rspamd/src/libserver/url.c

/*
 * Copyright 2024 Vsevolod Stakhov
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *    http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

#include "config.h"
#include "url.h"
#include "util.h"
#include "rspamd.h"
#include "message.h"
#include "multipattern.h"
#include "contrib/uthash/utlist.h"
#include "contrib/http-parser/http_parser.h"
#include <unicode/utf8.h>
#include <unicode/uchar.h>
#include <unicode/usprep.h>
#include <unicode/ucnv.h>

typedef struct url_match_s {
	const char *m_begin;
	gsize m_len;
	const char *pattern;
	const char *prefix;
	const char *newline_pos;
	const char *prev_newline_pos;
	gboolean add_prefix;
	char st;
} url_match_t;

#define URL_MATCHER_FLAG_NOHTML (1u << 0u)
#define URL_MATCHER_FLAG_TLD_MATCH (1u << 1u)
#define URL_MATCHER_FLAG_STAR_MATCH (1u << 2u)
#define URL_MATCHER_FLAG_REGEXP (1u << 3u)

struct url_callback_data;

static const struct {
	enum rspamd_url_protocol proto;
	const char *name;
	gsize len;
} rspamd_url_protocols[] = {
	{.proto = PROTOCOL_FILE,
	 .name = "file",
	 .len = 4},
	{.proto = PROTOCOL_FTP,
	 .name = "ftp",
	 .len = 3},
	{.proto = PROTOCOL_HTTP,
	 .name = "http",
	 .len = 4},
	{.proto = PROTOCOL_HTTPS,
	 .name = "https",
	 .len = 5},
	{.proto = PROTOCOL_MAILTO,
	 .name = "mailto",
	 .len = 6},
	{.proto = PROTOCOL_TELEPHONE,
	 .name = "tel",
	 .len = 3},
	{.proto = PROTOCOL_TELEPHONE,
	 .name = "callto",
	 .len = 3},
	{.proto = PROTOCOL_UNKNOWN,
	 .name = NULL,
	 .len = 0}};
struct url_matcher {
	const char *pattern;
	const char *prefix;

	gboolean (*start)(struct url_callback_data *cb,
					  const char *pos,
					  url_match_t *match);

	gboolean (*end)(struct url_callback_data *cb,
					const char *pos,
					url_match_t *match);

	int flags;
};

static gboolean url_file_start(struct url_callback_data *cb,
							   const char *pos,
							   url_match_t *match);

static gboolean url_file_end(struct url_callback_data *cb,
							 const char *pos,
							 url_match_t *match);

static gboolean url_web_start(struct url_callback_data *cb,
							  const char *pos,
							  url_match_t *match);

static gboolean url_web_end(struct url_callback_data *cb,
							const char *pos,
							url_match_t *match);

static gboolean url_tld_start(struct url_callback_data *cb,
							  const char *pos,
							  url_match_t *match);

static gboolean url_tld_end(struct url_callback_data *cb,
							const char *pos,
							url_match_t *match);

static gboolean url_email_start(struct url_callback_data *cb,
								const char *pos,
								url_match_t *match);

static gboolean url_email_end(struct url_callback_data *cb,
							  const char *pos,
							  url_match_t *match);

static gboolean url_tel_start(struct url_callback_data *cb,
							  const char *pos,
							  url_match_t *match);

static gboolean url_tel_end(struct url_callback_data *cb,
							const char *pos,
							url_match_t *match);

struct url_matcher static_matchers[] = {
	/* Common prefixes */
	{"file://", "", url_file_start, url_file_end,
	 0},
	{"file:\\\\", "", url_file_start, url_file_end,
	 0},
	{"ftp://", "", url_web_start, url_web_end,
	 0},
	{"ftp:\\\\", "", url_web_start, url_web_end,
	 0},
	{"sftp://", "", url_web_start, url_web_end,
	 0},
	{"http:", "", url_web_start, url_web_end,
	 0},
	{"https:", "", url_web_start, url_web_end,
	 0},
	{"news://", "", url_web_start, url_web_end,
	 0},
	{"nntp://", "", url_web_start, url_web_end,
	 0},
	{"telnet://", "", url_web_start, url_web_end,
	 0},
	{"tel:", "", url_tel_start, url_tel_end,
	 0},
	{"webcal://", "", url_web_start, url_web_end,
	 0},
	{"mailto:", "", url_email_start, url_email_end,
	 0},
	{"callto:", "", url_tel_start, url_tel_end,
	 0},
	{"h323:", "", url_web_start, url_web_end,
	 0},
	{"sip:", "", url_web_start, url_web_end,
	 0},
	{"www\\.[0-9a-z]", "http://", url_web_start, url_web_end,
	 URL_MATCHER_FLAG_REGEXP},
	{"ftp.", "ftp://", url_web_start, url_web_end,
	 0},
	/* Likely emails */
	{
		"@", "mailto://", url_email_start, url_email_end,
		0}};

struct rspamd_url_flag_name {
	const char *name;
	int flag;
	int hash;
} url_flag_names[] = {
	{"phished", RSPAMD_URL_FLAG_PHISHED, -1},
	{"numeric", RSPAMD_URL_FLAG_NUMERIC, -1},
	{"obscured", RSPAMD_URL_FLAG_OBSCURED, -1},
	{"redirected", RSPAMD_URL_FLAG_REDIRECTED, -1},
	{"html_displayed", RSPAMD_URL_FLAG_HTML_DISPLAYED, -1},
	{"text", RSPAMD_URL_FLAG_FROM_TEXT, -1},
	{"subject", RSPAMD_URL_FLAG_SUBJECT, -1},
	{"host_encoded", RSPAMD_URL_FLAG_HOSTENCODED, -1},
	{"schema_encoded", RSPAMD_URL_FLAG_SCHEMAENCODED, -1},
	{"path_encoded", RSPAMD_URL_FLAG_PATHENCODED, -1},
	{"query_encoded", RSPAMD_URL_FLAG_QUERYENCODED, -1},
	{"missing_slashes", RSPAMD_URL_FLAG_MISSINGSLASHES, -1},
	{"idn", RSPAMD_URL_FLAG_IDN, -1},
	{"has_port", RSPAMD_URL_FLAG_HAS_PORT, -1},
	{"has_user", RSPAMD_URL_FLAG_HAS_USER, -1},
	{"schemaless", RSPAMD_URL_FLAG_SCHEMALESS, -1},
	{"unnormalised", RSPAMD_URL_FLAG_UNNORMALISED, -1},
	{"zw_spaces", RSPAMD_URL_FLAG_ZW_SPACES, -1},
	{"url_displayed", RSPAMD_URL_FLAG_DISPLAY_URL, -1},
	{"image", RSPAMD_URL_FLAG_IMAGE, -1},
	{"query", RSPAMD_URL_FLAG_QUERY, -1},
	{"content", RSPAMD_URL_FLAG_CONTENT, -1},
	{"no_tld", RSPAMD_URL_FLAG_NO_TLD, -1},
	{"truncated", RSPAMD_URL_FLAG_TRUNCATED, -1},
	{"redirect_target", RSPAMD_URL_FLAG_REDIRECT_TARGET, -1},
	{"invisible", RSPAMD_URL_FLAG_INVISIBLE, -1},
	{"special", RSPAMD_URL_FLAG_SPECIAL, -1},
};


static inline khint_t rspamd_url_hash(struct rspamd_url *u);

static inline khint_t rspamd_url_host_hash(struct rspamd_url *u);
static inline bool rspamd_urls_cmp(struct rspamd_url *a, struct rspamd_url *b);
static inline bool rspamd_urls_host_cmp(struct rspamd_url *a, struct rspamd_url *b);

