Skip to content
/ httpc Public

HTTP client for embedded use - supports redirects and resume.

License

Notifications You must be signed in to change notification settings

howerj/httpc

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

60 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

% HTTPC(1) | A HTTP(S) Client For The Discerning Embedded Programmer

NAME

HTTPC - A HTTP(S) Client For The Discerning Embedded Programmer

SYNOPSIS

httpc -t

httpc example.com

httpc -o PUT example.com < file.txt

DESCRIPTION

Project:    Embeddable HTTP(S) Client
Author:     Richard James Howe
Email:      <mailto:[email protected]>
License:    The Unlicense
Repository: <https://github.com/howerj/httpc>

This library implements a small HTTP(S) client that can be used for downloading files and firmware on an embedded platform. The demonstration code targets linux and windows platforms.

The non-portable code for dealing with the socket and SSL/TLS layers must be provided by the user and are not contained within the HTTPC library. SSL/TLS support is entirely optional. The dependency on the C standard library is minimal (mostly string functions) and the allocator used by the HTTPC library can be provided by the user to aid in porting. The data is passed to a user supplied callback.

The library implements a retry mechanism and enough of HTTP 1.1 to allow partially downloaded files to be resumed. This allows for a more reliable file transfer over a bad network connection than HTTP 1.0 allows. A simple HTTP 1.0 client is almost trivial to implement, however it is not very reliable. Only certain operations are retried in full with a resume ("GET"/"HEAD"), others are not retried if a negative response is received ("DELETE"/"POST"/"POST").

There are other libraries that do something similar, however they tend to be larger, more complex and more complete. cURL is an example of one such program and library - if you do decided to use it, good luck porting it!

Use cases:

  • Downloading firmware over an unreliable connection.
  • Poking a web-server with data.
  • Talking to a known, trusted server.

EXAMPLE USAGE

Usage: ./httpc [-ht] *OR* -[1vy] -u URL *OR* -[1vy] URL

Options:

  • -h : print out a help message to stderr and exit successfully.

  • -u : specify the URL to attempt to download from

  • -t : run the internal tests, returning zero on success, and none zero on failure

  • -1 : perform a HTTP 1.0 request, with HTTP 1.1 responses still being dealt with

  • -v : add logging

  • -y : turn yielding on, this is only useful for debugging the library.

  • -H: add a custom header to the request

Examples:

./httpc example.com/index.html
./httpc http://example.com
./httpc https://example.com

C API

There are only a handful of functions required to be implemented in order to port this client. If your platform has a TCP/IP stack with a Berkeley sockets like API then porting should be trivial.

The functions all return negative on failure, zero on successful completion of an operation or session, and there are some positive values that are used to to signal special cases (such as the HTTP client yielding to the caller, "HTTPC_YIELD"), or the operation is finished but the connection is open ("HTTPC_REUSE" is returned). Both of these special return values are not encountered unless enabled, which by default they are not.

For each connection a "httpc_options_t" structure must be readied, which will contain the callbacks needed to connect to the web-server, and to allocate memory. Very little memory is allocated, non-blocking memory operations use allocate more on the heap, but blocking operations use more stack space. Allocation is only done during session setup usually, and should not grow during normal operation.

The following functions support non-blocking modes of operation, and reusing the HTTP connection after successful requests (unsuccessful requests cause the connection to be closed, and reopened, as do redirects).

int httpc_get(httpc_options_t *a, const char *url, httpc_callback fn, void *param);
int httpc_put(httpc_options_t *a, const char *url, httpc_callback fn, void *param);
int httpc_post(httpc_options_t *a, const char *url, httpc_callback fn, void *param);
int httpc_head(httpc_options_t *a, const char *url);
int httpc_delete(httpc_options_t *a, const char *url);
int httpc_trace(httpc_options_t *a, const char *url);
int httpc_options(httpc_options_t *a, const char *url);

The callback for the GET operation is called only when receiving the response body. Likewise the callback for the PUT operation is only called when sending the request body. All operations can optionally have both requests and response bodies according to the specification, but there is no way to deal with both (to keep things simple).

"httpc_head", "httpc_trace", and "httpc_options" are really only useful for interactive debugging purposes when logging is turned on. It is expected that they will not be used on an embedded platform.

