Skip to content

Latest commit

 

History

History
1384 lines (1038 loc) · 41.8 KB

README.md

File metadata and controls

1384 lines (1038 loc) · 41.8 KB

wolfSSL SSL/TLS Examples

Here are examples focused on TCP/IP connections. The simplest of which are the *-tcp.c files, which demonstrate a simple client/server TCP/IP connection without encryption. From there, the *-tls.c files demonstrate the same connection, but modified to utilize wolfSSL to establish a TLS 1.2 connection.

In general, the naming convention of these files mean that if a file is named in the form X-Y.c, then it's a copy of X.c intended to demonstrate Y. The exceptions being server-tls.c and client-tls.c, as noted above. Furthermore, the files is formatted such that using a diff tool such as vimdiff to compare X-Y.c to X.c should highlight only the relevant changes required to convert X.c into X-Y.c

The files in this directory are presented to you in hopes that they are useful, especially as a basic starting point. It is fully recognized that these examples are neither the most sophisticated nor robust. Instead, these examples are intended to be easy to follow and clear in demonstrating the basic procedure. It cannot be guaranteed that these programs will be free of memory leaks, especially in error conditions.

For Visual Studio users with the VisualGDB extension, there are additional example files in VisualGDB-tls.

Quick Start

This portion of the README will show you how to quickly build and run some of the examples in this directory. For more detail on the examples and further variations and features please see the Tutorial section.

Build wolfSSL:

./configure --enable-asynccrypt && make && sudo make install

In wolfssl-examples/tls:

make clean && make

To run simple TLS example, in separate terminals enter:

./server-tls
./client-tls 127.0.0.1

To run non-blocking / threaded TLS example, in separate terminals enter:

./server-tls-threaded
./client-tls-nonblocking 127.0.0.1

Tutorial

This portion of the README is dedicated to walking you through creating most of the files in this directory.

Before we begin, there are a few important things to note.

First, code will be referred to in terms of "blocks". Blocks are named by the comment at the top of them. Whenever we write code, we should also write the comments so that we can identify the blocks. For example, the following code contains two blocks:

    /* Hello, world! */
    printf("Hello, world!\n");

    /* check ret */
    ret = foo()
    if (ret == -1) {
        printf("foo() returned -1\n");
    }

When we refer to the "Hello, world!" block, we should look at the printf() statement. Similarly, when we refer to the "check ret" block, we should look at everything from the comment to the bottom of the if statement.

Second, all references to files are made relative to this tls/ directory. Any file reference should then be modified to point to the correct location.

Third, it is recommended that you make a new directory to write your files into.

Table of contents

  1. Running these examples

  2. Compiling these examples

  3. A simple TCP client/server pair

    1. Server
    2. Client
    3. Running
  4. Converting to use wolfSSL

    1. Server
    2. Client
    3. Running
  5. Using callbacks

    1. Running
  6. Using ECC

    1. Server
    2. Client
    3. Running
  7. Encrypted Client Hello

If the client is running on the same machine as the server, use 127.0.0.1 as the IPv4 address when calling the client.

Otherwise, if the server is on a remote machine, learn the IPv4 address of that machine and use that instead.

If you have a tool such as Wireshark, you can use it to inspect the connection.

If you have any troubles with compiling or running that do not seem to be due to the code, see the section on compiling below.

If you stick to the naming convention, copying the makefile Makefile (which can be found here) into the directory where you are working should allow you to simply use make to compile.

The makefile is written such that it knows when it is appropriate to include pthread support and if it is necessary to link against wolfSSL. You also have access to make clean to remove all of the compiled programs.

A simple call to make is equivalent to make all, meaning any file *.c will be compiled. If this is not desired, calling make X will only compile X.c.

Furthermore, for the TLS examples, you are assumed to have installed wolfSSL in /usr/local/. The makefile will take care of pointing to /usr/local/include/ to find the wolfSSL headers and /usr/local/lib/ to link against the wolfSSL library. This linking is dynamic, however, so /usr/local/lib/ must be in your linker's path. If you are receiving errors at runtime about loading shared libraries, run this command from the terminals you are running the TLS examples from:

export LD_LIBRARY_PATH=$(LD_LIBRARY_PATH):/usr/local/lib

If you have installed wolfSSL in another location, edit Makefile so that the LIB_PATH variable reflects this.

If you have configured wolfSSL to use static linking, comment out the line

LIBS+=$(DYN_LIB)

and uncomment the line

#LIBS+=$(STATIC_LIB)

to statically link against wolfSSL in these examples.

For client-tls-writedup and server-tls-writedup, it is required that wolfSSL be configured with the --enable-writedup flag. Remember to build and install wolfSSL after configuring it with this flag.

We'll begin by making a simple client/server pair that will communicate in plaintext over a TCP socket.

