#include "includes.h"
#include "db_wrap.h"
#include "lib/tdb/include/tdb.h"
-#include "lib/events/events.h"
+#include "lib/tevent/tevent.h"
#include "lib/util/dlinklist.h"
#include "system/network.h"
#include "system/filesys.h"
#include "system/wait.h"
-#include "../include/ctdb.h"
+#include "../include/ctdb_client.h"
#include "../include/ctdb_private.h"
#include <sys/socket.h>
-static void daemon_incoming_packet(void *, struct ctdb_req_header *);
-
-/*
- handler for when a node changes its flags
-*/
-static void flag_change_handler(struct ctdb_context *ctdb, uint64_t srvid,
- TDB_DATA data, void *private_data)
-{
- struct ctdb_node_flag_change *c = (struct ctdb_node_flag_change *)data.dptr;
-
- if (data.dsize != sizeof(*c) || !ctdb_validate_pnn(ctdb, c->pnn)) {
- DEBUG(DEBUG_CRIT,(__location__ "Invalid data in ctdb_node_flag_change\n"));
- return;
- }
-
- if (!ctdb_validate_pnn(ctdb, c->pnn)) {
- DEBUG(DEBUG_CRIT,("Bad pnn %u in flag_change_handler\n", c->pnn));
- return;
- }
-
- /* don't get the disconnected flag from the other node */
- ctdb->nodes[c->pnn]->flags =
- (ctdb->nodes[c->pnn]->flags&NODE_FLAGS_DISCONNECTED)
- | (c->new_flags & ~NODE_FLAGS_DISCONNECTED);
- DEBUG(DEBUG_INFO,("Node flags for node %u are now 0x%x\n", c->pnn, ctdb->nodes[c->pnn]->flags));
+struct ctdb_client_pid_list {
+ struct ctdb_client_pid_list *next, *prev;
+ struct ctdb_context *ctdb;
+ pid_t pid;
+ struct ctdb_client *client;
+};
- /* make sure we don't hold any IPs when we shouldn't */
- if (c->pnn == ctdb->pnn &&
- (ctdb->nodes[c->pnn]->flags & (NODE_FLAGS_INACTIVE|NODE_FLAGS_BANNED))) {
- ctdb_release_all_ips(ctdb);
- }
-}
+static void daemon_incoming_packet(void *, struct ctdb_req_header *);
static void print_exit_message(void)
{
DEBUG(DEBUG_NOTICE,("CTDB daemon shutting down\n"));
}
-
/* called when the "startup" event script has finished */
static void ctdb_start_transport(struct ctdb_context *ctdb)
{
/* Make sure we log something when the daemon terminates */
atexit(print_exit_message);
- /* a handler for when nodes are disabled/enabled */
- ctdb_register_message_handler(ctdb, ctdb, CTDB_SRVID_NODE_FLAGS_CHANGED,
- flag_change_handler, NULL);
-
/* start monitoring for connected/disconnected nodes */
ctdb_start_keepalive(ctdb);
/* start periodic update of tcp tickle lists */
ctdb_start_tcp_tickle_update(ctdb);
+
+ /* start listening for recovery daemon pings */
+ ctdb_control_recd_ping(ctdb);
}
static void block_signal(int signum)
static int daemon_queue_send(struct ctdb_client *client, struct ctdb_req_header *hdr)
{
client->ctdb->statistics.client_packets_sent++;
+ if (hdr->operation == CTDB_REQ_MESSAGE) {
+ if (ctdb_queue_length(client->queue) > client->ctdb->tunable.max_queue_depth_drop_msg) {
+ DEBUG(DEBUG_ERR,("CTDB_REQ_MESSAGE queue full - killing client connection.\n"));
+ talloc_free(client);
+ return -1;
+ }
+ }
return ctdb_queue_send(client->queue, (uint8_t *)hdr, hdr->length);
}
talloc_free(r);
}
-
/*
