nodemcu-firmware/app/modules/mqtt.c

1583 lines
50 KiB
C
Raw Normal View History

// Module for mqtt
//
#include "module.h"
#include "lauxlib.h"
#include "platform.h"
#include <string.h>
#include <stddef.h>
#include <stdint.h>
#include "mem.h"
#include "lwip/ip_addr.h"
#include "espconn.h"
2019-07-23 17:47:18 +02:00
#include "mqtt/mqtt_msg.h"
#include "mqtt/msg_queue.h"
#include "user_interface.h"
#define MQTT_BUF_SIZE 1460
#define MQTT_DEFAULT_KEEPALIVE 60
#define MQTT_MAX_CLIENT_LEN 64
#define MQTT_MAX_USER_LEN 64
#define MQTT_MAX_PASS_LEN 64
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
#define MQTT_SEND_TIMEOUT 5 /* seconds */
typedef enum {
MQTT_INIT,
MQTT_CONNECT_SENDING,
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
MQTT_CONNECT_SENT,
MQTT_DATA
} tConnState;
typedef struct mqtt_event_data_t
{
uint8_t type;
const char* topic;
const char* data;
uint16_t topic_length;
uint16_t data_length;
uint16_t data_offset;
} mqtt_event_data_t;
typedef enum {
MQTT_RECV_NORMAL,
MQTT_RECV_BUFFERING_SHORT,
MQTT_RECV_BUFFERING,
MQTT_RECV_SKIPPING,
} tReceiveState;
typedef struct mqtt_state_t
{
2015-03-29 18:24:09 +02:00
msg_queue_t* pending_msg_q;
uint16_t next_message_id;
uint8_t * recv_buffer; // heap buffer for multi-packet rx
uint8_t * recv_buffer_wp; // write pointer in multi-packet rx
union {
uint16_t recv_buffer_size; // size of recv_buffer
uint32_t recv_buffer_skip; // number of bytes left to skip, in skipping state
};
tReceiveState recv_buffer_state;
} mqtt_state_t;
typedef struct lmqtt_userdata
{
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
struct espconn pesp_conn;
int self_ref;
int cb_connect_ref;
int cb_connect_fail_ref;
int cb_disconnect_ref;
int cb_message_ref;
int cb_overflow_ref;
int cb_suback_ref;
2016-03-20 00:24:18 +01:00
int cb_unsuback_ref;
int cb_puback_ref;
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
/* Configuration options */
struct {
int client_id_ref;
int username_ref;
int password_ref;
int will_topic_ref;
int will_message_ref;
int keepalive;
uint16_t max_message_length;
struct {
unsigned will_qos : 2;
bool will_retain : 1;
bool clean_session : 1;
bool secure : 1;
} flags;
} conf;
mqtt_state_t mqtt_state;
2015-03-29 18:24:09 +02:00
bool connected; // indicate socket connected, not mqtt prot connected.
bool keepalive_sent;
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
bool sending; // data sent to network stack, awaiting local acknowledge
ETSTimer mqttTimer;
tConnState connState;
}lmqtt_userdata;
// How large MQTT messages to accept by default
#define DEFAULT_MAX_MESSAGE_LENGTH 1024
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
static sint8 mqtt_socket_do_connect(struct lmqtt_userdata *);
static void mqtt_socket_reconnected(void *arg, sint8_t err);
static void mqtt_socket_connected(void *arg);
static void mqtt_connack_fail(lmqtt_userdata * mud, int reason_code);
static uint16_t mqtt_next_message_id(lmqtt_userdata * mud)
{
mud->mqtt_state.next_message_id++;
if (mud->mqtt_state.next_message_id == 0)
mud->mqtt_state.next_message_id++;
return mud->mqtt_state.next_message_id;
}
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
static void mqtt_socket_cb_lua_noarg(lua_State *L, lmqtt_userdata *mud, int cb)
{
if (cb == LUA_NOREF)
return;
lua_rawgeti(L, LUA_REGISTRYINDEX, cb);
lua_rawgeti(L, LUA_REGISTRYINDEX, mud->self_ref);
lua_call(L, 1, 0);
}
static void mqtt_socket_disconnected(void *arg) // tcp only
{
NODE_DBG("enter mqtt_socket_disconnected.\n");
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
lmqtt_userdata *mud = arg;
if(mud == NULL)
return;
2015-03-29 18:24:09 +02:00
os_timer_disarm(&mud->mqttTimer);
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
while (mud->mqtt_state.pending_msg_q) {
msg_destroy(msg_dequeue(&(mud->mqtt_state.pending_msg_q)));
}
2015-03-29 18:24:09 +02:00
if(mud->mqtt_state.recv_buffer) {
free(mud->mqtt_state.recv_buffer);
mud->mqtt_state.recv_buffer = NULL;
}
mud->mqtt_state.recv_buffer_size = 0;
mud->mqtt_state.recv_buffer_state = MQTT_RECV_NORMAL;
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
if(mud->pesp_conn.proto.tcp)
free(mud->pesp_conn.proto.tcp);
mud->pesp_conn.proto.tcp = NULL;
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
int selfref = mud->self_ref;
mud->self_ref = LUA_NOREF;
lua_State *L = lua_getstate();
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if(mud->connected){ // call back only called when socket is from connection to disconnection.
mud->connected = false;
if((mud->cb_disconnect_ref != LUA_NOREF) && (selfref != LUA_NOREF)) {
lua_rawgeti(L, LUA_REGISTRYINDEX, mud->cb_disconnect_ref);
lua_rawgeti(L, LUA_REGISTRYINDEX, selfref);
lua_call(L, 1, 0);
}
}
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
// unref this, and the mqtt.socket userdata will delete it self
luaL_unref(L, LUA_REGISTRYINDEX, selfref);
NODE_DBG("leave mqtt_socket_disconnected.\n");
}
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
static void mqtt_socket_do_disconnect(struct lmqtt_userdata *mud)
{
#ifdef CLIENT_SSL_ENABLE
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if (mud->conf.flags.secure) {
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
espconn_secure_disconnect(&mud->pesp_conn);
} else
#endif
{
espconn_disconnect(&mud->pesp_conn);
}
}
static void mqtt_socket_reconnected(void *arg, sint8_t err)
{
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
NODE_DBG("enter mqtt_socket_reconnected (err=%d)\n", err);
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
lmqtt_userdata *mud = arg;
if(mud == NULL)
return;
os_timer_disarm(&mud->mqttTimer);
mqtt_connack_fail(mud, MQTT_CONN_FAIL_SERVER_NOT_FOUND);
mqtt_socket_disconnected(arg);
NODE_DBG("leave mqtt_socket_reconnected.\n");
}
static void deliver_publish(lmqtt_userdata * mud, uint8_t* message, uint16_t length, uint8_t is_overflow)
{
NODE_DBG("enter deliver_publish (len=%d, overflow=%d).\n", length, is_overflow);
if(mud == NULL)
return;
mqtt_event_data_t event_data;
event_data.topic_length = length;
event_data.topic = mqtt_get_publish_topic(message, &event_data.topic_length);
event_data.data_length = length;
event_data.data = mqtt_get_publish_data(message, &event_data.data_length);
int cb_ref = !is_overflow ? mud->cb_message_ref : mud->cb_overflow_ref;
if(cb_ref == LUA_NOREF)
return;
if(mud->self_ref == LUA_NOREF)
return;
lua_State *L = lua_getstate();
if(event_data.topic && (event_data.topic_length > 0)){
lua_rawgeti(L, LUA_REGISTRYINDEX, cb_ref);
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
lua_rawgeti(L, LUA_REGISTRYINDEX, mud->self_ref);
lua_pushlstring(L, event_data.topic, event_data.topic_length);
} else {
NODE_DBG("get wrong packet.\n");
return;
}
if(event_data.data && (event_data.data_length > 0)){
lua_pushlstring(L, event_data.data, event_data.data_length);
lua_call(L, 3, 0);
} else {
lua_call(L, 2, 0);
}
NODE_DBG("leave deliver_publish.\n");
}
static void mqtt_connack_fail(lmqtt_userdata * mud, int reason_code)
{
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
NODE_DBG("enter mqtt_connack_fail\n");
if(mud->cb_connect_fail_ref == LUA_NOREF || mud->self_ref == LUA_NOREF)
{
return;
}
lua_State *L = lua_getstate();
lua_rawgeti(L, LUA_REGISTRYINDEX, mud->cb_connect_fail_ref);
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
lua_rawgeti(L, LUA_REGISTRYINDEX, mud->self_ref);
lua_pushinteger(L, reason_code);
lua_call(L, 2, 0);
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
NODE_DBG("leave mqtt_connack_fail\n");
}
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
static sint8 mqtt_send_if_possible(struct lmqtt_userdata *mud)
{
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
/* Waiting for the local network stack to get back to us? Can't send. */
if (mud->sending)
return ESPCONN_OK;
sint8 espconn_status = ESPCONN_OK;
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
msg_queue_t *pending_msg = msg_peek(&(mud->mqtt_state.pending_msg_q));
if (pending_msg && !pending_msg->sent) {
pending_msg->sent = 1;
NODE_DBG("Sent: %d\n", pending_msg->msg.length);
#ifdef CLIENT_SSL_ENABLE
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if( mud->conf.flags.secure )
{
espconn_status = espconn_secure_send(&mud->pesp_conn, pending_msg->msg.data, pending_msg->msg.length );
}
else
#endif
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
{
espconn_status = espconn_send(&mud->pesp_conn, pending_msg->msg.data, pending_msg->msg.length );
}
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mud->sending = true;
os_timer_disarm(&mud->mqttTimer);
os_timer_arm(&mud->mqttTimer, MQTT_SEND_TIMEOUT * 1000, 0);
}
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
NODE_DBG("send_if_poss, queue size: %d\n", msg_size(&(mud->mqtt_state.pending_msg_q)));
return espconn_status;
}
static void mqtt_socket_received(void *arg, char *pdata, unsigned short len)
{
NODE_DBG("enter mqtt_socket_received (rxlen=%u).\n", len);
uint8_t msg_type;
uint8_t msg_qos;
uint16_t msg_id;
uint8_t *in_buffer = (uint8_t *)pdata;
uint16_t in_buffer_length = len;
uint8_t *continuation_buffer = NULL;
uint8_t *temp_pdata = NULL;
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
lmqtt_userdata *mud = arg;
if(mud == NULL)
return;
switch(mud->mqtt_state.recv_buffer_state) {
case MQTT_RECV_NORMAL:
// No previous buffer.