/* Hash table implementation */
__KHASH_IMPL(rspamd_url_hash, kh_inline, struct rspamd_url *, char, false,
			 rspamd_url_hash, rspamd_urls_cmp);
__KHASH_IMPL(rspamd_url_host_hash, kh_inline, struct rspamd_url *, char, false,
			 rspamd_url_host_hash, rspamd_urls_host_cmp);

struct url_callback_data {
	const char *begin;
	char *url_str;
	rspamd_mempool_t *pool;
	int len;
	enum rspamd_url_find_type how;
	gboolean prefix_added;
	unsigned int newline_idx;
	GArray *matchers;
	GPtrArray *newlines;
	const char *start;
	const char *fin;
	const char *end;
	const char *last_at;
	url_insert_function func;
	void *funcd;
};

struct url_match_scanner {
	GArray *matchers_full;
	GArray *matchers_strict;
	struct rspamd_multipattern *search_trie_full;
	struct rspamd_multipattern *search_trie_strict;
	bool has_tld_file;
};

struct url_match_scanner *url_scanner = NULL;

enum {
	IS_LWSP = (1 << 0),
	IS_DOMAIN = (1 << 1),
	IS_URLSAFE = (1 << 2),
	IS_MAILSAFE = (1 << 3),
	IS_DOMAIN_END = (1 << 4)
};

static const unsigned int url_scanner_table[256] = {
	0, 0, 0, 0, 0, 0, 0, 0, 0, IS_LWSP, IS_LWSP, IS_LWSP, IS_LWSP, IS_LWSP, 0, 0,
	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, IS_LWSP /*   */,
	IS_MAILSAFE /* ! */, IS_URLSAFE | IS_DOMAIN_END | IS_MAILSAFE /* " */,
	IS_MAILSAFE /* # */, IS_MAILSAFE /* $ */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* % */, 0 /* & */, IS_MAILSAFE /* ' */,
	0 /* ( */, 0 /* ) */, IS_MAILSAFE /* * */,
	IS_MAILSAFE /* + */, IS_MAILSAFE /* , */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* - */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* . */, IS_DOMAIN_END | IS_MAILSAFE /* / */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* 0 */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* 1 */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* 2 */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* 3 */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* 4 */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* 5 */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* 6 */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* 7 */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* 8 */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* 9 */, IS_DOMAIN_END /* : */,
	0 /* ; */, IS_URLSAFE | IS_DOMAIN_END /* < */, 0 /* = */,
	IS_URLSAFE | IS_DOMAIN_END /* > */, IS_DOMAIN_END /* ? */, 0 /* @ */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* A */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* B */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* C */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* D */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* E */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* F */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* G */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* H */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* I */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* J */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* K */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* L */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* M */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* N */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* O */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* P */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* Q */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* R */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* S */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* T */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* U */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* V */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* W */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* X */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* Y */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* Z */, 0 /* [ */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* \ */, 0 /* ] */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* ^ */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* _ */,
	IS_URLSAFE | IS_DOMAIN_END /* ` */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* a */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* b */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* c */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* d */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* e */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* f */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* g */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* h */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* i */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* j */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* k */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* l */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* m */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* n */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* o */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* p */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* q */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* r */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* s */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* t */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* u */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* v */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* w */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* x */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* y */,
	IS_URLSAFE | IS_DOMAIN | IS_MAILSAFE /* z */,
	IS_URLSAFE | IS_DOMAIN_END | IS_MAILSAFE /* { */,
	IS_URLSAFE | IS_DOMAIN_END | IS_MAILSAFE /* | */,
	IS_URLSAFE | IS_DOMAIN_END | IS_MAILSAFE /* } */,
	IS_URLSAFE | IS_DOMAIN_END | IS_MAILSAFE /* ~ */, 0, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN, IS_URLSAFE | IS_DOMAIN,
	IS_URLSAFE | IS_DOMAIN};

#define is_lwsp(x) ((url_scanner_table[(unsigned char) (x)] & IS_LWSP) != 0)
#define is_mailsafe(x) ((url_scanner_table[(unsigned char) (x)] & (IS_MAILSAFE)) != 0)
#define is_domain(x) ((url_scanner_table[(unsigned char) (x)] & IS_DOMAIN) != 0)
#define is_urlsafe(x) ((url_scanner_table[(unsigned char) (x)] & (IS_URLSAFE)) != 0)

const char *
rspamd_url_strerror(int err)
{
	switch (err) {
	case URI_ERRNO_OK:
		return "Parsing went well";
	case URI_ERRNO_EMPTY:
		return "The URI string was empty";
	case URI_ERRNO_INVALID_PROTOCOL:
		return "No protocol was found";
	case URI_ERRNO_BAD_FORMAT:
		return "Bad URL format";
	case URI_ERRNO_BAD_ENCODING:
		return "Invalid symbols encoded";
	case URI_ERRNO_INVALID_PORT:
		return "Port number is bad";
	case URI_ERRNO_TLD_MISSING:
		return "TLD part is not detected";
	case URI_ERRNO_HOST_MISSING:
		return "Host part is missing";
	case URI_ERRNO_TOO_LONG:
		return "URL is too long";
	}

	return NULL;
}

static gboolean
rspamd_url_parse_tld_file(const char *fname,
						  struct url_match_scanner *scanner)
{
	FILE *f;
	struct url_matcher m;
	char *linebuf = NULL, *p;
	gsize buflen = 0;
	gssize r;
	int flags;

	f = fopen(fname, "r");

	if (f == NULL) {
		msg_err("cannot open TLD file %s: %s", fname, strerror(errno));
		return FALSE;
	}

	m.end = url_tld_end;
	m.start = url_tld_start;
	m.prefix = "http://";

	while ((r = getline(&linebuf, &buflen, f)) > 0) {
		if (linebuf[0] == '/' || g_ascii_isspace(linebuf[0])) {
			/* Skip comment or empty line */
			continue;
		}

		g_strchomp(linebuf);

