/*
* Callback routine when the required locks are obtained.
* Called from parent context
*/
static void ctdb_lock_handler(struct tevent_context *ev,
struct tevent_fd *tfd,
uint16_t flags,
void *private_data)
{
struct lock_context *lock_ctx;
char c;
bool locked;
double t;
int id;
lock_ctx = talloc_get_type_abort(private_data, struct lock_context);
/* cancel the timeout event */
TALLOC_FREE(lock_ctx->ttimer);
t = timeval_elapsed(&lock_ctx->start_time);
id = lock_bucket_id(t);
/* Read the status from the child process */
if (sys_read(lock_ctx->fd[0], &c, 1) != 1) {
locked = false;
} else {
locked = (c == 0 ? true : false);
}
/* Update statistics */
CTDB_INCREMENT_STAT(lock_ctx->ctdb, locks.num_calls);
if (lock_ctx->ctdb_db) {
CTDB_INCREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_calls);
}
if (locked) {
if (lock_ctx->ctdb_db) {
CTDB_INCREMENT_STAT(lock_ctx->ctdb, locks.buckets[id]);
CTDB_UPDATE_LATENCY(lock_ctx->ctdb, lock_ctx->ctdb_db,
lock_type_str[lock_ctx->type], locks.latency,
lock_ctx->start_time);
CTDB_UPDATE_DB_LATENCY(lock_ctx->ctdb_db, lock_type_str[lock_ctx->type], locks.latency, t);
CTDB_INCREMENT_DB_STAT(lock_ctx->ctdb_db, locks.buckets[id]);
}
} else {
CTDB_INCREMENT_STAT(lock_ctx->ctdb, locks.num_failed);
if (lock_ctx->ctdb_db) {
CTDB_INCREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_failed);
}
}
process_callbacks(lock_ctx, locked);
}
/*
* Lock record / db depending on type
*/
static struct lock_request *ctdb_lock_internal(struct ctdb_context *ctdb,
struct ctdb_db_context *ctdb_db,
TDB_DATA key,
uint32_t priority,
void (*callback)(void *, bool),
void *private_data,
enum lock_type type,
bool auto_mark)
{
struct lock_context *lock_ctx = NULL;
struct lock_request *request;
if (callback == NULL) {
DEBUG(DEBUG_WARNING, ("No callback function specified, not locking\n"));
return NULL;
}
#if 0
/* Disable this optimization to ensure first-in-first-out fair
* scheduling of lock requests */
/* get a context for this key - search only the pending contexts,
* current contexts might in the middle of processing callbacks */
lock_ctx = find_lock_context(ctdb->lock_pending, ctdb_db, key, priority, type);
#endif
/* No existing context, create one */
if (lock_ctx == NULL) {
lock_ctx = talloc_zero(ctdb, struct lock_context);
if (lock_ctx == NULL) {
DEBUG(DEBUG_ERR, ("Failed to create a new lock context\n"));
return NULL;
}
lock_ctx->type = type;
lock_ctx->ctdb = ctdb;
lock_ctx->ctdb_db = ctdb_db;
lock_ctx->key.dsize = key.dsize;
if (key.dsize > 0) {
lock_ctx->key.dptr = talloc_memdup(lock_ctx, key.dptr, key.dsize);
} else {
lock_ctx->key.dptr = NULL;
}
lock_ctx->priority = priority;
lock_ctx->auto_mark = auto_mark;
lock_ctx->child = -1;
lock_ctx->block_child = -1;
DLIST_ADD_END(ctdb->lock_pending, lock_ctx, NULL);
ctdb->lock_num_pending++;
CTDB_INCREMENT_STAT(ctdb, locks.num_pending);
if (ctdb_db) {
CTDB_INCREMENT_DB_STAT(ctdb_db, locks.num_pending);
}
/* Start the timer when we activate the context */
lock_ctx->start_time = timeval_current();
}
/*
send a packet to a client
*/
static int daemon_queue_send(struct ctdb_client *client, struct ctdb_req_header *hdr)
{
CTDB_INCREMENT_STAT(client->ctdb, 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);
}
/*
handle a timeout of a control
*/
static void ctdb_control_timeout(struct event_context *ev, struct timed_event *te,
struct timeval t, void *private_data)
{
struct ctdb_control_state *state = talloc_get_type(private_data, struct ctdb_control_state);
TALLOC_CTX *tmp_ctx = talloc_new(ev);
CTDB_INCREMENT_STAT(state->ctdb, timeouts.control);
talloc_steal(tmp_ctx, state);
state->callback(state->ctdb, -1, tdb_null,
"ctdb_control timed out",
state->private_data);
talloc_free(tmp_ctx);
}
/*
called by the transport layer when a packet comes in
*/
static void ctdb_recv_pkt(struct ctdb_context *ctdb, uint8_t *data, uint32_t length)
{
struct ctdb_req_header *hdr = (struct ctdb_req_header *)data;
CTDB_INCREMENT_STAT(ctdb, node_packets_recv);
/* up the counter for this source node, so we know its alive */
if (ctdb_validate_pnn(ctdb, hdr->srcnode)) {
/* as a special case, redirected calls don't increment the rx_cnt */
if (hdr->operation != CTDB_REQ_CALL ||
((struct ctdb_req_call_old *)hdr)->hopcount == 0) {