break;
case MQTT_RECV_BUFFERING_SHORT:
// Last buffer had so few byte that we could not determine message length.
// Store in a local heap buffer and operate on this, as if was the regular pdata buffer.
// Avoids having to repeat message size/overflow logic.
temp_pdata = calloc(1,mud->mqtt_state.recv_buffer_size + len);
if(temp_pdata == NULL) {
NODE_DBG("MQTT[buffering-short]: Failed to allocate %u bytes, disconnecting...\n", mud->mqtt_state.recv_buffer_size + len);
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
mqtt_socket_do_disconnect(mud);
return;
}
NODE_DBG("MQTT[buffering-short]: Continuing with %u + %u bytes\n", mud->mqtt_state.recv_buffer_size, len);
memcpy(temp_pdata, mud->mqtt_state.recv_buffer, mud->mqtt_state.recv_buffer_size);
memcpy(temp_pdata + mud->mqtt_state.recv_buffer_size, pdata, len);
free(mud->mqtt_state.recv_buffer);
mud->mqtt_state.recv_buffer = NULL;
mud->mqtt_state.recv_buffer_state = MQTT_RECV_NORMAL;
in_buffer = temp_pdata;
in_buffer_length = mud->mqtt_state.recv_buffer_size + len;
break;
case MQTT_RECV_BUFFERING: {
// safe cast: we never allow longer buffer.
uint16_t current_length = (uint16_t) (mud->mqtt_state.recv_buffer_wp - mud->mqtt_state.recv_buffer);
NODE_DBG("MQTT[buffering]: appending %u bytes to previous recv buffer (%u out of wanted %u)\n",
in_buffer_length,
current_length,
mud->mqtt_state.recv_buffer_size);
// Copy from rx buffer to heap buffer. Smallest of [remainder of pending message] and [all of buffer]
uint16_t copy_length = LWIP_MIN(mud->mqtt_state.recv_buffer_size - current_length, in_buffer_length);
memcpy(mud->mqtt_state.recv_buffer_wp, pdata, copy_length);
mud->mqtt_state.recv_buffer_wp += copy_length;
in_buffer_length = (uint16_t) (mud->mqtt_state.recv_buffer_wp - mud->mqtt_state.recv_buffer);
if (in_buffer_length < mud->mqtt_state.recv_buffer_size) {
NODE_DBG("MQTT[buffering]: need %u more bytes, waiting for next rx.\n",
mud->mqtt_state.recv_buffer_size - in_buffer_length
);
goto RX_PACKET_FINISHED;
}
NODE_DBG("MQTT[buffering]: Full message received (%u). remainding bytes=%u\n",
mud->mqtt_state.recv_buffer_size,
len - copy_length);
// Point continuation_buffer to any additional data in pdata.
// May become 0 bytes, but used to trigger free!
continuation_buffer = pdata + copy_length;
len -= copy_length; // borrow len instead of having another variable..
in_buffer = mud->mqtt_state.recv_buffer;
// in_buffer_length was set above
mud->mqtt_state.recv_buffer_state = MQTT_RECV_NORMAL;
break;
}
case MQTT_RECV_SKIPPING:
// Last rx had a message which was too large to process, skip it.
if(mud->mqtt_state.recv_buffer_skip > in_buffer_length) {
NODE_DBG("MQTT[skipping]: skip=%u. Skipping full RX buffer (%u).\n",
mud->mqtt_state.recv_buffer_skip,
in_buffer_length
);
mud->mqtt_state.recv_buffer_skip -= in_buffer_length;
goto RX_PACKET_FINISHED;
}
NODE_DBG("MQTT[skipping]: skip=%u. Skipping partial RX buffer, continuing at %u\n",
mud->mqtt_state.recv_buffer_skip,
in_buffer_length
);
in_buffer += mud->mqtt_state.recv_buffer_skip;
in_buffer_length -= mud->mqtt_state.recv_buffer_skip;
mud->mqtt_state.recv_buffer_skip = 0;
mud->mqtt_state.recv_buffer_state = MQTT_RECV_NORMAL;
break;
}
READPACKET:
if(in_buffer_length <= 0)
goto RX_PACKET_FINISHED;
// MQTT publish message can in theory be 256Mb, while we do not support it we need to be
// able to do math on it.
int32_t message_length;
// temp buffer for control messages
uint8_t temp_buffer[MQTT_BUF_SIZE];
mqtt_message_buffer_t msgb;
mqtt_msg_init(&msgb, temp_buffer, MQTT_BUF_SIZE);
mqtt_message_t *temp_msg = NULL;
switch(mud->connState){
case MQTT_CONNECT_SENDING:
2015-03-29 18:24:09 +02:00
case MQTT_CONNECT_SENT:
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
os_timer_disarm(&mud->mqttTimer);
os_timer_arm(&mud->mqttTimer, mud->conf.keepalive * 1000, 0);
if(mqtt_get_type(in_buffer) != MQTT_MSG_TYPE_CONNACK){
NODE_DBG("MQTT: Invalid packet\r\n");
mud->connState = MQTT_INIT;
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
mqtt_socket_do_disconnect(mud);
mqtt_connack_fail(mud, MQTT_CONN_FAIL_NOT_A_CONNACK_MSG);
break;
} else if (mqtt_get_connect_ret_code(in_buffer) != MQTT_CONNACK_ACCEPTED) {
NODE_DBG("MQTT: CONNACK REFUSED (CODE: %d)\n", mqtt_get_connect_ret_code(in_buffer));
mud->connState = MQTT_INIT;
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
mqtt_socket_do_disconnect(mud);
mqtt_connack_fail(mud, mqtt_get_connect_ret_code(in_buffer));
break;
} else {
mud->connState = MQTT_DATA;
NODE_DBG("MQTT: Connected\r\n");
mud->keepalive_sent = 0;
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mqtt_socket_cb_lua_noarg(lua_getstate(), mud, mud->cb_connect_ref);
break;
}
break;
case MQTT_DATA:
message_length = mqtt_get_total_length(in_buffer, in_buffer_length);
msg_type = mqtt_get_type(in_buffer);
msg_qos = mqtt_get_qos(in_buffer);
msg_id = mqtt_get_id(in_buffer, in_buffer_length);
NODE_DBG("MQTT_DATA: msg length: %u, buffer length: %u\r\n",
message_length,
in_buffer_length);
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if (message_length > mud->conf.max_message_length) {
// The pending message length is larger than we was configured to allow
if(msg_qos > 0 && msg_id == 0) {
NODE_DBG("MQTT: msg too long, but not enough data to get msg_id: total=%u, deliver=%u\r\n", message_length, in_buffer_length);
// qos requested, but too short buffer to get a packet ID.
// Trigger the "short buffer" mode
message_length = -1;
// Drop through to partial message handling below.
} else {
NODE_DBG("MQTT: msg too long: total=%u, deliver=%u\r\n", message_length, in_buffer_length);
if (msg_type == MQTT_MSG_TYPE_PUBLISH) {
// In practice we should never get any other types..
deliver_publish(mud, in_buffer, in_buffer_length, 1);
// If qos specified, we should ACK it.
