mod_mime_magic.c

Apache HTTP Server 是一个功能强大的灵活的与HTTP/1.1相兼容的web服务器.这里给出的是Apache HTTP服务器的源码。

    /* these are easy, do them first */

    /*
     * for troff, look for . + letter + letter or .\"; this must be done to
     * disambiguate tar archives' ./file and other trash from real troff
     * input.
     */
    if (*buf == '.') {
	unsigned char *tp = buf + 1;

	while (apr_isspace(*tp))
	    ++tp;		/* skip leading whitespace */
	if ((apr_isalnum(*tp) || *tp == '\\') &&
	     (apr_isalnum(*(tp + 1)) || *tp == '"')) {
	    magic_rsl_puts(r, "application/x-troff");
	    return 1;
	}
    }
    if ((*buf == 'c' || *buf == 'C') && apr_isspace(*(buf + 1))) {
	/* Fortran */
	magic_rsl_puts(r, "text/plain");
	return 1;
    }

    /* look for tokens from names.h - this is expensive!, so we'll limit
     * ourselves to only SMALL_HOWMANY bytes */
    small_nbytes = (nbytes > SMALL_HOWMANY) ? SMALL_HOWMANY : nbytes;
    /* make a copy of the buffer here because apr_strtok() will destroy it */
    s = (unsigned char *) memcpy(nbuf, buf, small_nbytes);
    s[small_nbytes] = '\0';
    has_escapes = (memchr(s, '\033', small_nbytes) != NULL);
    while ((token = apr_strtok((char *) s, " \t\n\r\f", &strtok_state)) != NULL) {
	s = NULL;		/* make apr_strtok() keep on tokin' */
	for (p = names; p < names + NNAMES; p++) {
	    if (STREQ(p->name, token)) {
		magic_rsl_puts(r, types[p->type]);
		if (has_escapes)
		    magic_rsl_puts(r, " (with escape sequences)");
		return 1;
	    }
	}
    }

    switch (is_tar(buf, nbytes)) {
    case 1:
	/* V7 tar archive */
	magic_rsl_puts(r, "application/x-tar");
	return 1;
    case 2:
	/* POSIX tar archive */
	magic_rsl_puts(r, "application/x-tar");
	return 1;
    }

    /* all else fails, but it is ascii... */
    return 0;
}


/*
 * compress routines: zmagic() - returns 0 if not recognized, uncompresses
 * and prints information if recognized uncompress(s, method, old, n, newch)
 * - uncompress old into new, using method, return sizeof new
 */

static struct {
    char *magic;
    apr_size_t maglen;
    char *argv[3];
    int silent;
    char *encoding;	/* MUST be lowercase */
} compr[] = {

    /* we use gzip here rather than uncompress because we have to pass
     * it a full filename -- and uncompress only considers filenames
     * ending with .Z
     */
    {
	"\037\235", 2, {
	    "gzip", "-dcq", NULL
	}, 0, "x-compress"
    },
    {
	"\037\213", 2, {
	    "gzip", "-dcq", NULL
	}, 1, "x-gzip"
    },
    /*
     * XXX pcat does not work, cause I don't know how to make it read stdin,
     * so we use gzip
     */
    {
	"\037\036", 2, {
	    "gzip", "-dcq", NULL
	}, 0, "x-gzip"
    },
};

static int ncompr = sizeof(compr) / sizeof(compr[0]);

static int zmagic(request_rec *r, unsigned char *buf, apr_size_t nbytes)
{
    unsigned char *newbuf;
    int newsize;
    int i;

    for (i = 0; i < ncompr; i++) {
	if (nbytes < compr[i].maglen)
	    continue;
	if (memcmp(buf, compr[i].magic, compr[i].maglen) == 0)
	    break;
    }

    if (i == ncompr)
	return 0;

    if ((newsize = uncompress(r, i, &newbuf, nbytes)) > 0) {
	if (tryit(r, newbuf, newsize, 0) != OK) {
            return 0;
        }