this is called when the ctdb daemon received a ctdb request to
(unsigned long long)srvid));
}
- /* this is a hack for Samba - we now know the pid of the Samba client */
- if ((srvid & 0xFFFFFFFF) == srvid &&
- kill(srvid, 0) == 0) {
- client->pid = srvid;
- DEBUG(DEBUG_INFO,(__location__ " Registered PID %u for client %u\n",
- (unsigned)client->pid, client_id));
- }
return res;
}
*/
static int ctdb_client_destructor(struct ctdb_client *client)
{
+ struct ctdb_db_context *ctdb_db;
+
ctdb_takeover_client_destructor_hook(client);
ctdb_reqid_remove(client->ctdb, client->client_id);
- client->ctdb->statistics.num_clients--;
+ if (client->ctdb->statistics.num_clients) {
+ client->ctdb->statistics.num_clients--;
+ }
+
+ if (client->num_persistent_updates != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " Client disconnecting with %u persistent updates in flight. Starting recovery\n", client->num_persistent_updates));
+ client->ctdb->recovery_mode = CTDB_RECOVERY_ACTIVE;
+ }
+ ctdb_db = find_ctdb_db(client->ctdb, client->db_id);
+ if (ctdb_db) {
+ DEBUG(DEBUG_ERR, (__location__ " client exit while transaction "
+ "commit active. Forcing recovery.\n"));
+ client->ctdb->recovery_mode = CTDB_RECOVERY_ACTIVE;
+ ctdb_db->transaction_active = false;
+ }
+
return 0;
}
int res;
uint32_t length;
struct ctdb_client *client = dstate->client;
+ struct ctdb_db_context *ctdb_db = state->ctdb_db;
talloc_steal(client, dstate);
talloc_steal(dstate, dstate->call);
res = ctdb_daemon_call_recv(state, dstate->call);
if (res != 0) {
DEBUG(DEBUG_ERR, (__location__ " ctdbd_call_recv() returned error\n"));
- client->ctdb->statistics.pending_calls--;
- ctdb_latency(&client->ctdb->statistics.max_call_latency, dstate->start_time);
+ if (client->ctdb->statistics.pending_calls > 0) {
+ client->ctdb->statistics.pending_calls--;
+ }
+ ctdb_latency(ctdb_db, "call_from_client_cb 1", &client->ctdb->statistics.max_call_latency, dstate->start_time);
return;
}
length, struct ctdb_reply_call);
if (r == NULL) {
DEBUG(DEBUG_ERR, (__location__ " Failed to allocate reply_call in ctdb daemon\n"));
- client->ctdb->statistics.pending_calls--;
- ctdb_latency(&client->ctdb->statistics.max_call_latency, dstate->start_time);
+ if (client->ctdb->statistics.pending_calls > 0) {
+ client->ctdb->statistics.pending_calls--;
+ }
+ ctdb_latency(ctdb_db, "call_from_client_cb 2", &client->ctdb->statistics.max_call_latency, dstate->start_time);
return;
}
r->hdr.reqid = dstate->reqid;
memcpy(&r->data[0], dstate->call->reply_data.dptr, r->datalen);
res = daemon_queue_send(client, &r->hdr);
+ if (res == -1) {
+ /* client is dead - return immediately */
+ return;
+ }
if (res != 0) {
DEBUG(DEBUG_ERR, (__location__ " Failed to queue packet from daemon to client\n"));
}
- ctdb_latency(&client->ctdb->statistics.max_call_latency, dstate->start_time);
+ ctdb_latency(ctdb_db, "call_from_client_cb 3", &client->ctdb->statistics.max_call_latency, dstate->start_time);
talloc_free(dstate);
