HTTP Request-Line parser

Question

I am trying to write an HTTP/1.0 (RFC 1945) parser in C, so I started with the Request-Line. Ideally, the parser should return one of the three:

The request is valid.
The request is malformed, but appended data could make it valid.
The request is malformed and no appended data can make it valid.

The idea is to know whether there is still a point receiving data of a request from a connection. But internal errors can happen, and I have chosen some limitations, so error are also returned.

I present 3 files: the header file of the Request-Line parser (httpParse.h), the implementation (httpParse.c) and the file containing the error definitions (error.h). The library should not compile, because the functions is_http_method() and is_http_uri() are not here defined, since they are more about complex BNF. They'd be fined in httpChar.h.

httpParse.h

/* This file is of the project MarkServer, which is under the MIT licence. For
 * more information, see github.com/Schilive/MarkServer.
 *
 * This file was originally created in 2023/10/05
 */

/* This file declares functions to parse HTTP requests, but not responses.
 */

#ifndef MARKSERVER_HTTPPARSE_H
#define MARKSERVER_HTTPPARSE_H

#include <stdio.h>
#include "error.h"

#define HTTP_METHOD_MAX_LENGTH 255
#define HTTP_URI_MAX_LENGTH 1023

struct http_version {
    unsigned int major;
    unsigned int minor;
};

/* The parsed Request-Line. */
struct http_request_line {
    char method[HTTP_METHOD_MAX_LENGTH + 1];
    char uri[HTTP_URI_MAX_LENGTH];
    struct http_version http_version;
};

/* Parses the Request-Line in a string, checking the logic.
 *
 * @param [in]  fName   The filename of the file starting with the
 *                      Request-Line.
 * @param [out] pRes    The parsed Request-Line. Only changed if no
 *                      error is returned.
 * @return              The error value.
 * @except              ERROR_INVALID_PARAMETER.
 * @except              ERROR_REQUEST_INCOMPLETE
 * @except              ERROR_INTERNAL
 * @except              ERROR_REQUEST_TOO_LONG
 * @except              ERROR_BAD_REQUEST
 */
enum error parse_http_request_line(     const char *restrict fName,
        struct http_request_line *restrict pRes);

#endif /* MARKSERVER_HTTPPARSE_H */

httpParse.c

/* This file is of the project MarkServer, which is under the MIT licence. For
 * more information, see github.com/Schilive/MarkServer.
 *
 * This file was originally created in 2023/10/05
 */

/* This file implements the declarations on "httpParse.h".
 */

#include "httpParse.h"

#include <string.h>
#include <stdbool.h>
#include "httpChar.h"
#include <stdlib.h>
#include <limits.h>

#define HTTP_VERSION_TOKEN_MAX_LENGTH 63
#define HTTP_REQUEST_LINE_MAX_LENGTH (HTTP_METHOD_MAX_LENGTH + 1 + \
        HTTP_URI_MAX_LENGTH + 1 + HTTP_VERSION_TOKEN_MAX_LENGTH)

struct request_line_tokens {
    char method[HTTP_METHOD_MAX_LENGTH + 1];
    char uri[HTTP_URI_MAX_LENGTH + 1];
    char version[HTTP_VERSION_TOKEN_MAX_LENGTH + 1];
};

/* Returns the next character to be read, without changing the file pointer. */
static int seek_file(FILE *f)
{
        int ch = fgetc(f);

        if (ch != EOF)
                ungetc(ch, f);

        return ch;
}

/* Buffers the Request-Line of the start of the file, as a string, whether
 * the content of the fields are correct or not. That is, until CRLF.
 *
 * @param       f       The file to be read.
 * @param       reqLine The buffer.
 * @param       maxLen  The maximum length of the buffer. If the buffer is too
 *                      small, then ERROR_REQUEST_TOO_LONG is returned.
 * @except              ERROR_REQUEST_INCOMPLETE
 * @except              ERROR_INTERNAL
 * @except              ERROR_REQUEST_TOO_LONG
 */
static enum error read_request_line(FILE *restrict f, char *restrict reqLine)
{
        char reqLineTxt[HTTP_REQUEST_LINE_MAX_LENGTH];
        size_t reqLineLen = 0;

        while (1) {
                int ch = fgetc(f);

                if (ch == EOF)
                        return  ferror(f)
                                ? ERROR_REQUEST_INCOMPLETE
                                : ERROR_INTERNAL;
                if (reqLineLen == HTTP_REQUEST_LINE_MAX_LENGTH)
                        return ERROR_REQUEST_TOO_LONG;
                if (ch == '\r' && seek_file(f) == '\n')
                        break;