	/* set encoding type in the request record */
	r->content_encoding = compr[i].encoding;
    }
    return 1;
}


struct uncompress_parms {
    request_rec *r;
    int method;
};

static int create_uncompress_child(struct uncompress_parms *parm, apr_pool_t *cntxt,
                                   apr_file_t **pipe_in)
{
    int rc = 1;
    const char *new_argv[4];
    const char *const *env;
    request_rec *r = parm->r;
    apr_pool_t *child_context = cntxt;
    apr_procattr_t *procattr;
    apr_proc_t *procnew;

    /* XXX missing 1.3 logic: 
     *
     * what happens when !compr[parm->method].silent?
     * Should we create the err pipe, read it, and copy to the log?
     */
        
    env = (const char *const *)ap_create_environment(child_context, r->subprocess_env);

    if ((apr_procattr_create(&procattr, child_context) != APR_SUCCESS) ||
        (apr_procattr_io_set(procattr, APR_FULL_BLOCK, 
                           APR_FULL_BLOCK, APR_NO_PIPE)   != APR_SUCCESS) ||
        (apr_procattr_dir_set(procattr, r->filename)        != APR_SUCCESS) ||
        (apr_procattr_cmdtype_set(procattr, APR_PROGRAM)    != APR_SUCCESS)) {
        /* Something bad happened, tell the world. */
        ap_log_rerror(APLOG_MARK, APLOG_ERR, APR_ENOPROC, r,
               "couldn't setup child process: %s", r->filename);
    }
    else {
        new_argv[0] = compr[parm->method].argv[0];
        new_argv[1] = compr[parm->method].argv[1];
        new_argv[2] = r->filename;
        new_argv[3] = NULL;

        procnew = apr_pcalloc(child_context, sizeof(*procnew));
        rc = apr_proc_create(procnew, compr[parm->method].argv[0],
                               new_argv, env, procattr, child_context);

        if (rc != APR_SUCCESS) {
            /* Bad things happened. Everyone should have cleaned up. */
            ap_log_rerror(APLOG_MARK, APLOG_ERR, APR_ENOPROC, r,
                          MODNAME ": could not execute `%s'.",
                          compr[parm->method].argv[0]);
        }
        else {
            apr_pool_note_subprocess(child_context, procnew, APR_KILL_AFTER_TIMEOUT);
            *pipe_in = procnew->out;
        }
    }

    return (rc);
}

static int uncompress(request_rec *r, int method, 
		      unsigned char **newch, apr_size_t n)
{
    struct uncompress_parms parm;
    apr_file_t *pipe_out = NULL;
    apr_pool_t *sub_context;
    apr_status_t rv;

    parm.r = r;
    parm.method = method;

    /* We make a sub_pool so that we can collect our child early, otherwise
     * there are cases (i.e. generating directory indicies with mod_autoindex)
     * where we would end up with LOTS of zombies.
     */
    if (apr_pool_create(&sub_context, r->pool) != APR_SUCCESS)
        return -1;

    if ((rv = create_uncompress_child(&parm, sub_context, &pipe_out)) != APR_SUCCESS) {
	ap_log_rerror(APLOG_MARK, APLOG_ERR, rv, r,
		    MODNAME ": couldn't spawn uncompress process: %s", r->uri);
	return -1;
    }

    *newch = (unsigned char *) apr_palloc(r->pool, n);
    rv = apr_file_read(pipe_out, *newch, &n);
    if (n == 0) {
	apr_pool_destroy(sub_context);
	ap_log_rerror(APLOG_MARK, APLOG_ERR, rv, r,
	    MODNAME ": read failed %s", r->filename);
	return -1;
    }
    apr_pool_destroy(sub_context);
    return n;
}

/*
 * is_tar() -- figure out whether file is a tar archive.
 *
 * Stolen (by author of file utility) from the public domain tar program: Public
 * Domain version written 26 Aug 1985 John Gilmore (ihnp4!hoptoad!gnu).
 *
 * @(#)list.c 1.18 9/23/86 Public Domain - gnu $Id: mod_mime_magic.c,v 1.7
 * 1997/06/24 00:41:02 ikluft Exp ikluft $
 *
 * Comments changed and some code/comments reformatted for file command by Ian
 * Darwin.
 */