- client->ctdb->statistics.pending_calls--;
+ if (client->ctdb->statistics.pending_calls > 0) {
+ client->ctdb->statistics.pending_calls--;
+ }
}
struct ctdb_daemon_packet_wrap {
struct ctdb_daemon_packet_wrap *w;
ctdb->statistics.total_calls++;
- ctdb->statistics.pending_calls++;
+ if (client->ctdb->statistics.pending_calls > 0) {
+ ctdb->statistics.pending_calls++;
+ }
ctdb_db = find_ctdb_db(client->ctdb, c->db_id);
if (!ctdb_db) {
DEBUG(DEBUG_ERR, (__location__ " Unknown database in request. db_id==0x%08x",
c->db_id));
- ctdb->statistics.pending_calls--;
+ if (client->ctdb->statistics.pending_calls > 0) {
+ ctdb->statistics.pending_calls--;
+ }
return;
}
+ if (ctdb_db->unhealthy_reason) {
+ /*
+ * this is just a warning, as the tdb should be empty anyway,
+ * and only persistent databases can be unhealthy, which doesn't
+ * use this code patch
+ */
+ DEBUG(DEBUG_WARNING,("warn: db(%s) unhealty in daemon_request_call_from_client(): %s\n",
+ ctdb_db->db_name, ctdb_db->unhealthy_reason));
+ }
+
key.dptr = c->data;
key.dsize = c->keylen;
daemon_incoming_packet_wrap, w, True);
if (ret == -2) {
/* will retry later */
- ctdb->statistics.pending_calls--;
+ if (client->ctdb->statistics.pending_calls > 0) {
+ ctdb->statistics.pending_calls--;
+ }
return;
}
if (ret != 0) {
DEBUG(DEBUG_ERR,(__location__ " Unable to fetch record\n"));
- ctdb->statistics.pending_calls--;
+ if (client->ctdb->statistics.pending_calls > 0) {
+ ctdb->statistics.pending_calls--;
+ }
return;
}
dstate = talloc(client, struct daemon_call_state);
if (dstate == NULL) {
- ctdb_ltdb_unlock(ctdb_db, key);
+ ret = ctdb_ltdb_unlock(ctdb_db, key);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
+ }
+
DEBUG(DEBUG_ERR,(__location__ " Unable to allocate dstate\n"));
- ctdb->statistics.pending_calls--;
+ if (client->ctdb->statistics.pending_calls > 0) {
+ ctdb->statistics.pending_calls--;
+ }
return;
}
dstate->start_time = timeval_current();
call = dstate->call = talloc_zero(dstate, struct ctdb_call);
if (call == NULL) {
- ctdb_ltdb_unlock(ctdb_db, key);
+ ret = ctdb_ltdb_unlock(ctdb_db, key);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
+ }
+
DEBUG(DEBUG_ERR,(__location__ " Unable to allocate call\n"));
- ctdb->statistics.pending_calls--;
- ctdb_latency(&ctdb->statistics.max_call_latency, dstate->start_time);
+ if (client->ctdb->statistics.pending_calls > 0) {
+ ctdb->statistics.pending_calls--;
+ }
+ ctdb_latency(ctdb_db, "call_from_client 1", &ctdb->statistics.max_call_latency, dstate->start_time);
return;
}
state = ctdb_daemon_call_send_remote(ctdb_db, call, &header);
}
- ctdb_ltdb_unlock(ctdb_db, key);
+ ret = ctdb_ltdb_unlock(ctdb_db, key);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
+ }
if (state == NULL) {
DEBUG(DEBUG_ERR,(__location__ " Unable to setup call send\n"));
- ctdb->statistics.pending_calls--;
- ctdb_latency(&ctdb->statistics.max_call_latency, dstate->start_time);
+ if (client->ctdb->statistics.pending_calls > 0) {
+ ctdb->statistics.pending_calls--;