We'll write the server first. We'll be rewriting server-tcp.c from scratch, so there's no reason to copy anything. The finished version can be found here.

To start, let's write up a quick skeleton for this file like this:

/* the usual suspects */
#include <stdlib.h>
#include <stdio.h>
#include <string.h>

/* socket includes */
#include <sys/socket.h>
#include <arpa/inet.h>
#include <netinet/in.h>
#include <unistd.h>

#define DEFAULT_PORT 11111



int main()
{
    int                sockfd;
    int                connd;
    struct sockaddr_in servAddr;
    struct sockaddr_in clientAddr;
    socklen_t          size = sizeof(clientAddr);
    char               buff[256];
    size_t             len;
    int                shutdown = 0;



    /* We'll fill in our work here */



    /* Cleanup and return */
    return 0;               /* Return reporting a success               */
}

Let's go over all that.

"The usual suspects" are some of the usual includes you'd expect to see in a C file. Beyond them, the "socket includes" are all of the includes we need to do socket operations. sys/sockets gives us our standard definitions for sockets, and arpa/inet and netinet/in.h each give us a few functions for dealing with internet sockets.

DEFAULT_PORT is a quick definition for which port to bind to.

There are quite a few variables here. We'll talk about them a bit more in depth when we get to using them. As a quick description, however, sockfd and connd are file descriptors; servAddr, clientAddr, and size are used to describe the address of the server and client; buff and len are for I/O; and finally shutdown is for flow control.

Now we'll set up the sockets.

The next step is to get ahold of a socket for our server. Replace the "We'll fill in our work here" comment with these lines:

    /* Create a socket that uses an internet IPv4 address,
     * Sets the socket to be stream based (TCP),
     * 0 means choose the default protocol. */
    if ((sockfd = socket(AF_INET, SOCK_STREAM, 0)) == -1) {
        fprintf(stderr, "ERROR: failed to create the socket\n");
        return -1;
    }

The comment quickly explains the arguments passed to socket(). Otherwise, socket() returns a -1 on error.

This is also an opportunity to introduce the error convention for these examples: print an error message and end execution. It's important to realise that this is not a graceful way to do this. In the event of an error, we'll often end up ending execution with quite a bit of memory allocated. For the purpose of these examples, however, it was decided that dealing with these errors gracefully would obscure the purposes of the examples. As such, we make a nod at the importance of error handling without complicating the matter by putting in the full amount of effort.

Anyway, now that we've opened a socket, we should update the "Cleanup and return" section at the bottom of main() to close it when we're done. It should look a bit like this now:

    /* Cleanup and return */
    close(sockfd);          /* Close the socket listening for clients   */
    return 0;               /* Return reporting a success               */

Now that we have a socket, let's fill out our address. Just after the "Create a socket [...]" block, add these lines:

    /* Initialize the server address struct with zeros */
    memset(&servAddr, 0, sizeof(servAddr));

    /* Fill in the server address */
    servAddr.sin_family      = AF_INET;             /* using IPv4      */
    servAddr.sin_port        = htons(DEFAULT_PORT); /* on DEFAULT_PORT */
    servAddr.sin_addr.s_addr = INADDR_ANY;          /* from anywhere   */

That "Initialize the server address struct wit zeros" step is not strictly necessary, but it's usually a good idea, and it doesn't complicate the example too much.

With the address all filled out, the final stage of setting up the server is to bind our socket to a port and start listening for connections. In total, this should look like this:

    /* Bind the server socket to our port */
    if (bind(sockfd, (struct sockaddr*)&servAddr, sizeof(servAddr)) == -1) {
        fprintf(stderr, "ERROR: failed to bind\n");
        return -1;
    }

    /* Listen for a new connection, allow 5 pending connections */
    if (listen(sockfd, 5) == -1) {
        fprintf(stderr, "ERROR: failed to listen\n");
        return -1;
    }

bind() makes it such that our server on sockfd is now visible at the location described by servAddr, and listen() causes us to start listening to sockfd for incoming client connections.

And now the setup is complete. Now we just need to deal with I/O.

We're going to keep accepting clients until one of them tells us "shutdown". To start, let's write up a quick skeleton for this part of the code:

    while (!shutdown) {
        printf("Waiting for a connection...\n");

        /* Accept clients here */

        printf("Client connected successfully\n");



        /* Do communication here */



        /* Cleanup after this connection */
    }

    printf("Shutdown complete\n");

We'll deal with accepting clients first. Replace the "Accept clients here" comment with these lines:

        /* Accept client connections */
        if ((connd = accept(sockfd, (struct sockaddr*)&clientAddr, &size))
            == -1) {
            fprintf(stderr, "ERROR: failed to accept the connection\n\n");
            return -1;
        }

This call will block execution until a client connects to our server. At which point, we'll get a connection to the client through connd. Now that we've opened a new file, we should close it when we're done with it. Update the "Cleanup after this connection" section at the bottom of the loop to close connd. It should look a bit like this now:

        /* Cleanup after this connection */
        close(connd);           /* Close the connection to the client   */

Now that we have a connection to the client, we can do communication.

