static void ctdb_run_startup(struct event_context *ev, struct timed_event *te,
struct timeval t, void *private_data)
{
struct ctdb_context *ctdb = talloc_get_type(private_data,
struct ctdb_context);
int ret;
/* This is necessary to avoid the "startup" event colliding
* with the "ipreallocated" event from the takeover run
* following the first recovery. We might as well serialise
* these things if we can.
*/
if (ctdb->runstate < CTDB_RUNSTATE_STARTUP) {
DEBUG(DEBUG_NOTICE,
("Not yet in startup runstate. Wait one more second\n"));
event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
timeval_current_ofs(1, 0),
ctdb_run_startup, ctdb);
return;
}
DEBUG(DEBUG_NOTICE,("Running the \"startup\" event.\n"));
ret = ctdb_event_script_callback(ctdb,
ctdb->monitor->monitor_context,
ctdb_startup_callback,
ctdb, CTDB_EVENT_STARTUP, "%s", "");
if (ret != 0) {
DEBUG(DEBUG_ERR,("Unable to launch startup event script\n"));
event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
timeval_current_ofs(5, 0),
ctdb_run_startup, ctdb);
}
}
/*
main program
*/
int main(int argc, const char *argv[])
{
struct ctdb_context *ctdb;
struct ctdb_db_context *ctdb_db;
struct poptOption popt_options[] = {
POPT_AUTOHELP
POPT_CTDB_CMDLINE
{ "num-records", 'r', POPT_ARG_INT, &num_records, 0, "num_records", "integer" },
{ "base-rec", 'b', POPT_ARG_INT, &base_rec, 0, "base_rec", "integer" },
{ "delete-pct", 'p', POPT_ARG_INT, &delete_pct, 0, "delete_pct", "integer" },
POPT_TABLEEND
};
int opt;
const char **extra_argv;
int extra_argc = 0;
poptContext pc;
struct event_context *ev;
pc = poptGetContext(argv[0], argc, argv, popt_options, POPT_CONTEXT_KEEP_FIRST);
while ((opt = poptGetNextOpt(pc)) != -1) {
switch (opt) {
default:
fprintf(stderr, "Invalid option %s: %s\n",
poptBadOption(pc, 0), poptStrerror(opt));
exit(1);
}
}
/* setup the remaining options for the main program to use */
extra_argv = poptGetArgs(pc);
if (extra_argv) {
extra_argv++;
while (extra_argv[extra_argc]) extra_argc++;
}
ev = event_context_init(NULL);
ctdb = ctdb_cmdline_client(ev, timeval_current_ofs(3, 0));
if (ctdb == NULL) {
printf("failed to connect to daemon\n");
exit(1);
}
/* attach to a specific database */
ctdb_db = ctdb_attach(ctdb, timeval_current_ofs(2, 0), "test.tdb",
false, 0);
if (!ctdb_db) {
printf("ctdb_attach failed - %s\n", ctdb_errstr(ctdb));
exit(1);
}
store_records(ctdb, ev);
return 0;
}
static bool ldapsrv_call_read_next(struct ldapsrv_connection *conn)
{
struct tevent_req *subreq;
if (timeval_is_zero(&conn->limits.endtime)) {
conn->limits.endtime =
timeval_current_ofs(conn->limits.initial_timeout, 0);
} else {
conn->limits.endtime =
timeval_current_ofs(conn->limits.conn_idle_time, 0);
}
/*
* The minimun size of a LDAP pdu is 7 bytes
*
* dumpasn1 -hh ldap-unbind-min.dat
*
* <30 05 02 01 09 42 00>
* 0 5: SEQUENCE {
* <02 01 09>
* 2 1: INTEGER 9
* <42 00>
* 5 0: [APPLICATION 2]
* : Error: Object has zero length.
* : }
*
* dumpasn1 -hh ldap-unbind-windows.dat
*
* <30 84 00 00 00 05 02 01 09 42 00>
* 0 5: SEQUENCE {
* <02 01 09>
* 6 1: INTEGER 9
* <42 00>
* 9 0: [APPLICATION 2]
* : Error: Object has zero length.
* : }
*
* This means using an initial read size
* of 7 is ok.