+ }
+ ctdb_latency(ctdb_db, "call_from_client 2", &ctdb->statistics.max_call_latency, dstate->start_time);
return;
}
talloc_steal(state, dstate);
daemon_incoming_packet(client, hdr);
}
+
+static int ctdb_clientpid_destructor(struct ctdb_client_pid_list *client_pid)
+{
+ if (client_pid->ctdb->client_pids != NULL) {
+ DLIST_REMOVE(client_pid->ctdb->client_pids, client_pid);
+ }
+
+ return 0;
+}
+
+
static void ctdb_accept_client(struct event_context *ev, struct fd_event *fde,
uint16_t flags, void *private_data)
{
- struct sockaddr_in addr;
+ struct sockaddr_un addr;
socklen_t len;
int fd;
struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
struct ctdb_client *client;
+ struct ctdb_client_pid_list *client_pid;
#ifdef _AIX
struct peercred_struct cr;
socklen_t crl = sizeof(struct peercred_struct);
set_nonblocking(fd);
set_close_on_exec(fd);
+ DEBUG(DEBUG_DEBUG,(__location__ " Created SOCKET FD:%d to connected child\n", fd));
+
client = talloc_zero(ctdb, struct ctdb_client);
#ifdef _AIX
if (getsockopt(fd, SOL_SOCKET, SO_PEERID, &cr, &crl) == 0) {
#else
if (getsockopt(fd, SOL_SOCKET, SO_PEERCRED, &cr, &crl) == 0) {
#endif
- talloc_asprintf(client, "struct ctdb_client: pid:%u", (unsigned)cr.pid);
+ DEBUG(DEBUG_INFO,("Connected client with pid:%u\n", (unsigned)cr.pid));
}
client->ctdb = ctdb;
client->fd = fd;
client->client_id = ctdb_reqid_new(ctdb, client);
- ctdb->statistics.num_clients++;
+ client->pid = cr.pid;
+
+ client_pid = talloc(client, struct ctdb_client_pid_list);
+ if (client_pid == NULL) {
+ DEBUG(DEBUG_ERR,("Failed to allocate client pid structure\n"));
+ close(fd);
+ talloc_free(client);
+ return;
+ }
+ client_pid->ctdb = ctdb;
+ client_pid->pid = cr.pid;
+ client_pid->client = client;
+
+ DLIST_ADD(ctdb->client_pids, client_pid);
client->queue = ctdb_queue_setup(ctdb, client, fd, CTDB_DS_ALIGNMENT,
- ctdb_daemon_read_cb, client);
+ ctdb_daemon_read_cb, client,
+ "client-%u", client->pid);
talloc_set_destructor(client, ctdb_client_destructor);
+ talloc_set_destructor(client_pid, ctdb_clientpid_destructor);
+ ctdb->statistics.num_clients++;
}
}
- if (listen(ctdb->daemon.sd, 10) != 0) {
+ if (listen(ctdb->daemon.sd, 100) != 0) {
DEBUG(DEBUG_CRIT,("Unable to listen on ctdb socket '%s'\n", ctdb->daemon.name));
goto failed;
}
return -1;
}
-/*
- delete the socket on exit - called on destruction of autofree context
- */
-static int unlink_destructor(const char *name)
+static void sig_child_handler(struct event_context *ev,
+ struct signal_event *se, int signum, int count,
+ void *dont_care,
+ void *private_data)
{
- unlink(name);
- return 0;
+// struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
+ int status;
+ pid_t pid = -1;
+
+ while (pid != 0) {
+ pid = waitpid(-1, &status, WNOHANG);
+ if (pid == -1) {
+ DEBUG(DEBUG_ERR, (__location__ " waitpid() returned error. errno:%d\n", errno));
+ return;
+ }
+ if (pid > 0) {
+ DEBUG(DEBUG_DEBUG, ("SIGCHLD from %d\n", (int)pid));
+ }
+ }
+}
+