		/* TODO: add support for ! patterns */
		if (linebuf[0] == '!') {
			msg_debug("skip '!' patterns from parsing for now: %s", linebuf);
			continue;
		}

		flags = URL_MATCHER_FLAG_NOHTML | URL_MATCHER_FLAG_TLD_MATCH;

		if (linebuf[0] == '*') {
			flags |= URL_MATCHER_FLAG_STAR_MATCH;
			p = strchr(linebuf, '.');

			if (p == NULL) {
				msg_err("got bad star line, skip it: %s", linebuf);
				continue;
			}
			p++;
		}
		else {
			p = linebuf;
		}

		m.flags = flags;
		rspamd_multipattern_add_pattern(url_scanner->search_trie_full, p,
										RSPAMD_MULTIPATTERN_TLD | RSPAMD_MULTIPATTERN_ICASE | RSPAMD_MULTIPATTERN_UTF8);
		m.pattern = rspamd_multipattern_get_pattern(url_scanner->search_trie_full,
													rspamd_multipattern_get_npatterns(url_scanner->search_trie_full) - 1);

		g_array_append_val(url_scanner->matchers_full, m);
	}

	free(linebuf);
	fclose(f);

	return TRUE;
}

static void
rspamd_url_add_static_matchers(struct url_match_scanner *sc)
{
	int n = G_N_ELEMENTS(static_matchers), i;

	for (i = 0; i < n; i++) {
		if (static_matchers[i].flags & URL_MATCHER_FLAG_REGEXP) {
			rspamd_multipattern_add_pattern(url_scanner->search_trie_strict,
											static_matchers[i].pattern,
											RSPAMD_MULTIPATTERN_ICASE | RSPAMD_MULTIPATTERN_UTF8 |
												RSPAMD_MULTIPATTERN_RE);
		}
		else {
			rspamd_multipattern_add_pattern(url_scanner->search_trie_strict,
											static_matchers[i].pattern,
											RSPAMD_MULTIPATTERN_ICASE | RSPAMD_MULTIPATTERN_UTF8);
		}
	}

	g_array_append_vals(sc->matchers_strict, static_matchers, n);

	if (sc->matchers_full) {
		for (i = 0; i < n; i++) {
			if (static_matchers[i].flags & URL_MATCHER_FLAG_REGEXP) {
				rspamd_multipattern_add_pattern(url_scanner->search_trie_full,
												static_matchers[i].pattern,
												RSPAMD_MULTIPATTERN_ICASE | RSPAMD_MULTIPATTERN_UTF8 |
													RSPAMD_MULTIPATTERN_RE);
			}
			else {
				rspamd_multipattern_add_pattern(url_scanner->search_trie_full,
												static_matchers[i].pattern,
												RSPAMD_MULTIPATTERN_ICASE | RSPAMD_MULTIPATTERN_UTF8);
			}
		}
		g_array_append_vals(sc->matchers_full, static_matchers, n);
	}
}

void rspamd_url_deinit(void)
{
	if (url_scanner != NULL) {
		if (url_scanner->search_trie_full) {
			rspamd_multipattern_destroy(url_scanner->search_trie_full);
			g_array_free(url_scanner->matchers_full, TRUE);
		}

		rspamd_multipattern_destroy(url_scanner->search_trie_strict);
		g_array_free(url_scanner->matchers_strict, TRUE);
		g_free(url_scanner);

		url_scanner = NULL;
	}
}

void rspamd_url_init(const char *tld_file)
{
	GError *err = NULL;
	gboolean ret = TRUE;
	int mp_compile_flags = 0;

	if (url_scanner != NULL) {
		rspamd_url_deinit();
	}

	url_scanner = g_malloc(sizeof(struct url_match_scanner));

	url_scanner->matchers_strict = g_array_sized_new(FALSE, TRUE,
													 sizeof(struct url_matcher), G_N_ELEMENTS(static_matchers));
	url_scanner->search_trie_strict = rspamd_multipattern_create_sized(
		G_N_ELEMENTS(static_matchers),
		RSPAMD_MULTIPATTERN_ICASE | RSPAMD_MULTIPATTERN_UTF8);

	if (tld_file) {
		/* Reserve larger multipattern */
		url_scanner->matchers_full = g_array_sized_new(FALSE, TRUE,
													   sizeof(struct url_matcher), 13000);
		url_scanner->search_trie_full = rspamd_multipattern_create_sized(13000,
																		 RSPAMD_MULTIPATTERN_ICASE | RSPAMD_MULTIPATTERN_UTF8);
		url_scanner->has_tld_file = true;
	}
	else {
		url_scanner->matchers_full = NULL;
		url_scanner->search_trie_full = NULL;
		url_scanner->has_tld_file = false;
		mp_compile_flags |= RSPAMD_MULTIPATTERN_COMPILE_NO_FS;
	}

	rspamd_url_add_static_matchers(url_scanner);

	if (tld_file != NULL) {
		ret = rspamd_url_parse_tld_file(tld_file, url_scanner);
	}

	if (url_scanner->matchers_full && url_scanner->matchers_full->len > 1000) {
		msg_info("start compiling of %d TLD suffixes; it might take a long time",
				 url_scanner->matchers_full->len);
	}

	if (!rspamd_multipattern_compile(url_scanner->search_trie_strict, mp_compile_flags, &err)) {
		msg_err("cannot compile url matcher static patterns, fatal error: %e", err);
		abort();
	}

	if (url_scanner->search_trie_full) {
		if (!rspamd_multipattern_compile(url_scanner->search_trie_full, mp_compile_flags, &err)) {
			msg_err("cannot compile tld patterns, url matching will be "
					"incomplete: %e",
					err);
			g_error_free(err);
			ret = FALSE;
		}
	}

	if (tld_file != NULL) {
		if (ret) {
			msg_info("initialized %ud url match suffixes from '%s'",
					 url_scanner->matchers_full->len - url_scanner->matchers_strict->len,
					 tld_file);
		}
		else {
			msg_err("failed to initialize url tld suffixes from '%s', "
					"use %ud internal match suffixes",
					tld_file,
					url_scanner->matchers_strict->len);
		}
	}

	/* Generate hashes for flags */
	for (int i = 0; i < G_N_ELEMENTS(url_flag_names); i++) {
		url_flag_names[i].hash =
			rspamd_cryptobox_fast_hash_specific(RSPAMD_CRYPTOBOX_HASHFAST_INDEPENDENT,
												url_flag_names[i].name,
												strlen(url_flag_names[i].name), 0);
	}
	/* Ensure that we have no hashes collisions O(N^2) but this array is small */
	for (int i = 0; i < G_N_ELEMENTS(url_flag_names) - 1; i++) {
		for (int j = i + 1; j < G_N_ELEMENTS(url_flag_names); j++) {
			if (url_flag_names[i].hash == url_flag_names[j].hash) {
				msg_err("collision: both %s and %s map to %d",
						url_flag_names[i].name, url_flag_names[j].name,
						url_flag_names[i].hash);
				abort();
			}
		}
	}
}