#define isodigit(c) (((unsigned char)(c) >= '0') && ((unsigned char)(c) <= '7'))

/*
 * Return 0 if the checksum is bad (i.e., probably not a tar archive), 1 for
 * old UNIX tar file, 2 for Unix Std (POSIX) tar file.
 */

static int is_tar(unsigned char *buf, apr_size_t nbytes)
{
    register union record *header = (union record *) buf;
    register int i;
    register long sum, recsum;
    register char *p;

    if (nbytes < sizeof(union record))
	       return 0;

    recsum = from_oct(8, header->header.chksum);

    sum = 0;
    p = header->charptr;
    for (i = sizeof(union record); --i >= 0;) {
	/*
	 * We can't use unsigned char here because of old compilers, e.g. V7.
	 */
	sum += 0xFF & *p++;
    }

    /* Adjust checksum to count the "chksum" field as blanks. */
    for (i = sizeof(header->header.chksum); --i >= 0;)
	sum -= 0xFF & header->header.chksum[i];
    sum += ' ' * sizeof header->header.chksum;

    if (sum != recsum)
	return 0;		/* Not a tar archive */

    if (0 == strcmp(header->header.magic, TMAGIC))
	return 2;		/* Unix Standard tar archive */

    return 1;			/* Old fashioned tar archive */
}


/*
 * Quick and dirty octal conversion.
 *
 * Result is -1 if the field is invalid (all blank, or nonoctal).
 */
static long from_oct(int digs, char *where)
{
    register long value;

    while (apr_isspace(*where)) {	/* Skip spaces */
	where++;
	if (--digs <= 0)
	    return -1;		/* All blank field */
    }
    value = 0;
    while (digs > 0 && isodigit(*where)) {	/* Scan til nonoctal */
	value = (value << 3) | (*where++ - '0');
	--digs;
    }

    if (digs > 0 && *where && !apr_isspace(*where))
	return -1;		/* Ended on non-space/nul */

    return value;
}

/*
 * Check for file-revision suffix
 *
 * This is for an obscure document control system used on an intranet.
 * The web representation of each file's revision has an @1, @2, etc
 * appended with the revision number.  This needs to be stripped off to
 * find the file suffix, which can be recognized by sending the name back
 * through a sub-request.  The base file name (without the @num suffix)
 * must exist because its type will be used as the result.
 */
static int revision_suffix(request_rec *r)
{
    int suffix_pos, result;
    char *sub_filename;
    request_rec *sub;

#if MIME_MAGIC_DEBUG
    ap_log_rerror(APLOG_MARK, APLOG_DEBUG, 0, r,
		MODNAME ": revision_suffix checking %s", r->filename);
#endif /* MIME_MAGIC_DEBUG */

    /* check for recognized revision suffix */
    suffix_pos = strlen(r->filename) - 1;
    if (!apr_isdigit(r->filename[suffix_pos])) {
	return 0;
    }
    while (suffix_pos >= 0 && apr_isdigit(r->filename[suffix_pos]))
	suffix_pos--;
    if (suffix_pos < 0 || r->filename[suffix_pos] != '@') {
	return 0;
    }

    /* perform sub-request for the file name without the suffix */
    result = 0;
    sub_filename = apr_pstrndup(r->pool, r->filename, suffix_pos);
#if MIME_MAGIC_DEBUG
    ap_log_rerror(APLOG_MARK, APLOG_DEBUG, 0, r,
		MODNAME ": subrequest lookup for %s", sub_filename);
#endif /* MIME_MAGIC_DEBUG */
    sub = ap_sub_req_lookup_file(sub_filename, r, NULL);

    /* extract content type/encoding/language from sub-request */
    if (sub->content_type) {
	ap_set_content_type(r, apr_pstrdup(r->pool, sub->content_type));
#if MIME_MAGIC_DEBUG
	ap_log_rerror(APLOG_MARK, APLOG_DEBUG, 0, r,
		    MODNAME ": subrequest %s got %s",
		    sub_filename, r->content_type);
#endif /* MIME_MAGIC_DEBUG */
	if (sub->content_encoding)
	    r->content_encoding =
		apr_pstrdup(r->pool, sub->content_encoding);
        if (sub->content_languages) {
            int n;
	    r->content_languages = apr_array_copy(r->pool, 
                                                  sub->content_languages);
            for (n = 0; n < r->content_languages->nelts; ++n) {
                char **lang = ((char **)r->content_languages->elts) + n;
                *lang = apr_pstrdup(r->pool, *lang);
            }
        }
	result = 1;
    }