+static void ctdb_setup_event_callback(struct ctdb_context *ctdb, int status,
+ void *private_data)
+{
+ if (status != 0) {
+ ctdb_fatal(ctdb, "Failed to run setup event\n");
+ return;
+ }
+ ctdb_run_notification_script(ctdb, "setup");
+
+ /* tell all other nodes we've just started up */
+ ctdb_daemon_send_control(ctdb, CTDB_BROADCAST_ALL,
+ 0, CTDB_CONTROL_STARTUP, 0,
+ CTDB_CTRL_FLAG_NOREPLY,
+ tdb_null, NULL, NULL);
}
/*
start the protocol going as a daemon
*/
-int ctdb_start_daemon(struct ctdb_context *ctdb, bool do_fork)
+int ctdb_start_daemon(struct ctdb_context *ctdb, bool do_fork, bool use_syslog)
{
int res, ret = -1;
struct fd_event *fde;
const char *domain_socket_name;
+ struct signal_event *se;
/* get rid of any old sockets */
unlink(ctdb->daemon.name);
}
block_signal(SIGPIPE);
- if (ctdb->do_setsched) {
- /* try to set us up as realtime */
- ctdb_set_scheduler(ctdb);
- }
+ ctdbd_pid = getpid();
+
+
+ DEBUG(DEBUG_ERR, ("Starting CTDBD as pid : %u\n", ctdbd_pid));
+
+ ctdb_high_priority(ctdb);
/* ensure the socket is deleted on exit of the daemon */
domain_socket_name = talloc_strdup(talloc_autofree_context(), ctdb->daemon.name);
- talloc_set_destructor(domain_socket_name, unlink_destructor);
+ if (domain_socket_name == NULL) {
+ DEBUG(DEBUG_ALERT,(__location__ " talloc_strdup failed.\n"));
+ exit(12);
+ }
ctdb->ev = event_context_init(NULL);
+ tevent_loop_allow_nesting(ctdb->ev);
ctdb_set_child_logging(ctdb);
ctdb_fatal(ctdb, "transport failed to initialise");
}
- /* attach to any existing persistent databases */
- if (ctdb_attach_persistent(ctdb) != 0) {
- ctdb_fatal(ctdb, "Failed to attach to persistent databases\n");
+ /* attach to existing databases */
+ if (ctdb_attach_databases(ctdb) != 0) {
+ ctdb_fatal(ctdb, "Failed to attach to databases\n");
+ }
+
+ ret = ctdb_event_script(ctdb, CTDB_EVENT_INIT);
+ if (ret != 0) {
+ ctdb_fatal(ctdb, "Failed to run init event\n");
}
+ ctdb_run_notification_script(ctdb, "init");
/* start frozen, then let the first election sort things out */
- if (!ctdb_blocking_freeze(ctdb)) {
+ if (ctdb_blocking_freeze(ctdb)) {
ctdb_fatal(ctdb, "Failed to get initial freeze\n");
}
/* now start accepting clients, only can do this once frozen */
fde = event_add_fd(ctdb->ev, ctdb, ctdb->daemon.sd,
- EVENT_FD_READ|EVENT_FD_AUTOCLOSE,
+ EVENT_FD_READ,
ctdb_accept_client, ctdb);
-
- /* tell all other nodes we've just started up */
- ctdb_daemon_send_control(ctdb, CTDB_BROADCAST_ALL,
- 0, CTDB_CONTROL_STARTUP, 0,
- CTDB_CTRL_FLAG_NOREPLY,
- tdb_null, NULL, NULL);
+ tevent_fd_set_auto_close(fde);
/* release any IPs we hold from previous runs of the daemon */
ctdb_release_all_ips(ctdb);
/* start the transport going */
ctdb_start_transport(ctdb);
- /* we dont want any SIGCHLD */
- signal(SIGCHLD, SIG_IGN);
+ /* set up a handler to pick up sigchld */
+ se = event_add_signal(ctdb->ev, ctdb,
+ SIGCHLD, 0,
+ sig_child_handler,
+ ctdb);
+ if (se == NULL) {
+ DEBUG(DEBUG_CRIT,("Failed to set up signal handler for SIGCHLD\n"));
