatus);
        messaging_register(winbind_messaging_context(), NULL,
                           MSG_DUMP_EVENT_LIST, child_msg_dump_event_list);
        messaging_register(winbind_messaging_context(), NULL,
                           MSG_DEBUG, debug_message);
        
        primary_domain = find_our_domain();

        if (primary_domain == NULL) {
                smb_panic("no primary domain found");
        }

        /* It doesn't matter if we allow cache login,
         * try to bring domain online after fork. */
        if ( child->domain ) {
                child->domain->startup = True;
                child->domain->startup_time = time(NULL);
                /* we can be in primary domain or in trusted domain
                 * If we are in trusted domain, set the primary domain
                 * in start-up mode */
                if (!(child->domain->internal)) {
                        set_domain_online_request(child->domain);
                        if (!(child->domain->primary)) {
                                primary_domain->startup = True;
                                primary_domain->startup_time = time(NULL);
                                set_domain_online_request(primary_domain);
                        }
                }
        }
        
        /*
         * We are in idmap child, make sure that we set the
         * check_online_event to bring primary domain online.
         */
        if (child == idmap_child()) {
                set_domain_online_request(primary_domain);
        }

        /* We might be in the idmap child...*/
        if (child->domain && !(child->domain->internal) &&
            lp_winbind_offline_logon()) {

                set_domain_online_request(child->domain);

                if (primary_domain && (primary_domain != child->domain)) {
                        /* We need to talk to the primary
                         * domain as well as the trusted
                         * domain inside a trusted domain
                         * child.
                         * See the code in :
                         * set_dc_type_and_flags_trustinfo()
                         * for details.
                         */
                        set_domain_online_request(primary_domain);
                }

                child->lockout_policy_event = event_add_timed(
                        winbind_event_context(), NULL, timeval_zero(),
                        "account_lockout_policy_handler",
                        account_lockout_policy_handler,
                        child);
        }

        while (1) {

                int ret;
                fd_set read_fds;
                struct timeval t;
                struct timeval *tp;
                struct timeval now;
                TALLOC_CTX *frame = talloc_stackframe();

                /* check for signals */
                winbind_check_sigterm(false);
                winbind_check_sighup(override_logfile ? NULL :
                                child->logfilename);

                run_events(winbind_event_context(), 0, NULL, NULL);

                GetTimeOfDay(&now);

                if (child->domain && child->domain->startup &&
                                (now.tv_sec > child->domain->startup_time + 30)) {
                        /* No longer in "startup" mode. */
                        DEBUG(10,("fork_domain_child: domain %s no longer in 'startup' mode.\n",
                                child->domain->name ));
                        child->domain->startup = False;
                }

                tp = get_timed_events_timeout(winbind_event_context(), &t);
                if (tp) {
                        DEBUG(11,("select will use timeout of %u.%u seconds\n",
                                (unsigned int)tp->tv_sec, (unsigned int)tp->tv_usec ));
                }

                /* Handle messages */

                message_dispatch(winbind_messaging_context());

                FD_ZERO(&read_fds);
                FD_SET(state.sock, &read_fds);

                ret = sys_select(state.sock + 1, &read_fds, NULL, NULL, tp);

                if (ret == 0) {
                        DEBUG(11,("nothing is ready yet, continue\n"));
                        TALLOC_FREE(frame);
                        continue;
                }

                if (ret == -1 && errno == EINTR) {
                        /* We got a signal - continue. */
                        TALLOC_FREE(frame);
                        continue;
                }

                if (ret == -1 && errno != EINTR) {
                        DEBUG(0,("select error occured\n"));
                        TALLOC_FREE(frame);
                        perror("select");
                        _exit(1);
                }

                /* fetch a request from the main daemon */
                child_read_request(&state);

                if (state.finished) {
                        /* we lost contact with our parent */
                        _exit(0);
                }

                DEBUG(4,("child daemon request %d\n", (int)state.request.cmd));

                ZERO_STRUCT(state.response);
                state.request.null_term = '\0';
                child_process_request(child, &state);

                SAFE_FREE(state.request.extra_data.data);

                cache_store_response(sys_getpid(), &state.response);

                SAFE_FREE(state.response.extra_data.data);

                /* We just send the result code back, the result
                 * structure needs to be fetched via the
                 * winbindd_cache. Hmm. That needs fixing... */

                if (write_data(state.sock,
                               (const char *)&state.response.result,
                               sizeof(state.response.result)) !=
                    sizeof(state.response.result)) {
                        DEBUG(0, ("Could not write result\n"));
                        exit(1);
                }
                TALLOC_FREE(frame);
        }
}