*/
subreq = tstream_read_pdu_blob_send(conn,
conn->connection->event.ctx,
conn->sockets.active,
7, /* initial_read_size */
ldap_full_packet,
conn);
if (subreq == NULL) {
ldapsrv_terminate_connection(conn, "ldapsrv_call_read_next: "
"no memory for tstream_read_pdu_blob_send");
return false;
}
tevent_req_set_endtime(subreq,
conn->connection->event.ctx,
conn->limits.endtime);
tevent_req_set_callback(subreq, ldapsrv_call_read_done, conn);
return true;
}
/*
see if the event scripts think we are healthy
*/
static void ctdb_check_health(struct event_context *ev, struct timed_event *te,
struct timeval t, void *private_data)
{
struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
int ret = 0;
if (ctdb->recovery_mode != CTDB_RECOVERY_NORMAL ||
(ctdb->monitor->monitoring_mode == CTDB_MONITORING_DISABLED && ctdb->done_startup)) {
event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
timeval_current_ofs(ctdb->monitor->next_interval, 0),
ctdb_check_health, ctdb);
return;
}
if (!ctdb->done_startup) {
ret = ctdb_event_script_callback(ctdb,
ctdb->monitor->monitor_context, ctdb_startup_callback,
ctdb, false,
CTDB_EVENT_STARTUP, "%s", "");
} else {
int i;
int skip_monitoring = 0;
if (ctdb->recovery_mode != CTDB_RECOVERY_NORMAL) {
skip_monitoring = 1;
DEBUG(DEBUG_ERR,("Skip monitoring during recovery\n"));
}
for (i=1; i<=NUM_DB_PRIORITIES; i++) {
if (ctdb->freeze_handles[i] != NULL) {
DEBUG(DEBUG_ERR,("Skip monitoring since databases are frozen\n"));
skip_monitoring = 1;
break;
}
}
if (skip_monitoring != 0) {
event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
timeval_current_ofs(ctdb->monitor->next_interval, 0),
ctdb_check_health, ctdb);
return;
} else {
ret = ctdb_event_script_callback(ctdb,
ctdb->monitor->monitor_context, ctdb_health_callback,
ctdb, false,
CTDB_EVENT_MONITOR, "%s", "");
}
}
if (ret != 0) {
DEBUG(DEBUG_ERR,("Unable to launch monitor event script\n"));
ctdb->monitor->next_interval = 5;
event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
timeval_current_ofs(5, 0),
ctdb_check_health, ctdb);
}
}
/*
complete an async reconnect
*/
static void reopen_connection_complete(struct composite_context *ctx)
{
struct benchopen_state *state = (struct benchopen_state *)ctx->async.private_data;
NTSTATUS status;
struct smb_composite_connect *io = &state->reconnect;
status = smb_composite_connect_recv(ctx, state->mem_ctx);
if (!NT_STATUS_IS_OK(status)) {
talloc_free(state->te);
state->te = event_add_timed(state->ev, state->mem_ctx,
timeval_current_ofs(1,0),
reopen_connection, state);
return;
}
state->tree = io->out.tree;
num_connected++;
DEBUG(0,("[%u] reconnect to %s finished (%u connected)\n",
state->client_num, state->dest_host, num_connected));
state->open_fnum = -1;
state->close_fnum = -1;
next_open(state);
}
static void report_rate(struct tevent_context *ev, struct tevent_timer *te,
struct timeval t, void *private_data)
{
struct benchopen_state *state = talloc_get_type(private_data,
struct benchopen_state);
int i;
for (i=0;i<nprocs;i++) {
printf("%5u ", (unsigned)(state[i].count - state[i].lastcount));
state[i].lastcount = state[i].count;
}
printf("\r");
fflush(stdout);
report_te = event_add_timed(ev, state, timeval_current_ofs(1, 0),
report_rate, state);
/* send an echo on each interface to ensure it stays alive - this helps
with IP takeover */
for (i=0;i<nprocs;i++) {
struct smb_echo p;
struct smbcli_request *req;
if (!state[i].tree) {
continue;
}
p.in.repeat_count = 1;
p.in.size = 0;
p.in.data = NULL;
req = smb_raw_echo_send(state[i].tree->session->transport, &p);
req->async.private_data = &state[i];
req->async.fn = echo_completion;
}
}
static void aio_linux_housekeeping(struct tevent_context *event_ctx,
struct tevent_timer *te,
struct timeval now,
void *private_data)
{
/* Remove this timed event handler. */
TALLOC_FREE(te);
if ((num_busy != 0) || used) {
used = false;
/* Still busy. Look again in 30 seconds. */
(void)tevent_add_timer(event_ctx,
NULL,
timeval_current_ofs(30, 0),
aio_linux_housekeeping,
NULL);
return;
}
/* No activity for 30 seconds. Close out kernel resources. */
io_queue_release(io_ctx);
memset(&io_ctx, '\0', sizeof(io_ctx));
if (event_fd != -1) {
close(event_fd);
event_fd = -1;
}
TALLOC_FREE(aio_read_event);
}
/*
handle a socket write event
*/
static void messaging_send_handler(struct messaging_context *msg)
{
while (msg->pending) {
struct messaging_rec *rec = msg->pending;
NTSTATUS status;
status = try_send(rec);
if (NT_STATUS_EQUAL(status, STATUS_MORE_ENTRIES)) {
rec->retries++;
if (rec->retries > 3) {
/* we're getting continuous write errors -
backoff this record */
DLIST_REMOVE(msg->pending, rec);
DLIST_ADD_END(msg->retry_queue, rec,
struct messaging_rec *);
if (msg->retry_te == NULL) {
msg->retry_te =
event_add_timed(msg->event.ev, msg,
timeval_current_ofs(1, 0),
msg_retry_timer, msg);
}
}
break;
}
rec->retries = 0;
if (!NT_STATUS_IS_OK(status)) {
DEBUG(1,("messaging: Lost message from %s to %s of type %u - %s\n",
cluster_id_string(debug_ctx(), rec->header->from),
cluster_id_string(debug_ctx(), rec->header->to),
rec->header->msg_type,
nt_errstr(status)));
}
DLIST_REMOVE(msg->pending, rec);
talloc_free(rec);
}
static void add_oplock_timeout_handler(files_struct *fsp)
{
struct smbd_server_connection *sconn = fsp->conn->sconn;
struct kernel_oplocks *koplocks = sconn->oplocks.kernel_ops;
/*
* If kernel oplocks already notifies smbds when an oplock break times
* out, just return.