#define SET_U(u, field)                             \
	do {                                            \
		if ((u) != NULL) {                          \
			(u)->field_set |= 1 << (field);         \
			(u)->field_data[(field)].len = p - c;   \
			(u)->field_data[(field)].off = c - str; \
		}                                           \
	} while (0)

static bool
is_url_start(char c)
{
	if (c == '(' ||
		c == '{' ||
		c == '[' ||
		c == '<' ||
		c == '\'') {
		return TRUE;
	}

	return FALSE;
}

static bool
is_url_end(char c)
{
	if (c == ')' ||
		c == '}' ||
		c == ']' ||
		c == '>' ||
		c == '\'') {
		return TRUE;
	}

	return FALSE;
}

static bool
is_domain_start(int p)
{
	if (g_ascii_isalnum(p) ||
		p == '[' ||
		p == '%' ||
		p == '_' ||
		(p & 0x80)) {
		return TRUE;
	}

	return FALSE;
}

static const unsigned int max_domain_length = 253;
static const unsigned int max_dns_label = 63;
static const unsigned int max_email_user = 64;

static int
rspamd_mailto_parse(struct http_parser_url *u,
					const char *str, gsize len,
					char const **end,
					enum rspamd_url_parse_flags parse_flags, unsigned int *flags)
{
	const char *p = str, *c = str, *last = str + len;
	char t;
	int ret = 1;
	enum {
		parse_mailto,
		parse_slash,
		parse_slash_slash,
		parse_semicolon,
		parse_prefix_question,
		parse_destination,
		parse_equal,
		parse_user,
		parse_at,
		parse_domain,
		parse_suffix_question,
		parse_query
	} st = parse_mailto;

	if (u != NULL) {
		memset(u, 0, sizeof(*u));
	}

	while (p < last) {
		t = *p;

		if (p - str > max_email_user + max_domain_length + 1) {
			goto out;
		}

		switch (st) {
		case parse_mailto:
			if (t == ':') {
				st = parse_semicolon;
				SET_U(u, UF_SCHEMA);
			}
			p++;
			break;
		case parse_semicolon:
			if (t == '/' || t == '\\') {
				st = parse_slash;
				p++;
			}
			else {
				*flags |= RSPAMD_URL_FLAG_MISSINGSLASHES;
				st = parse_slash_slash;
			}
			break;
		case parse_slash:
			if (t == '/' || t == '\\') {
				st = parse_slash_slash;
			}
			else {
				goto out;
			}
			p++;
			break;
		case parse_slash_slash:
			if (t == '?') {
				st = parse_prefix_question;
				p++;
			}
			else if (t != '/' && t != '\\') {
				c = p;
				st = parse_user;
			}
			else {
				/* Skip multiple slashes */
				p++;
			}
			break;
		case parse_prefix_question:
			if (t == 't') {
				/* XXX: accept only to= */
				st = parse_destination;
			}
			else {
				goto out;
			}
			break;
		case parse_destination:
			if (t == '=') {
				st = parse_equal;
			}
			p++;
			break;
		case parse_equal:
			c = p;
			st = parse_user;
			break;
		case parse_user:
			if (t == '@') {
				if (p - c == 0) {
					goto out;
				}
				SET_U(u, UF_USERINFO);
				st = parse_at;
			}
			else if (!is_mailsafe(t)) {
				goto out;
			}
			else if (p - c > max_email_user) {
				goto out;
			}
			p++;
			break;
		case parse_at:
			c = p;
			st = parse_domain;
			break;
		case parse_domain:
			if (t == '?') {
				SET_U(u, UF_HOST);
				st = parse_suffix_question;
			}
			else if (!is_domain(t) && t != '.' && t != '_') {
				goto out;
			}
			else if (p - c > max_domain_length) {
				goto out;
			}
			p++;
			break;
		case parse_suffix_question:
			c = p;
			st = parse_query;
			break;
		case parse_query:
			if (t == '#') {
				if (p - c != 0) {
					SET_U(u, UF_QUERY);
				}
				c = p + 1;
				ret = 0;

				goto out;
			}
			else if (!(parse_flags & RSPAMD_URL_PARSE_HREF) && is_url_end(t)) {
				ret = 0;
				goto out;
			}
			else if (is_lwsp(t)) {
				if (!(parse_flags & RSPAMD_URL_PARSE_CHECK)) {
					if (g_ascii_isspace(t)) {
						ret = 0;
					}
					goto out;
				}
				else {
					goto out;
				}
			}
			p++;
			break;
		}
	}

	if (st == parse_domain) {
		if (p - c != 0) {
			SET_U(u, UF_HOST);
			ret = 0;
		}
	}
	else if (st == parse_query) {
		if (p - c > 0) {
			SET_U(u, UF_QUERY);
		}

		ret = 0;
	}

out:
	if (end != NULL) {
		*end = p;
	}

	if ((parse_flags & RSPAMD_URL_PARSE_CHECK)) {
		return 0;
	}

	return ret;
}

static int
rspamd_telephone_parse(struct http_parser_url *u,
					   const char *str, gsize len,
					   char const **end,
					   enum rspamd_url_parse_flags parse_flags,
					   unsigned int *flags)
{
	enum {
		parse_protocol,
		parse_semicolon,
		parse_slash,
		parse_slash_slash,
		parse_spaces,
		parse_plus,
		parse_phone_start,
		parse_phone,
	} st = parse_protocol;

	const char *p = str, *c = str, *last = str + len;
	char t;
	int ret = 1, i;
	UChar32 uc;

	if (u != NULL) {
		memset(u, 0, sizeof(*u));
	}

	while (p < last) {
		t = *p;

		if (p - str > max_email_user) {
			goto out;
		}

		switch (st) {
		case parse_protocol:
			if (t == ':') {
				st = parse_semicolon;
				SET_U(u, UF_SCHEMA);
			}
			p++;
			break;
		case parse_semicolon:
			if (t == '/' || t == '\\') {
				st = parse_slash;
				p++;
			}
			else {
				st = parse_slash_slash;
			}
			break;
		case parse_slash:
			if (t == '/' || t == '\\') {
				st = parse_slash_slash;
			}
			else {
				goto out;
			}
			p++;
			break;
		case parse_slash_slash:
			if (g_ascii_isspace(t)) {
				st = parse_spaces;
				p++;
			}
			else if (t == '+') {
				c = p;
				st = parse_plus;
			}
			else if (t == '/') {
				/* Skip multiple slashes */
				p++;
			}
			else {
				st = parse_phone_start;
				c = p;
			}
			break;
		case parse_spaces:
			if (t == '+') {
				c = p;
				st = parse_plus;
			}
			else if (!g_ascii_isspace(t)) {
				st = parse_phone_start;
				c = p;
			}
			else {
				p++;
			}
			break;
		case parse_plus:
			c = p;
			p++;
			st = parse_phone_start;
			break;
		case parse_phone_start:
			if (*p == '%' || *p == '(' || g_ascii_isdigit(*p)) {
				st = parse_phone;
				p++;
			}
			else {
				goto out;
			}
			break;
		case parse_phone:
			i = p - str;
			U8_NEXT(str, i, len, uc);
			p = str + i;