ctdb->nodes[hdr->srcnode]->rx_cnt++;
}
}
ctdb_input_pkt(ctdb, hdr);
}
/*
called when we need to process a packet. This can be a requeued packet
after a lockwait, or a real packet from another node
*/
void ctdb_input_pkt(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
{
TALLOC_CTX *tmp_ctx;
/* place the packet as a child of the tmp_ctx. We then use
talloc_free() below to free it. If any of the calls want
to keep it, then they will steal it somewhere else, and the
talloc_free() will only free the tmp_ctx */
tmp_ctx = talloc_new(ctdb);
talloc_steal(tmp_ctx, hdr);
DEBUG(DEBUG_DEBUG,(__location__ " ctdb request %u of type %u length %u from "
"node %u to %u\n", hdr->reqid, hdr->operation, hdr->length,
hdr->srcnode, hdr->destnode));
switch (hdr->operation) {
case CTDB_REQ_CALL:
case CTDB_REPLY_CALL:
case CTDB_REQ_DMASTER:
case CTDB_REPLY_DMASTER:
/* we don't allow these calls when banned */
if (ctdb->nodes[ctdb->pnn]->flags & NODE_FLAGS_BANNED) {
DEBUG(DEBUG_DEBUG,(__location__ " ctdb operation %u"
" request %u"
" length %u from node %u to %u while node"
" is banned\n",
hdr->operation, hdr->reqid,
hdr->length,
hdr->srcnode, hdr->destnode));
goto done;
}
/* for ctdb_call inter-node operations verify that the
remote node that sent us the call is running in the
same generation instance as this node
*/
if (ctdb->vnn_map->generation != hdr->generation) {
DEBUG(DEBUG_DEBUG,(__location__ " ctdb operation %u"
" request %u"
" length %u from node %u to %u had an"
" invalid generation id:%u while our"
" generation id is:%u\n",
hdr->operation, hdr->reqid,
hdr->length,
hdr->srcnode, hdr->destnode,
hdr->generation, ctdb->vnn_map->generation));
goto done;
}
}
switch (hdr->operation) {
case CTDB_REQ_CALL:
CTDB_INCREMENT_STAT(ctdb, node.req_call);
ctdb_request_call(ctdb, hdr);
break;
case CTDB_REPLY_CALL:
CTDB_INCREMENT_STAT(ctdb, node.reply_call);
ctdb_reply_call(ctdb, hdr);
break;
case CTDB_REPLY_ERROR:
CTDB_INCREMENT_STAT(ctdb, node.reply_error);
ctdb_reply_error(ctdb, hdr);
break;
case CTDB_REQ_DMASTER:
CTDB_INCREMENT_STAT(ctdb, node.req_dmaster);
ctdb_request_dmaster(ctdb, hdr);
break;
case CTDB_REPLY_DMASTER:
CTDB_INCREMENT_STAT(ctdb, node.reply_dmaster);
ctdb_reply_dmaster(ctdb, hdr);
break;
case CTDB_REQ_MESSAGE:
CTDB_INCREMENT_STAT(ctdb, node.req_message);
ctdb_request_message(ctdb, hdr);
break;
case CTDB_REQ_CONTROL:
CTDB_INCREMENT_STAT(ctdb, node.req_control);
ctdb_request_control(ctdb, hdr);
break;
case CTDB_REPLY_CONTROL:
CTDB_INCREMENT_STAT(ctdb, node.reply_control);
ctdb_reply_control(ctdb, hdr);
break;
case CTDB_REQ_KEEPALIVE:
CTDB_INCREMENT_STAT(ctdb, keepalive_packets_recv);
break;
default:
DEBUG(DEBUG_CRIT,("%s: Packet with unknown operation %u\n",
__location__, hdr->operation));
break;
}
done:
talloc_free(tmp_ctx);
}
/*
* Schedule a new lock child process
* Set up callback handler and timeout handler
*/
static void ctdb_lock_schedule(struct ctdb_context *ctdb)
{
struct lock_context *lock_ctx, *next_ctx, *active_ctx;
int ret;
TALLOC_CTX *tmp_ctx;
const char *helper = BINDIR "/ctdb_lock_helper";
static const char *prog = NULL;
char **args;
if (prog == NULL) {
const char *t;
t = getenv("CTDB_LOCK_HELPER");
if (t != NULL) {
prog = talloc_strdup(ctdb, t);
} else {
prog = talloc_strdup(ctdb, helper);
}
CTDB_NO_MEMORY_VOID(ctdb, prog);
}
if (ctdb->lock_pending == NULL) {
return;
}
/* Find a lock context with requests */
lock_ctx = ctdb->lock_pending;
while (lock_ctx != NULL) {
next_ctx = lock_ctx->next;
if (! lock_ctx->req_queue) {
DEBUG(DEBUG_INFO, ("Removing lock context without lock requests\n"));
DLIST_REMOVE(ctdb->lock_pending, lock_ctx);
ctdb->lock_num_pending--;
CTDB_DECREMENT_STAT(ctdb, locks.num_pending);
if (lock_ctx->ctdb_db) {
CTDB_DECREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_pending);
}
talloc_free(lock_ctx);
} else {
active_ctx = find_lock_context(ctdb->lock_current, lock_ctx->ctdb_db,
lock_ctx->key, lock_ctx->priority,
lock_ctx->type);
if (active_ctx == NULL) {
if (lock_ctx->ctdb_db == NULL ||
lock_ctx->ctdb_db->lock_num_current < MAX_LOCK_PROCESSES_PER_DB) {
/* Found a lock context with lock requests */
break;
}
}
/* There is already a child waiting for the
* same key. So don't schedule another child
* just yet.