*/
if (koplocks &&
(koplocks->flags & KOPLOCKS_TIMEOUT_NOTIFICATION)) {
return;
}
if (fsp->oplock_timeout != NULL) {
DEBUG(0, ("Logic problem -- have an oplock event hanging "
"around\n"));
}
fsp->oplock_timeout =
tevent_add_timer(fsp->conn->sconn->ev_ctx, fsp,
timeval_current_ofs(OPLOCK_BREAK_TIMEOUT, 0),
oplock_timeout_handler, fsp);
if (fsp->oplock_timeout == NULL) {
DEBUG(0, ("Could not add oplock timeout handler\n"));
}
}
static void wait_replies(struct tevent_context *ev_ctx,
struct messaging_context *msg_ctx,
bool multiple_replies)
{
struct tevent_timer *te;
bool timed_out = False;
te = tevent_add_timer(ev_ctx, NULL,
timeval_current_ofs(timeout, 0),
smbcontrol_timeout, (void *)&timed_out);
if (te == NULL) {
DEBUG(0, ("tevent_add_timer failed\n"));
return;
}
while (!timed_out) {
int ret;
if (num_replies > 0 && !multiple_replies)
break;
ret = tevent_loop_once(ev_ctx);
if (ret != 0) {
break;
}
}
}
/*
establish a new connection to the web server
*/
static void websrv_accept(struct stream_connection *conn)
{
struct web_server_data *wdata = talloc_get_type(conn->private_data, struct web_server_data);
struct websrv_context *web;
struct socket_context *tls_socket;
web = talloc_zero(conn, struct websrv_context);
if (web == NULL) goto failed;
web->task = wdata->task;
web->conn = conn;
conn->private_data = web;
talloc_set_destructor(web, websrv_destructor);
event_add_timed(conn->event.ctx, web,
timeval_current_ofs(HTTP_TIMEOUT, 0),
websrv_timeout, web);
/* Overwrite the socket with a (possibly) TLS socket */
tls_socket = tls_init_server(wdata->tls_params, conn->socket,
conn->event.fde, "GPHO");
/* We might not have TLS, or it might not have initilised */
if (tls_socket) {
talloc_unlink(conn, conn->socket);
talloc_steal(conn, tls_socket);
conn->socket = tls_socket;
} else {
DEBUG(3, ("TLS not available for web_server connections\n"));
}
return;
failed:
talloc_free(conn);
}
static void remove_child_pid(struct smbd_parent_context *parent,
pid_t pid,
bool unclean_shutdown)
{
struct smbd_child_pid *child;
struct server_id child_id;
int ret;
child_id = pid_to_procid(pid);
ret = messaging_cleanup(parent->msg_ctx, pid);
if ((ret != 0) && (ret != ENOENT)) {
DEBUG(10, ("%s: messaging_cleanup returned %s\n",
__func__, strerror(ret)));
}
smbprofile_cleanup(pid);
for (child = parent->children; child != NULL; child = child->next) {
if (child->pid == pid) {
struct smbd_child_pid *tmp = child;
DLIST_REMOVE(parent->children, child);
TALLOC_FREE(tmp);
parent->num_children -= 1;
break;
}
}
if (child == NULL) {
/* not all forked child processes are added to the children list */
DEBUG(2, ("Could not find child %d -- ignoring\n", (int)pid));
return;
}
if (unclean_shutdown) {
/* a child terminated uncleanly so tickle all
processes to see if they can grab any of the
pending locks
*/
DEBUG(3,(__location__ " Unclean shutdown of pid %u\n",
(unsigned int)pid));
if (parent->cleanup_te == NULL) {
/* call the cleanup timer, but not too often */
int cleanup_time = lp_parm_int(-1, "smbd", "cleanuptime", 20);
parent->cleanup_te = tevent_add_timer(parent->ev_ctx,
parent,
timeval_current_ofs(cleanup_time, 0),
cleanup_timeout_fn,
parent);
DEBUG(1,("Scheduled cleanup of brl and lock database after unclean shutdown\n"));
}
}
if (!serverid_deregister(child_id)) {
DEBUG(1, ("Could not remove pid %d from serverid.tdb\n",
(int)pid));
}
}
/*
perform the send side of a async dcerpc request
*/
static struct rpc_request *dcerpc_request_send(struct dcerpc_pipe *p,
const struct GUID *object,
uint16_t opnum,
bool async,
DATA_BLOB *stub_data)
{
struct rpc_request *req;
p->conn->transport.recv_data = dcerpc_recv_data;
req = talloc(p, struct rpc_request);
