@madpilot makes

Up until now I’ve been testing the door opener using a pigtail off the 5V line on the FTDI cable – the board has a LM317T regulator setup to output the 3.3V that the ESP8266 requires. The installation point for this board is a 12V output from the garage door controller. Those of you with a bit more experience with voltage regulators would realise where this is going… I plugged the board in to a 12V supply and noticed that the regulator was getting really hot.

On my lunch break today, I made my way to Jaycar to grab a heatsink. I was thinking about the little clip on ones, which I looked up on the website while I was on the tram, and I came across “thermal resistance” in the specs.

I had no idea what this was, but I knew that it was probably an interesting specification to look up. After googling it, I discovered it is a measure of how many degrees the heatsink will heat up per watt it is dissipating. Some more googling and I found this page around heatsinking the LM317T, which had all the magic formulas that I needed.

So, the power dissipation is:

P = (Vin – Vout) * I

I had two parts of that equation: Vin = 12V input, and Vout = 3.3V, so I needed to work out the power draw of the ESP8266.

More Googling…

I found these (anecdotal) results, which sounded reasonable. So taking the peak current of 320mA, the power is:

P = (12 – 3.3) * 0.32
P = 2.784W

Apparently, the maximum Thermal Resistance the regulator can tolerate is calculated by:

T(rmax) = (60 – roomtemp) / P

So, let’s make roomtemp = 20 (Yes, I’m in Australia, but I’m also in Melbourne, so that is a good average), giving us:

T(rmax) = (60 – 20) / 2.784
T(rmax) = 14.37 C/W

The little clip on heatsinks have a thermal resistance of 30 C/W, which is double what we need.

Boo.

By this stage I had arrived at the store, so went to the heatsink section to see what they had. They had this guy, which has a thermal resistance of 12 C/W. Perfect! It was quite large though, and I knew they board was tight, so I had to be a little creative in fitting it.

Don’t judge me.

I’m a little annoyed that it doesn’t sit on the board properly, but I couldn’t stand it up, as there was limited room above the board, so it was the best I could do without redesigning an new board.

It’ll do.

So, it turns out the ESP8266-01 (at least the one I have) supports 1MB of storage. I had been butting up against the 512kB I thought I had, and it turns out I needn’t worry.

Running this command:

ESP.getFlashChipSizeByChipId()

Will tell you what your chip supports.
This means I can probably incorporate Over-the-air updates, which will be handy. I still need to keep the image under 512kB (OTA needs the space to upload the new image).

w00t!

I’ve mentioned before that I plan on using mDNS to resolve the name of my server. While the ESP8266 Arduino library can broadcast a mDNS name, it doesn’t query mDNS when resolving names. I found mrdunk’s mdns on Github that implements enough of the mDNS protocol, that I should be able to hack mDNS name queries into the project.

I spun up a quick proof of concept sketch to see how it all works.

// This sketch will display mDNS (multicast DNS) data seen on the network.

#include <ESP8266WiFi.h>
#include "mdns.h"

// When an mDNS packet gets parsed this optional callback gets called once per Query.
// See mdns.h for definition of mdns::Answer.
void answerCallback(const mdns::Answer* answer){
  if(strcmp(answer->name_buffer, "mqtt.local") == 0) {
    Serial.print("Name: ");
    Serial.println(answer->name_buffer);
    Serial.print("Answer: ");
    Serial.println(answer->rdata_buffer);
    Serial.print("TTL: ");
    Serial.println(answer->rrttl);
  }
}

// Initialise MDns.
// If you don't want the optional callbacks, just provide a NULL pointer as the callback.
mdns::MDns my_mdns(NULL, NULL, answerCallback);

void setup() {
  // Open serial communications and wait for port to open:
  Serial.begin(115200);

  // setting up Station AP
  WiFi.begin("[ssid]", "[password]");

  Serial.print("Connecting to WiFi");
  // Wait for connect to AP
  int tries = 0;
  while (WiFi.status() != WL_CONNECTED) {
    delay(500);
    Serial.print(".");
    tries++;
    if (tries > 30) {
      break;
    }
  }
  Serial.println();
}

void query(const char *name) {
  mdns::Query q;

  int len = strlen(name);

  for(int i = 0; i < len; i++) {
    q.qname_buffer[i] = name[i];
  }
  q.qname_buffer[len] = '&#92;&#48;';

  q.qtype = 0x01;
  q.qclass = 0x01;
  q.unicast_response = 0;
  q.valid = 0;

  my_mdns.Clear();
  my_mdns.AddQuery(q);
  my_mdns.Send();
}

int send = 0;
void loop() {
  if(send == 0) {
    query("mqtt.local");
    send = 1;
  }
  my_mdns.Check();
}

The sketch sends a query out for mqtt.local. We are building up a query struct with the requested host name, and a qtype of 0x01 (host address query), and qclass of 0x01 (internet).

When a response is received, the name is checked against the name we requested, and if it matches, the IP address and Time-to-Live (TTL) are printed to the Serial console. The name check is required, because that answerCallback will get called every time a mDNS packet is received, regardless of who sends it – It can get quite chatty.

I found that I needed to call Clear() before adding the query, otherwise the packet was filled with garbage – Clear seems to initialize all the required buffers.

For name resolution, all I really need is the IP address and TTL. My plan is to have an array of names that I need to resolve (for the moment, it’ll be a maximum of two – one for the MQTT server, and one for the log file server). If either of those names end with .local, I’ll resolve the name using mDNS.

On the first request, I’ll cache the result speeding up subsequent requests. I can use the TTL to expire the cached version, and re-query the network when required. A little clunky (it would be nice if the underlying network stack automatically did this), but it should work.

mDNS: A Teardown.