// In theory it might be wrong to ack it before we received all TCP packets, but this avoids
// buffering and special code to handle this corner-case. Server will most likely have
// written all to OS socket anyway, and not be aware that we "should" not have received it all yet.
if(msg_qos == 1){
temp_msg = mqtt_msg_puback(&msgb, msg_id);
msg_enqueue(&(mud->mqtt_state.pending_msg_q), temp_msg,
msg_id, MQTT_MSG_TYPE_PUBACK, (int)mqtt_get_qos(temp_msg->data) );
}
else if(msg_qos == 2){
temp_msg = mqtt_msg_pubrec(&msgb, msg_id);
msg_enqueue(&(mud->mqtt_state.pending_msg_q), temp_msg,
msg_id, MQTT_MSG_TYPE_PUBREC, (int)mqtt_get_qos(temp_msg->data) );
}
if(msg_qos == 1 || msg_qos == 2){
NODE_DBG("MQTT: Queue response QoS: %d\r\n", msg_qos);
}
}
if (message_length > in_buffer_length) {
// Ignore bytes in subsequent packet(s) too.
NODE_DBG("MQTT: skipping into next rx\n");
mud->mqtt_state.recv_buffer_state = MQTT_RECV_SKIPPING;
mud->mqtt_state.recv_buffer_skip = (uint32_t) message_length - in_buffer_length;
break;
} else {
NODE_DBG("MQTT: Skipping message\n");
mud->mqtt_state.recv_buffer_state = MQTT_RECV_NORMAL;
goto RX_MESSAGE_PROCESSED;
}
}
}
if (message_length == -1 || message_length > in_buffer_length) {
// Partial message in buffer, need to store on heap until next RX. Allocate size for full message directly,
// instead of potential reallocs, to avoid fragmentation.
// If message_length is indicated as -1, we do not have enough data to determine the length properly.
// Just put what we have on heap, and place in state BUFFERING_SHORT.
NODE_DBG("MQTT: Partial message received (%u of %d). Buffering\r\n",
in_buffer_length,
message_length);
// although message_length is 32bit, it should never go above 16bit since
// max_message_length is 16bit.
uint16_t alloc_size = message_length > 0 ? (uint16_t)message_length : in_buffer_length;
mud->mqtt_state.recv_buffer = calloc(1,alloc_size);
if (mud->mqtt_state.recv_buffer == NULL) {
NODE_DBG("MQTT: Failed to allocate %u bytes, disconnecting...\n", alloc_size);
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
mqtt_socket_do_disconnect(mud);
return;
}
memcpy(mud->mqtt_state.recv_buffer, in_buffer, in_buffer_length);
mud->mqtt_state.recv_buffer_wp = mud->mqtt_state.recv_buffer + in_buffer_length;
mud->mqtt_state.recv_buffer_state = message_length > 0 ? MQTT_RECV_BUFFERING : MQTT_RECV_BUFFERING_SHORT;
mud->mqtt_state.recv_buffer_size = alloc_size;
NODE_DBG("MQTT: Wait for next recv\n");
break;
}
2015-03-31 17:38:28 +02:00
msg_queue_t *pending_msg = msg_peek(&(mud->mqtt_state.pending_msg_q));
NODE_DBG("MQTT_DATA: type: %d, qos: %d, msg_id: %d, pending_id: %d, msg length: %u, buffer length: %u\r\n",
msg_type,
msg_qos,
msg_id,
(pending_msg)?pending_msg->msg_id:0,
message_length,
in_buffer_length);
2015-03-29 18:24:09 +02:00
switch(msg_type)
{
case MQTT_MSG_TYPE_SUBACK:
2015-03-29 18:24:09 +02:00
if(pending_msg && pending_msg->msg_type == MQTT_MSG_TYPE_SUBSCRIBE && pending_msg->msg_id == msg_id){
NODE_DBG("MQTT: Subscribe successful\r\n");
2015-03-29 18:24:09 +02:00
msg_destroy(msg_dequeue(&(mud->mqtt_state.pending_msg_q)));
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mqtt_socket_cb_lua_noarg(lua_getstate(), mud, mud->cb_suback_ref);
2015-03-29 18:24:09 +02:00
}
break;
case MQTT_MSG_TYPE_UNSUBACK:
2015-03-29 18:24:09 +02:00
if(pending_msg && pending_msg->msg_type == MQTT_MSG_TYPE_UNSUBSCRIBE && pending_msg->msg_id == msg_id){
NODE_DBG("MQTT: UnSubscribe successful\r\n");
2015-03-29 18:24:09 +02:00
msg_destroy(msg_dequeue(&(mud->mqtt_state.pending_msg_q)));
2016-03-20 00:24:18 +01:00
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mqtt_socket_cb_lua_noarg(lua_getstate(), mud, mud->cb_unsuback_ref);
2015-03-29 18:24:09 +02:00
}
break;
case MQTT_MSG_TYPE_PUBLISH:
2015-03-29 18:24:09 +02:00
if(msg_qos == 1){
temp_msg = mqtt_msg_puback(&msgb, msg_id);
msg_enqueue(&(mud->mqtt_state.pending_msg_q), temp_msg,
msg_id, MQTT_MSG_TYPE_PUBACK, (int)mqtt_get_qos(temp_msg->data) );
2015-03-29 18:24:09 +02:00
}
else if(msg_qos == 2){
temp_msg = mqtt_msg_pubrec(&msgb, msg_id);
msg_enqueue(&(mud->mqtt_state.pending_msg_q), temp_msg,
msg_id, MQTT_MSG_TYPE_PUBREC, (int)mqtt_get_qos(temp_msg->data) );
2015-03-29 18:24:09 +02:00
}
if(msg_qos == 1 || msg_qos == 2){
NODE_DBG("MQTT: Queue response QoS: %d\r\n", msg_qos);
}
deliver_publish(mud, in_buffer, (uint16_t)message_length, 0);
break;
case MQTT_MSG_TYPE_PUBACK:
2015-03-29 18:24:09 +02:00
if(pending_msg && pending_msg->msg_type == MQTT_MSG_TYPE_PUBLISH && pending_msg->msg_id == msg_id){
NODE_DBG("MQTT: Publish with QoS = 1 successful\r\n");
2015-03-29 18:24:09 +02:00
msg_destroy(msg_dequeue(&(mud->mqtt_state.pending_msg_q)));
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mqtt_socket_cb_lua_noarg(lua_getstate(), mud, mud->cb_puback_ref);
}
break;
case MQTT_MSG_TYPE_PUBREC:
2015-03-31 17:38:28 +02:00
if(pending_msg && pending_msg->msg_type == MQTT_MSG_TYPE_PUBLISH && pending_msg->msg_id == msg_id){
NODE_DBG("MQTT: Publish with QoS = 2 Received PUBREC\r\n");
// Note: actually, should not destroy the msg until PUBCOMP is received.
2015-03-31 17:38:28 +02:00
msg_destroy(msg_dequeue(&(mud->mqtt_state.pending_msg_q)));
temp_msg = mqtt_msg_pubrel(&msgb, msg_id);
msg_enqueue(&(mud->mqtt_state.pending_msg_q), temp_msg,
msg_id, MQTT_MSG_TYPE_PUBREL, (int)mqtt_get_qos(temp_msg->data) );
NODE_DBG("MQTT: Response PUBREL\r\n");
2015-03-31 17:38:28 +02:00
}
break;
case MQTT_MSG_TYPE_PUBREL:
2015-03-31 17:38:28 +02:00
if(pending_msg && pending_msg->msg_type == MQTT_MSG_TYPE_PUBREC && pending_msg->msg_id == msg_id){
msg_destroy(msg_dequeue(&(mud->mqtt_state.pending_msg_q)));
temp_msg = mqtt_msg_pubcomp(&msgb, msg_id);
msg_enqueue(&(mud->mqtt_state.pending_msg_q), temp_msg,
msg_id, MQTT_MSG_TYPE_PUBCOMP, (int)mqtt_get_qos(temp_msg->data) );
NODE_DBG("MQTT: Response PUBCOMP\r\n");
2015-03-31 17:38:28 +02:00
}
break;
case MQTT_MSG_TYPE_PUBCOMP:
2015-03-31 17:38:28 +02:00
if(pending_msg && pending_msg->msg_type == MQTT_MSG_TYPE_PUBREL && pending_msg->msg_id == msg_id){
NODE_DBG("MQTT: Publish with QoS = 2 successful\r\n");
2015-03-29 18:24:09 +02:00
msg_destroy(msg_dequeue(&(mud->mqtt_state.pending_msg_q)));
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mqtt_socket_cb_lua_noarg(lua_getstate(), mud, mud->cb_puback_ref);
}
break;
case MQTT_MSG_TYPE_PINGREQ:
temp_msg = mqtt_msg_pingresp(&msgb);
msg_enqueue(&(mud->mqtt_state.pending_msg_q), temp_msg,
msg_id, MQTT_MSG_TYPE_PINGRESP, (int)mqtt_get_qos(temp_msg->data) );
2015-03-29 18:24:09 +02:00
NODE_DBG("MQTT: Response PINGRESP\r\n");
break;
case MQTT_MSG_TYPE_PINGRESP:
// Ignore
mud->keepalive_sent = 0;
2015-03-29 18:24:09 +02:00
NODE_DBG("MQTT: PINGRESP received\r\n");
break;
}
RX_MESSAGE_PROCESSED:
if(continuation_buffer != NULL) {
NODE_DBG("MQTT[buffering]: buffered message finished. Continuing with rest of rx buffer (%u)\n",
len);
free(mud->mqtt_state.recv_buffer);
mud->mqtt_state.recv_buffer = NULL;
in_buffer = continuation_buffer;
in_buffer_length = len;
continuation_buffer = NULL;
}else{
// Message have been fully processed (or ignored). Move pointer ahead
// and continue with next message, if any.