if (req == NULL) {
return NULL;
}
req->p = p;
req->call_id = next_call_id(p->conn);
req->status = NT_STATUS_OK;
req->state = RPC_REQUEST_QUEUED;
req->payload = data_blob(NULL, 0);
req->flags = 0;
req->fault_code = 0;
req->async_call = async;
req->ignore_timeout = false;
req->async.callback = NULL;
req->async.private_data = NULL;
req->recv_handler = NULL;
if (object != NULL) {
req->object = (struct GUID *)talloc_memdup(req, (const void *)object, sizeof(*object));
if (req->object == NULL) {
talloc_free(req);
return NULL;
}
} else {
req->object = NULL;
}
req->opnum = opnum;
req->request_data.length = stub_data->length;
req->request_data.data = talloc_reference(req, stub_data->data);
if (req->request_data.length && req->request_data.data == NULL) {
return NULL;
}
DLIST_ADD_END(p->conn->request_queue, req, struct rpc_request *);
talloc_set_destructor(req, dcerpc_req_dequeue);
dcerpc_ship_next_request(p->conn);
if (p->request_timeout) {
event_add_timed(dcerpc_event_context(p), req,
timeval_current_ofs(p->request_timeout, 0),
dcerpc_timeout_handler, req);
}
return req;
}
int32_t ctdb_control_set_ban_state(struct ctdb_context *ctdb, TDB_DATA indata)
{
struct ctdb_ban_time *bantime = (struct ctdb_ban_time *)indata.dptr;
DEBUG(DEBUG_INFO,("SET BAN STATE\n"));
if (bantime->pnn != ctdb->pnn) {
if (bantime->pnn < 0 || bantime->pnn >= ctdb->num_nodes) {
DEBUG(DEBUG_ERR,(__location__ " ERROR: Invalid ban request. PNN:%d is invalid. Max nodes %d\n", bantime->pnn, ctdb->num_nodes));
return -1;
}
if (bantime->time == 0) {
DEBUG(DEBUG_INFO,("unbanning node %d\n", bantime->pnn));
ctdb->nodes[bantime->pnn]->flags &= ~NODE_FLAGS_BANNED;
} else {
DEBUG(DEBUG_INFO,("banning node %d\n", bantime->pnn));
if (ctdb->tunable.enable_bans == 0) {
DEBUG(DEBUG_INFO,("Bans are disabled - ignoring ban of node %u\n", bantime->pnn));
return 0;
}
ctdb->nodes[bantime->pnn]->flags |= NODE_FLAGS_BANNED;
}
return 0;
}
if (ctdb->banning_ctx != NULL) {
talloc_free(ctdb->banning_ctx);
ctdb->banning_ctx = NULL;
}
if (bantime->time == 0) {
DEBUG(DEBUG_ERR,("Unbanning this node\n"));
ctdb->nodes[bantime->pnn]->flags &= ~NODE_FLAGS_BANNED;
return 0;
}
if (ctdb->tunable.enable_bans == 0) {
DEBUG(DEBUG_ERR,("Bans are disabled - ignoring ban of node %u\n", bantime->pnn));
return 0;
}
ctdb->banning_ctx = talloc(ctdb, struct ctdb_ban_time);
if (ctdb->banning_ctx == NULL) {
DEBUG(DEBUG_CRIT,(__location__ " ERROR Failed to allocate new banning state\n"));
return -1;
}
*((struct ctdb_ban_time *)(ctdb->banning_ctx)) = *bantime;
DEBUG(DEBUG_ERR,("Banning this node for %d seconds\n", bantime->time));
ctdb->nodes[bantime->pnn]->flags |= NODE_FLAGS_BANNED;
event_add_timed(ctdb->ev, ctdb->banning_ctx, timeval_current_ofs(bantime->time,0), ctdb_ban_node_event, ctdb);
return 0;
}
/*
called when a wins name register has completed
*/
static void nbtd_wins_register_handler(struct composite_context *c)
{
NTSTATUS status;
struct nbt_name_register_wins io;
struct nbtd_iface_name *iname = talloc_get_type(c->async.private_data,
struct nbtd_iface_name);
TALLOC_CTX *tmp_ctx = talloc_new(iname);
status = nbt_name_register_wins_recv(c, tmp_ctx, &io);
if (NT_STATUS_EQUAL(status, NT_STATUS_IO_TIMEOUT)) {
/* none of the WINS servers responded - try again
periodically */
int wins_retry_time = lp_parm_int(iname->iface->nbtsrv->task->lp_ctx, NULL, "nbtd", "wins_retry", 300);
event_add_timed(iname->iface->nbtsrv->task->event_ctx,
iname,
timeval_current_ofs(wins_retry_time, 0),
nbtd_wins_register_retry,
iname);
talloc_free(tmp_ctx);
return;
}
if (!NT_STATUS_IS_OK(status)) {
DEBUG(1,("Name register failure with WINS for %s - %s\n",
nbt_name_string(tmp_ctx, &iname->name), nt_errstr(status)));