                reqLineTxt[reqLineLen] = (char)ch;
                reqLineLen++;
        }

        memcpy(reqLine, reqLineTxt, reqLineLen);
        reqLine[reqLineLen] = 0;

        return ERROR_SUCCESS;
}

/* Tokenizes the Request-Line string.
 *
 * @except ERROR_BAD_REQUEST
 */
static enum error tokenize_request_line_str(const char *restrict reqLine,
        struct request_line_tokens *restrict pRes)
{
        char *sp1 = strchr(reqLine, ' ');
        if (!sp1)
                return ERROR_BAD_REQUEST;
        char *sp2 = strchr(sp1 + 1, ' ');
        if (!sp2)
                return ERROR_BAD_REQUEST;

        size_t methodLen = sp1 - reqLine;
        size_t uriLen = sp2 - (sp1 + 1);
        size_t versLen = strlen(sp2 + 1);

        memcpy(pRes->method, reqLine, methodLen);
        memcpy(pRes->uri, sp1 + 1, uriLen);
        memcpy(pRes->version, sp2 + 1, versLen);

        pRes->method[methodLen] = 0;
        pRes->uri[uriLen] = 0;
        pRes->version[versLen] = 0;

        return ERROR_SUCCESS;
}

/* Returns whether the method format is correct. */
static bool is_valid_method(const char *str)
{
        return is_http_method(str, strlen(str));
}

/* Returns whether the URI format is correct. */
static bool is_valid_uri(const char *str)
{
        return is_http_requestURI(str, strlen(str));
}

/* Parses the HTTP version string, verifying that its format is correct.
 * Returns the major and minor versions.
 *
 * @except ERROR_BAD_REQUEST
 */
static enum error parse_http_version(const char *restrict vers,
                          unsigned long *restrict pMajor,
                          unsigned long *restrict pMinor)
{
        char prefix[] = "HTTP/";
        size_t prefixLen = sizeof(prefix) - 1;

        if (strncmp(vers, prefix, 5) != 0)
                return ERROR_BAD_REQUEST;

        char *pPeriod = strchr(vers, '.');
        if (!pPeriod)
                return ERROR_BAD_REQUEST;

        char *next;
        errno = 0;
        unsigned long ulMajor = strtoul(vers + prefixLen, &next, 10);
        if (errno)
                return ERROR_BAD_REQUEST;

        if (next != pPeriod || next == vers)
                return ERROR_BAD_REQUEST;

        unsigned long ulMinor = strtoul(pPeriod + 1, &next, 10);
        if (errno)
                return ERROR_BAD_REQUEST;

        if (*next != 0 || next == pPeriod + 1)
                return ERROR_BAD_REQUEST;

        if (ulMajor >= ULONG_MAX || ulMinor >= ULONG_MAX)
                return ERROR_BAD_REQUEST;

        *pMajor = (unsigned int)ulMajor;
        *pMinor = (unsigned int)ulMinor;
        return ERROR_SUCCESS;
}

/* Parse tokens into the struct 'request_line'
 *
 * @except ERROR_BAD_REQUEST
 */
static enum error parse_request_line_tokens(
        const struct request_line_tokens *restrict pReqLineToks,
        struct http_request_line *restrict pRes)
{
        if (!is_valid_method(pReqLineToks->method))
                return ERROR_BAD_REQUEST;
        if (!is_valid_uri(pReqLineToks->uri))
                return ERROR_BAD_REQUEST;

        unsigned long major, minor;
        enum error err =
                parse_http_version(pReqLineToks->version, &major, &minor);
        if (err)
                return err;

        strcpy(pRes->method, pReqLineToks->method);
        strcpy(pRes->uri, pReqLineToks->uri);
        pRes->http_version.major = major;
        pRes->http_version.minor = minor;

        return ERROR_SUCCESS;
}

enum error parse_http_request_line(const char *restrict fName,
        struct http_request_line *restrict pRes)
{
        if (!fName || !pRes)
                return ERROR_INVALID_PARAMETER;

        FILE *f = fopen(fName, "rb");
        if (!f)
                return ERROR_INVALID_PARAMETER;

        char reqLineStr[HTTP_REQUEST_LINE_MAX_LENGTH + 1];
        enum error err = read_request_line(f, reqLineStr);
        if (err)
                return err;

        struct request_line_tokens reqLineToks;
        err = tokenize_request_line_str(reqLineStr, &reqLineToks);
        if (err)
                return err;

        err = parse_request_line_tokens(&reqLineToks, pRes);
        if (err)
                return err;

        return ERROR_SUCCESS;
}

error.h

/* This file is of the project MarkServer, which is under the MIT licence. For
 * more information, see github.com/Schilive/MarkServer.
 *
 * This file was originally created in 2023/10/05
 */