in_buffer_length -= message_length;
in_buffer += message_length;
}
if(in_buffer_length > 0)
{
NODE_DBG("Get another published message\r\n");
goto READPACKET;
}
break;
}
RX_PACKET_FINISHED:
if(temp_pdata != NULL) {
free(temp_pdata);
}
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
mqtt_send_if_possible(mud);
NODE_DBG("leave mqtt_socket_received\n");
return;
}
static void mqtt_socket_sent(void *arg)
{
NODE_DBG("enter mqtt_socket_sent.\n");
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
lmqtt_userdata *mud = arg;
if(mud == NULL)
return;
if(!mud->connected)
2015-03-29 18:24:09 +02:00
return;
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mud->sending = false;
os_timer_disarm(&mud->mqttTimer);
2015-03-29 18:24:09 +02:00
if(mud->connState == MQTT_CONNECT_SENDING){
mud->connState = MQTT_CONNECT_SENT;
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
os_timer_arm(&mud->mqttTimer, MQTT_SEND_TIMEOUT * 1000, 0);
// MQTT_CONNECT not queued.
return;
2015-03-29 18:24:09 +02:00
}
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
/* Ready for timeout */
os_timer_arm(&mud->mqttTimer, mud->conf.keepalive * 1000, 0);
2015-03-31 17:38:28 +02:00
NODE_DBG("sent1, queue size: %d\n", msg_size(&(mud->mqtt_state.pending_msg_q)));
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
2015-03-31 17:38:28 +02:00
msg_queue_t *node = msg_peek(&(mud->mqtt_state.pending_msg_q));
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if (node) {
switch (node->msg_type) {
case MQTT_MSG_TYPE_PUBLISH:
// qos = 0, publish and forget. Run the callback now because we
// won't get a puback from the server and it's not clear when else
// we should tell the user the message drained from the egress queue
if (node->publish_qos == 0) {
msg_destroy(msg_dequeue(&(mud->mqtt_state.pending_msg_q)));
mqtt_socket_cb_lua_noarg(lua_getstate(), mud, mud->cb_puback_ref);
mqtt_send_if_possible(mud);
}
break;
case MQTT_MSG_TYPE_PUBACK:
/* FALLTHROUGH */
case MQTT_MSG_TYPE_PUBCOMP:
/* FALLTHROUGH */
case MQTT_MSG_TYPE_PINGREQ:
msg_destroy(msg_dequeue(&(mud->mqtt_state.pending_msg_q)));
mqtt_send_if_possible(mud);
break;
case MQTT_MSG_TYPE_DISCONNECT:
msg_destroy(msg_dequeue(&(mud->mqtt_state.pending_msg_q)));
mqtt_socket_do_disconnect(mud);
break;
2016-02-02 04:29:32 +01:00
}
}
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
2015-03-31 17:38:28 +02:00
NODE_DBG("sent2, queue size: %d\n", msg_size(&(mud->mqtt_state.pending_msg_q)));
NODE_DBG("leave mqtt_socket_sent.\n");
}
static void mqtt_socket_connected(void *arg)
{
NODE_DBG("enter mqtt_socket_connected.\n");
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
lmqtt_userdata *mud = arg;
if(mud == NULL)
return;
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
struct espconn *pesp_conn = &mud->pesp_conn;
mud->connected = true;
espconn_regist_recvcb(pesp_conn, mqtt_socket_received);
espconn_regist_sentcb(pesp_conn, mqtt_socket_sent);
espconn_regist_disconcb(pesp_conn, mqtt_socket_disconnected);
uint8_t temp_buffer[MQTT_BUF_SIZE];
mqtt_message_buffer_t msgb;
mqtt_msg_init(&msgb, temp_buffer, MQTT_BUF_SIZE);
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
/* Build the mqtt_connect_info on the stack and assemble the message */
lua_State *L = lua_getstate();
int ltop = lua_gettop(L);
struct mqtt_connect_info mci;
mci.clean_session = mud->conf.flags.clean_session;
mci.will_retain = mud->conf.flags.will_retain;
mci.will_qos = mud->conf.flags.will_qos;
mci.keepalive = mud->conf.keepalive;
lua_rawgeti(L, LUA_REGISTRYINDEX, mud->conf.client_id_ref);
mci.client_id = lua_tostring(L, -1);
lua_rawgeti(L, LUA_REGISTRYINDEX, mud->conf.username_ref);
mci.username = lua_tostring(L, -1);
lua_rawgeti(L, LUA_REGISTRYINDEX, mud->conf.password_ref);
mci.password = lua_tostring(L, -1);
lua_rawgeti(L, LUA_REGISTRYINDEX, mud->conf.will_topic_ref);
mci.will_topic = lua_tostring(L, -1);
lua_rawgeti(L, LUA_REGISTRYINDEX, mud->conf.will_message_ref);
mci.will_message = lua_tostring(L, -1);
mqtt_message_t* temp_msg = mqtt_msg_connect(&msgb, &mci);
NODE_DBG("Send MQTT connection infomation, data len: %d, d[0]=%d \r\n", temp_msg->length, temp_msg->data[0]);
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
lua_settop(L, ltop); /* Done with strings, restore the stack */
// not queue this message. should send right now. or should enqueue this before head.
#ifdef CLIENT_SSL_ENABLE
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if(mud->conf.flags.secure)
{
espconn_secure_send(pesp_conn, temp_msg->data, temp_msg->length);
}
else
#endif
{
espconn_send(pesp_conn, temp_msg->data, temp_msg->length);
}
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mud->sending = true;
os_timer_disarm(&mud->mqttTimer);
os_timer_arm(&mud->mqttTimer, MQTT_SEND_TIMEOUT * 1000, 0);
mud->connState = MQTT_CONNECT_SENDING;
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
NODE_DBG("leave mqtt_socket_connected, heap = %u.\n", system_get_free_heap_size());
return;
}
void mqtt_socket_timer(void *arg)
{
2015-03-31 17:38:28 +02:00
NODE_DBG("enter mqtt_socket_timer.\n");
lmqtt_userdata *mud = (lmqtt_userdata*) arg;
2015-03-29 18:24:09 +02:00
if(mud == NULL)
return;
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
if(mud->pesp_conn.proto.tcp == NULL){
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
NODE_DBG("MQTT not connected\n");
return;
}
2015-03-31 17:38:28 +02:00
NODE_DBG("timer, queue size: %d\n", msg_size(&(mud->mqtt_state.pending_msg_q)));
2015-03-29 18:24:09 +02:00
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if(mud->connState == MQTT_CONNECT_SENDING){ // MQTT_CONNECT send time out.
NODE_DBG("sSend MQTT_CONNECT failed.\n");
mud->connState = MQTT_INIT;
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
mqtt_socket_do_disconnect(mud);
mqtt_connack_fail(mud, MQTT_CONN_FAIL_TIMEOUT_SENDING);
} else if(mud->connState == MQTT_CONNECT_SENT) { // wait for CONACK time out.
NODE_DBG("MQTT_CONNECT timeout.\n");
mud->connState = MQTT_INIT;
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
mqtt_socket_do_disconnect(mud);
mqtt_connack_fail(mud, MQTT_CONN_FAIL_TIMEOUT_RECEIVING);
} else if(mud->connState == MQTT_DATA){
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if(!msg_peek(&(mud->mqtt_state.pending_msg_q))) {
// no queued event.
