uri.c

一个很有名的浏览器

	int retries = 0;

	/* Strip starting spaces */
	while (*url == ' ') url++;
	if (!*url) return NULL;

	newurl = expand_tilde(url); /* XXX: Post data copy. */
	if (!newurl) return NULL;

parse_uri:
	/* Yay a goto loop. If we get some URI parse error and try to
	 * fix it we go back to here and try again. */
	/* Ordinary parse */
	uri_errno = parse_uri(&uri, newurl);

	/* Bail out if the same error occurs twice */
	if (uri_errno == prev_errno || retries++ > MAX_TRANSLATION_ATTEMPTS) {
		if (retries > MAX_TRANSLATION_ATTEMPTS) {
			ERROR("Maximum number of parsing attempts exceeded "
			      "for %s.", url);
		}
		mem_free(newurl);
		return NULL;
	}

	prev_errno = uri_errno;

	switch (uri_errno) {
	case URI_ERRNO_OK:
		/* Fix translation of 1.2.3.4:5 so IP address part won't be
		 * interpreted as the protocol name. */
		if (uri.protocol == PROTOCOL_UNKNOWN) {
			unsigned char *ipscan = newurl;

			/* FIXME: This is looking more and more like the code
			 * for the URI_ERRNO_INVALID_PROTOCOL label and
			 * find_uri_protocol(). Unite? --jonas */

			/* Check if it could be just some local file with ':'
			 * in its name. */
			if (check_uri_file(newurl) >= 0) {
				struct string str;

				if (!init_string(&str)) return NULL;

				add_to_string(&str, "file://");
				if (!dir_sep(*newurl))
					add_to_string(&str, "./");
				encode_file_uri_string(&str, newurl);

				mem_free(newurl);
				newurl = str.source;

				/* Work around the infinite loop prevention */
				prev_errno = URI_ERRNO_EMPTY;
				goto parse_uri;
			}

			/* Check if it's <IP-address>:<port> */
			while (isdigit(*ipscan) || *ipscan == '.')
				ipscan++;

			if (!*ipscan || *ipscan == ':' || *ipscan == '/') {
				insert_in_string(&newurl, 0, "http://", 7);
				/* Work around the infinite loop prevention */
				prev_errno = URI_ERRNO_EMPTY;
				goto parse_uri;
			}
		}

		/* If file:// URI is transformed we need to reparse. */
		if (uri.protocol == PROTOCOL_FILE && cwd && *cwd
		    && transform_file_url(&uri, cwd))
			return normalize_uri_reparse(struri(&uri));

		/* Translate the proxied URI too if proxy:// */
		if (uri.protocol == PROTOCOL_PROXY) {
			unsigned char *data = translate_url(uri.data, cwd);
			int pos = uri.data - struri(&uri);

			if (!data) break;
			struri(&uri)[pos] = 0;
			insert_in_string(&struri(&uri), pos, data, strlen(data));
			mem_free(data);
			return normalize_uri_reparse(struri(&uri));
		}

		return normalize_uri_noparse(&uri);

	case URI_ERRNO_TOO_MANY_SLASHES:
	{
		unsigned char *from, *to;

		assert(uri.string[uri.protocollen] == ':'
		       && uri.string[uri.protocollen + 1] == '/'
		       && uri.string[uri.protocollen + 2] == '/');

		from = to = uri.string + uri.protocollen + 3;
		while (*from == '/') from++;

		assert(to < from);
		memmove(to, from, strlen(from) + 1);
		goto parse_uri;
	}
	case URI_ERRNO_NO_SLASHES:
	{
		/* Try prefix:some.url -> prefix://some.url.. */
		int slashes = 2;

		/* Check if only one '/' is needed. */
		if (uri.string[uri.protocollen + 1] == '/')
			slashes--;

		insert_in_string(&newurl, uri.protocollen + 1, "//", slashes);
		goto parse_uri;
	}
	case URI_ERRNO_TRAILING_DOTS:
	{
		/* Trim trailing '.'s */
		unsigned char *from = uri.host + uri.hostlen;
		unsigned char *to = from;

		assert(uri.host < to && to[-1] == '.' && *from != '.');

		while (uri.host < to && to[-1] == '.') to--;

		assert(to < from);
		memmove(to, from, strlen(from) + 1);
		goto parse_uri;
	}
	case URI_ERRNO_NO_PORT_COLON:
		assert(uri.portlen == 0
		       && uri.string < uri.port
		       && uri.port[-1] == ':');

		memmove(uri.port - 1, uri.port, strlen(uri.port) + 1);
		goto parse_uri;