Trying to get mDNS queries working hasn’t quite been as straight forward as I was hoping. I mentioned in a previous log that I found a library, but it was a little overkill for what I need, so I did what any silly software developer does: started rolling my own.

How am I justifying this? Well, I’m fast running out of space. The GitHub version of the ESP8266 hardware definition I’m using is significantly larger than the distributed version, so I only have about 10k of program storage left – as a result I’m being more weary of how much source code I’m uploading. Since I don’t need to respond to mDNS questions (the ESP libs have one built in), I can skip question parsing, and since I’m only interested in name browsing, I can ignore a all bar two classes of responses. And sometimes learning a protocol can be fun. Sometimes.

My completed library can be found here.

The first thing I did was work out how mDNS works. It’s a pretty clever hack – It reuses standard DNS packets, but rather than ask a specific DNS server to answer a query, mDNS clients just broadcast a UDP packet to anyone who will listen. If another listener can answer the question, it broadcasts the answer. You can see more information about the packet format on the mDNS Wikipedia page.

Looking at the packet structure, and looking through the code from this mrdunk’s library, as well as some packet sniffing using wireshark, I was able to generate questions and parse answers. This was very much proof of concept code that was embedded in my project, and it worked, though it lacked formal testing, and this would definitely be something I would want to reuse in the future.

ime to break it out in to another library.

Now, my C++ isn’t particularly strong, so this sounded like a good opportunity to learn more about C++ classes. I had a niggling concern around code size, memory usage and speed when it came to using C++ for embedded systems, so I did a bit of research around best practices, and found this article. What I found particularly useful in the article was the explanation of how C++ achieves what is does in terms of the equivalent C code. This clicked in to place what C++ was doing and allowed me to make some better decisions around structuring my library. Though I did get tripped up. A lot

The basic rules are:

• Don’t use exceptions (you can’t on Arduino), so most functions return a status code indicating success or error states.
• Avoid virtual functions where possible, because you end up with a vtable which takes up memory and requires extra cycles to lookup where the address of the required function
• Avoid dynamic memory
• If you need dynamic memory, free it as soon as possible, or make it super long living to avoid fragmentation
• Don’t include unnecessary libraries. I thought I would do the right thing and use std::string which abstracts string handling. This made my library’s object file an order of magnitude larger. Reverting back to regular char* string keped everything nice and small, and as I didn’t need to do much actual string manipulation, the trade of was totally worth it.

Testing

I’m a big believer in automated testing. Especially for something as low level as a library where you rely on certain types of network packets. Automated testing on an Arduino is probably possible, but since I’m not doing anything too Arduino specific, I was able to build a test suite that ran on my laptop. This made the test cycle much quicker as I didn’t have to wait for my code to upload. The downside is I need to mock out a few objects, but with some clever code organisation I managed to avoid too many mocks.

I found Catch which is a header-only, lightweight testing framework for C++. It was pretty easy to setup, and works with TravisCI, so every time I commit a change the tests run automatically.

Travis was a bit of a pain to setup, as the C++ they run is ancient, so I needed to setup a custom toolchain in travis.yml.

Once I had everything written and tested, I was able to drop it in to the Garagedoor sketch, and now the sketch can find the mqtt server!

Up to this point, all of the settings have just been stored in the source code. This will make changing these setting pretty difficult, so I’ll need some way of setting, reading and storing them.

Being a web developer, the first thing I thought of was JSON. There is an Arduino JSON library, and it’s supposedly pretty efficient. Memory is still pretty tight though, and there is some parsing involved, so I started looking at something a little more lo-fi.

The easiest way to go about this would be to create a class with members that match the required settings, but I wanted to make something a bit more re-usable. I knocked up a quick class that allows me to define the layout dynamically.

Config.h

#ifndef Config_h
#define Config_h

#define config_result           uint8_t
#define E_CONFIG_OK             0
#define E_CONFIG_FS_ACCESS      1
#define E_CONFIG_FILE_NOT_FOUND 2
#define E_CONFIG_FILE_OPEN      3
#define E_CONFIG_PARSE_ERROR    4
#define E_CONFIG_MAX            5

#define CONFIG_MAX_OPTIONS      15

#include <EEPROM.h>
#include <FS.h>

class ConfigOption {
  public:
    ConfigOption(const char *key, const char *value, int maxLength);
    const char *getKey();
    const char *getValue();
    int getLength();
    void setValue(const char *value);

  private:
    const char *_key;
    char *_value;
    int _maxLength;
};

class Config {
  public:
    Config();
    config_result addKey(const char *key, int maxLength);
    config_result addKey(const char *key, const char *value, int maxLength);
    config_result read();
    config_result write();

    ConfigOption *get(const char *key);
    bool *set(const char *key, const char *value);
  private:
    int _optionCount;
    ConfigOption *_options[CONFIG_MAX_OPTIONS];
    void reset();
};

#endif

Config.cpp

#include "Config.h"

#define CONFIG_FILE_PATH "/config.dat"

ConfigOption::ConfigOption(const char *key, const char *value, int maxLength) {
  _key = key;
  _maxLength = maxLength;
  setValue(value);
}

const char *ConfigOption::getKey() {
  return _key;
}

const char *ConfigOption::getValue() {
  return _value;
}

int ConfigOption::getLength() {
  return _maxLength;
}

void ConfigOption::setValue(const char *value) {
  _value = (char *)malloc(sizeof(char) * (_maxLength + 1));