+ exit(1);
+ }
+
+ ret = ctdb_event_script_callback(ctdb,
+ ctdb,
+ ctdb_setup_event_callback,
+ ctdb,
+ false,
+ CTDB_EVENT_SETUP,
+ "");
+ if (ret != 0) {
+ DEBUG(DEBUG_CRIT,("Failed to set up 'setup' event\n"));
+ exit(1);
+ }
+ if (use_syslog) {
+ if (start_syslog_daemon(ctdb)) {
+ DEBUG(DEBUG_CRIT, ("Failed to start syslog daemon\n"));
+ exit(10);
+ }
+ }
+
+ ctdb_lockdown_memory(ctdb);
+
/* go into a wait loop to allow other nodes to complete */
event_loop_wait(ctdb->ev);
struct ctdb_client *client = state->client;
struct ctdb_reply_control *r;
size_t len;
+ int ret;
/* construct a message to send to the client containing the data */
len = offsetof(struct ctdb_reply_control, data) + data.dsize;
memcpy(&r->data[r->datalen], errormsg, r->errorlen);
}
- daemon_queue_send(client, &r->hdr);
-
- talloc_free(state);
+ ret = daemon_queue_send(client, &r->hdr);
+ if (ret != -1) {
+ talloc_free(state);
+ }
}
/*
struct ctdb_req_message *r;
int len;
+ if (ctdb->methods == NULL) {
+ DEBUG(DEBUG_ERR,(__location__ " Failed to send message. Transport is DOWN\n"));
+ return -1;
+ }
+
/* see if this is a message to ourselves */
if (pnn == ctdb->pnn) {
return ctdb_local_message(ctdb, srvid, data);
return 0;
}
+
+
+struct ctdb_client_notify_list {
+ struct ctdb_client_notify_list *next, *prev;
+ struct ctdb_context *ctdb;
+ uint64_t srvid;
+ TDB_DATA data;
+};
+
+
+static int ctdb_client_notify_destructor(struct ctdb_client_notify_list *nl)
+{
+ int ret;
+
+ DEBUG(DEBUG_ERR,("Sending client notify message for srvid:%llu\n", (unsigned long long)nl->srvid));
+
+ ret = ctdb_daemon_send_message(nl->ctdb, CTDB_BROADCAST_CONNECTED, (unsigned long long)nl->srvid, nl->data);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,("Failed to send client notify message\n"));
+ }
+
+ return 0;
+}
+
+int32_t ctdb_control_register_notify(struct ctdb_context *ctdb, uint32_t client_id, TDB_DATA indata)
+{
+ struct ctdb_client_notify_register *notify = (struct ctdb_client_notify_register *)indata.dptr;
+ struct ctdb_client *client = ctdb_reqid_find(ctdb, client_id, struct ctdb_client);
+ struct ctdb_client_notify_list *nl;
+
+ DEBUG(DEBUG_INFO,("Register srvid %llu for client %d\n", (unsigned long long)notify->srvid, client_id));
+
+ if (indata.dsize < offsetof(struct ctdb_client_notify_register, notify_data)) {
+ DEBUG(DEBUG_ERR,(__location__ " Too little data in control : %d\n", (int)indata.dsize));
+ return -1;
+ }
+
+ if (indata.dsize != (notify->len + offsetof(struct ctdb_client_notify_register, notify_data))) {
+ DEBUG(DEBUG_ERR,(__location__ " Wrong amount of data in control. Got %d, expected %d\n", (int)indata.dsize, (int)(notify->len + offsetof(struct ctdb_client_notify_register, notify_data))));
+ return -1;
+ }
+
+
+ if (client == NULL) {
+ DEBUG(DEBUG_ERR,(__location__ " Could not find client parent structure. You can not send this control to a remote node\n"));
+ return -1;
+ }
+
+ for(nl=client->notify; nl; nl=nl->next) {