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if (mud->keepalive_sent) {
// Oh dear -- keepalive timer expired and still no ack of previous message
mqtt_socket_do_disconnect(mud);
} else {
uint8_t temp_buffer[MQTT_BUF_SIZE];
mqtt_message_buffer_t msgb;
mqtt_msg_init(&msgb, temp_buffer, MQTT_BUF_SIZE);
NODE_DBG("\r\nMQTT: Send keepalive packet\r\n");
mqtt_message_t* temp_msg = mqtt_msg_pingreq(&msgb);
msg_queue_t *node = msg_enqueue( &(mud->mqtt_state.pending_msg_q), temp_msg,
0, MQTT_MSG_TYPE_PINGREQ, (int)mqtt_get_qos(temp_msg->data) );
mud->keepalive_sent = 1;
mqtt_send_if_possible(mud);
2015-03-29 18:24:09 +02:00
}
}
}
2015-03-31 17:38:28 +02:00
NODE_DBG("leave mqtt_socket_timer.\n");
}
// Lua: mqtt.Client(clientid, keepalive, user, pass, clean_session, max_message_length)
static int mqtt_socket_client( lua_State* L )
{
NODE_DBG("enter mqtt_socket_client.\n");
lmqtt_userdata *mud;
int stack = 1;
int top = lua_gettop(L);
// create a object
mud = (lmqtt_userdata *)lua_newuserdata(L, sizeof(lmqtt_userdata));
memset(mud, 0, sizeof(*mud));
// pre-initialize it, in case of errors
mud->self_ref = LUA_NOREF;
mud->cb_connect_ref = LUA_NOREF;
mud->cb_connect_fail_ref = LUA_NOREF;
mud->cb_disconnect_ref = LUA_NOREF;
mud->cb_message_ref = LUA_NOREF;
mud->cb_overflow_ref = LUA_NOREF;
mud->cb_suback_ref = LUA_NOREF;
mud->cb_unsuback_ref = LUA_NOREF;
mud->cb_puback_ref = LUA_NOREF;
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mud->conf.client_id_ref = LUA_NOREF;
mud->conf.username_ref = LUA_NOREF;
mud->conf.password_ref = LUA_NOREF;
mud->conf.will_topic_ref = LUA_NOREF;
mud->conf.will_message_ref = LUA_NOREF;
mud->connState = MQTT_INIT;
// set its metatable
luaL_getmetatable(L, "mqtt.socket");
lua_setmetatable(L, -2);
if( lua_isstring(L,stack) ) // deal with the clientid string
{
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
lua_pushvalue(L, stack);
stack++;
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
} else {
char tempid[20] = {0};
sprintf(tempid, "%s%x", "NodeMCU_", system_get_chip_id() );
lua_pushstring(L, tempid);
}
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mud->conf.client_id_ref = luaL_ref(L, LUA_REGISTRYINDEX);
if(lua_isnumber( L, stack ))
{
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mud->conf.keepalive = luaL_checkinteger( L, stack);
stack++;
}
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if(mud->conf.keepalive == 0) {
mud->conf.keepalive = MQTT_DEFAULT_KEEPALIVE;
}
if(lua_isstring( L, stack )){
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
lua_pushvalue(L, stack);
mud->conf.username_ref = luaL_ref(L, LUA_REGISTRYINDEX);
stack++;
}
if(lua_isstring( L, stack )){
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
lua_pushvalue(L, stack);
mud->conf.password_ref = luaL_ref(L, LUA_REGISTRYINDEX);
stack++;
}
2015-09-05 20:54:13 +02:00
if(lua_isnumber( L, stack ))
{
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mud->conf.flags.clean_session = !!luaL_checkinteger(L, stack);
2015-09-05 20:54:13 +02:00
stack++;
}
if(lua_isnumber( L, stack ))
{
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mud->conf.max_message_length = luaL_checkinteger( L, stack);
stack++;
}
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if(mud->conf.max_message_length == 0) {
mud->conf.max_message_length = DEFAULT_MAX_MESSAGE_LENGTH;
}
2015-03-29 18:24:09 +02:00
mud->mqtt_state.pending_msg_q = NULL;
mud->mqtt_state.recv_buffer = NULL;
mud->mqtt_state.recv_buffer_size = 0;
mud->mqtt_state.recv_buffer_state = MQTT_RECV_NORMAL;
2015-03-29 18:24:09 +02:00
NODE_DBG("leave mqtt_socket_client.\n");
return 1;
}
// Lua: mqtt.delete( socket )
// call close() first
// socket: unref everything
static int mqtt_delete( lua_State* L )
{
NODE_DBG("enter mqtt_delete.\n");
lmqtt_userdata *mud = (lmqtt_userdata *)luaL_checkudata(L, 1, "mqtt.socket");
os_timer_disarm(&mud->mqttTimer);
2015-03-29 18:24:09 +02:00
mud->connected = false;
// ---- alloc-ed in mqtt_socket_connect()
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
if(mud->pesp_conn.proto.tcp)
free(mud->pesp_conn.proto.tcp);
mud->pesp_conn.proto.tcp = NULL;
while(mud->mqtt_state.pending_msg_q) {
msg_destroy(msg_dequeue(&(mud->mqtt_state.pending_msg_q)));
}
//--------- alloc-ed in mqtt_socket_received()
if(mud->mqtt_state.recv_buffer) {
free(mud->mqtt_state.recv_buffer);
mud->mqtt_state.recv_buffer = NULL;
}
// ----
// free (unref) callback ref
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_connect_ref);
mud->cb_connect_ref = LUA_NOREF;
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_connect_fail_ref);
mud->cb_connect_fail_ref = LUA_NOREF;
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_disconnect_ref);
mud->cb_disconnect_ref = LUA_NOREF;
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_message_ref);
mud->cb_message_ref = LUA_NOREF;
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_overflow_ref);
mud->cb_overflow_ref = LUA_NOREF;
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_suback_ref);
mud->cb_suback_ref = LUA_NOREF;
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_unsuback_ref);
mud->cb_unsuback_ref = LUA_NOREF;
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_puback_ref);
mud->cb_puback_ref = LUA_NOREF;
2020-04-05 02:25:57 +02:00
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
luaL_unref(L, LUA_REGISTRYINDEX, mud->conf.client_id_ref);
mud->conf.client_id_ref = LUA_NOREF;
luaL_unref(L, LUA_REGISTRYINDEX, mud->conf.username_ref);
mud->conf.username_ref = LUA_NOREF;
luaL_unref(L, LUA_REGISTRYINDEX, mud->conf.password_ref);
mud->conf.password_ref = LUA_NOREF;
luaL_unref(L, LUA_REGISTRYINDEX, mud->conf.will_topic_ref);
mud->conf.will_topic_ref = LUA_NOREF;
luaL_unref(L, LUA_REGISTRYINDEX, mud->conf.will_message_ref);
mud->conf.will_message_ref = LUA_NOREF;
2020-04-05 02:25:57 +02:00
int selfref = mud->self_ref;
mud->self_ref = LUA_NOREF;
2020-04-05 02:25:57 +02:00
luaL_unref(L, LUA_REGISTRYINDEX, mud->self_ref);
NODE_DBG("leave mqtt_delete.\n");
return 0;
}
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
static sint8 mqtt_socket_do_connect(struct lmqtt_userdata *mud)
{
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
NODE_DBG("enter mqtt_socket_do_connect.\n");
sint8 espconn_status;
mud->connState = MQTT_INIT;
#ifdef CLIENT_SSL_ENABLE
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if(mud->conf.flags.secure)
{
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
NODE_DBG("mqtt_socket_do_connect using espconn_secure\n");
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
espconn_status = espconn_secure_connect(&mud->pesp_conn);
}
else
#endif
{
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
espconn_status = espconn_connect(&mud->pesp_conn);
}
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
NODE_DBG("leave mqtt_socket_do_connect espconn_status=%d\n", espconn_status);
return espconn_status;
}
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
static void socket_dns_found(const char *name, ip_addr_t *ipaddr, void *arg)
{
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
lmqtt_userdata *mud = arg;
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if((ipaddr == NULL) || (ipaddr->addr == 0))
{
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
NODE_DBG("socket_dns_found failure\n");
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
mqtt_connack_fail(mud, MQTT_CONN_FAIL_DNS);
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
// although not connected, but fire disconnect callback to release every thing.