  // NULL out the string, including a terminator
  for(int i = 0; i < _maxLength + 1; i++) {
    _value[i] = '&#92;&#48;';
  }

  if(value != NULL) {
    strncpy(_value, value, _maxLength);
  }
}

Config::Config() {
  _optionCount = 0;
};

config_result Config::addKey(const char *key, int maxLength) {
  return addKey(key, NULL, maxLength);
}

config_result Config::addKey(const char *key, const char *value, int maxLength) {
  if(_optionCount == CONFIG_MAX_OPTIONS) {
    return E_CONFIG_MAX;
  }
  _options[_optionCount] = new ConfigOption(key, value, maxLength);
  _optionCount += 1;

  return E_CONFIG_OK;
}

config_result Config::read() {
  if (SPIFFS.begin()) {
    if (SPIFFS.exists(CONFIG_FILE_PATH)) {
      File configFile = SPIFFS.open(CONFIG_FILE_PATH, "r");

      if (configFile) {
        int i = 0;
        int offset = 0;
        int length = 0;

        for(i = 0; i < _optionCount; i++) {
          length += _options[i]->getLength();
        }

        if(length != configFile.size()) {
          return E_CONFIG_PARSE_ERROR;
        }

        uint8_t *content = (uint8_t *)malloc(sizeof(uint8_t) * length);
        configFile.read(content, length);

        for(i = 0; i < _optionCount; i++) {
          // Because we know the right number of bytes gets copied,
          // and it gets null terminated,
          // we can just pass in an offset pointer to save a temporary variable
          _options[i]->setValue((const char *)(content + offset));
          offset += _options[i]->getLength();
        }

        configFile.close();
        free(content);

        return E_CONFIG_OK;
      } else {
        configFile.close();
        return E_CONFIG_FILE_OPEN;
      }
    } else {
      return E_CONFIG_FILE_NOT_FOUND;
    }
  } else {
    return E_CONFIG_FS_ACCESS;
  }
}

config_result Config::write() {
  if (SPIFFS.begin()) {
    int i = 0;
    int offset = 0;
    int length = 0;

    for(i = 0; i < _optionCount; i++) {
      length += _options[i]->getLength();
    }

    File configFile = SPIFFS.open(CONFIG_FILE_PATH, "w+");
    if(configFile) {
      uint8_t *content = (uint8_t *)malloc(sizeof(uint8_t) * length);
      for(i = 0; i < _optionCount; i++) {
        memcpy(content + offset, _options[i]->getValue(), _options[i]->getLength());
        offset += _options[i]->getLength();
      }

      configFile.write(content, length);
      configFile.close();

      free(content);
      return E_CONFIG_OK;
    } else {
      return E_CONFIG_FILE_OPEN;
    }
  }

  return E_CONFIG_FS_ACCESS;
}

/*
 * Returns the config option that maps to the supplied key.
 * Returns NULL if not found
 */
ConfigOption *Config::get(const char *key) {
  for(int i = 0; i < _optionCount; i++) {
    if(strcmp(_options[i]->getKey(), key) == 0) {
      return _options[i];
    }
  }

  return NULL;
}

It consists of two classes: ConfigOption and Config. Config has an array (up to CONFIG_MAX_OPTIONS) of ConfigOptions.

The addKey method adds a ConfigOption to the Config Option. It takes a maximum length, so that we can pre-allocate memory for each string. This makes each configuration length a known quantity, reducing the change of buffer overflows.

There is also a read and write method that reads and writes a blob of data to the SPIFFS flash area. The file format is very simple: a set of concatenated strings of a fixed length – because the class knows the length of each string, it knows exactly where to read each option from. This does make removing or changing the length of an option difficult, though.

Setup is fairly easy:

Config config;

void configSetup() {
  config.addKey("ssid", "", 32);
  config.addKey("passkey", "", 32);
  config.addKey("encryption", "0", 1);

  config.addKey("mqttDeviceName", "garage", 32);

  config.addKey("mqttServer", "", 128);
  config.addKey("mqttPort", "1883", 5);

  config.addKey("mqttAuthMode", "0", 1);
  config.addKey("mqttTLS", TLS_NO, 1);

  config.addKey("mqttUsername", "", 32);
  config.addKey("mqttPassword", "", 32);

  config.addKey("mqttFingerprint", "", 64);

  config.addKey("syslog", "0", 1);
  config.addKey("syslogHost", "", 128);
  config.addKey("syslogPort", "514", 5);
  config.addKey("syslogLevel", "6", 1);

  switch(config.read()) {
    case E_CONFIG_OK:
      Serial.println("Config read");
      return;
    case E_CONFIG_FS_ACCESS:
      Serial.println("E_CONFIG_FS_ACCESS: Couldn't access file system");
      return;
    case E_CONFIG_FILE_NOT_FOUND:
      Serial.println("E_CONFIG_FILE_NOT_FOUND: File not found");
      return;
    case E_CONFIG_FILE_OPEN:
      Serial.println("E_CONFIG_FILE_OPEN: Couldn't open file");
      return;
    case E_CONFIG_PARSE_ERROR:
      Serial.println("E_CONFIG_PARSE_ERROR: File was not parsable");
      return;
  }
}

This sets up each of the keys with a name, a default value and a max length. It then reads in the configuration file from the flash.

To read the options, you can do this:

int authMode = atoi(config.get("mqttAuthMode")->getValue());
int tls = atoi(config.get("mqttTLS")->getValue());

Note: every config comes back as a string, so they need to be cast if required.

The beauty of using a config file read from flash means I can build the configuration externally and upload it, which means I don’t have to have a configuration system built yet – and it means I know exactly what config options I’ll need when I finally build the configuration system.

