Embedding CivetWeb

Embedded C/C++ web server


Embedding CivetWeb

CivetWeb is primarily designed so applications can easily add HTTP and HTTPS server as well as WebSocket (WS and WSS) server functionality. For example, a C/C++ application could use CivetWeb to enable a web service and configuration interface, to add a HTML5 data visualization interface, for automation or remote control, as a protocol gateway or as a HTTP/WebSocket client for firewall traversal. Often the easiest way to embedd CivetWeb is to add the civetweb.c file into your existing C project (see below).

CivetWeb can also be used as a stand-alone executable. It can deliver static files and offers built-in server side Lua, JavaScript and CGI support. Some instructions how to build the stand-alone server can be found in Building.md.

Files

There is just a small set of files to compile in to the application, but if a library is desired, see Building.md

Regarding the INL file extension

The INL file extension represents code that is statically included inline in a source file. Slightly different from C++ where it means “inline” code which is technically not the same as static code. CivetWeb overloads this extension for the sake of clarity as opposed to having .c extensions on files that should not be directly compiled.

HTTP Server Source Files

These files constitute the CivetWeb library. They do not contain a main function, but all functions required to run a HTTP server.

  • HTTP server API
    • include/civetweb.h
  • C implementation
    • src/civetweb.c
    • src/md5.inl (MD5 calculation)
    • src/sha1.inl (SHA calculation)
    • src/handle_form.inl (HTML form handling functions)
    • src/response.inl (helper for generating HTTP response headers)
    • src/timer.inl (optional timer support)
    • src/http2.inl (optional HTTP2 support)
  • Optional: C++ wrapper
    • include/CivetServer.h (C++ interface)
    • src/CivetServer.cpp (C++ wrapper implementation)
  • Optional: Third party components
    • src/third_party/* (third party components, mainly used for the standalone server)
    • src/mod_*.inl (modules to access third party components from civetweb)

Note: The C++ wrapper uses the official C interface (civetweb.h) without adding any features to the server itself. Several features available in the C interface are missing in the C++ interface. While all features should be accessible using the C interface, this is not a design goal of the C++ interface. New code is advised to use the C interface, since this is unit tested and new API functions are often only added there.

Additional Source Files for Executables

These files can be used to build a server executable. They contain a main function starting the HTTP server.

  • Stand-alone C server
    • src/main.c
  • Reference embedded C server
    • examples/embedded_c/embedded_c.c
  • Reference embedded C++ server
    • examples/embedded_cpp/embedded_cpp.cpp

Note: The “embedded” example is actively maintained, updated, extended and tested. Other examples in the examples/ folder might be outdated and remain there for reference.

Quick Start

By default, the server will automatically serve up files like a normal HTTP server. An embedded server is most likely going to overload this functionality.

C

  • Include the C interface civetweb.h.
  • Use mg_start() to start the server.
    • Use options to select the port and document root among other things.
    • Use callbacks to add your own hooks.
  • Use mg_set_request_handler() to easily add your own request handlers.
  • Use mg_stop() to stop the server.

C++

  • Note that CivetWeb is Clean C, and C++ interface CivetServer.h is only a wrapper layer around the C interface. Not all CivetWeb features available in C are also available in C++.
  • Create CivetHandlers for each URI.
  • Register the handlers with CivetServer::addHandler()
  • CivetServer starts on construction and stops on destruction.
  • Use constructor options to select the port and document root among other things.
  • Use constructor callbacks to add your own hooks.

Alternative quick start: Have a look at the examples embedded_c and embedded_cpp

Feature Selection

CivetWeb is highly customizable at build time, in addition to configuration at start time.

start time options

Start time options are passed to mg_start. They are documented in the UserManual.md.

callbacks

Pointers to callback functions are passed to mg_start as well. They are documented in civetweb.h and the callbacks API documentation.