talloc_free(tmp_ctx);
return;
}
if (io.out.rcode != 0) {
DEBUG(1,("WINS server %s rejected name register of %s - %s\n",
io.out.wins_server, nbt_name_string(tmp_ctx, &iname->name),
nt_errstr(nbt_rcode_to_ntstatus(io.out.rcode))));
iname->nb_flags |= NBT_NM_CONFLICT;
talloc_free(tmp_ctx);
return;
}
/* success - start a periodic name refresh */
iname->nb_flags |= NBT_NM_ACTIVE;
if (iname->wins_server) {
/*
* talloc_free() would generate a warning,
* so steal it into the tmp context
*/
talloc_steal(tmp_ctx, iname->wins_server);
}
iname->wins_server = talloc_steal(iname, io.out.wins_server);
iname->registration_time = timeval_current();
nbtd_wins_start_refresh_timer(iname);
DEBUG(3,("Registered %s with WINS server %s\n",
nbt_name_string(tmp_ctx, &iname->name), iname->wins_server));
talloc_free(tmp_ctx);
}
static void each_second(struct event_context *ev, struct timed_event *te,
struct timeval t, void *private_data)
{
struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
print_counters();
event_add_timed(ev, ctdb, timeval_current_ofs(1, 0), each_second, ctdb);
}
static bool init_aio_linux(struct vfs_handle_struct *handle)
{
struct tevent_timer *te = NULL;
if (event_fd != -1) {
/* Already initialized. */
return true;
}
/* Schedule a shutdown event for 30 seconds from now. */
te = tevent_add_timer(handle->conn->sconn->ev_ctx,
NULL,
timeval_current_ofs(30, 0),
aio_linux_housekeeping,
NULL);
if (te == NULL) {
goto fail;
}
event_fd = eventfd(0, EFD_NONBLOCK | EFD_CLOEXEC);
if (event_fd == -1) {
goto fail;
}
aio_read_event = tevent_add_fd(server_event_context(),
NULL,
event_fd,
TEVENT_FD_READ,
aio_linux_done,
NULL);
if (aio_read_event == NULL) {
goto fail;
}
if (io_queue_init(lp_aio_max_threads(), &io_ctx)) {
goto fail;
}
DEBUG(10,("init_aio_linux: initialized with up to %d events\n",
(int)lp_aio_max_threads()));
return true;
fail:
DEBUG(10,("init_aio_linux: initialization failed\n"));
TALLOC_FREE(te);
TALLOC_FREE(aio_read_event);
if (event_fd != -1) {
close(event_fd);
event_fd = -1;
}
memset(&io_ctx, '\0', sizeof(io_ctx));
return false;
}
/*
setup a timer to destroy a open search after a inactivity period
*/
static void pvfs_search_setup_timer(struct pvfs_search_state *search)
{
struct tevent_context *ev = search->pvfs->ntvfs->ctx->event_ctx;
if (search->handle == INVALID_SEARCH_HANDLE) return;
talloc_free(search->te);
search->te = event_add_timed(ev, search,
timeval_current_ofs(search->pvfs->search.inactivity_time, 0),
pvfs_search_timer, search);
}
bool fsp_lease_update(struct share_mode_lock *lck,
const struct GUID *client_guid,
struct fsp_lease *lease)
{
struct share_mode_data *d = lck->data;
int idx;
struct share_mode_lease *l = NULL;
idx = find_share_mode_lease(d, client_guid, &lease->lease.lease_key);
if (idx != -1) {
l = &d->leases[idx];
}
if (l == NULL) {
DEBUG(1, ("%s: Could not find lease entry\n", __func__));
TALLOC_FREE(lease->timeout);
lease->lease.lease_state = SMB2_LEASE_NONE;
lease->lease.lease_epoch += 1;
lease->lease.lease_flags = 0;
return false;
}
DEBUG(10,("%s: refresh lease state\n", __func__));
/* Ensure we're in sync with current lease state. */
if (lease->lease.lease_epoch != l->epoch) {
DEBUG(10,("%s: cancel outdated timeout\n", __func__));
TALLOC_FREE(lease->timeout);
}
lease->lease.lease_epoch = l->epoch;
lease->lease.lease_state = l->current_state;
if (l->breaking) {
lease->lease.lease_flags |= SMB2_LEASE_FLAG_BREAK_IN_PROGRESS;
if (lease->timeout == NULL) {
struct timeval t = timeval_current_ofs(OPLOCK_BREAK_TIMEOUT, 0);
DEBUG(10,("%s: setup timeout handler\n", __func__));
lease->timeout = tevent_add_timer(lease->sconn->ev_ctx,
lease, t,
lease_timeout_handler,