First, we'll read from the client. Replace the "Do communication here" comment with these lines:

        /* Read the client data into our buff array */
        memset(buff, 0, sizeof(buff));
        if (read(connd, buff, sizeof(buff)-1) == -1) {
            fprintf(stderr, "ERROR: failed to read\n");
            return -1;
        }

        /* Print to stdout any data the client sends */
        printf("Client: %s\n", buff);

This zeros out the buffer, then gets a message from the client and prints it to stdout so we can see it. Recall that we want to shutdown when the client tells us "shutdown". To accomplish this, add these lines after we print the message:

        /* Check for server shutdown command */
        if (strncmp(buff, "shutdown", 8) == 0) {
            printf("Shutdown command issued!\n");
            shutdown = 1;
        }

Note that this block doesn't end execution or break out of the loop right away. We still want to respond to the client with our own message.

After reading the message from the client, we first write our message into the buffer like this:

        /* Write our reply into buff */
        memset(buff, 0, sizeof(buff));
        memcpy(buff, reply, strlen(reply));
        len = strnlen(buff, sizeof(buff));

And then we send it to the client like this:

        /* Reply back to the client */
        if (write(connd, buff, len) != len) {
            fprintf(stderr, "ERROR: failed to write\n");
            return -1;
        }

And we're done.

We've set up a server on a TCP socket and dealt with a quick back-and-forth with a client.

Now we just need a client.

Now we'll write the client. We'll be rewriting client-tcp.c from scratch, so there's no reason to copy anything. The finished version can be found here.

Once more, we'll start with a skeleton:

/* the usual suspects */
#include <stdlib.h>
#include <stdio.h>
#include <string.h>

/* socket includes */
#include <sys/socket.h>
#include <arpa/inet.h>
#include <netinet/in.h>
#include <unistd.h>

#define DEFAULT_PORT 11111



int main(int argc, char** argv)
{
    int                sockfd;
    struct sockaddr_in servAddr;
    char               buff[256];
    size_t             len;



    /* We'll fill in our work here */



    /* Cleanup and return */
    return 0;               /* Return reporting a success               */
}

It looks quite similar to the server's skeleton, but as you can see, there are way fewer variables. Similarly, we're taking in command line arguments this time rather than ignoring them.

This'll be our first step: verify that the program has been called correctly. There are more sophisticated ways of doing this, but we'll use a simple solution. Replace the "We'll fill in our work here" comment with the following lines:

    /* Check for proper calling convention */
    if (argc != 2) {
        printf("usage: %s <IPv4 address>\n", argv[0]);
        return 0;
    }

We expect exactly two arguments: the program's name and an IPv4 address. We won't be confirming that the second argument is a well formed IPv4 address, though. If it's not, we'll error out midway through.

Now that we know the program has been called more-or-less correctly, we'll open a socket to connect to the server through.

    /* Create a socket that uses an internet IPv4 address,
     * Sets the socket to be stream based (TCP),
     * 0 means choose the default protocol. */
    if ((sockfd = socket(AF_INET, SOCK_STREAM, 0)) == -1) {
        fprintf(stderr, "ERROR: failed to create the socket\n");
        return -1;
    }

This is just like the call we make in the server code. Similarly, we should remember to close the socket at the bottom of main(). Make the "Cleanup and return" section look a bit like this:

    /* Cleanup and return */
    close(sockfd);          /* Close the connection to the server       */
    return 0;               /* Return reporting a success               */

Now we can fill out the address of the server we want to connect to. After the "Create a socket [...]" block, add these lines:

    /* Initialize the server address struct with zeros */
    memset(&servAddr, 0, sizeof(servAddr));

    /* Fill in the server address */
    servAddr.sin_family = AF_INET;             /* using IPv4      */
    servAddr.sin_port   = htons(DEFAULT_PORT); /* on DEFAULT_PORT */

    /* Get the server IPv4 address from the command line call */
    if (inet_pton(AF_INET, argv[1], &servAddr.sin_addr) != 1) {
        fprintf(stderr, "ERROR: invalid address\n");
        return -1;
    }

Once more, this is quite similar to server code. This time, however, rather than setting servAddr.sin_addr.s_addr to INADDR_ANY, we're going to make a call to inet_pton() to read argv[1] as an IPv4 address and assign it to the right place in servAddr. If argv[1] is well formed, inet_pton() will return a 1. Otherwise, we have an error.