#ifndef MARKSERVER_ERROR_H
#define MARKSERVER_ERROR_H

enum error {
    ERROR_SUCCESS,
    ERROR_REQUEST_INCOMPLETE,
    ERROR_INVALID_PARAMETER,
    ERROR_REQUEST_TOO_LONG,     /* Something of the HTTP request was considered
                                 * too long. */
    ERROR_INTERNAL,
    ERROR_BAD_REQUEST           /* Request is malformed by the standard. */
};

#endif /* MARKSERVER_ERROR_H */

This is a great question, but can you please add error.h to the code to be reviewed? — pacmaninbw
– pacmaninbw ♦, Commented Oct 19, 2023 at 12:54
@pacmaninbw, I will, And thank you. I think I was going to add it and forgot. — Schilive
– Schilive, Commented Oct 19, 2023 at 17:00

vnp · Accepted Answer · 2023-10-19 23:35:19Z

read_request_line is too complicated. There is no need to read it character by character. You are reading from FILE *, so a simple fgets will do the job. Just test that the line ends with \r\n, and you are done.
Too many copies made.

I don't see a need for recLineTxt. Reading directly into reqLine is perfectly OK.

I don't see a need for struct http_request_line to contain actual strings. Pointers into reqLine will work (mutating reqLine is perfectly OK too), e.g.
```
  pRes->method = reqLine;
  sp = strchr(reqLine, ' ');
  if (!sp) {
      return ERROR_BAD_REQUEST;
  }
  *sp++ = 0;
  // etc
```
Ditto for reqLineToks. You may parse the line directly into pRes, so parse_request_line_tokens wouldn't need all strcpys and assignments.
parse_http_version can be streamlined. There is no need to compute pPeriod separately. Consider strtoul(vers + prefixLen, &pPeriod, 10) test for *pPeriod == '.', and have ulMinor = strtoul(pPeriod + 1, &next, 10)

Also, the cast of unsigned long to unsigned int, followed by an assignmebt to unsigned long is weird.
is_http_requestURI, which in my opinion is the most challenging and interesting part, is missing.

Edit: a concrete example. This is how I would approach it (forgive the broad strokes, error checking is omitted for clarity):

struct http_request {
    char * method;
    char * uri;
    unsigned long major;
    unsigned long minor;
};      

static enum error parse_http_request_line(char * line, struct http_request * req) 
{   
    req->method = line;
    
    char * sp = strchr(line, ' ');
    *sp++ = 0; 
    req->uri = sp;
    
    sp = strchr(line, ' ');
    *sp++ = 0
    parse_http_version(sp, req);

    return ERROR_SUCCESS;
}

enum error handle_request_line(FILE * src, struct http_request_line * req)
{
    char request[HTTP_REQUEST_LINE_MAX_LENGTH + 1];
    fgets(request, HTTP_REQUEST_LINE_MAX_LENGTH + 1, src);
    struct http_request req;
    parse_http_request_line(line, & req);
    // Actual busyness logic here. The caller only sets up a
    // file pointer, and does not care about the request.
}

Thank you for the review! If you'd like to see is_http_requestURI(), you can check httpChar.c/h on here. It is not pretty though. About the copiesin functions, it is true: they are static functions, not interface, so there is no need. About fgets(), I did not remember it is -- very useful. — Schilive
– Schilive, Commented Oct 19, 2023 at 18:11
I just did not understand what you said about pointers about http_request_line. Do you mean having points to the start of the field of the Request-Line? — Schilive
– Schilive, Commented Oct 19, 2023 at 18:16
vnp, thank you for the example. But wouldn't I have to malloc() the string to do it? — Schilive
– Schilive, Commented Oct 19, 2023 at 19:02
No, you don't have to. Essentially, split reqLine into 3 fragments in-place, by placing nul bytes where the spaces were, and keep the pointers in the struct http_request_line. If I am still unclear, let me know, and I will give a more complete example. — vnp
– vnp, Commented Oct 19, 2023 at 19:40

Stack Exchange Network

HTTP Request-Line parser

httpParse.h

httpParse.c

error.h

1 Answer 1

You must log in to answer this question.

Hot Network Questions

HTTP Request-Line parser

httpParse.h

httpParse.c

error.h

1 Answer 1

You must log in to answer this question.

Related

Hot Network Questions