			if (u_isdigit(uc) || uc == '(' || uc == ')' || uc == '[' || uc == ']' || u_isspace(uc) || uc == '%') {
				/* p is already incremented by U8_NEXT! */
			}
			else if (uc <= 0 || is_url_end(uc)) {
				ret = 0;
				goto set;
			}
			break;
		}
	}

set:
	if (st == parse_phone) {
		if (p - c != 0) {
			SET_U(u, UF_HOST);
			ret = 0;
		}
	}

out:
	if (end != NULL) {
		*end = p;
	}

	if ((parse_flags & RSPAMD_URL_PARSE_CHECK)) {
		return 0;
	}

	return ret;
}

static int
rspamd_web_parse(struct http_parser_url *u, const char *str, gsize len,
				 char const **end,
				 enum rspamd_url_parse_flags parse_flags,
				 unsigned int *flags)
{
	const char *p = str, *c = str, *last = str + len, *slash = NULL,
			   *password_start = NULL, *user_start = NULL;
	char t = 0;
	UChar32 uc;
	glong pt;
	int ret = 1;
	gboolean user_seen = FALSE;
	enum {
		parse_protocol,
		parse_slash,
		parse_slash_slash,
		parse_semicolon,
		parse_user,
		parse_at,
		parse_multiple_at,
		parse_password_start,
		parse_password,
		parse_domain_start,
		parse_domain,
		parse_ipv6,
		parse_port_password,
		parse_port,
		parse_suffix_slash,
		parse_path,
		parse_query,
		parse_part
	} st = parse_protocol;

	if (u != NULL) {
		memset(u, 0, sizeof(*u));
	}

	while (p < last) {
		t = *p;

		switch (st) {
		case parse_protocol:
			if (t == ':') {
				st = parse_semicolon;
				SET_U(u, UF_SCHEMA);
			}
			else if (!g_ascii_isalnum(t) && t != '+' && t != '-') {
				if ((parse_flags & RSPAMD_URL_PARSE_CHECK) && p > c) {
					/* We might have some domain, but no protocol */
					st = parse_domain_start;
					p = c;
					slash = c;
					break;
				}
				else {
					goto out;
				}
			}
			p++;
			break;
		case parse_semicolon:
			if (t == '/' || t == '\\') {
				st = parse_slash;
				p++;
			}
			else {
				st = parse_slash_slash;
				*(flags) |= RSPAMD_URL_FLAG_MISSINGSLASHES;
			}
			break;
		case parse_slash:
			if (t == '/' || t == '\\') {
				st = parse_slash_slash;
			}
			else {
				goto out;
			}
			p++;
			break;
		case parse_slash_slash:

			if (t != '/' && t != '\\') {
				c = p;
				slash = p;
				st = parse_domain_start;

				/*
				 * Unfortunately, due to brain damage of the RFC 3986 authors,
				 * we have to distinguish two possibilities here:
				 * authority = [ userinfo "@" ] host [ ":" port ]
				 * So if we have @ somewhere before hostname then we must process
				 * with the username state. Otherwise, we have to process via
				 * the hostname state. Unfortunately, there is no way to distinguish
				 * them aside of running NFA or two DFA or performing lookahead.
				 * Lookahead approach looks easier to implement.
				 */

				const char *tp = p;
				while (tp < last) {
					if (*tp == '@') {
						user_seen = TRUE;
						st = parse_user;
						break;
					}
					else if (*tp == '/' || *tp == '#' || *tp == '?') {
						st = parse_domain_start;
						break;
					}

					tp++;
				}

				if (st == parse_domain_start && *p == '[') {
					st = parse_ipv6;
					p++;
					c = p;
				}
			}
			else {
				/* Skip multiple slashes */
				p++;
			}
			break;
		case parse_ipv6:
			if (t == ']') {
				if (p - c == 0) {
					goto out;
				}
				SET_U(u, UF_HOST);
				p++;

				if (*p == ':') {
					st = parse_port;
					c = p + 1;
				}
				else if (*p == '/' || *p == '\\') {
					st = parse_path;
					c = p + 1;
				}
				else if (*p == '?') {
					st = parse_query;
					c = p + 1;
				}
				else if (*p == '#') {
					st = parse_part;
					c = p + 1;
				}
				else if (p != last) {
					goto out;
				}
			}
			else if (!g_ascii_isxdigit(t) && t != ':' && t != '.') {
				goto out;
			}
			p++;
			break;
		case parse_user:
			if (t == ':') {
				if (p - c == 0) {
					goto out;
				}
				user_start = c;
				st = parse_password_start;
			}
			else if (t == '@') {
				/* No password */
				if (p - c == 0) {
					/* We have multiple at in fact */
					st = parse_multiple_at;
					user_seen = TRUE;
					*flags |= RSPAMD_URL_FLAG_OBSCURED;

					continue;
				}

				SET_U(u, UF_USERINFO);
				*flags |= RSPAMD_URL_FLAG_HAS_USER;
				st = parse_at;
			}
			else if (!g_ascii_isgraph(t)) {
				goto out;
			}
			else if (p - c > max_email_user) {
				goto out;
			}

			p++;
			break;
		case parse_multiple_at:
			if (t != '@') {
				if (p - c == 0) {
					goto out;
				}