Now we can connect to the server. Add these lines next:

    /* Connect to the server */
    if (connect(sockfd, (struct sockaddr*) &servAddr, sizeof(servAddr))
        == -1) {
        fprintf(stderr, "ERROR: failed to connect\n");
        return -1;
    }

connect() will block until the connection is successful. If something goes wrong, it returns a -1 that we'll catch as a fatal error.

After that, we're done with setup. We can now have our back-and-forth with the server. Just to be fun, let's quickly get a message from stdin like this:

    /* Get a message for the server from stdin */
    printf("Message for server: ");
    memset(buff, 0, sizeof(buff));
    fgets(buff, sizeof(buff), stdin);
    len = strnlen(buff, sizeof(buff));

We know that our server is expecting us to say something first, so let's send the message like this:

    /* Send the message to the server */
    if (write(sockfd, buff, len) != len) {
        fprintf(stderr, "ERROR: failed to write\n");
        return -1;
    }

We also know that our server will then send us a reply. Let's write that down and print it to stdout so we can see it:

    /* Read the server data into our buff array */
    memset(buff, 0, sizeof(buff));
    if (read(sockfd, buff, sizeof(buff)-1) == -1) {
        fprintf(stderr, "ERROR: failed to read\n");
        return -1;
    }

    /* Print to stdout any data the server sends */
    printf("Server: %s\n", buff);

And that's everything! Our client will just be a quick one-and-done thing.

server-tcp can be connected to by the following:

  • client-tcp

client-tcp can connect to the following:

  • server-tcp

Now that we have a TCP client/server pair, we can modify them to use TLS 1.2 to secure their communication.

We'll modify the server first. Copy server-tcp.c to a new file, server-tls.c, that we will modify. The finished version can be found here.

The first thing to do is include the wolfSSL header. Just below the "socket includes" block, add these lines:

/* wolfSSL */
#include <wolfssl/ssl.h>

We're also going to need a few more defines for our file paths. Just below the definition of DEFAULT_PORT add these lines:

#define CERT_FILE "../certs/server-cert.pem"
#define KEY_FILE  "../certs/server-key.pem"

Before continuing on, make sure that these are the correct paths to those files.

Now, inside of main, we need two more variables. At the top of main(), add these lines after the other variable declarations:

    /* declare wolfSSL objects */
    WOLFSSL_CTX* ctx;
    WOLFSSL*     ssl;

And now we can start hooking up wolfSSL. The first thing to do is initialize the wolfSSL library. Beneath the variable declarations add the lines:

    /* Initialize wolfSSL */
    wolfSSL_Init();

Next we'll set up the wolfSSL context. After the "Create a socket [...]" block these lines:

    /* Create and initialize WOLFSSL_CTX */
    if ((ctx = wolfSSL_CTX_new(wolfTLSv1_2_server_method())) == NULL) {
        fprintf(stderr, "ERROR: failed to create WOLFSSL_CTX\n");
        return -1;
    }

The call to wolfTLSv1_2_server_method() inside the call to wolfSSL_CTX_new() is what makes us use the TLS 1.2 protocol.

Just like we have to close sockfd when we're done, we'll have to cleanup wolfSSL. Edit the "Cleanup and return" section at the bottom of main() to look something like this:

    /* Cleanup and return */
    wolfSSL_CTX_free(ctx);  /* Free the wolfSSL context object          */
    wolfSSL_Cleanup();      /* Cleanup the wolfSSL environment          */
    close(sockfd);          /* Close the socket listening for clients   */
    return 0;               /* Return reporting a success               */

Though we're not done with ctx yet: we have to load in our certificates. After the "Create and initialize WOLFSSL_CTX" block, add these lines:

    /* Load server certificates into WOLFSSL_CTX */
    if (wolfSSL_CTX_use_certificate_file(ctx, CERT_FILE, SSL_FILETYPE_PEM)
        != SSL_SUCCESS) {
        fprintf(stderr, "ERROR: failed to load %s, please check the file.\n",
                CERT_FILE);
        return -1;
    }

Recall that CERT_FILE is one of our own defines.

What the above does is load the certificate file CERT_FILE so that the server can verify itself to clients. This certificate is known to be in the PEM format (indicated by SSL_FILETYPE_PEM), and this information is then kept in ctx.

Next we need the server to load its private key:

    /* Load server key into WOLFSSL_CTX */
    if (wolfSSL_CTX_use_PrivateKey_file(ctx, KEY_FILE, SSL_FILETYPE_PEM)
        != SSL_SUCCESS) {
        fprintf(stderr, "ERROR: failed to load %s, please check the file.\n",
                KEY_FILE);
        return -1;
    }

Recall that KEY_FILE is one of our own defines.

Similar to the call to wolfSSL_CTX_use_certificate_file() previously, this loads a private key from KEY_FILE of the format SSL_FILETYPE_PEM, and stores the information into ctx.

And that's everything for the setup phase. Now we just need to add a few things to how we handle clients.