I created a super small program in C++ that compiles with G++. The header file is exactly the same as one above. The CPP file looks like this:

#include <string.h>
#include <stdlib.h>
#include <stdio.h>

#include "config.h"

ConfigOption::ConfigOption(const char *key, const char *value, int maxLength) {
  _key = key;
  _maxLength = maxLength;
  setValue(value);
}

const char *ConfigOption::getKey() {
  return _key;
}

char *ConfigOption::getValue() {
  return _value;
}

int ConfigOption::getLength() {
  return _maxLength;
}

void ConfigOption::setValue(const char *value) {
  if(_value) {
    free(_value);
  }

  _value = (char *)malloc(sizeof(char) * (_maxLength + 1));

  // NULL out the string, including a terminator
  for(int i = 0; i < _maxLength + 1; i++) {
    _value[i] = '&#92;&#48;';
  }

  if(value != NULL) {
    strncpy(_value, value, _maxLength);
  }
}

Config::Config() {
  _optionCount = 0;
};

config_result Config::addKey(const char *key, int maxLength) {
  return addKey(key, NULL, maxLength);
}

config_result Config::addKey(const char *key, const char *value, int maxLength) {
  if(_optionCount == CONFIG_MAX_OPTIONS) {
    return E_CONFIG_MAX;
  }
  _options[_optionCount] = new ConfigOption(key, value, maxLength);
  _optionCount += 1;

  return E_CONFIG_OK;
}

config_result Config::read() {
  int i;
  int offset = 0;
	int length = 0;

  for(i = 0; i < _optionCount; i++) {
    length += _options[i]->getLength();
  }

  char *content = (char *)malloc(sizeof(char) * length);

  FILE *f = fopen("config.dat", "r");
  fread(content, sizeof(char), length, f);
  fclose(f);

  for(i = 0; i < _optionCount; i++) {
    // Because we know the right number of bytes gets copied,
    // and it gets null terminated,
    // we can just pass in an offset pointer to save a temporary variable
    _options[i]->setValue(content + offset);
    offset += _options[i]->getLength();
  }

  free(content);
}

config_result Config::write() {
	int i;
  int offset = 0;
	int length = 0;

  for(i = 0; i < _optionCount; i++) {
    length += _options[i]->getLength();
  }

  char *content = (char *)malloc(sizeof(char) * length);
  for(i = 0; i < _optionCount; i++) {
    printf("%s\n", _options[i]->getKey());
    memcpy(content + offset, _options[i]->getValue(), _options[i]->getLength());
    offset += _options[i]->getLength();
  }

  FILE *f = fopen("config.dat", "w");
  fwrite(content, sizeof(char), length, f);
  fclose(f);
  printf("Length: %d\n", length);
}

ConfigOption *Config::get(const char *key) {
  for(int i = 0; i < _optionCount; i++) {
    if(strcmp(key, _options[i]->getKey()) == 0) {
      return _options[i];
    }
  }
  return NULL;
}

bool Config::set(const char *key, const char *value) {
  for(int i = 0; i < _optionCount; i++) {
    if(strcmp(key, _options[i]->getKey()) == 0) {
      _options[i]->setValue(value);
      return true;
    }
  }
  return false;
}

int main(int argc, char **argv) {
  Config config;

  config.addKey("ssid", "", 32);
  config.addKey("passkey", "", 32);
  config.addKey("encryption", "0", 1);

  config.addKey("mqttDeviceName", "garage", 32);

  config.addKey("mqttServer", "", 128);
  config.addKey("mqttPort", "1883", 5);

  config.addKey("mqttAuthMode", "0", 1);
  config.addKey("mqttTLS", "0", 1);

  config.addKey("mqttUsername", "", 32);
  config.addKey("mqttPassword", "", 32);

  config.addKey("mqttFingerprint", "", 64);

  config.addKey("syslog", "0", 1);
  config.addKey("syslogHost", "", 128);
  config.addKey("syslogPort", "514", 5);
  config.addKey("syslogLevel", "6", 1);

  config.set("ssid", "[ssid]");
  config.set("passkey", "[passkey]");
  config.set("encryption", "2");
  config.set("mqttServer", "[mqtt server name]");
  config.set("mqttPort", "8883");
  config.set("mqttAuthMode", "2");
  config.set("mqttTLS", "1");
  config.set("mqttFingerprint", "[fingerprint]");

  config.set("syslog", "1");
  config.set("syslogHost", "[log server name]");
  config.set("syslogPort", "514");
  config.set("syslogLevel", "7");

  config.write();
  return 0;
}

Some sensitive settings have been redacted.

Compile it with:

g++ config.cpp -o config

And then run it. It’ll generate a config.dat file, that I just copy to the data directory of the Arduino project. I then upload it using the ESP8266 Sketch Upload tool.

I’m not 100% sold on this, I wonder whether I should just create a class with static members, rather than dynamically building it. Ironically, it might be better to build the strings dynamically so only the memory required is used (up to a maximum length to avoid those scary buffer overflows).

If I start hitting memory limits, I’ll revisit.

While I want to do full CA verification, I’m waiting on some of the bugs to get ironed out of the ESP8266 Arduino library, so I’ll take a shortcut for the moment, and use fingerprinting to verify the server certificate (It should be pretty easy to move to CA verification down the track).

To do this, we need three pieces of information:

1. Client certificate
2. Client secret key
3. Server fingerprint

We have already generated the certificate and the secret key, so let”s generate a fingerprint for the server certificate. We can use OpenSSL to do this:

openssl x509 -in mosquitto/etc/certs/mqtt.local.crt.pem -sha1 -noout -fingerprint

This will generate a 160bit sha1 string, that may look a little like this:

SHA1 Fingerprint=DA:39:A3:EE:5E:6B:4B:0D:32:55:BF:EF:95:60:18:90:AF:D8:07:09

Convert PEM to DER

I was reading through the docs, and while the axTLS says it supports PEM, I found that I needed to convert the PEM format certificate (which is ASCII based) to DER which is binary.