+ if (nl->srvid == notify->srvid) {
+ break;
+ }
+ }
+ if (nl != NULL) {
+ DEBUG(DEBUG_ERR,(__location__ " Notification for srvid:%llu already exists for this client\n", (unsigned long long)notify->srvid));
+ return -1;
+ }
+
+ nl = talloc(client, struct ctdb_client_notify_list);
+ CTDB_NO_MEMORY(ctdb, nl);
+ nl->ctdb = ctdb;
+ nl->srvid = notify->srvid;
+ nl->data.dsize = notify->len;
+ nl->data.dptr = talloc_size(nl, nl->data.dsize);
+ CTDB_NO_MEMORY(ctdb, nl->data.dptr);
+ memcpy(nl->data.dptr, notify->notify_data, nl->data.dsize);
+
+ DLIST_ADD(client->notify, nl);
+ talloc_set_destructor(nl, ctdb_client_notify_destructor);
+
+ return 0;
+}
+
+int32_t ctdb_control_deregister_notify(struct ctdb_context *ctdb, uint32_t client_id, TDB_DATA indata)
+{
+ struct ctdb_client_notify_deregister *notify = (struct ctdb_client_notify_deregister *)indata.dptr;
+ struct ctdb_client *client = ctdb_reqid_find(ctdb, client_id, struct ctdb_client);
+ struct ctdb_client_notify_list *nl;
+
+ DEBUG(DEBUG_INFO,("Deregister srvid %llu for client %d\n", (unsigned long long)notify->srvid, client_id));
+
+ if (client == NULL) {
+ DEBUG(DEBUG_ERR,(__location__ " Could not find client parent structure. You can not send this control to a remote node\n"));
+ return -1;
+ }
+
+ for(nl=client->notify; nl; nl=nl->next) {
+ if (nl->srvid == notify->srvid) {
+ break;
+ }
+ }
+ if (nl == NULL) {
+ DEBUG(DEBUG_ERR,(__location__ " No notification for srvid:%llu found for this client\n", (unsigned long long)notify->srvid));
+ return -1;
+ }
+
+ DLIST_REMOVE(client->notify, nl);
+ talloc_set_destructor(nl, NULL);
+ talloc_free(nl);
+
+ return 0;
+}
+
+struct ctdb_client *ctdb_find_client_by_pid(struct ctdb_context *ctdb, pid_t pid)
+{
+ struct ctdb_client_pid_list *client_pid;
+
+ for (client_pid = ctdb->client_pids; client_pid; client_pid=client_pid->next) {
+ if (client_pid->pid == pid) {
+ return client_pid->client;
+ }
+ }
+ return NULL;
+}
+
+
+/* This control is used by samba when probing if a process (of a samba daemon)
+ exists on the node.
+ Samba does this when it needs/wants to check if a subrecord in one of the
+ databases is still valied, or if it is stale and can be removed.
+ If the node is in unhealthy or stopped state we just kill of the samba
+ process holding htis sub-record and return to the calling samba that
+ the process does not exist.
+ This allows us to forcefully recall subrecords registered by samba processes
+ on banned and stopped nodes.
+*/
+int32_t ctdb_control_process_exists(struct ctdb_context *ctdb, pid_t pid)
+{
+ struct ctdb_client *client;
+
+ if (ctdb->nodes[ctdb->pnn]->flags & (NODE_FLAGS_BANNED|NODE_FLAGS_STOPPED)) {
+ client = ctdb_find_client_by_pid(ctdb, pid);
+ if (client != NULL) {
+ DEBUG(DEBUG_NOTICE,(__location__ " Killing client with pid:%d on banned/stopped node\n", (int)pid));
+ talloc_free(client);
+ }
+ return -1;
+ }
+
+ return kill(pid, 0);
+}
git.samba.org / metze / ctdb / wip.git / blobdiff