lease);
if (lease->timeout == NULL) {
DEBUG(0, ("%s: Could not add lease timeout handler\n",
__func__));
}
}
} else {
lease->lease.lease_flags &= ~SMB2_LEASE_FLAG_BREAK_IN_PROGRESS;
TALLOC_FREE(lease->timeout);
}
return true;
}
/*
see if any nodes are dead
*/
static void ctdb_check_for_dead_nodes(struct tevent_context *ev,
struct tevent_timer *te,
struct timeval t, void *private_data)
{
struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
int i;
/* send a keepalive to all other nodes, unless */
for (i=0;i<ctdb->num_nodes;i++) {
struct ctdb_node *node = ctdb->nodes[i];
if (node->flags & NODE_FLAGS_DELETED) {
continue;
}
if (node->pnn == ctdb->pnn) {
continue;
}
if (node->flags & NODE_FLAGS_DISCONNECTED) {
/* it might have come alive again */
if (node->rx_cnt != 0) {
ctdb_node_connected(node);
}
continue;
}
if (node->rx_cnt == 0) {
node->dead_count++;
} else {
node->dead_count = 0;
}
node->rx_cnt = 0;
if (node->dead_count >= ctdb->tunable.keepalive_limit) {
DEBUG(DEBUG_NOTICE,("dead count reached for node %u\n", node->pnn));
ctdb_node_dead(node);
ctdb_send_keepalive(ctdb, node->pnn);
/* maybe tell the transport layer to kill the
sockets as well?
*/
continue;
}
DEBUG(DEBUG_DEBUG,("sending keepalive to %u\n", node->pnn));
ctdb_send_keepalive(ctdb, node->pnn);
node->tx_cnt = 0;
}
tevent_add_timer(ctdb->ev, ctdb->keepalive_ctx,
timeval_current_ofs(ctdb->tunable.keepalive_interval, 0),
ctdb_check_for_dead_nodes, ctdb);
}
static void wreplsrv_out_pull_reschedule(struct wreplsrv_partner *partner, uint32_t interval)
{
NTSTATUS status;
partner->pull.next_run = timeval_current_ofs(interval, 0);
status = wreplsrv_periodic_schedule(partner->service, interval);
if (!NT_STATUS_IS_OK(status)) {
DEBUG(0,("wreplsrv_periodic_schedule() failed\n"));
}
}
static bool recalc_brl_timeout(void)
{
struct blocking_lock_record *blr;
struct timeval next_timeout;
TALLOC_FREE(brl_timeout);
next_timeout = timeval_zero();
for (blr = blocking_lock_queue; blr; blr = blr->next) {
if (timeval_is_zero(&blr->expire_time)) {
/*
* If we're blocked on pid 0xFFFFFFFF this is
* a POSIX lock, so calculate a timeout of
* 10 seconds into the future.
*/
if (blr->blocking_pid == 0xFFFFFFFF) {
struct timeval psx_to = timeval_current_ofs(10, 0);
next_timeout = timeval_min(&next_timeout, &psx_to);
}
continue;
}
if (timeval_is_zero(&next_timeout)) {
next_timeout = blr->expire_time;
}
else {
next_timeout = timeval_min(&next_timeout,
&blr->expire_time);
}
}
if (timeval_is_zero(&next_timeout)) {
DEBUG(10, ("Next timeout = Infinite.\n"));
return True;
}
if (DEBUGLVL(10)) {
struct timeval cur, from_now;
cur = timeval_current();
from_now = timeval_until(&cur, &next_timeout);
DEBUG(10, ("Next timeout = %d.%d seconds from now.\n",
(int)from_now.tv_sec, (int)from_now.tv_usec));
}
if (!(brl_timeout = event_add_timed(smbd_event_context(), NULL,
next_timeout,
brl_timeout_fn, NULL))) {
return False;
}
return True;
}
/*
* Callback routine when required locks are not obtained within timeout
* Called from parent context
*/
static void ctdb_lock_timeout_handler(struct tevent_context *ev,
struct tevent_timer *ttimer,
struct timeval current_time,
void *private_data)
{
static const char * debug_locks = NULL;
struct lock_context *lock_ctx;
struct ctdb_context *ctdb;
pid_t pid;
lock_ctx = talloc_get_type_abort(private_data, struct lock_context);
ctdb = lock_ctx->ctdb;
if (lock_ctx->type == LOCK_RECORD || lock_ctx->type == LOCK_DB) {
DEBUG(DEBUG_WARNING,
("Unable to get %s lock on database %s for %.0lf seconds\n",
(lock_ctx->type == LOCK_RECORD ? "RECORD" : "DB"),
lock_ctx->ctdb_db->db_name,
timeval_elapsed(&lock_ctx->start_time)));
} else {
DEBUG(DEBUG_WARNING,