mqtt_socket_disconnected(arg);
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
return;
}
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
memcpy(&mud->pesp_conn.proto.tcp->remote_ip, &(ipaddr->addr), 4);
NODE_DBG("socket_dns_found success: ");
NODE_DBG(IPSTR, IP2STR(&(ipaddr->addr)));
NODE_DBG("\n");
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mqtt_socket_do_connect(mud);
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
}
#include "pm/swtimer.h"
// Lua: mqtt:connect( host, port, secure, function(client), function(client, connect_return_code) )
static int mqtt_socket_connect( lua_State* L )
{
NODE_DBG("enter mqtt_socket_connect.\n");
lmqtt_userdata *mud = NULL;
unsigned port = 1883;
size_t il;
ip_addr_t ipaddr;
const char *domain;
int stack = 1;
int top = lua_gettop(L);
mud = (lmqtt_userdata *)luaL_checkudata(L, stack, "mqtt.socket");
stack++;
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
struct espconn *pesp_conn = &mud->pesp_conn;
2015-02-07 16:45:01 +01:00
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if(mud->connected){
return luaL_error(L, "already connected");
}
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
2015-03-29 18:24:09 +02:00
mud->connected = false;
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mud->sending = false;
if( (stack<=top) && lua_isstring(L,stack) ) // deal with the domain string
{
domain = luaL_checklstring( L, stack, &il );
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if (!domain)
return luaL_error(L, "invalid domain");
stack++;
}
if ( (stack<=top) && lua_isnumber(L, stack) )
{
port = lua_tointeger(L, stack);
stack++;
}
if ( (stack<=top) && (lua_isnumber(L, stack) || lua_isboolean(L, stack)) )
{
if (lua_isnumber(L, stack)) {
platform_print_deprecation_note("mqtt.connect secure parameter as integer","in the future");
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mud->conf.flags.secure = !!lua_tointeger(L, stack);
} else {
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mud->conf.flags.secure = lua_toboolean(L, stack);
}
stack++;
}
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
#ifndef CLIENT_SSL_ENABLE
if ( mud->conf.flags.secure )
{
return luaL_error(L, "ssl not available");
}
#endif
// call back function when a connection is obtained, tcp only
if ((stack<=top) && (lua_isfunction(L, stack))){
lua_pushvalue(L, stack); // copy argument (func) to the top of stack
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_connect_ref);
mud->cb_connect_ref = luaL_ref(L, LUA_REGISTRYINDEX);
}
stack++;
// call back function when a connection fails
if ((stack<=top) && (lua_isfunction(L, stack))){
lua_pushvalue(L, stack); // copy argument (func) to the top of stack
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_connect_fail_ref);
mud->cb_connect_fail_ref = luaL_ref(L, LUA_REGISTRYINDEX);
}
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if (!pesp_conn->proto.tcp)
pesp_conn->proto.tcp = (esp_tcp *)calloc(1,sizeof(esp_tcp));
if(!pesp_conn->proto.tcp) {
return luaL_error(L, "not enough memory");
}
luaL_unref(L, LUA_REGISTRYINDEX, mud->self_ref);
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
lua_pushvalue(L, 1); // copy userdata and persist it in the registry
mud->self_ref = luaL_ref(L, LUA_REGISTRYINDEX);
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
pesp_conn->type = ESPCONN_TCP;
pesp_conn->state = ESPCONN_NONE;
pesp_conn->proto.tcp->remote_port = port;
if (pesp_conn->proto.tcp->local_port == 0)
pesp_conn->proto.tcp->local_port = espconn_port();
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
espconn_regist_connectcb(pesp_conn, mqtt_socket_connected);
espconn_regist_reconcb(pesp_conn, mqtt_socket_reconnected);
os_timer_disarm(&mud->mqttTimer);
os_timer_setfn(&mud->mqttTimer, (os_timer_func_t *)mqtt_socket_timer, mud);
SWTIMER_REG_CB(mqtt_socket_timer, SWTIMER_RESUME);
//I assume that mqtt_socket_timer connects to the mqtt server, but I'm not really sure what impact light_sleep will have on it.
//My guess: If in doubt, resume the timer
// timer started in socket_connect()
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
ip_addr_t host_ip;
switch (dns_gethostbyname(domain, &host_ip, socket_dns_found, mud))
{
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
case ERR_OK:
socket_dns_found(domain, &host_ip, mud); // ip is returned in host_ip.
break;
case ERR_INPROGRESS:
break;
default:
// Something has gone wrong; bail out?
mqtt_connack_fail(mud, MQTT_CONN_FAIL_DNS);
}
NODE_DBG("leave mqtt_socket_connect.\n");
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
return 0;
}
// Lua: mqtt:close()
// client disconnect and unref itself
static int mqtt_socket_close( lua_State* L )
{
NODE_DBG("enter mqtt_socket_close.\n");
lmqtt_userdata *mud = NULL;
mud = (lmqtt_userdata *)luaL_checkudata(L, 1, "mqtt.socket");
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
sint8 espconn_status = 0;
if (mud->connected) {
uint8_t temp_buffer[MQTT_BUF_SIZE];
mqtt_message_buffer_t msgb;
mqtt_msg_init(&msgb, temp_buffer, MQTT_BUF_SIZE);
// Send disconnect message
mqtt_message_t* temp_msg = mqtt_msg_disconnect(&msgb);
NODE_DBG("Send MQTT disconnect infomation, data len: %d, d[0]=%d \r\n", temp_msg->length, temp_msg->data[0]);
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
if (!msg_enqueue(&(mud->mqtt_state.pending_msg_q), temp_msg, 0,
MQTT_MSG_TYPE_DISCONNECT, 0))
goto err;
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mqtt_send_if_possible(mud);
}
NODE_DBG("leave mqtt_socket_close.\n");
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
return 0;
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
err:
/* OOM while trying to build the disconnect message. Fail the connection */
mqtt_socket_do_disconnect(mud);
NODE_DBG("leave mqtt_socket_close on error path\n");
return 0;
}
// Lua: mqtt:on( "method", function() )
static int mqtt_socket_on( lua_State* L )
{
NODE_DBG("enter mqtt_socket_on.\n");
lmqtt_userdata *mud;
size_t sl;
mud = (lmqtt_userdata *)luaL_checkudata(L, 1, "mqtt.socket");
luaL_checktype(L, 3, LUA_TFUNCTION);
lua_pushvalue(L, 3); // copy argument (func) to the top of stack
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
static const char * const cbnames[] = {
"connect", "connfail", "offline",
"message", "overflow",
"puback", "suback", "unsuback",
NULL
};
switch (luaL_checkoption(L, 2, NULL, cbnames)) {
case 0:
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_connect_ref);
mud->cb_connect_ref = luaL_ref(L, LUA_REGISTRYINDEX);
break;
case 1:
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_connect_fail_ref);
mud->cb_connect_fail_ref = luaL_ref(L, LUA_REGISTRYINDEX);
break;
case 2:
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_disconnect_ref);
mud->cb_disconnect_ref = luaL_ref(L, LUA_REGISTRYINDEX);
break;
case 3:
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_message_ref);
mud->cb_message_ref = luaL_ref(L, LUA_REGISTRYINDEX);
break;
case 4:
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_overflow_ref);
mud->cb_overflow_ref = luaL_ref(L, LUA_REGISTRYINDEX);
break;
case 5:
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_puback_ref);
mud->cb_puback_ref = luaL_ref(L, LUA_REGISTRYINDEX);
break;
case 6:
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_suback_ref);
mud->cb_suback_ref = luaL_ref(L, LUA_REGISTRYINDEX);
break;
case 7:
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_unsuback_ref);
mud->cb_unsuback_ref = luaL_ref(L, LUA_REGISTRYINDEX);
break;
}
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
NODE_DBG("leave mqtt_socket_on.\n");
return 0;
}
2016-03-20 00:24:18 +01:00
// Lua: bool = mqtt:unsubscribe(topic, function())
static int mqtt_socket_unsubscribe( lua_State* L ) {
NODE_DBG("enter mqtt_socket_unsubscribe.\n");
uint8_t stack = 1;
uint16_t msg_id;
2016-03-20 00:24:18 +01:00
const char *topic;
size_t il;
lmqtt_userdata *mud;
mud = (lmqtt_userdata *) luaL_checkudata( L, stack, "mqtt.socket" );
stack++;
if(!mud->connected){
luaL_error( L, "not connected" );
lua_pushboolean(L, 0);
return 1;
}
uint8_t temp_buffer[MQTT_BUF_SIZE];
mqtt_message_buffer_t msgb;
mqtt_msg_init(&msgb, temp_buffer, MQTT_BUF_SIZE);
2016-03-20 00:24:18 +01:00
mqtt_message_t *temp_msg = NULL;
if( lua_istable( L, stack ) ) {
NODE_DBG("unsubscribe table\n");
lua_pushnil( L ); /* first key */
int topic_count = 0;
uint8_t overflow = 0;
while( lua_next( L, stack ) != 0 ) {
topic = luaL_checkstring( L, -2 );
if (topic_count == 0) {
msg_id = mqtt_next_message_id(mud);
temp_msg = mqtt_msg_unsubscribe_init( &msgb, msg_id );
2016-03-20 00:24:18 +01:00
}
temp_msg = mqtt_msg_unsubscribe_topic( &msgb, topic );
2016-03-20 00:24:18 +01:00
topic_count++;
NODE_DBG("topic: %s - length: %d\n", topic, temp_msg->length);
if (temp_msg->length == 0) {
lua_pop(L, 1);
overflow = 1;
break; // too long message for the outbuffer.