compiler defines

Several features can be turned “on” or “off” by setting compile defines. CivetWeb builds with a reasonable default feature set. Optional features not including in the default can be added by adding a USE\_<feature> define. Default features can be removed by adding a NO_<feature> define. E.g., to build with Lua support, set #define USE_LUA (-DUSE_LUA), to build without CGI support set #define NO_CGI (-DNO_CGI). A list of feature defines is available in Building.md - some versions may have additional, undocumented feature defines. Undocumented defines may become unavailable in future versions without notice.

externally provided functions

In some special cases, it might be meaningful to completely replace an internal function in civetweb.c with your own implementation. Since CivetWeb is free and open source software covered by the MIT license, you can feel free to just edit civetweb.c according to your needs. However, this might be annoying when updating the server, pulling new features or bug fixes from the main repository. For some selected functions, it is possible to provide your own implementation using a MG_EXTERNAL_FUNCTION_<internal_function_name> define. For details on this mechanism, please look directly into the source code civetweb.c. Interfaces and even names of internal functions may change without notice - when you use these defines, you have to check this every time you update CivetWeb. It might still be less effort than to apply your patches every time. This customization option is currently in an evaluation phase. In case you need additional function defines, please create an issue on GitHub explaining your use case, to discuss if this would be an appropriate solution - in general, other customization options are preferred.

Stack Sizes

Stack sizes are usually nothing you need to worry about when running on a system with sufficient memory, such as a desktop PC running Windows or Linux. CivetWeb will use the default stack size on those system (typically 1 MB per Windows, up to 8 MB for Linux) - for each thread. CivetWeb uses one thread for each HTTP connection, so in case a request handler uses a blocking, only this connection will be blocked, while others are not affected. The number of threads can be configured (see UserManual.md). This number also defines the limit for the number of threads that can be handled simultaneously - additional requests can be queues, but are not processed. Since HTTP clients as web browsers tend to open multiple connections to the same server when loading one page, using less then five to ten threads may cause delays, even when there is only one user using one browser. The total amount of virtual memory required for all stacks is num_threads multiplied by the stack size per thread (e.g., 50 MB for 50 threads on Windows). This virtual memory (more precisely: reserved virtual address space) will only require “real” physical memory, when the stack is really used up to this level. If one stack in the entire process exceeds its virtual address space limit, the entire process crashes. Thus, for a system with sufficient resources, using large stacks with big reserves is advisable.

However, for small embedded devices 50 MB may already be a lot, in particular if they immediately commit physical memory for all virtual memory allocations. To limit the stack size, the define USE_STACK_SIZE can be set. Note that stack (as all memory) comes in pages of 4096 bytes (for X86 CPUs and may others), so USE_STACK_SIZE must always be an integer multiple of this page size. In case no additional features are used, 4 pages (16 kB) of stack for each thread could be sufficient. Using lower stack sizes may require to reduce some buffer sizes (e.g. MG_BUF_LEN). (Note: There is no warranty for these numbers, neither for this, nor for any past or future versions.)

Lua Support

Lua is a server side include functionality. Files ending in .lua will be processed with Lua.

Add the following CFLAGS
  • -DLUA_COMPAT_ALL
  • -DUSE_LUA
  • -DUSE_LUA_SQLITE3
  • -DUSE_LUA_FILE_SYSTEM
Add the following sources
  • src/mod_lua.inl
  • src/third_party/lua-5.2.4/src
    • lapi.c
    • lauxlib.c
    • lbaselib.c
    • lbitlib.c
    • lcode.c
    • lcorolib.c
    • lctype.c
    • ldblib.c
    • ldebug.c
    • ldo.c
    • ldump.c
    • lfunc.c
    • lgc.c
    • linit.c
    • liolib.c
    • llex.c
    • lmathlib.c
    • lmem.c
    • loadlib.c
    • lobject.c
    • lopcodes.c
    • loslib.c
    • lparser.c
    • lstate.c
    • lstring.c
    • lstrlib.c
    • ltable.c
    • ltablib.c
    • ltm.c
    • lundump.c
    • lvm.c
    • lzio.c
  • src/third_party/sqlite3.c
  • src/third_party/sqlite3.h
  • src/third_party/lsqlite3.c
  • src/third_party/lfs.c
  • src/third_party/lfs.h

This build is valid for Lua version Lua 5.2. It is also possible to build with Lua 5.1 (including LuaJIT), Lua 5.3 or Lua 5.4.