Again, OpenSSL to the rescue:

openssl x509 -outform der -in garage.local.csr.pem -out garage.local.csr.der
openssl x509 -outform der -in garage.local.key.pem -out garage.local.key.der

Loading the certificates

Down the track, I’ll be able to upload the certificate and key via the administration console (certificates expire, and need to be updated), which means storing them in the code makes little sense. Instead I’m going to store them on the flash as a file.

Thankfully there is a tool that makes uploading files to the SPIFFS filesystem super simple. Gratuitously stolen from the Filesystem Read Me:

ESP8266FS is a tool which integrates into the Arduino IDE. It adds a menu item to Tools menu for uploading the contents of sketch data directory into ESP8266 flash file system.

• Download the tool: https://github.com/esp8266/arduino-esp8266fs-plugin/releases/download/0.2.0/ESP8266FS-0.2.0.zip.
• In your Arduino sketchbook directory, create tools directory if it doesn’t exist yet
• Unpack the tool into tools directory (the path will look like_<home_dir>/Arduino/tools/ESP8266FS/tool/esp8266fs.jar_)
• Restart Arduino IDE
• Open a sketch (or create a new one and save it)
• Go to sketch directory (choose Sketch > Show Sketch Folder)
• Create a directory named data and any files you want in the file system there
• Make sure you have selected a board, port, and closed Serial Monitor
• Select Tools > ESP8266 Sketch Data Upload. This should start uploading the files into ESP8266 flash file system. When done, IDE status bar will display SPIFFS Image Uploaded message.

Copy the two der files in to the data directory, and run the data upload function (I renamed them client.key.der and client.crt.der because that kind of makes more sense in this context). To load the files, the code looks a little like this:

// We need to use WiFiCLientSecure to the encryption works
WiFiClientSecure wifiClient

SPIFFS.begin();
File ca = SPIFFS.open("/client.crt.der", "r");
if(!ca) {
  Serial.println("Couldn't load cert");
  return;  
}

if(wifiClient.loadCertificate(ca)) {
  Serial.println("Loaded Cert");
} else {
  Serial.println("Didn't load cert");
  return;
}
  
File key = SPIFFS.open("/client.key.der", "r");
if(!key) {
  Serial.println("Couldn't load key");
  return;  
}

if(wifiClient.loadPrivateKey(key)) {
  Serial.println("Loaded Key");
} else {
  Serial.println("Didn't load Key");
}

The device will connect at this point, and the mosquitto server will be happy. It will use the name on the certificate as the client name, and the connection will be encrypted.

However, I not verifing the server yet. I will do that using the verify function

PubSubClient client("mqtt.local", 8883, callback, wifiClient); //set  MQTT port number to 8883 as per //standard

wifiClient.connect("mqtt.local", 8883);
  
if(wifiClient.verify("DA 39 A3 EE 5E 6B 4B 0D 32 55 BF EF 95 60 18 90 AF D8 07 09", "mqtt.local")) {
  Serial.println("connection checks out");
  wifiClient.stop();

  if(client.connect("garage")) {
    Serial.println("Connected");
    client.subscribe("test");
    client.publish("test", "hello world");    
  } else {
    Serial.println("Not connected");
  }
} else {
  wifiClient.stop();
  Serial.println("connection doesn't check out");
}

Notice that all of the colons have been replace with spaces.

Because of the way the PubSubClient works, I need to make a (pre?)connection to the server to do the verification. It’s pretty simple – connect, verify and disconnect, then continue with the PubSub connection.

There is a problem here – the mDNS name won’t actually resolve at this point. Because the DNS resolver doesn’t support it.

Ok, that isn’t strictly true – every DNS resolver CAN resolve mDNS, but you need to hit a special DNS server and port (224.0.0.251 on port 5353), which I haven’t worked out how to do with the Arduino library yet…

I was digging around the ESP8266 github page, as there was an announcement that it now supports CA verification (it ALMOST does – there was a regression bug that means it’s still not working), but I noticed that the last release (2.3.0 at time of writing) is actually quite old – it was released in June.

I was going through recent commit messages, and noticed there had been quite a bit of work around integrating axTLS, which is a tiny TLS library specifically designed for computers with small memory footprints (AKA microcontrollers).

This piqued my interest.

It has been said that the ESP8266 doesn’t support TLSv1.2, because of a buggy implementation in the firmware, which I had verified earlier in testing and hence configuring MQTT to force TLSv1.1.

By including axTLS as part of the Arduino library, and not relying on the ESP8266 API, I wondered if TLS1.2 was now supported.

Turns out it is!

It does mean we’ll have to use a “unstable” version of ESP8266 library. This means other stuff might not work, we need to install it using git, and it’s up to us to keep everything up to day.

Thankfully, it isn’t that difficult to install the library using git.

First, remove the existing ESP8266 library using the board manager – find it using the search function, then hit remove.

Now, you can manually install – Instructions are here. Don’t forget to re-select the right board and to set the CPU frequency to 160Mhz!

Now, that regression bug. It stops client certificates being sent for verification, which is bad for us, so we’ll need to roll back a couple of commits. To do that:

cd ~/Arduino/hardware/esp8266com/esp8266
git checkout -b pre-axtls-2 d6e38f0abd2e1bf796a32e8b1a24d37fdc7daaf8

This creates a new branch based on an older revision that seems to still work ok.

Modify mosquitto to use TLSv1.2

This one’s pretty easy: Remove the tls_version line

#tls_version tlsv1.1

Then remove

"--tls-version", "tlsv1.1", 
from both mosquitto-client/Dockerfile.pub and mosquitto-client/Dockerfile.sub.