				/* For now, we ignore all that stuff as it is bogus */
				/* Off by one */
				p--;
				SET_U(u, UF_USERINFO);
				p++;
				*flags |= RSPAMD_URL_FLAG_HAS_USER;
				st = parse_at;
			}
			else {
				p++;
			}
			break;
		case parse_password_start:
			if (t == '@') {
				/* Empty password */
				SET_U(u, UF_USERINFO);
				if (u != NULL && u->field_data[UF_USERINFO].len > 0) {
					/* Eat semicolon */
					u->field_data[UF_USERINFO].len--;
				}
				*flags |= RSPAMD_URL_FLAG_HAS_USER;
				st = parse_at;
			}
			else {
				c = p;
				password_start = p;
				st = parse_password;
			}
			p++;
			break;
		case parse_password:
			if (t == '@') {
				/* XXX: password is not stored */
				if (u != NULL) {
					if (u->field_data[UF_USERINFO].len == 0 && password_start && user_start && password_start > user_start + 1) {
						*flags |= RSPAMD_URL_FLAG_HAS_USER;
						u->field_set |= 1u << (UF_USERINFO);
						u->field_data[UF_USERINFO].len =
							password_start - user_start - 1;
						u->field_data[UF_USERINFO].off =
							user_start - str;
					}
				}
				st = parse_at;
			}
			else if (!g_ascii_isgraph(t)) {
				goto out;
			}
			else if (p - c > max_domain_length) {
				goto out;
			}
			p++;
			break;
		case parse_at:
			c = p;

			if (t == '@') {
				*flags |= RSPAMD_URL_FLAG_OBSCURED;
				p++;
			}
			else if (t == '[') {
				st = parse_ipv6;
				p++;
				c = p;
			}
			else {
				st = parse_domain_start;
			}
			break;
		case parse_domain_start:
			if (is_domain_start(t)) {
				st = parse_domain;
			}
			else {
				goto out;
			}
			break;
		case parse_domain:
			if (p - c > max_domain_length) {
				/* Too large domain */
				goto out;
			}
			if (t == '/' || t == '\\' || t == ':' || t == '?' || t == '#') {
				if (p - c == 0) {
					goto out;
				}
				if (t == '/' || t == '\\') {
					SET_U(u, UF_HOST);
					st = parse_suffix_slash;
				}
				else if (t == '?') {
					SET_U(u, UF_HOST);
					st = parse_query;
					c = p + 1;
				}
				else if (t == '#') {
					SET_U(u, UF_HOST);
					st = parse_part;
					c = p + 1;
				}
				else if (t == ':' && !user_seen) {
					/*
					 * Here we can have both port and password, hence we need
					 * to apply some heuristic here
					 */
					st = parse_port_password;
				}
				else {
					/*
					 * We can go only for parsing port here
					 */
					SET_U(u, UF_HOST);
					st = parse_port;
					c = p + 1;
				}
				p++;
			}
			else {
				if (is_url_end(t) || is_url_start(t)) {
					goto set;
				}
				else if (*p == '@' && !user_seen) {
					/* We need to fallback and test user */
					p = slash;
					user_seen = TRUE;
					st = parse_user;
				}
				else if (*p != '.' && *p != '-' && *p != '_' && *p != '%') {
					if (*p & 0x80) {
						unsigned int i = 0;

						U8_NEXT(((const unsigned char *) p), i, last - p, uc);

						if (uc < 0) {
							/* Bad utf8 */
							goto out;
						}

						if (!u_isalnum(uc)) {
							/* Bad symbol */
							if (IS_ZERO_WIDTH_SPACE(uc)) {
								(*flags) |= RSPAMD_URL_FLAG_ZW_SPACES;
							}
							else {
								if (!u_isgraph(uc)) {
									if (!(parse_flags & RSPAMD_URL_PARSE_CHECK)) {
										goto out;
									}
									else {
										goto set;
									}
								}
							}
						}
						else {
							(*flags) |= RSPAMD_URL_FLAG_IDN;
						}

						p = p + i;
					}
					else if (is_urlsafe(*p)) {
						p++;
					}
					else {
						if (parse_flags & RSPAMD_URL_PARSE_HREF) {
							/* We have to use all shit we are given here */
							p++;
							(*flags) |= RSPAMD_URL_FLAG_OBSCURED;
						}
						else {
							if (!(parse_flags & RSPAMD_URL_PARSE_CHECK)) {
								goto out;
							}
							else {
								goto set;
							}
						}
					}
				}
				else {
					p++;
				}
			}
			break;
		case parse_port_password:
			if (g_ascii_isdigit(t)) {
				const char *tmp = p;

				while (tmp < last) {
					if (!g_ascii_isdigit(*tmp)) {
						if (*tmp == '/' || *tmp == '#' || *tmp == '?' ||
							is_url_end(*tmp) || g_ascii_isspace(*tmp)) {
							/* Port + something */
							st = parse_port;
							c = slash;
							p--;
							SET_U(u, UF_HOST);
							p++;
							c = p;
							break;
						}
						else {
							/* Not a port, bad character at the end */
							break;
						}
					}
					tmp++;
				}

				if (tmp == last) {
					/* Host + port only */
					st = parse_port;
					c = slash;
					p--;
					SET_U(u, UF_HOST);
					p++;
					c = p;
				}

				if (st != parse_port) {
					/* Fallback to user:password */
					p = slash;
					c = slash;
					user_seen = TRUE;
					st = parse_user;
				}
			}
			else {
				/* Rewind back */
				p = slash;
				c = slash;
				user_seen = TRUE;
				st = parse_user;
			}
			break;
		case parse_port:
			if (t == '/' || t == '\\') {
				pt = strtoul(c, NULL, 10);
				if (pt == 0 || pt > 65535) {
					goto out;
				}
				if (u != NULL) {
					u->port = pt;
					*flags |= RSPAMD_URL_FLAG_HAS_PORT;
				}
				st = parse_suffix_slash;
			}
			else if (t == '?') {
				pt = strtoul(c, NULL, 10);
				if (pt == 0 || pt > 65535) {
					goto out;
				}
				if (u != NULL) {
					u->port = pt;
					*flags |= RSPAMD_URL_FLAG_HAS_PORT;
				}

				c = p + 1;
				st = parse_query;
			}
			else if (t == '#') {
				pt = strtoul(c, NULL, 10);
				if (pt == 0 || pt > 65535) {
					goto out;
				}
				if (u != NULL) {
					u->port = pt;
					*flags |= RSPAMD_URL_FLAG_HAS_PORT;
				}