*/
}
lock_ctx = next_ctx;
}
if (lock_ctx == NULL) {
return;
}
lock_ctx->child = -1;
ret = pipe(lock_ctx->fd);
if (ret != 0) {
DEBUG(DEBUG_ERR, ("Failed to create pipe in ctdb_lock_schedule\n"));
return;
}
set_close_on_exec(lock_ctx->fd[0]);
/* Create data for child process */
tmp_ctx = talloc_new(lock_ctx);
if (tmp_ctx == NULL) {
DEBUG(DEBUG_ERR, ("Failed to allocate memory for helper args\n"));
close(lock_ctx->fd[0]);
close(lock_ctx->fd[1]);
return;
}
/* Create arguments for lock helper */
args = lock_helper_args(tmp_ctx, lock_ctx, lock_ctx->fd[1]);
if (args == NULL) {
DEBUG(DEBUG_ERR, ("Failed to create lock helper args\n"));
close(lock_ctx->fd[0]);
close(lock_ctx->fd[1]);
talloc_free(tmp_ctx);
return;
}
lock_ctx->child = vfork();
if (lock_ctx->child == (pid_t)-1) {
DEBUG(DEBUG_ERR, ("Failed to create a child in ctdb_lock_schedule\n"));
close(lock_ctx->fd[0]);
close(lock_ctx->fd[1]);
talloc_free(tmp_ctx);
return;
}
/* Child process */
if (lock_ctx->child == 0) {
ret = execv(prog, args);
if (ret < 0) {
DEBUG(DEBUG_ERR, ("Failed to execute helper %s (%d, %s)\n",
prog, errno, strerror(errno)));
}
_exit(1);
}
/* Parent process */
ctdb_track_child(ctdb, lock_ctx->child);
close(lock_ctx->fd[1]);
talloc_set_destructor(lock_ctx, ctdb_lock_context_destructor);
talloc_free(tmp_ctx);
/* Set up timeout handler */
lock_ctx->ttimer = tevent_add_timer(ctdb->ev,
lock_ctx,
timeval_current_ofs(10, 0),
ctdb_lock_timeout_handler,
(void *)lock_ctx);
if (lock_ctx->ttimer == NULL) {
ctdb_kill(ctdb, lock_ctx->child, SIGKILL);
lock_ctx->child = -1;
talloc_set_destructor(lock_ctx, NULL);
close(lock_ctx->fd[0]);
return;
}
/* Set up callback */
lock_ctx->tfd = tevent_add_fd(ctdb->ev,
lock_ctx,
lock_ctx->fd[0],
EVENT_FD_READ,
ctdb_lock_handler,
(void *)lock_ctx);
if (lock_ctx->tfd == NULL) {
TALLOC_FREE(lock_ctx->ttimer);
ctdb_kill(ctdb, lock_ctx->child, SIGKILL);
lock_ctx->child = -1;
talloc_set_destructor(lock_ctx, NULL);
close(lock_ctx->fd[0]);
return;
}
tevent_fd_set_auto_close(lock_ctx->tfd);
/* Move the context from pending to current */
DLIST_REMOVE(ctdb->lock_pending, lock_ctx);
ctdb->lock_num_pending--;
DLIST_ADD_END(ctdb->lock_current, lock_ctx, NULL);
if (lock_ctx->ctdb_db) {
lock_ctx->ctdb_db->lock_num_current++;
CTDB_INCREMENT_STAT(lock_ctx->ctdb, locks.num_current);
CTDB_INCREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_current);
}
}