Run

docker-compose build
and start the server with
docker-compose up

You should see “Hello world” in there somewhere. Next up, the Arduino side of things…

While having a username and password is better than an open MQTT server, the credentials are still being sent in plain text, so anyone with a packet sniffer could intercept them and then gain access to our server.

To avoid this, let’s setup TLS (commonly, but maybe not correctly known as SSL) to encrypt the connection and kept the credentials secret.

TLS 101

TLS provides a chain of trust, allowing two computers that know nothing about each other to trust each other.

Let’s look at how this works in the context of a browser hitting a secure website, as this is an example we should all be familiar with.

At a really high level, both your computer and the server have a Certificate Authority (CA) that they both implicitly trust (there is actually more than one, but let’s keep it simple). When you hit a website that is secured the server sends it’s certificate to the browser, which does some complicated maths and verifies the certificate against the CA – if the numbers work out, your browser can be confident that the certificate is legitimate and that the secure site is who it says it is.

Web servers generally don’t care who the client is, so they rarely ask for a client certificate, but in other applications (like this garage door opener) the server will also verify the client against it’s CAs.

After both computers are happy they are talking to the right people, they will exchange a set of symmetric keys that both computers will use to communicate. Symmetric keys are much faster to encrypt and decrypt that asymmetric keys, but it requires both sides to have the same key.

To do that securely, a symmetric key is generated, encrypted using the other computer’s public key, sent securely across the wire and then decrypted – now both computers have the key and they can start talking.

Setting up our own trust network

Enough theory! The first thing to do is generate a CA. If you have OpenSSL installed on your computer (Linux and OSX users more than likely will), you can generate a CA on the command line:

openssl req -new -x509 -days 3650 -extensions v3_ca -keyout ca.key.pem -out ca.crt.pem

You’ll be asked for a PEM password – make this something difficult to guess and then store it in your password manager.

This is literally the key to the city. If someone gets hold of your ca.crt.pem and ca.key.pem, and can guess your password, then they can generate their own valid certificates – not good. Fill in the details as you are asked – these details aren’t super important as this is a personal CA, but make the values something you will recognise. The common name can be whatever you want. Now that you have your CA certificate (ca.crt.pem) and your CA.key (ca.key.pem) you can generate some certificates.

An aside: If you have a password manager that supports notes, I’d recommend saving these files in the there and deleting them from your file system after you have generated any certificates that you need.
Generate a server certificate

For this to work, it’s best to use actual domain names for servers and clients – the easiest way to do this is setting up mDNS, via avahi (linux) or bonjour (OSX) – Windows probably has something too – let’s assume that the name of the computer running MQTT is mqtt.local and the name of the garage door opener will be garage.local.

Generate a server certificate

For this to work, it’s best to use actual domain names for servers and clients – the easiest way to do this is setting up mDNS, via avahi (linux) or bonjour (OSX) – Windows probably has something too – let’s assume that the name of the computer running MQTT is mqtt.local and the name of the garage door opener will be garage.local.

# Generate a key
openssl genrsa -out mqtt.local.key.pem 2048
# Create a Certificate signing request
openssl req -out mqtt.local.csr.pem -key mqtt.local.key.pem -new

Again, fill in the details as you are asked.

The MOST important one is Common Name. IT MUST MATCH THE DOMAIN. So in our case: mqtt.local

# Sign the certificate
openssl x509 -req -in mqtt.local.csr.pem -CA ca.crt.pem -CAkey ca.key.pem -CAcreateserial -out mqtt.local.crt.pem -days 365

You will be asked for the password you entered when you created your CA.

Note that the certificate is valid for 365 days – you’ll have to generate a new one in a years time. You can decide if you want to make it longer or shorter.

Generate the client certificate

Generating a client certificate looks a lot like the generation of a server certificate – just with different filenames.

# Generate a key
openssl genrsa -out garage.local.key.pem 2048
# Create a Certificate signing request
openssl req -out garage.local.csr.pem -key garage.local.key.pem -new

The common name this time should be garage.local

# Sign the certificate
openssl x509 -req -in garage.local.csr.pem -CA ca.crt.pem -CAkey ca.key.pem -CAcreateserial -out garage.local.crt.pem -days 365

Setting up Mosquitto

Now we have all of the certificates that we need, let’s setup mosquitto to use it.

Create a ca_certificates and certs directory in the mosquitto/etc docker folder

mkdir mosquitto/etc/ca_certificates
mkdir mosquitto/etc/certs

and copy ca.crt.pem to the ca_certificates folder and mqtt.local.crt.pem and mqtt.local.key.pem to the certs folder. Finally update the etc/mosquitto.conf file to look something like this:

persistence true
persistence_location /var/lib/mosquitto/
password_file /etc/mosquitto/passwd
allow_anonymous false
cafile /etc/mosquitto/ca_certificates/ca.crt.pem
certfile /etc/mosquitto/certs/mqtt.local.crt.pem
keyfile /etc/mosquitto/certs/mqtt.local.key.pem
port 8883
tls_version tlsv1.1
include_dir /etc/mosquitto/conf.d

Notice we have changed the port to 8883, which is the standard port for secure MQTT. We are also still authenticating via username and password.

We are explicitly forcing TLS version 1.1 because the ESP8266 implementation of 1.2 is buggy, and will fail.

Save the file, and run

docker-compose build mosquitto

Because we aren’t verifying the certificates, we can use the garage.local certificate on the command line to test everything out.

Create the ca_certificates and certs directories, this time in mosquitto-client

mkdir mosquitto-client/ca_certificates
mkdir mosquitto-client/certs

Copy ca.crt.pem to the ca_certifications directory and both the garage.local.crt.pem and garage.local.key.pem files to the certs directory.