				c = p + 1;
				st = parse_part;
			}
			else if (is_url_end(t)) {
				goto set;
			}
			else if (!g_ascii_isdigit(t)) {
				if (!(parse_flags & RSPAMD_URL_PARSE_CHECK) ||
					!g_ascii_isspace(t)) {
					goto out;
				}
				else {
					goto set;
				}
			}
			p++;
			break;
		case parse_suffix_slash:
			if (t != '/' && t != '\\') {
				c = p;
				st = parse_path;
			}
			else {
				/* Skip extra slashes */
				p++;
			}
			break;
		case parse_path:
			if (t == '?') {
				if (p - c != 0) {
					SET_U(u, UF_PATH);
				}
				c = p + 1;
				st = parse_query;
			}
			else if (t == '#') {
				/* No query, just fragment */
				if (p - c != 0) {
					SET_U(u, UF_PATH);
				}
				c = p + 1;
				st = parse_part;
			}
			else if (!(parse_flags & RSPAMD_URL_PARSE_HREF) && is_url_end(t)) {
				goto set;
			}
			else if (is_lwsp(t)) {
				if (!(parse_flags & RSPAMD_URL_PARSE_CHECK)) {
					if (g_ascii_isspace(t)) {
						goto set;
					}
					goto out;
				}
				else {
					goto set;
				}
			}
			p++;
			break;
		case parse_query:
			if (t == '#') {
				if (p - c != 0) {
					SET_U(u, UF_QUERY);
				}
				c = p + 1;
				st = parse_part;
			}
			else if (!(parse_flags & RSPAMD_URL_PARSE_HREF) && is_url_end(t)) {
				goto set;
			}
			else if (is_lwsp(t)) {
				if (!(parse_flags & RSPAMD_URL_PARSE_CHECK)) {
					if (g_ascii_isspace(t)) {
						goto set;
					}
					goto out;
				}
				else {
					goto set;
				}
			}
			p++;
			break;
		case parse_part:
			if (!(parse_flags & RSPAMD_URL_PARSE_HREF) && is_url_end(t)) {
				goto set;
			}
			else if (is_lwsp(t)) {
				if (!(parse_flags & RSPAMD_URL_PARSE_CHECK)) {
					if (g_ascii_isspace(t)) {
						goto set;
					}
					goto out;
				}
				else {
					goto set;
				}
			}
			p++;
			break;
		}
	}

set:
	/* Parse remaining */
	switch (st) {
	case parse_domain:
		if (p - c == 0 || !is_domain(*(p - 1)) || !is_domain(*c)) {
			goto out;
		}
		SET_U(u, UF_HOST);
		ret = 0;

		break;
	case parse_port:
		pt = strtoul(c, NULL, 10);
		if (pt == 0 || pt > 65535) {
			goto out;
		}
		if (u != NULL) {
			u->port = pt;
		}

		ret = 0;
		break;
	case parse_suffix_slash:
		/* Url ends with '/' */
		ret = 0;
		break;
	case parse_path:
		if (p - c > 0) {
			SET_U(u, UF_PATH);
		}
		ret = 0;
		break;
	case parse_query:
		if (p - c > 0) {
			SET_U(u, UF_QUERY);
		}
		ret = 0;
		break;
	case parse_part:
		if (p - c > 0) {
			SET_U(u, UF_FRAGMENT);
		}
		ret = 0;
		break;
	case parse_ipv6:
		if (t != ']') {
			ret = 1;
		}
		else {
			/* e.g. http://[::] */
			ret = 0;
		}
		break;
	default:
		/* Error state */
		ret = 1;
		break;
	}
out:
	if (end != NULL) {
		*end = p;
	}

	return ret;
}

#undef SET_U

static int
rspamd_tld_trie_callback(struct rspamd_multipattern *mp,
						 unsigned int strnum,
						 int match_start,
						 int match_pos,
						 const char *text,
						 gsize len,
						 void *context)
{
	struct url_matcher *matcher;
	const char *start, *pos, *p;
	struct rspamd_url *url = context;
	int ndots;

	matcher = &g_array_index(url_scanner->matchers_full, struct url_matcher,
							 strnum);
	ndots = 1;

	if (matcher->flags & URL_MATCHER_FLAG_STAR_MATCH) {
		/* Skip one more tld component */
		ndots++;
	}

	pos = text + match_start;
	p = pos - 1;
	start = rspamd_url_host_unsafe(url);

	if (*pos != '.' || match_pos != (int) url->hostlen) {
		/* Something weird has been found */
		if (match_pos == (int) url->hostlen - 1) {
			pos = rspamd_url_host_unsafe(url) + match_pos;
			if (*pos == '.') {
				/* This is dot at the end of domain */
				url->hostlen--;
			}
			else {
				return 0;
			}
		}
		else {
			return 0;
		}
	}

	/* Now we need to find top level domain */
	pos = start;
	while (p >= start && ndots > 0) {
		if (*p == '.') {
			ndots--;
			pos = p + 1;
		}
		else {
			pos = p;
		}

		p--;
	}

	if ((ndots == 0 || p == start - 1) &&
		url->tldlen < rspamd_url_host_unsafe(url) + url->hostlen - pos) {
		url->tldshift = (pos - url->string);
		url->tldlen = rspamd_url_host_unsafe(url) + url->hostlen - pos;
	}

	return 0;
}

static void
rspamd_url_regen_from_inet_addr(struct rspamd_url *uri, const void *addr, int af,
								rspamd_mempool_t *pool)
{
	char *strbuf, *p;
	const char *start_offset;
	gsize slen = uri->urllen - uri->hostlen;
	goffset r = 0;

	if (af == AF_INET) {
		slen += INET_ADDRSTRLEN;
	}
	else {
		slen += INET6_ADDRSTRLEN;
	}

	if (uri->flags & RSPAMD_URL_FLAG_HAS_PORT) {
		slen += sizeof("65535") - 1;
	}

	/* Allocate new string to build it from IP */
	strbuf = rspamd_mempool_alloc(pool, slen + 1);
	r += rspamd_snprintf(strbuf + r, slen - r, "%*s",
						 (int) (uri->hostshift),
						 uri->string);

	uri->hostshift = r;
	uri->tldshift = r;
	start_offset = strbuf + r;
	inet_ntop(af, addr, strbuf + r, slen - r + 1);
	uri->hostlen = strlen(start_offset);
	r += uri->hostlen;
	uri->tldlen = uri->hostlen;
	uri->flags |= RSPAMD_URL_FLAG_NUMERIC;

	/* Reconstruct URL */
	if (uri->flags & RSPAMD_URL_FLAG_HAS_PORT && uri->ext) {
		p = strbuf + r;
		start_offset = p + 1;
		r += rspamd_snprintf(strbuf + r, slen - r, ":%ud",
							 (unsigned int) uri->ext->port);
	}
	if (uri->datalen > 0) {
		p = strbuf + r;
		start_offset = p + 1;
		r += rspamd_snprintf(strbuf + r, slen - r, "/%*s",
							 (int) uri->datalen,
							 rspamd_url_data_unsafe(uri));
		uri->datashift = start_offset - strbuf;
	}
	else {
		/* Add trailing slash if needed */
		if (uri->hostlen + uri->hostshift < uri->urllen &&
			*(rspamd_url_host_unsafe(uri) + uri->hostlen) == '/') {
			r += rspamd_snprintf(strbuf + r, slen - r, "/");
		}
	}

	if (uri->querylen > 0) {
		p = strbuf + r;
		start_offset = p + 1;
		r += rspamd_snprintf(strbuf + r, slen - r, "?%*s",
							 (int) uri->querylen,
							 rspamd_url_query_unsafe(uri));
		uri->queryshift = start_offset - strbuf;
	}
	if (uri->fragmentlen > 0) {
		p = strbuf + r;
		start_offset = p + 1;
		r += rspamd_snprintf(strbuf + r, slen - r, "#%*s",
							 (int) uri->fragmentlen,
							 rspamd_url_fragment_unsafe(uri));
		uri->fragmentshift = start_offset - strbuf;
	}