	case URI_ERRNO_NO_HOST_SLASH:
	{
		int offset = uri.port
			   ? uri.port + uri.portlen - struri(&uri)
			   : uri.host + uri.hostlen - struri(&uri) + uri.ipv6 /* ']' */;

		assertm(uri.host, "uri.host not set after no host slash error");
		insert_in_string(&newurl, offset, "/", 1);
		goto parse_uri;
	}
	case URI_ERRNO_INVALID_PROTOCOL:
	{
		/* No protocol name */
		enum protocol protocol = find_uri_protocol(newurl);
		struct string str;

		if (!init_string(&str)) return NULL;

		switch (protocol) {
			case PROTOCOL_FTP:
				add_to_string(&str, "ftp://");
				encode_uri_string(&str, newurl, 0);
				break;

			case PROTOCOL_HTTP:
				add_to_string(&str, "http://");
				add_to_string(&str, newurl);
				break;

			case PROTOCOL_FILE:
			default:
				add_to_string(&str, "file://");
				if (!dir_sep(*newurl))
					add_to_string(&str, "./");

				encode_file_uri_string(&str, newurl);
		}

		mem_free(newurl);
		newurl = str.source;

		goto parse_uri;
	}
	case URI_ERRNO_EMPTY:
	case URI_ERRNO_IPV6_SECURITY:
	case URI_ERRNO_NO_HOST:
	case URI_ERRNO_INVALID_PORT:
	case URI_ERRNO_INVALID_PORT_RANGE:
		/* None of these can be handled properly. */
		break;
	}

	mem_free(newurl);
	return NULL;
}


struct uri *
get_composed_uri(struct uri *uri, enum uri_component components)
{
	unsigned char *string;

	assert(uri);
	if_assert_failed return NULL;

	string = get_uri_string(uri, components);
	if (!string) return NULL;

	uri = get_uri(string, 0);
	mem_free(string);

	return uri;
}

struct uri *
get_translated_uri(unsigned char *uristring, unsigned char *cwd)
{
	struct uri *uri;

	uristring = translate_url(uristring, cwd);
	if (!uristring) return NULL;

	uri = get_uri(uristring, 0);
	mem_free(uristring);

	return uri;
}


unsigned char *
get_extension_from_uri(struct uri *uri)
{
	unsigned char *extension = NULL;
	int afterslash = 1;
	unsigned char *pos = uri->data;

	assert(pos);

	for (; *pos && !end_of_dir(*pos); pos++) {
		if (!afterslash && !extension && *pos == '.') {
			extension = pos;
		} else if (is_uri_dir_sep(uri, *pos)) {
			extension = NULL;
			afterslash = 1;
		} else {
			afterslash = 0;
		}
	}

	if (extension && extension < pos)
		return memacpy(extension, pos - extension);

	return NULL;
}

/* URI encoding, escaping unallowed characters. */
static inline int
safe_char(unsigned char c)
{
	/* RFC 2396, Page 8, Section 2.3 ;-) */
	return isident(c) || c == '.' || c == '!' || c == '~'
	       || c == '*' || c == '\''|| c == '(' || c == ')';
}

void
encode_uri_string(struct string *string, unsigned char *name,
		  int convert_slashes)
{
	unsigned char n[4];

	n[0] = '%';
	n[3] = '\0';

	for (; *name; name++) {
#if 0
		/* This is probably correct only for query part of URI..? */
		if (*name == ' ') add_char_to_string(data, len, '+');
		else
#endif
		if (safe_char(*name) || (!convert_slashes && *name == '/')) {
			add_char_to_string(string, *name);
		} else {
			/* Hex it. */
			n[1] = hx((((int) *name) & 0xF0) >> 4);
			n[2] = hx(((int) *name) & 0xF);
			add_bytes_to_string(string, n, sizeof(n) - 1);
		}
	}
}

/* This function is evil, it modifies its parameter. */
/* XXX: but decoded string is _never_ longer than encoded string so it's an
 * efficient way to do that, imho. --Zas */
void
decode_uri(unsigned char *src)
{
	unsigned char *dst = src;
	unsigned char c;

	do {
		c = *src++;

		if (c == '%') {
			int x1 = unhx(*src);

			if (x1 >= 0) {
				int x2 = unhx(*(src + 1));

				if (x2 >= 0) {
					x1 = (x1 << 4) + x2;
					if (x1 != 0) { /* don't allow %00 */
						c = (unsigned char) x1;
						src += 2;
					}
				}
			}