After we accept a new client, we need to make a wolfSSL object. Just after the "Accept client connections" block, add these lines:

        /* Create a WOLFSSL object */
        if ((ssl = wolfSSL_new(ctx)) == NULL) {
            fprintf(stderr, "ERROR: failed to create WOLFSSL object\n");
            return -1;
        }

        /* Attach wolfSSL to the socket */
        wolfSSL_set_fd(ssl, connd);

This allocates a new wolfSSL object. It will inherit all of the files we registered to ctx. After that, we give ssl the file descriptor connd to hold onto.

We can't forget to free this object either. Update the "Cleanup after this connection" section at the bottom of the loop to look like this:

        /* Cleanup after this connection */
        wolfSSL_free(ssl);      /* Free the wolfSSL object              */
        close(connd);           /* Close the connection to the client   */

From here on, ssl is our new connd for purposes of communicating with clients.

First, this means that we need to replace calls to read() with calls to wolfSSL_read() and replace argument connd with ssl. The "Read the client data [...]" block now should look like this:

        /* Read the client data into our buff array */
        memset(buff, 0, sizeof(buff));
        if (wolfSSL_read(ssl, buff, sizeof(buff)-1) == -1) {
            fprintf(stderr, "ERROR: failed to read\n");
            return -1;
        }

Second, we need to do something similar to our calls to write(). The "Reply back to the client" block now should look like this:

        /* Reply back to the client */
        if (wolfSSL_write(ssl, buff, len) != len) {
            fprintf(stderr, "ERROR: failed to write\n");
            return -1;
        }

And after this, we've successfully modified our TCP server to use wolfSSL for TLS 1.2 connections. Now we just need a client to connect with.

Now we'll write the client. Copy client-tcp.c to a new file, client-tls.c, that we will modify. The finished version can be found here.

Like the server, the first thing to do is include the wolfSSL header. Just below the "socket includes" block, add these lines:

/* wolfSSL */
#include <wolfssl/ssl.h>

We're going to need a different definition for CERT_FILE, however. Just below the definition of DEFAULT_PORT add this line:

#define CERT_FILE "../certs/ca-cert.pem"

Before continuing on, make sure that this is the correct path to that file.

Now, inside of main, we need two more variables. At the top of main(), add these lines after the other variable declarations:

    /* declare wolfSSL objects */
    WOLFSSL_CTX* ctx;
    WOLFSSL*     ssl;

And now we can start hooking up wolfSSL. The first thing to do is initialize the wolfSSL library. This time we'll do it after we check our calling convention. Beneath the "Check for proper calling convention" block, add the lines:

    /* Initialize wolfSSL */
    wolfSSL_Init();

Next we set up ctx. After the "Create a socket [...]" block, add these lines:

    /* Create and initialize WOLFSSL_CTX */
    if ((ctx = wolfSSL_CTX_new(wolfTLSv1_2_client_method())) == NULL) {
        fprintf(stderr, "ERROR: failed to create WOLFSSL_CTX\n");
        return -1;
    }

    /* Load client certificates into WOLFSSL_CTX */
    if (wolfSSL_CTX_load_verify_locations(ctx, CERT_FILE, NULL)
        != SSL_SUCCESS) {
        fprintf(stderr, "ERROR: failed to load %s, please check the file.\n",
                CERT_FILE);
        return -1;
    }

That "Create and initialize WOLFSSL_CTX" block should be familiar from the server code.

The "Load client certificates into WOLFSSL_CTX" block is how we load CERT_FILE into ctx so that we can verify the certificate of the servers we connect to. It should be noted that all files are assumed to be in the PEM format.

Before continuing, we should remember to update our "Cleanup and return" block at the bottom of main() to free ctx and cleanup wolfSSL. It should now look like this:

    /* Cleanup and return */
    wolfSSL_CTX_free(ctx);  /* Free the wolfSSL context object          */
    wolfSSL_Cleanup();      /* Cleanup the wolfSSL environment          */
    close(sockfd);          /* Close the connection to the server       */
    return 0;               /* Return reporting a success               */

And this concludes the setup. Now we just need to deal with the connection to the server. After the "Connect to the server" block, add these lines:

    /* Create a WOLFSSL object */
    if ((ssl = wolfSSL_new(ctx)) == NULL) {
        fprintf(stderr, "ERROR: failed to create WOLFSSL object\n");
        return -1;
    }

    /* Attach wolfSSL to the socket */
    wolfSSL_set_fd(ssl, sockfd);

    /* Connect to wolfSSL on the server side */
    if (wolfSSL_connect(ssl) != SSL_SUCCESS) {
        fprintf(stderr, "ERROR: failed to connect to wolfSSL\n");
        return -1;
    }

Those first two blocks should be familiar from the server code: we're making an SSL object from the context and giving it our connection to the server.