Now, modify the Dockerfile.sub ENTRYPOINT to look like this:

ENTRYPOINT [ "mosquitto_sub", "-h", "mosquitto", "-p", "8883", "--tls-version", "tlsv1.1", "--cafile", "/ca_certificates/ca.crt.pem", "--insecure", "--cert", "/certs/garage.local.crt.pem", "--key", "/certs/garage.local.key.pem" ]

and the Dockerfile.pub ENTRYPOINT to look like:

ENTRYPOINT [ "mosquitto_pub", "-h", "mosquitto", "-p", "8883", "--tls-version", "tlsv1.1", "--cafile", "/ca_certificates/ca.crt.pem", "--insecure", "--cert", "/certs/garage.local.crt.pem", "--key", "/certs/garage.local.key.pem" ]

Finally, run

docker-compose build

and

docker-compose up

And everything should connect again, but this time securely!
All of these instructions gratuitously stolen from: https://mosquitto.org/man/mosquitto-tls-7.html

Now we have the ESP8266 talking to the MQTT broker, let’s have a look at adding some authentication. The simplest form of authentication is a username and password, which Mosquitto supports. It uses a password file that has a list of all the usernames allowed to login as well as hashes of their passwords.

We will need to create a configuration file to tell mosquitto where to find that file – create an etc directory in the mosquitto directory that is in the root of the docker project:

mkdir mosquitto/etc

create a file called passwd and enter the following:

password_file /etc/mosquitto/passwd
allow_anonymous false

We are simply telling Mosquitto to

• look for the passwd file in /etc/mosquitto, and
• we only want to allow users that have authenticated.

Next, let’s use docker to run the mosquitto_passwd command

docker-compose mosquitto run /bin/bash -c "touch /tmp/passwd && mosquitto_passwd -b /tmp/passwd [username] [password] && cat /tmp/passwd && rm /tmp/passwd" > mosquitto/etc/passwd

Let’s break that monstrosity down.

Since we haven’t set up a docker volume, we can’t easily fetch the file from the container, so instead we run bash and pass it a mini script that:

1. creates a temporary passwd file,
2. runs mosquitto_passwd (Don’t forget to change [username] and [password] to real values),
3. prints out the file, then
4. deletes the temporary file.
5. Finally, (back on the host computer) we pipe the result in to a file mosquotto/etc/passwd.

If that worked, you should see a new file called passwd in the mosquitto/etc folder.

Next, we need to tell docker to copy our new config file and password file into the container when it builds. Open mosquitto/Dockerfile and add the following lines before the EXPOSE command

COPY ./etc/mosquitto.conf /etc/mosquitto/mosquitto.conf
COPY ./etc/passwd /etc/mosquitto/passwd

We should also update the client containers so they can login when we bring docker-compose up

The ENTRYPOINT line should like this for mosquitto-client/Dockerfile.pub

ENTRYPOINT [ "mosquitto_pub", "-h", "mosquitto", "-u", "[username]", "-P", "[password]" ]

and for mosquitto-client/Dockerfile.sub

ENTRYPOINT [ "mosquitto_sub", "-h", "mosquitto", "-u", "[username]", "-P", "[password]" ]

Those are capital Ps – and don’t forget to update [username] and [password] to match the values you saved in the passwd file!

Rebuild, and run docker, and you should see the two clients successfully connect, and “Hello World” being printed to the console.

docker-compose build
docker-compose run

Setup Ardiuno

This is a fairly easy modification. Find the line that looks like:

if(!pubSubClient.connect(MQTT_NAME)) {

and change it to

if(!pubSubClient.connect(MQTT_NAME, "[username]", "[password]")) {

Push the code to the ESP8266, run it, and you should see it connect just like it did before! Except this time, it will be using a username and password.You can verify that by using an incorrect username or password – the ESP8266 will never be able to connect.

So, now we have an MQTT server, lets see if we can get the ESP8266 to connect to it.

I’m going to use Nick O’Leary’s pubsubclient library. It’s pretty easy to install – find Manage Libraries in the Sketch Menu, search for pubsubclient and it should appear. Hit install.

After we physically connected to the WIFI (Which we’ll do via via hard coding the SSID and passkey for the moment – we’ll do that properly later), there are two things we need to do:

1. Connect to the MQTT server – no security for the moment, we’ll work up to that
2. Subscribe to a message. We provide the library with a callback function that gets called every time the server receives a message topic that we are subscribed to.

Here is the some proof of concept code that does all of that:

#include <ESP8266WiFi.h>
#include <ESP8266mDNS.h>
#include <PubSubClient.h>

#define SSID "your-wifi-ssid"
#define PASSKEY "your-wifi-password"
#define MQTT_IP "ip-address-of-the-mqtt-server"
#define MQTT_PORT 1883
#define MQTT_NAME "garage"

#define TOPIC "test"

#define QOS_LEVEL 0

// PubSub client
WiFiClient espClient;
PubSubClient pubSubClient(espClient);
void PubSubCallback(char* topic, byte* payload, unsigned int length);
long lastPubSubConnectionAttempt = 0;

void PubSubSetup() {
  pubSubClient.setServer(MQTT_IP, MQTT_PORT);
  pubSubClient.setCallback(PubSubCallback);
}

boolean PubSubConnect() {
  Serial.print("Connecting to MQTT server...");

  if(!pubSubClient.connect(MQTT_NAME)) {
    Serial.println("\nCouldn't connect to MQTT server. Will try again in 5 seconds.");
    return false;
  }

  if(!pubSubClient.subscribe(TOPIC, QOS_LEVEL)) {
    Serial.print("\nUnable to subscribe to ");
    Serial.println(TOPIC);
    pubSubClient.disconnect();
    return false;
  }