#if 0
		} else if (c == '+') {
			/* As the comment in encode_uri_string suggests, '+'
			 * should only be decoded in the query part of a URI
			 * (should that be 'URL'?). I'm not bold enough to
			 * disable this code, tho. -- Miciah */
			c = ' ';
#endif
		}

		*dst++ = c;
	} while (c != '\0');
}

void
decode_uri_string(struct string *string)
{
	decode_uri(string->source);
	string->length = strlen(string->source);
}

void
decode_uri_for_display(unsigned char *src)
{
	decode_uri(src);

	for (; *src; src++)
		if (!isprint(*src) || iscntrl(*src))
			*src = '*';
}

void
decode_uri_string_for_display(struct string *string)
{
	decode_uri_for_display(string->source);
	string->length = strlen(string->source);
}


/* URI list */

#define URI_LIST_GRANULARITY 0x3

#define realloc_uri_list(list) \
	mem_align_alloc(&(list)->uris, (list)->size, (list)->size + 1, \
			struct uri *, URI_LIST_GRANULARITY)

struct uri *
add_to_uri_list(struct uri_list *list, struct uri *uri)
{
	if (!realloc_uri_list(list))
		return NULL;

	list->uris[list->size++] = get_uri_reference(uri);

	return uri;
};

void
free_uri_list(struct uri_list *list)
{
	struct uri *uri;
	int index;

	if (!list->uris) return;

	foreach_uri (uri, index, list) {
		done_uri(uri);
	}

	mem_free(list->uris);
	list->size = 0;
}

/* URI cache */

struct uri_cache_entry {
	struct uri uri;
	unsigned char string[1];
};

struct uri_cache {
	struct hash *map;
	struct object object;
};

static struct uri_cache uri_cache;

#ifdef CONFIG_DEBUG
static inline void
check_uri_sanity(struct uri *uri)
{
	int pos;

	for (pos = 0; pos < uri->protocollen; pos++)
		if (isupper(uri->string[pos])) goto error;

	if (uri->hostlen)
		for (pos = 0; pos < uri->hostlen; pos++)
			if (isupper(uri->host[pos])) goto error;
	return;
error:
	INTERNAL("Uppercase letters detected in protocol or host part (%s).", struri(uri));
}
#else
#define check_uri_sanity(uri)
#endif

static inline struct uri_cache_entry *
get_uri_cache_entry(unsigned char *string, int length)
{
	struct uri_cache_entry *entry;
	struct hash_item *item;

	assert(string && length > 0);
	if_assert_failed return NULL;

	item = get_hash_item(uri_cache.map, string, length);
	if (item) return item->value;

	/* Setup a new entry */

	entry = mem_calloc(1, sizeof(*entry) + length);
	if (!entry) return NULL;

	object_nolock(&entry->uri, "uri");
	memcpy(&entry->string, string, length);
	string = entry->string;

	if (parse_uri(&entry->uri, string) != URI_ERRNO_OK
	    || !add_hash_item(uri_cache.map, string, length, entry)) {
		mem_free(entry);
		return NULL;
	}

	object_lock(&uri_cache);

	return entry;
}

struct uri *
get_uri(unsigned char *string, enum uri_component components)
{
	struct uri_cache_entry *entry;

	assert(string);

	if (components) {
		struct uri uri;

		if (parse_uri(&uri, string) != URI_ERRNO_OK)
			return NULL;

		return get_composed_uri(&uri, components);
	}

	if (!is_object_used(&uri_cache)) {
		uri_cache.map = init_hash(hash_size(3), strhash);
		if (!uri_cache.map) return NULL;
		object_nolock(&uri_cache, "uri_cache");
	}

	entry = get_uri_cache_entry(string, strlen(string));
	if (!entry) {
		if (!is_object_used(&uri_cache))
			free_hash(uri_cache.map);
		return NULL;
	}

	check_uri_sanity(&entry->uri);
	object_nolock(&entry->uri, "uri");
	object_lock(&entry->uri);

	return &entry->uri;
}

void
done_uri(struct uri *uri)
{
	unsigned char *string = struri(uri);
	int length = strlen(string);
	struct hash_item *item;
	struct uri_cache_entry *entry;

	assert(is_object_used(&uri_cache));

	object_unlock(uri);
	if (is_object_used(uri)) return;

	item = get_hash_item(uri_cache.map, string, length);
	entry = item ? item->value : NULL;

	assertm(entry, "Releasing unknown URI [%s]", string);
	del_hash_item(uri_cache.map, item);
	mem_free(entry);

	/* Last URI frees the cache */
	object_unlock(&uri_cache);
	if (!is_object_used(&uri_cache))
		free_hash(uri_cache.map);
}