The third block, "Connect to wolfSSL on the server side", is new. It is technically optional; If we don't call it, the first time we try to read or write though ssl it will be invoked anyway. We're going to call it ourself to make things easier when modifying this file to add different features in the future.

Before continuing, we should remember to update our "Cleanup and return" block at the bottom of main() to free ssl. It should now look like this:

    /* Cleanup and return */
    wolfSSL_free(ssl);      /* Free the wolfSSL object                  */
    wolfSSL_CTX_free(ctx);  /* Free the wolfSSL context object          */
    wolfSSL_Cleanup();      /* Cleanup the wolfSSL environment          */
    close(sockfd);          /* Close the connection to the server       */
    return 0;               /* Return reporting a success               */

Now all we have to do is update our write() and read() calls to use wolfSSL. The "Send the message to the server" block should now look like this:

    /* Send the message to the server */
    if (wolfSSL_write(ssl, buff, len) != len) {
        fprintf(stderr, "ERROR: failed to write\n");
        return -1;
    }

And the "Read the server data [...]" block should look like this:

    /* Read the server data into our buff array */
    memset(buff, 0, sizeof(buff));
    if (wolfSSL_read(ssl, buff, sizeof(buff)-1) == -1) {
        fprintf(stderr, "ERROR: failed to read\n");
        return -1;
    }

And we're done. We should now have a fully functional TLS client.

server-tls can be connected to by the following:

  • client-tls
  • client-tls-callback
  • client-tls-nonblocking
  • client-tls-resume
  • client-tls-writedup

client-tls can connect to the following:

  • server-tls
  • server-tls-callback
  • server-tls-nonblocking
  • server-tls-threaded
  • server-tls-writedup

The edits required to make wolfSSL use custom functions for getting I/O from a socket are identical between the client and server code. As such, pick whichever *-tls.c you want and copy it to *-tls-callback.c. The finished version of the server can be found here, and the finished version of the client can be found here.

The first step is to add errno.h to the list of "usual suspects". This block should now look like this:

/* the usual suspects */
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <errno.h>

Next, we're going to need write our callbacks functions. These functions will be called by wolfSSL whenever it need to read or write data through the sockets. These callbacks can be quite sophisticated, but we're going to write simple callbacks that tell us how many bytes they read or write.

First, the read callback:

int my_IORecv(WOLFSSL* ssl, char* buff, int sz, void* ctx)
{
    /* By default, ctx will be a pointer to the file descriptor to read from.
     * This can be changed by calling wolfSSL_SetIOReadCtx(). */
    int sockfd = *(int*)ctx;
    int recvd;


    /* Receive message from socket */
    if ((recvd = recv(sockfd, buff, sz, 0)) == -1) {
        /* error encountered. Be responsible and report it in wolfSSL terms */

        fprintf(stderr, "IO RECEIVE ERROR: ");
        switch (errno) {
        #if EAGAIN != EWOULDBLOCK
        case EAGAIN: /* EAGAIN == EWOULDBLOCK on some systems, but not others */
        #endif
        case EWOULDBLOCK:
            if (!wolfSSL_dtls(ssl) || wolfSSL_get_using_nonblock(ssl)) {
                fprintf(stderr, "would block\n");
                return WOLFSSL_CBIO_ERR_WANT_READ;
            }
            else {
                fprintf(stderr, "socket timeout\n");
                return WOLFSSL_CBIO_ERR_TIMEOUT;
            }
        case ECONNRESET:
            fprintf(stderr, "connection reset\n");
            return WOLFSSL_CBIO_ERR_CONN_RST;
        case EINTR:
            fprintf(stderr, "socket interrupted\n");
            return WOLFSSL_CBIO_ERR_ISR;
        case ECONNREFUSED:
            fprintf(stderr, "connection refused\n");
            return WOLFSSL_CBIO_ERR_WANT_READ;
        case ECONNABORTED:
            fprintf(stderr, "connection aborted\n");
            return WOLFSSL_CBIO_ERR_CONN_CLOSE;
        default:
            fprintf(stderr, "general error\n");
            return WOLFSSL_CBIO_ERR_GENERAL;
        }
    }
    else if (recvd == 0) {
        printf("Connection closed\n");
        return WOLFSSL_CBIO_ERR_CONN_CLOSE;
    }

    /* successful receive */
    printf("my_IORecv: received %d bytes from %d\n", sz, sockfd);
    return recvd;
}

Next the write callback:

int my_IOSend(WOLFSSL* ssl, char* buff, int sz, void* ctx)
{
    /* By default, ctx will be a pointer to the file descriptor to write to.
     * This can be changed by calling wolfSSL_SetIOWriteCtx(). */
    int sockfd = *(int*)ctx;
    int sent;