JavaScript Support

CivetWeb can be built with server side JavaScript support by including the Duktape library.

CivetWeb internals

CivetWeb is multithreaded web server. mg_start() function allocates web server context (struct mg_context), which holds all information about web server instance:

  • configuration options. Note that CivetWeb makes internal copies of passed options.
  • SSL context, if any
  • user-defined callbacks
  • opened listening sockets
  • a queue for accepted sockets
  • mutexes and condition variables for inter-thread synchronization

When mg_start() returns, all initialization is guaranteed to be complete (e.g. listening ports are opened, SSL is initialized, etc). mg_start() starts some threads: a master thread, that accepts new connections, and several worker threads, that process accepted connections. The number of worker threads is configurable via num_threads configuration option. That number puts a limit on number of simultaneous requests that can be handled by CivetWeb. If you embed CivetWeb into a program that uses SSL outside CivetWeb as well, you may need to initialize SSL before calling mg_start(), and set the pre- processor define SSL_ALREADY_INITIALIZED. This is not required if SSL is used only within CivetWeb.

When master thread accepts new a connection, a new accepted socket (described by struct socket) it placed into the accepted sockets queue, which has size of MGSQLEN (default 20). Any idle worker thread can grab accepted sockets from that queue. If all worker threads are busy, master thread can accept and queue up to 20 more TCP connections, filling up the queue. In the attempt to queue even more accepted connection, the master thread blocks until there is space in the queue. When the master thread is blocked on a full queue, the operating system can also queue incoming connection. The number is limited by the listen() call parameter, which is SOMAXCONN and depends on the platform.

Worker threads are running in an infinite loop, which in a simplified form looks something like this:

    static void *worker_thread() {
      while (consume_socket()) {
        process_new_connection();
      }
    }

Function consume_socket() gets a new accepted socket from the CivetWeb socket queue, atomically removing it from the queue. If the queue is empty, consume_socket() blocks and waits until a new socket is placed in the queue by the master thread.

process_new_connection() actually processes the connection, i.e. reads the request, parses it, and performs appropriate action depending on the parsed request.

Master thread uses poll() and accept() to accept new connections on listening sockets. poll() is used to avoid FD_SETSIZE limitation of select(). Since there are only a few listening sockets, there is no reason to use hi-performance alternatives like epoll() or kqueue(). Worker threads use blocking IO on accepted sockets for reading and writing data. All accepted sockets have SO_RCVTIMEO and SO_SNDTIMEO socket options set (controlled by the request_timeout_ms CivetWeb option, 30 seconds default) which specifies a read/write timeout on client connections.

A minimal example

Initializing a HTTP server

{
    /* Server context handle */
    struct mg_context *ctx;

    /* Initialize the library */
    mg_init_library(0);

    /* Start the server */
    ctx = mg_start(NULL, 0, NULL);

    /* Add some handler */
    mg_set_request_handler(ctx, "/hello", handler, "Hello world");

    ... Run the application ...
    
    /* Stop the server */
    mg_stop(ctx);

    /* Un-initialize the library */
    mg_exit_library();
}

A simple callback (new structure supporting HTTP/1.x and HTTP/2):

static int
handler(struct mg_connection *conn, void *ignored)
{
	const char *msg = "Hello world";
	unsigned long len = (unsigned long)strlen(msg);

	mg_send_http_ok(conn, "text/plain", len);

	mg_write(conn, msg, len);

	return 200; /* HTTP state 200 = OK */
}

A simple callback (old structure supporting HTTP/1.x only):

static int
handler(struct mg_connection *conn, void *ignored)
{
	const char *msg = "Hello world";
	unsigned long len = (unsigned long)strlen(msg);

	mg_printf(conn,
	          "HTTP/1.1 200 OK\r\n"
	          "Content-Length: %lu\r\n"
	          "Content-Type: text/plain\r\n"
	          "Connection: close\r\n\r\n",
	          len);

	mg_write(conn, msg, len);

	return 200;
}