}
lua_pop( L, 1 );
}
if (topic_count == 0){
return luaL_error( L, "no topics found" );
}
if (overflow != 0){
return luaL_error( L, "buffer overflow, can't enqueue all unsubscriptions" );
}
temp_msg = mqtt_msg_unsubscribe_fini( &msgb );
2016-03-20 00:24:18 +01:00
if (temp_msg->length == 0) {
return luaL_error( L, "buffer overflow, can't enqueue all unsubscriptions" );
}
stack++;
} else {
NODE_DBG("unsubscribe string\n");
topic = luaL_checklstring( L, stack, &il );
stack++;
if( topic == NULL ){
return luaL_error( L, "need topic name" );
}
msg_id = mqtt_next_message_id(mud);
temp_msg = mqtt_msg_unsubscribe( &msgb, topic, msg_id );
2016-03-20 00:24:18 +01:00
}
if (lua_isfunction(L, stack)) { // TODO: this will overwrite the previous one.
lua_pushvalue( L, stack ); // copy argument (func) to the top of stack
luaL_unref( L, LUA_REGISTRYINDEX, mud->cb_unsuback_ref );
2016-03-20 00:24:18 +01:00
mud->cb_unsuback_ref = luaL_ref( L, LUA_REGISTRYINDEX );
}
msg_queue_t *node = msg_enqueue( &(mud->mqtt_state.pending_msg_q), temp_msg,
msg_id, MQTT_MSG_TYPE_UNSUBSCRIBE, (int)mqtt_get_qos(temp_msg->data) );
NODE_DBG("topic: %s - id: %d - qos: %d, length: %d\n", topic, node->msg_id, node->publish_qos, node->msg.length);
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
NODE_DBG("msg_size: %d\n", msg_size(&(mud->mqtt_state.pending_msg_q)));
2016-03-20 00:24:18 +01:00
sint8 espconn_status = ESPCONN_IF;
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
espconn_status = mqtt_send_if_possible(mud);
2016-03-20 00:24:18 +01:00
if(!node || espconn_status != ESPCONN_OK){
lua_pushboolean(L, 0);
} else {
lua_pushboolean(L, 1); // enqueued succeed.
}
NODE_DBG("unsubscribe, queue size: %d\n", msg_size(&(mud->mqtt_state.pending_msg_q)));
NODE_DBG("leave mqtt_socket_unsubscribe.\n");
return 1;
}
2015-03-29 18:24:09 +02:00
// Lua: bool = mqtt:subscribe(topic, qos, function())
static int mqtt_socket_subscribe( lua_State* L ) {
NODE_DBG("enter mqtt_socket_subscribe.\n");
uint8_t stack = 1, qos = 0;
uint16_t msg_id;
const char *topic;
size_t il;
lmqtt_userdata *mud;
mud = (lmqtt_userdata *) luaL_checkudata( L, stack, "mqtt.socket" );
stack++;
if(!mud->connected){
luaL_error( L, "not connected" );
2015-03-29 18:24:09 +02:00
lua_pushboolean(L, 0);
return 1;
}
uint8_t temp_buffer[MQTT_BUF_SIZE];
mqtt_message_buffer_t msgb;
mqtt_msg_init(&msgb, temp_buffer, MQTT_BUF_SIZE);
mqtt_message_t *temp_msg = NULL;
if( lua_istable( L, stack ) ) {
NODE_DBG("subscribe table\n");
lua_pushnil( L ); /* first key */
int topic_count = 0;
uint8_t overflow = 0;
2015-03-29 18:24:09 +02:00
while( lua_next( L, stack ) != 0 ) {
topic = luaL_checkstring( L, -2 );
qos = luaL_checkinteger( L, -1 );
if (topic_count == 0) {
msg_id = mqtt_next_message_id(mud);
temp_msg = mqtt_msg_subscribe_init( &msgb, msg_id );
}
temp_msg = mqtt_msg_subscribe_topic( &msgb, topic, qos );
topic_count++;
NODE_DBG("topic: %s - qos: %d, length: %d\n", topic, qos, temp_msg->length);
2015-03-29 18:24:09 +02:00
if (temp_msg->length == 0) {
2015-03-29 18:24:09 +02:00
lua_pop(L, 1);
overflow = 1;
2015-03-29 18:24:09 +02:00
break; // too long message for the outbuffer.
}
lua_pop( L, 1 );
}
if (topic_count == 0){
return luaL_error( L, "no topics found" );
2015-03-29 18:24:09 +02:00
}
if (overflow != 0){
return luaL_error( L, "buffer overflow, can't enqueue all subscriptions" );
}
temp_msg = mqtt_msg_subscribe_fini( &msgb );
if (temp_msg->length == 0) {
return luaL_error( L, "buffer overflow, can't enqueue all subscriptions" );
}
2015-03-29 18:24:09 +02:00
stack++;
} else {
NODE_DBG("subscribe string\n");
topic = luaL_checklstring( L, stack, &il );
stack++;
if( topic == NULL ){
return luaL_error( L, "need topic name" );
2015-03-29 18:24:09 +02:00
}
qos = luaL_checkinteger( L, stack );
msg_id = mqtt_next_message_id(mud);
temp_msg = mqtt_msg_subscribe( &msgb, topic, qos, msg_id );
stack++;
}
if (lua_isfunction(L, stack)) { // TODO: this will overwrite the previous one.
2015-03-29 18:24:09 +02:00
lua_pushvalue( L, stack ); // copy argument (func) to the top of stack
luaL_unref( L, LUA_REGISTRYINDEX, mud->cb_suback_ref );
2015-03-29 18:24:09 +02:00
mud->cb_suback_ref = luaL_ref( L, LUA_REGISTRYINDEX );
}
msg_queue_t *node = msg_enqueue( &(mud->mqtt_state.pending_msg_q), temp_msg,
msg_id, MQTT_MSG_TYPE_SUBSCRIBE, (int)mqtt_get_qos(temp_msg->data) );
2015-03-29 18:24:09 +02:00
NODE_DBG("topic: %s - id: %d - qos: %d, length: %d\n", topic, node->msg_id, node->publish_qos, node->msg.length);
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
NODE_DBG("msg_size: %d\n", msg_size(&(mud->mqtt_state.pending_msg_q)));
sint8 espconn_status = ESPCONN_IF;
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
espconn_status = mqtt_send_if_possible(mud);
2015-03-29 18:24:09 +02:00
if(!node || espconn_status != ESPCONN_OK){
2015-03-29 18:24:09 +02:00
lua_pushboolean(L, 0);
} else {
lua_pushboolean(L, 1); // enqueued succeed.