    /* Receive message from socket */
    if ((sent = send(sockfd, buff, sz, 0)) == -1) {
        /* error encountered. Be responsible and report it in wolfSSL terms */

        fprintf(stderr, "IO SEND ERROR: ");
        switch (errno) {
        #if EAGAIN != EWOULDBLOCK
        case EAGAIN: /* EAGAIN == EWOULDBLOCK on some systems, but not others */
        #endif
        case EWOULDBLOCK:
            fprintf(stderr, "would block\n");
            return WOLFSSL_CBIO_ERR_WANT_WRITE;
        case ECONNRESET:
            fprintf(stderr, "connection reset\n");
            return WOLFSSL_CBIO_ERR_CONN_RST;
        case EINTR:
            fprintf(stderr, "socket interrupted\n");
            return WOLFSSL_CBIO_ERR_ISR;
        case EPIPE:
            fprintf(stderr, "socket EPIPE\n");
            return WOLFSSL_CBIO_ERR_CONN_CLOSE;
        default:
            fprintf(stderr, "general error\n");
            return WOLFSSL_CBIO_ERR_GENERAL;
        }
    }
    else if (sent == 0) {
        printf("Connection closed\n");
        return 0;
    }

    /* successful send */
    printf("my_IOSend: sent %d bytes to %d\n", sz, sockfd);
    return sent;
}

The most complex part of these callbacks are probably their error reporting. wolfSSL expects more error information that we've been doing in these examples until now. As such, to be responsible we have to translate the errno value into terms that wolfSSL expects form its I/O functions.

But that's the hard part. Now all that's left is to register these functions with wolfSSL. We can do this at any time between calling wolfSSL_CTX_new() to get ctx and wolfSSL_new() to get ssl. After we've handled loading certificate or key files but before the "Initialize the server address struct with zeros" block is a good place. After this, add these lines:

    /* Register callbacks */
    wolfSSL_SetIORecv(ctx, my_IORecv);
    wolfSSL_SetIOSend(ctx, my_IOSend);

And just like that wolfSSL will use our functions to send and receive data. Now when this program is run we should see a number of "my_OISend: sent" and "my_IORecv: received" lines in our output.

server-tls-callback can be connected to by the following:

  • client-tls
  • client-tls-callback
  • client-tls-nonblocking
  • client-tls-resume
  • client-tls-writedup

client-tls-callback can connect to the following:

  • server-tls
  • server-tls-callback
  • server-tls-nonblocking
  • server-tls-threaded
  • server-tls-writedup

We can also have wolfSSL use ECC keys.

We'll modify the server first. Copy server-tls.c to a new file, server-tls-ecdhe.c, that we will modify. The finished version can be found here.

The first change will be to our file locations. Change the defines for CERT_FILE and KEY_FILE to the following:

#define CERT_FILE "../certs/server-ecc.pem"
#define KEY_FILE  "../certs/ecc-key.pem"

Furthermore, we'll want to define CIPHER_LIST. This will be a list of all ciphers we support on this server. The define will look like this:

#define CIPHER_LIST "ECDHE-ECDSA-CHACHA20-POLY1305"

Finally, all we need to do is set this cipher list. Just after the "Load server key into WOLFSSL_CTX" block, add these lines:

    /* Set cipher list */
    if (wolfSSL_CTX_set_cipher_list(ctx, CIPHER_LIST) != SSL_SUCCESS) {
        fprintf(stderr, "ERROR: failed to set cipher list\n");
        return -1;
    }

And just like that, we're done!

This is one of those cases where we have more to change about the client than the server. Copy client-tls.c to a new file, client-tls-ecdhe.c, that we will modify. The finished version can be found here.

We start by changing the definition of CERT_FILE. It should now look like this:

#define CERT_FILE "../certs/server-ecc.pem"

After this, we need two more files and a cipher list. In total, these new defines will look like this:

#define ECC_FILE    "../certs/client-ecc-cert.pem"
#define KEY_FILE    "../certs/ecc-client-key.pem"
#define CIPHER_LIST "ECDHE-ECDSA-CHACHA20-POLY1305"

With these defined, we just need to load them into ctx. We're already doing this for CERT_FILE in the "Load client certificates into WOLFSSL_CTX" block, so add these blocks just after that:

    /* Load client ecc certificates into WOLFSSL_CTX */
    if (wolfSSL_CTX_use_certificate_chain_file(ctx, ECC_FILE) != SSL_SUCCESS) {
        fprintf(stderr, "ERROR: failed to load %s, please check the file.\n",
                ECC_FILE);
        return -1;
    }

    /* Load client ecc key into WOLFSSL_CTX */
    if (wolfSSL_CTX_use_PrivateKey_file(ctx, KEY_FILE, SSL_FILETYPE_PEM)
        != SSL_SUCCESS) {
        fprintf(stderr, "ERROR: failed to load %s, please check the file.\n",
                KEY_FILE);
        return -1;
    }

    /* Set cipher list */
    if (wolfSSL_CTX_set_cipher_list(ctx, CIPHER_LIST) != SSL_SUCCESS) {
        fprintf(stderr, "ERROR: failed to set cipher list\n");
        return -1;
    }

And now the client is set up.

server-tls-ecdhe can be connected to by the following:

  • client-tls-ecdhe

client-tls-ecdhe can connect to the following:

  • server-tls-ecdhe

Crypto Callbacks

See the client-tls-cryptocb.c example for demonstrating the --enable-cryptocb feature for allowing custom cryptographic algorithm offload.