These functions support blocking operations only, and do not support reusing the connection:

int httpc_get_buffer(httpc_options_t *a, const char *url, char *buffer, size_t *length);
int httpc_put_buffer(httpc_options_t *a, const char *url, char *buffer, size_t length);
int httpc_post_buffer(httpc_options_t *a, const char *url, char *buffer, size_t length);

The above functions are for reading into a buffer (GET) writing from a buffer (PUT, POST). GET expects the length of the buffer to be passed into 'length', and it returns the actual length of the data retrieved in 'length'. If the buffer is not big enough an error is returned.

The "httpc_tests" function executes a series of built in unit tests, they may be compiled out with the right define to save on space.

int httpc_tests(httpc_options_t *a);

You should not normally need to run this function. Negative is returned on error, as usual.

The function "httpc_end_session";

int httpc_end_session(httpc_options_t *a);

Is only needed when using the yield option and you want to end a HTTP operation prematurely, or when using the keep-alive options and you want to finally close the connection. It is safe to call this function multiple times, or if those options are not set.

BUILDING

To build you will need a C99 compiler and GNU Make. There are minimal system dependencies on the C standard library, the other dependencies needed are for the SSL/TLS library used which can be compiled out, and the systems TCP/IP stack.

To build:

make

To build without SSL/TLS support:

make DEFINES="-DUSE_SSL=0"

To run the internal tests:

make test

To install:

make install

The source code for the program is quite small, at the top of the 'httpc.c' file there are a number of configurable options that can be specified by the build system (and do not require you to modify the source itself).

The code is most likely to have been recently tested on Linux, it may work on other Unixen, and it will be less frequently tested on Windows. Likewise, it is more likely to have been tested on a 64-bit platform. The SSL library, if used at all, should provide an openSSL interface on either Windows or Linux.

RETURN CODES

This program will return a non-zero value on failure and zero on success. Like pretty much every Unix program ever.

LICENSE

This program is licensed under the The Unlicense, do what thou wilt.

LIMITATIONS

Known limitations include:

  • If reallocation is disabled (is is enabled by default) then the default buffer size (of 128, which can be changed) will place limitations of URL length and line length (each field in the HTTP header is a single line).
  • File sizes are probably limited up to 2GiB on many platforms.
  • The library has a non-blocking mechanism, however what this really means is 'less-blocky' and may block until a time out on a socket. This is likely to improve and yield more often as time goes on. However turning this library into a fully non-blocking version is a non-trivial task. See https://www.chiark.greenend.org.uk/~sgtatham/coroutines.html for both a description of the problem and a possible solution in C.
  • The library should be fuzzed, and has not been. Preferably with something like AFL.
  • The library would be better expressed as a header only library as that is what all of the cool kids are using.
  • The library is quite thirsty for stack size for a library meant to be run on an embedded microcontroller. It will require at least 1KiB of stack, and possibly more depending on how your callbacks are written.
  • The client is more liberal in what it will accept from the server than it should be, allowing newlines to be terminated by a single CR with no LF, and comparisons are done in a case-insensitive manner.
  • HTTP Status Code 206 (partial content) is not handled, which could be fairly easily.
  • A C++ wrapper, and wrappers for other language, would not go amiss.
  • The socket and SSL settings are provided by a series of callbacks - this allows you to set things like timeouts and keep-alive settings, the defaults may not suite you.
  • The set of functions provided by this library should suite the common cases, however somethings are not supported, for example a GET request can have a body.
  • Entropy cannot be reversed, meaning all acts of man no matter how great will eventually be rendered futile, the best one can hope for is eternal return, are you proud of what you have achieved? Or will you die like a dog in the face of heat-death?
  • This library does not implement WebSockets, but looking through the specification, they should be fairly easy to implement.

BUGS

For any bugs please contact the author at mailto:[email protected]. Please include as much information as possible including, but not limited to: information on what platform you are compiling your code on (OS, 32/64-bit, ...), tracing information (for example, valgrind output), a minimal test case, a (small) blood sacrifice, three bushels of wheat, thoughts, comments and general rants. Alternatively shout into a buck or pray to your gods.