}
2015-03-31 17:38:28 +02:00
NODE_DBG("subscribe, queue size: %d\n", msg_size(&(mud->mqtt_state.pending_msg_q)));
NODE_DBG("leave mqtt_socket_subscribe.\n");
return 1;
}
2015-03-29 18:24:09 +02:00
// Lua: bool = mqtt:publish( topic, payload, qos, retain, function() )
static int mqtt_socket_publish( lua_State* L )
{
NODE_DBG("enter mqtt_socket_publish.\n");
lmqtt_userdata *mud;
size_t l;
uint8_t stack = 1;
2015-03-29 18:24:09 +02:00
uint16_t msg_id = 0;
mud = (lmqtt_userdata *)luaL_checkudata(L, stack, "mqtt.socket");
stack++;
if(!mud->connected){
return luaL_error( L, "not connected" );
}
const char *topic = luaL_checklstring( L, stack, &l );
stack ++;
2015-03-29 18:24:09 +02:00
if (topic == NULL){
return luaL_error( L, "need topic" );
2015-03-29 18:24:09 +02:00
}
const char *payload = luaL_checklstring( L, stack, &l );
stack ++;
uint8_t qos = luaL_checkinteger( L, stack);
stack ++;
uint8_t retain = luaL_checkinteger( L, stack);
stack ++;
if (qos != 0) {
msg_id = mqtt_next_message_id(mud);
}
uint8_t temp_buffer[MQTT_BUF_SIZE];
mqtt_message_buffer_t msgb;
mqtt_msg_init(&msgb, temp_buffer, MQTT_BUF_SIZE);
mqtt_message_t *temp_msg = mqtt_msg_publish(&msgb,
topic, payload, l,
qos, retain,
msg_id);
2015-03-29 18:24:09 +02:00
if (lua_isfunction(L, stack)){
lua_pushvalue(L, stack); // copy argument (func) to the top of stack
luaL_unref(L, LUA_REGISTRYINDEX, mud->cb_puback_ref);
mud->cb_puback_ref = luaL_ref(L, LUA_REGISTRYINDEX);
}
msg_queue_t *node = msg_enqueue(&(mud->mqtt_state.pending_msg_q), temp_msg,
2015-03-29 18:24:09 +02:00
msg_id, MQTT_MSG_TYPE_PUBLISH, (int)qos );
2016-02-02 04:29:32 +01:00
sint8 espconn_status = ESPCONN_OK;
Networking rampage and accumulated fixes (#3060) * espconn: remove unused espconn code, take 1 This is the easiest part of https://github.com/nodemcu/nodemcu-firmware/issues/3004 . It removes a bunch of functions that were never called in our tree. * espconn: De-orbit espconn_gethostbyname Further work on https://github.com/nodemcu/nodemcu-firmware/issues/3004 While here, remove `mqtt`'s charming DNS-retry logic (which is neither shared with nor duplicated in other modules) and update its :connect() return value behavior and documentation. * espconn: remove scary global pktinfo A write-only global! How about that. * net: remove deprecated methods All the TLS stuff moved over there a long time ago, and net_createUDPSocket should just do what it says on the tin. * espconn_secure: remove ESPCONN_SERVER support We can barely function as a TLS client; being a TLS server seems like a real stretch. This code was never called from Lua anyway. * espconn_secure: more code removal * espconn_secure: simplify ssl options structure There is nothing "ssl_packet" about this structure. Get rid of the terrifying "pbuffer" pointer. Squash two structure types together and eliminate an unused field. * espconn_secure: refactor mbedtls_msg_info_load Split out espconn_mbedtls_parse, which we can use as part of our effort towards addressing https://github.com/nodemcu/nodemcu-firmware/issues/3032 * espconn_secure: introduce TLS cert/key callbacks The new feature part of https://github.com/nodemcu/nodemcu-firmware/issues/3032 Subsequent work will remove the old mechanism. * tls: add deprecation warnings * luacheck: net.ifinfo is a thing now * tls: remove use of espconn->reverse * mqtt: stop using espconn->reverse Instead, just place the espconn structure itself at the top of the user data. This enlarges the structure somewhat but removes one more layer of dynamic heap usage and NULL checks. While here, simplify the code a bit. * mqtt: remove redundant pointer to connect_info Everywhere we have the mqtt_state_t we also have the lmqtt_userdata. * mqtt: doc fixes * mqtt: note bug * tls: allow :on(...,nil) to unregister a callback
2020-04-07 14:06:27 +02:00
espconn_status = mqtt_send_if_possible(mud);
2015-03-29 18:24:09 +02:00
if(!node || espconn_status != ESPCONN_OK){
2015-03-29 18:24:09 +02:00
lua_pushboolean(L, 0);
} else {
lua_pushboolean(L, 1); // enqueued succeed.
}
2015-03-31 17:38:28 +02:00
NODE_DBG("publish, queue size: %d\n", msg_size(&(mud->mqtt_state.pending_msg_q)));
NODE_DBG("leave mqtt_socket_publish.\n");
2015-03-29 18:24:09 +02:00
return 1;
}
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
// Lua: mqtt:lwt( topic, message, [qos, [retain]])
static int mqtt_socket_lwt( lua_State* L )
{
NODE_DBG("enter mqtt_socket_lwt.\n");
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
lmqtt_userdata *mud = luaL_checkudata( L, 1, "mqtt.socket" );
luaL_argcheck( L, lua_isstring(L, 2), 2, "need lwt topic");
luaL_argcheck( L, lua_isstring(L, 3), 3, "need lwt message");
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
lua_pushvalue(L, 2);
luaL_reref(L, LUA_REGISTRYINDEX, &mud->conf.will_topic_ref);
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
lua_pushvalue(L, 3);
luaL_reref(L, LUA_REGISTRYINDEX, &mud->conf.will_message_ref);
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
int stack = 4;
if ( lua_isnumber(L, stack) )
{
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mud->conf.flags.will_qos = lua_tointeger(L, stack) & 0x3;
stack++;
}
if ( lua_isnumber(L, stack) )
{
First round of MQTT fixes (#3360) * mqtt: remove concept of connection timeout Just rely on the network stack to tell us when things have gone south. * mqtt: remove write-only mqtt_state.port field * mqtt: drop useless conditional * mqtt: decouple message sent flag from timer * mqtt: reconnect callback does not need to hang up The network stack has certainly done that for us at this point. Similarly, since we're about to call mqtt_socket_disconnected, don't bother unregistering the timer here, either. * mqtt: don't tick once per second Set the timer for the duration of the wait and cancel it on the other side. * mqtt: defer message queue destruction to _disconnect We're going to want to publish a disconnect message for real, so doing this in _close does no one any favors * mqtt: miscellaneous cleanups No functional change intended * mqtt: close() should send disconnect message for real This means waiting for _sent() to fire again before telling the network stack to disconnect. * mqtt: tidy connect and dns - Push the self-ref to after all allocations and error returns - Don't try to extract IPv4 from the domain string ourselves, let the resolver, since it can - Don't try to connect to localhost. That can't possibly work. * mqtt: common up some callback invocations * mqtt: don't retransmit messages on timeout There's no point in retransmitting messages on timeout; the network stack will be trying to do it for us anyway. * mqtt: remove unnecessary NULL udata checks * mqtt: hold strings in Lua, not C Eliminates a host of C-side allocations. While here, move the rest of the mqtt_connect_info structure out to its own thing, and pack some flags using a bitfield. * mqtt: mqtt_socket_on use lua_checkoption * mqtt: slightly augment debug messages These changes have made some debugging ever so slightly easier.
2021-01-05 12:07:09 +01:00
mud->conf.flags.will_retain = !!lua_tointeger(L, stack);
stack++;
}
NODE_DBG("leave mqtt_socket_lwt.\n");
return 0;
}
// Module function map
LROT_BEGIN(mqtt_socket, NULL, LROT_MASK_GC_INDEX)
LROT_FUNCENTRY( __gc, mqtt_delete )
LROT_TABENTRY( __index, mqtt_socket )
LROT_FUNCENTRY( connect, mqtt_socket_connect )
LROT_FUNCENTRY( close, mqtt_socket_close )
LROT_FUNCENTRY( publish, mqtt_socket_publish )
LROT_FUNCENTRY( subscribe, mqtt_socket_subscribe )
LROT_FUNCENTRY( unsubscribe, mqtt_socket_unsubscribe )
LROT_FUNCENTRY( lwt, mqtt_socket_lwt )
LROT_FUNCENTRY( on, mqtt_socket_on )
LROT_END(mqtt_socket, NULL, LROT_MASK_GC_INDEX)
LROT_BEGIN(mqtt, NULL, 0)
LROT_FUNCENTRY( Client, mqtt_socket_client )
LROT_NUMENTRY( CONN_FAIL_SERVER_NOT_FOUND, MQTT_CONN_FAIL_SERVER_NOT_FOUND )
LROT_NUMENTRY( CONN_FAIL_NOT_A_CONNACK_MSG, MQTT_CONN_FAIL_NOT_A_CONNACK_MSG )
LROT_NUMENTRY( CONN_FAIL_DNS, MQTT_CONN_FAIL_DNS )
LROT_NUMENTRY( CONN_FAIL_TIMEOUT_RECEIVING, MQTT_CONN_FAIL_TIMEOUT_RECEIVING )
LROT_NUMENTRY( CONN_FAIL_TIMEOUT_SENDING, MQTT_CONN_FAIL_TIMEOUT_SENDING )
LROT_NUMENTRY( CONNACK_ACCEPTED, MQTT_CONNACK_ACCEPTED )
LROT_NUMENTRY( CONNACK_REFUSED_PROTOCOL_VER, MQTT_CONNACK_REFUSED_PROTOCOL_VER )
LROT_NUMENTRY( CONNACK_REFUSED_ID_REJECTED, MQTT_CONNACK_REFUSED_ID_REJECTED )
LROT_NUMENTRY( CONNACK_REFUSED_SERVER_UNAVAILABLE, MQTT_CONNACK_REFUSED_SERVER_UNAVAILABLE )
LROT_NUMENTRY( CONNACK_REFUSED_BAD_USER_OR_PASS, MQTT_CONNACK_REFUSED_BAD_USER_OR_PASS )
LROT_NUMENTRY( CONNACK_REFUSED_NOT_AUTHORIZED, MQTT_CONNACK_REFUSED_NOT_AUTHORIZED )
LROT_END(mqtt, NULL, 0)
int luaopen_mqtt( lua_State *L )
{
luaL_rometatable(L, "mqtt.socket", LROT_TABLEREF(mqtt_socket));
return 0;
}
NODEMCU_MODULE(MQTT, "mqtt", mqtt, luaopen_mqtt);