TLS v1.3 Wireshark Logging

Build wolfSSL with HAVE_SECRET_CALLBACK included:

./configure --enable-tls13 CFLAGS="-DHAVE_SECRET_CALLBACK" && make && sudo make install

In wolfssl-examples/tls:

make clean && make
./server-tls13 &
./client-tls13 127.0.0.1

# Execute client-tls13 again with the message "shutdown" in order to end the execution of the server.

Wireshark can decode traffic using the created "sslkeylog.log". To configure in Wireshark Prferences go to Protocols -> TLS. In the "(Pre)-Master-Secret log filename" choose the "sslkeylog.log" file in this directory. Capture TLS traffic and all packets will be decrypted (handshake and application data). To see application data decrypted you may have to right-click on packet click "Follow" -> "TLS Stream".

TLS over UART Example

Test setup uses two USB UART adapters wired in cross-over to PC.

In separate terminals run:

./server-tls-uart /dev/ttyUSB0 ./client-tls-uart /dev/ttyUSB1

Example output:

./server-tls-uart /dev/ttyUSB0
Waiting for client
TLS Accept handshake done
Read (0): Testing 1, 2 and 3
Sent (0): Testing 1, 2 and 3
$ ./client-tls-uart /dev/ttyUSB1
TLS Connect handshake done
Sending test string
Sent (0): Testing 1, 2 and 3
Read (0): Testing 1, 2 and 3

TLS Example with PK Callbacks and optionally Async

See server-tls-pkcallback.c and client-tls-pkcallback.c.

The top of both files have an optional build setting to gate use of ECC/RSA and TLS v1.2 or TLS v1.3:

#define USE_ECDHE_ECDSA
#define USE_TLSV13

For this example it uses the --enable-pkcallbacks or HAVE_PK_CALLBACKS feature. Optionally if using the --enable-asynccrypt option and wolfSSL Async code this also shows being able to return WC_PENDING_E while the asymmetric key operation is ocurring.

./configure --enable-pkcallbacks [--enable-asynccrypt]
make
sudo make install

Example Output:

./server-tls-pkcallback
Waiting for a connection...
PK ECC Sign: inSz 32, keySz 121
PK ECC Sign: Async Simulate
PK ECC Sign: inSz 32, keySz 121
PK ECC Sign: Curve ID 7
PK ECC Sign: ret 0 outSz 72
Client connected successfully
Client: test

Shutdown complete
Waiting for a connection...
% ./client-tls-pkcallback 127.0.0.1
PK ECC Sign: inSz 32, keySz 121
PK ECC Sign: Async Simulate
PK ECC Sign: inSz 32, keySz 121
PK ECC Sign: Curve ID 7
PK ECC Sign: ret 0 outSz 71
Message for server: test
Server: I hear ya fa shizzle!

To generate your own cert text, see the DER to C script.


Encrypted Client Hello (ECH) encrypts sensitive fields in the client hello step of the TLS handshake. The client-ech example connects to a cloudflare server that is setup to test different TLS options including ECH. To build wolfSSL for this ech example run ./configure --enable-ech && make && sudo make install.

This test is successful if the cloudflare http response shows that sni=encrypted.

make
./client-ech 
HTTP/1.1 200 OK
Access-Control-Allow-Origin: *
Cache-Control: no-cache
Cf-Ray: 77c3e3e937c6b08e-ATL
Content-Type: text/plain
Expires: Thu, 01 Jan 1970 00:00:01 GMT
Server: cloudflare
X-Content-Type-Options: nosniff
X-Frame-Options: DENY
Date: Mon, 19 Dec 2022 23:24:11 GMT
Transfer-Encoding: chunked

106
fl=507f46
h=crypto.cloudflare.com
ip=173.93.184.37
ts=1671492251.082
visit_scheme=https
uag=Mozilla/5.0 (X11; Linux x86_64; rv:105.0) Gecko/20100101 Firefox/105.0
colo=ATL
sliver=none
http=http/1.1
loc=US
tls=TLSv1.3
sni=encrypted
warp=off
gateway=off
kex=P-256

0

Support

Please contact wolfSSL at support@wolfssl.com with any questions, bug fixes, or suggested feature additions.