  Serial.println(" Connected.");
  return true;
}

void PubSubLoop() {
  if(!pubSubClient.connected()) {
    long now = millis();

    if(now - lastPubSubConnectionAttempt > 5000) {
      lastPubSubConnectionAttempt = now;
      if(PubSubConnect()) {
        lastPubSubConnectionAttempt = 0;
      }
    }
  } else {
    pubSubClient.loop();
  }
}

void PubSubCallback(char* topic, byte* payload, unsigned int length) {
  char *p = (char *)malloc((length + 1) * sizeof(char *));
  strncpy(p, (char *)payload, length);
  p[length] = '&#92;&#48;';

  Serial.print("Message received: ");
  Serial.print(topic);
  Serial.print(" - ");
  Serial.println(p);

  free(p);
}

void connectWifi(const char* ssid, const char* password) {
  int WiFiCounter = 0;
  // We start by connecting to a WiFi network
  Serial.print("Connecting to ");
  Serial.println(ssid);

  WiFi.disconnect();
  WiFi.mode(WIFI_STA);
  WiFi.begin(ssid, password);
  while (WiFi.status() != WL_CONNECTED &amp;&amp; WiFiCounter &amp;amp;lt; 30) {
    delay(1000);
    WiFiCounter++;
    Serial.print(".");
  }

  Serial.println("");
  Serial.println("WiFi connected");
  Serial.print("IP address: ");
  Serial.println(WiFi.localIP());
}

void setup() {
  Serial.begin(115200);
  connectWifi(SSID, PASSKEY);
  PubSubSetup();
}

void loop() {
  PubSubLoop();
}

To run this, first fire up our MQTT server using docker-compose up, next load up the above code on the ESP8266, replacing the constants at the top of the file with the relevant settings.

If you are running Docker on Linux, or using Docker for Mac; or Docker for Windows the IP address of the server will be the same as the IP address of your computer.

If you are running docker-machine, you will need to run docker-machine ip to find out the local IP address.

Open up the serial monitor and you should see something like this:

Connecting to [what ever your ssid is].
WiFi connected
IP address: [some IP address]
Connecting to MQTT server... Connected.

Now, if we publish a message on to the queue

docker-compose run mosquitto-publisher -t "test" -m "Hello Arduino!"

We should see the Arduino consuming the message:

Message received: test - Hello Arduino!

If you don’t see the message, there should be hints in the serial monitor about what went wrong.

Things to look out for:

• If you can’t connect to the WiFI, check your SSID and passkey
• If you can’t connect to the MQTT server, check your IP address. Also, check the docker output – you should see a message when the Arduino connects.

One of the design goals for this project was a secure system – this device can open my garage, and I don’t really want just any person to be able to do that!

Disclaimer: I’m not a security expert! If you find holes in my logic, please let me know!

The most obvious way to interface with the controller was to setup some sort of server that a controller (eg an iPhone app) talks to. The problem with this is that an open network port is an attack vector – if a good guy can connect to a network port, so can a bad guy. While I could set up a username or password, having a port open potentially means buffer overflows, and since I’m not really a C/C++ guy this a big problem for me – I mean a lot of really smart people avoid writing servers in C…

After some research, I found this paper (PDF) talking about the use of the publish/subscriber pattern (AKA pub/sub), where a device connects to as server and subscribes to event notifications. This means there is no port open on the device, so even if an attacker found it’s IP address, there is no way for them to connect to it. It seems MQTT is the IOT pub/sub system of choice (it drives a lot of public IOT platforms).

This of course is not without other issues (not exhaustive – please suggest more if you can think of others):

1. Problem: If an attacker got on the network, they could just connect to the MQTT server and send a “open” command. Solution: Access control lists (ACL): Only allow certain MQTT clients send certain commands. We would need some way for the server to verify who the client is…
2. Problem: What if an attacker managed to spoof the IP or MAC address (perhaps through ARP poisoning) they could become a legitimate MQTT server, allowing them to send an open command. Solution: Verify the identity of the server. TLS certificates can help with that. Using a certificate signed by a custom Certificate Authority (CA) means we can guarantee the server is who it says it is. This also fixes the client verification problem from point 1.

After a quick search, I found the mosquitto project – an open source MQTT server which looked pretty easy to setup, supports ACLs and TLS.

Docker all the things

I decided to set it up a test environment using docker, as it makes dependency management much easier, and also makes the setup scriptable, so when it comes time to deploy to “production” I know exactly what I need to do.

You can clone my git repository to get started (I’ve tagged each of the steps so you can play at home – we’ll start with mqtt-1):

git clone https://github.com/madpilot/home-automation
cd home-automation
git checkout -b mqtt-1 mqtt-1
docker-compose build
docker-compose up

Congratulations! You now have a working MQTT server running on port 1883. Because it hasn’t configured it yet, the server has no authentication or authorisation – that’s fine for the moment – let’s get the baseline working, then we can add the security layers later.The easiest way to test the server is to use the mosquitto command line client. I’ve included these as part of the docker-compose cluster – in fact, when you ran docker-compose, a subscriber was automatically created (subscribed to the “test” channel) and publisher publishes the first message.

Look through the output on the screen. If you see:

mosquitto-subscriber_1  | Hello World

Everything has worked!You can try it out yourself – run the following command:

docker-compose run mosquitto-publisher -t "test" -m "Hello again"

And you should see

mosquitto-subscriber_1  | Hello again

I’ve been a bit tricky by using a docker entry point to hide away the full command to make sending messages much easier. The full command that gets called is

mosquitto_pub -h mosquitto -t test -m "Hello again"

Which tells mosquittopub to connect to the server with a host name of _mosquitto (Which is set by docker), publish to the “test” channel, and send the message “Hello again”.Now that we have a working MQTT server and a way of testing things, we can start looking at the Arduino side of things.