    /* clean up */
    ap_destroy_sub_req(sub);

    return result;
}

/*
 * initialize the module
 */
static int magic_init(apr_pool_t *p, apr_pool_t *plog, apr_pool_t *ptemp, server_rec *main_server)
{
    int result;
    magic_server_config_rec *conf;
    magic_server_config_rec *main_conf;
    server_rec *s;
#if MIME_MAGIC_DEBUG
    struct magic *m, *prevm;
#endif /* MIME_MAGIC_DEBUG */

    main_conf = ap_get_module_config(main_server->module_config, &mime_magic_module);
    for (s = main_server; s; s = s->next) {
	conf = ap_get_module_config(s->module_config, &mime_magic_module);
	if (conf->magicfile == NULL && s != main_server) {
	    /* inherits from the parent */
	    *conf = *main_conf;
	}
	else if (conf->magicfile) {
	    result = apprentice(s, p);
	    if (result == -1)
		return OK;
#if MIME_MAGIC_DEBUG
	    prevm = 0;
	    ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, s,
			MODNAME ": magic_init 1 test");
	    for (m = conf->magic; m; m = m->next) {
		if (apr_isprint((((unsigned long) m) >> 24) & 255) &&
		    apr_isprint((((unsigned long) m) >> 16) & 255) &&
		    apr_isprint((((unsigned long) m) >> 8) & 255) &&
		    apr_isprint(((unsigned long) m) & 255)) {
		    ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, s,
				MODNAME ": magic_init 1: POINTER CLOBBERED! "
				"m=\"%c%c%c%c\" line=%d",
				(((unsigned long) m) >> 24) & 255,
				(((unsigned long) m) >> 16) & 255,
				(((unsigned long) m) >> 8) & 255,
				((unsigned long) m) & 255,
				prevm ? prevm->lineno : -1);
		    break;
		}
		prevm = m;
	    }
#endif
	}
    }
    return OK;
}

/*
 * Find the Content-Type from any resource this module has available
 */

static int magic_find_ct(request_rec *r)
{
    int result;
    magic_server_config_rec *conf;

    /* the file has to exist */
    if (r->finfo.filetype == 0 || !r->filename) {
	return DECLINED;
    }

    /* was someone else already here? */
    if (r->content_type) {
	return DECLINED;
    }

    conf = ap_get_module_config(r->server->module_config, &mime_magic_module);
    if (!conf || !conf->magic) {
	return DECLINED;
    }

    /* initialize per-request info */
    if (!magic_set_config(r)) {
	return HTTP_INTERNAL_SERVER_ERROR;
    }

    /* try excluding file-revision suffixes */
    if (revision_suffix(r) != 1) {
	/* process it based on the file contents */
	if ((result = magic_process(r)) != OK) {
	    return result;
	}
    }

    /* if we have any results, put them in the request structure */
    return magic_rsl_to_request(r);
}

static void register_hooks(apr_pool_t *p)
{
    static const char * const aszPre[]={ "mod_mime.c", NULL };

    /* mod_mime_magic should be run after mod_mime, if at all. */

    ap_hook_type_checker(magic_find_ct, aszPre, NULL, APR_HOOK_MIDDLE);
    ap_hook_post_config(magic_init, NULL, NULL, APR_HOOK_FIRST);
}

/*
 * Apache API module interface
 */

module AP_MODULE_DECLARE_DATA mime_magic_module =
{
    STANDARD20_MODULE_STUFF,
    NULL,                      /* dir config creator */
    NULL,                      /* dir merger --- default is to override */
    create_magic_server_config,        /* server config */
    merge_magic_server_config, /* merge server config */
    mime_magic_cmds,           /* command apr_table_t */
    register_hooks              /* register hooks */
};