("Unable to get ALLDB locks for %.0lf seconds\n",
timeval_elapsed(&lock_ctx->start_time)));
}
/* Fire a child process to find the blocking process. */
if (debug_locks == NULL) {
debug_locks = getenv("CTDB_DEBUG_LOCKS");
if (debug_locks == NULL) {
debug_locks = talloc_asprintf(ctdb,
"%s/debug_locks.sh",
getenv("CTDB_BASE"));
}
}
if (debug_locks != NULL) {
pid = vfork();
if (pid == 0) {
execl(debug_locks, debug_locks, NULL);
_exit(0);
}
ctdb_track_child(ctdb, pid);
} else {
DEBUG(DEBUG_WARNING,
(__location__
" Unable to setup lock debugging - no memory?\n"));
}
/* reset the timeout timer */
// talloc_free(lock_ctx->ttimer);
lock_ctx->ttimer = tevent_add_timer(ctdb->ev,
lock_ctx,
timeval_current_ofs(10, 0),
ctdb_lock_timeout_handler,
(void *)lock_ctx);
}
static void report_rate(struct tevent_context *ev, struct tevent_timer *te,
struct timeval t, void *private_data)
{
struct offline_state *state = talloc_get_type(private_data,
struct offline_state);
int i;
uint32_t total=0, total_offline=0, total_online=0;
for (i=0;i<numstates;i++) {
total += state[i].count - state[i].lastcount;
if (timeval_elapsed(&state[i].tv_start) > latencies[state[i].op]) {
latencies[state[i].op] = timeval_elapsed(&state[i].tv_start);
}
state[i].lastcount = state[i].count;
total_online += state[i].online_count;
total_offline += state[i].offline_count;
}
printf("ops/s=%4u offline=%5u online=%4u set_lat=%.1f/%.1f get_lat=%.1f/%.1f save_lat=%.1f/%.1f load_lat=%.1f/%.1f\n",
total, total_offline, total_online,
latencies[OP_SETOFFLINE],
worst_latencies[OP_SETOFFLINE],
latencies[OP_GETOFFLINE],
worst_latencies[OP_GETOFFLINE],
latencies[OP_SAVEFILE],
worst_latencies[OP_SAVEFILE],
latencies[OP_LOADFILE],
worst_latencies[OP_LOADFILE]);
fflush(stdout);
event_add_timed(ev, state, timeval_current_ofs(1, 0), report_rate, state);
for (i=0;i<OP_ENDOFLIST;i++) {
if (latencies[i] > worst_latencies[i]) {
worst_latencies[i] = latencies[i];
}
latencies[i] = 0;
}
/* send an echo on each interface to ensure it stays alive - this helps
with IP takeover */
for (i=0;i<numstates;i++) {
struct smb_echo p;
struct smbcli_request *req;
if (!state[i].tree) {
continue;
}
p.in.repeat_count = 1;
p.in.size = 0;
p.in.data = NULL;
req = smb_raw_echo_send(state[i].tree->session->transport, &p);
req->async.private_data = &state[i];
req->async.fn = echo_completion;
}
}
/*
called when the startup event script finishes
*/
static void ctdb_startup_callback(struct ctdb_context *ctdb, int status, void *p)
{
if (status != 0) {
DEBUG(DEBUG_ERR,("startup event failed\n"));
tevent_add_timer(ctdb->ev, ctdb->monitor->monitor_context,
timeval_current_ofs(5, 0),
ctdb_run_startup, ctdb);
return;
}
DEBUG(DEBUG_NOTICE,("startup event OK - enabling monitoring\n"));
ctdb_set_runstate(ctdb, CTDB_RUNSTATE_RUNNING);
ctdb->monitor->next_interval = 2;
ctdb_run_notification_script(ctdb, "startup");
ctdb->monitor->monitoring_mode = CTDB_MONITORING_ACTIVE;
tevent_add_timer(ctdb->ev, ctdb->monitor->monitor_context,
timeval_current_ofs(ctdb->monitor->next_interval, 0),
ctdb_check_health, ctdb);
}
static bool test_event_context_threaded(struct torture_context *test,
const void *test_data)
{
struct tevent_context *ev;
struct tevent_timer *te;
struct tevent_fd *fde;
pthread_t poll_thread;
int fds[2];
int ret;
char c = 0;
ev = tevent_context_init_byname(test, "poll_mt");
torture_assert(test, ev != NULL, "poll_mt not supported");
tevent_set_trace_callback(ev, test_event_threaded_trace, NULL);
te = tevent_add_timer(ev, ev, timeval_current_ofs(5, 0),
test_event_threaded_timer, NULL);
torture_assert(test, te != NULL, "Could not add timer");
ret = pthread_create(&poll_thread, NULL, test_event_poll_thread, ev);
torture_assert(test, ret == 0, "Could not create poll thread");
ret = pipe(fds);
torture_assert(test, ret == 0, "Could not create pipe");
poll(NULL, 0, 100);
test_event_threaded_lock();
fde = tevent_add_fd(ev, ev, fds[0], TEVENT_FD_READ,
test_event_threaded_read_handler, &fds[0]);
torture_assert(test, fde != NULL, "Could not add fd event");
test_event_threaded_unlock();
poll(NULL, 0, 100);
write(fds[1], &c, 1);
poll(NULL, 0, 100);
test_event_threaded_lock();
do_shutdown = true;
test_event_threaded_unlock();
write(fds[1], &c, 1);
ret = pthread_join(poll_thread, NULL);
torture_assert(test, ret == 0, "pthread_join failed");
return true;
}
static void ctdb_time_tick(struct event_context *ev, struct timed_event *te,
struct timeval t, void *private_data)
{
struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
if (getpid() != ctdb->ctdbd_pid) {
return;
}
event_add_timed(ctdb->ev, ctdb,
timeval_current_ofs(1, 0),
ctdb_time_tick, ctdb);
}
/*
see if the event scripts think we are healthy
*/
static void ctdb_check_health(struct tevent_context *ev,
struct tevent_timer *te,
struct timeval t, void *private_data)
{
struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
bool skip_monitoring = false;
int ret = 0;
if (ctdb->recovery_mode != CTDB_RECOVERY_NORMAL ||
ctdb->monitor->monitoring_mode == CTDB_MONITORING_DISABLED) {
skip_monitoring = true;
} else {
if (ctdb_db_all_frozen(ctdb)) {
DEBUG(DEBUG_ERR,
("Skip monitoring since databases are frozen\n"));
skip_monitoring = true;
}
}
if (skip_monitoring) {
tevent_add_timer(ctdb->ev, ctdb->monitor->monitor_context,
timeval_current_ofs(ctdb->monitor->next_interval, 0),
ctdb_check_health, ctdb);
return;
}
ret = ctdb_event_script_callback(ctdb,
ctdb->monitor->monitor_context,
ctdb_health_callback,
ctdb, CTDB_EVENT_MONITOR, "%s", "");
if (ret != 0) {
DEBUG(DEBUG_ERR,("Unable to launch monitor event script\n"));
ctdb->monitor->next_interval = 5;
tevent_add_timer(ctdb->ev, ctdb->monitor->monitor_context,
timeval_current_ofs(5, 0),
ctdb_check_health, ctdb);
}
}
/*
startup a client only ctdb context
*/
struct ctdb_context *ctdb_cmdline_client(struct event_context *ev)
{
struct ctdb_context *ctdb;
char *socket_name;
int ret;
/* initialise ctdb */
ctdb = ctdb_init(ev);
if (ctdb == NULL) {
fprintf(stderr, "Failed to init ctdb\n");
exit(1);
}
/* tell ctdb the socket address */
socket_name = getenv("CTDB_SOCKET");
if (socket_name != NULL) {
ret = ctdb_set_socketname(ctdb, socket_name);
if (ret == -1) {
printf("ctdb_set_socketname failed - %s\n",
ctdb_errstr(ctdb));
exit(1);
}
}
if (ctdb_cmdline.socketname != NULL) {
ret = ctdb_set_socketname(ctdb, ctdb_cmdline.socketname);
if (ret == -1) {
fprintf(stderr, "ctdb_set_socketname failed - %s\n",
ctdb_errstr(ctdb));
exit(1);
}
}
ret = ctdb_socket_connect(ctdb);
if (ret != 0) {
fprintf(stderr, __location__ " Failed to connect to daemon\n");
talloc_free(ctdb);
return NULL;
}
/* get our pnn */
ctdb->pnn = ctdb_ctrl_getpnn(ctdb, timeval_current_ofs(3, 0), CTDB_CURRENT_NODE);
if (ctdb->pnn == (uint32_t)-1) {
DEBUG(DEBUG_CRIT,(__location__ " Failed to get ctdb pnn\n"));
talloc_free(ctdb);
return NULL;
}
return ctdb;
}
/*
start watching for nodes that might be dead
*/
void ctdb_wait_for_first_recovery(struct ctdb_context *ctdb)
{
ctdb_set_runstate(ctdb, CTDB_RUNSTATE_FIRST_RECOVERY);
ctdb->monitor = talloc(ctdb, struct ctdb_monitor_state);
CTDB_NO_MEMORY_FATAL(ctdb, ctdb->monitor);
ctdb->monitor->monitor_context = talloc_new(ctdb->monitor);
CTDB_NO_MEMORY_FATAL(ctdb, ctdb->monitor->monitor_context);
tevent_add_timer(ctdb->ev, ctdb->monitor->monitor_context,
timeval_current_ofs(1, 0),
ctdb_wait_until_recovered, ctdb);
}