	uri->string = strbuf;
	uri->urllen = r;
}

static gboolean
rspamd_url_maybe_regenerate_from_ip(struct rspamd_url *uri, rspamd_mempool_t *pool)
{
	const char *p, *end, *c;
	char *errstr;
	struct in_addr in4;
	struct in6_addr in6;
	gboolean ret = FALSE, check_num = TRUE;
	uint32_t n, dots, t = 0, i = 0, shift, nshift;

	p = rspamd_url_host_unsafe(uri);
	end = p + uri->hostlen;

	if (*p == '[' && *(end - 1) == ']') {
		p++;
		end--;
	}

	while (*(end - 1) == '.' && end > p) {
		end--;
	}

	if (end - p == 0 || end - p > INET6_ADDRSTRLEN) {
		return FALSE;
	}

	if (rspamd_str_has_8bit(p, end - p)) {
		return FALSE;
	}

	if (rspamd_parse_inet_address_ip4(p, end - p, &in4)) {
		rspamd_url_regen_from_inet_addr(uri, &in4, AF_INET, pool);
		ret = TRUE;
	}
	else if (rspamd_parse_inet_address_ip6(p, end - p, &in6)) {
		rspamd_url_regen_from_inet_addr(uri, &in6, AF_INET6, pool);
		ret = TRUE;
	}
	else {
		/* Heuristics for broken urls */
		char buf[INET6_ADDRSTRLEN + 1];
		/* Try also numeric notation */
		c = p;
		n = 0;
		dots = 0;
		shift = 0;

		while (p <= end && check_num) {
			if (shift < 32 &&
				((*p == '.' && dots < 3) || (p == end && dots <= 3))) {
				if (p - c + 1 >= (int) sizeof(buf)) {
					msg_debug_pool("invalid numeric url %*.s...: too long",
								   INET6_ADDRSTRLEN, c);
					return FALSE;
				}

				rspamd_strlcpy(buf, c, p - c + 1);
				c = p + 1;

				if (p < end && *p == '.') {
					dots++;
				}

				glong long_n = strtol(buf, &errstr, 0);

				if ((errstr == NULL || *errstr == '\0') && long_n >= 0) {

					t = long_n; /* Truncate as windows does */
					/*
					 * Even if we have zero, we need to shift by 1 octet
					 */
					nshift = (t == 0 ? shift + 8 : shift);

					/*
					 * Here we count number of octets encoded in this element
					 */
					for (i = 0; i < 4; i++) {
						if ((t >> (8 * i)) > 0) {
							nshift += 8;
						}
						else {
							break;
						}
					}
					/*
					 * Here we need to find the proper shift of the previous
					 * components, so we check possible cases:
					 * 1) 1 octet - just use it applying shift
					 * 2) 2 octets - convert to big endian 16 bit number
					 * 3) 3 octets - convert to big endian 24 bit number
					 * 4) 4 octets - convert to big endian 32 bit number
					 */
					switch (i) {
					case 4:
						t = GUINT32_TO_BE(t);
						break;
					case 3:
						t = (GUINT32_TO_BE(t & 0xFFFFFFU)) >> 8;
						break;
					case 2:
						t = GUINT16_TO_BE(t & 0xFFFFU);
						break;
					default:
						t = t & 0xFF;
						break;
					}

					if (p != end) {
						n |= t << shift;

						shift = nshift;
					}
				}
				else {
					check_num = FALSE;
				}
			}

			p++;
		}

		/* The last component should be last according to url normalization:
		 * 192.168.1 -> 192.168.0.1
		 * 192 -> 0.0.0.192
		 * 192.168 -> 192.0.0.168
		 */
		shift = 8 * (4 - i);

		if (shift < 32) {
			n |= t << shift;
		}

		if (check_num) {
			if (dots <= 4) {
				memcpy(&in4, &n, sizeof(in4));
				rspamd_url_regen_from_inet_addr(uri, &in4, AF_INET, pool);
				uri->flags |= RSPAMD_URL_FLAG_OBSCURED;
				ret = TRUE;
			}
			else if (end - c > (int) sizeof(buf) - 1) {
				rspamd_strlcpy(buf, c, end - c + 1);

				if (inet_pton(AF_INET6, buf, &in6) == 1) {
					rspamd_url_regen_from_inet_addr(uri, &in6, AF_INET6, pool);
					uri->flags |= RSPAMD_URL_FLAG_OBSCURED;
					ret = TRUE;
				}
			}
		}
	}

	return ret;
}

static void
rspamd_url_shift(struct rspamd_url *uri, gsize nlen,
				 enum http_parser_url_fields field)
{
	unsigned int old_shift, shift = 0;
	int remain;

	/* Shift remaining data */
	switch (field) {
	case UF_SCHEMA:
		if (nlen >= uri->protocollen) {
			return;
		}
		else {
			shift = uri->protocollen - nlen;
		}

		old_shift = uri->protocollen;
		uri->protocollen -= shift;
		remain = uri->urllen - uri->protocollen;
		g_assert(remain >= 0);
		memmove(uri->string + uri->protocollen, uri->string + old_shift,
				remain);
		uri->urllen -= shift;
		uri->flags |= RSPAMD_URL_FLAG_SCHEMAENCODED;
		break;
	case UF_HOST:
		if (nlen >= uri->hostlen) {
			return;
		}
		else {
			shift = uri->hostlen - nlen;
		}

		old_shift = uri->hostlen;
		uri->hostlen -= shift;
		remain = (uri->urllen - (uri->hostshift)) - old_shift;
		g_assert(remain >= 0);
		memmove(rspamd_url_host_unsafe(uri) + uri->hostlen,
				rspamd_url_host_unsafe(uri) + old_shift,
				remain);
		uri->urllen -= shift;
		uri->flags |= RSPAMD_URL_FLAG_HOSTENCODED;
		break;
	case UF_PATH:
		if (nlen >= uri->datalen) {
			return;
		}
		else {
			shift = uri->datalen - nlen;
		}

		old_shift = uri->datalen;
		uri->datalen -= shift;
		remain = (uri->urllen - (uri->datashift)) - old_shift;
		g_assert(remain >= 0);
		memmove(rspamd_url_data_unsafe(uri) + uri->datalen,
				rspamd_url_data_unsafe(uri) + old_shift,
				remain);
		uri->urllen -= shift;
		uri->flags |= RSPAMD_URL_FLAG_PATHENCODED;
		break;
	case UF_QUERY:
		if (nlen >= uri->querylen) {
			return;
		}
		else {
			shift = uri->querylen - nlen;
		}

		old_shift = uri->querylen;
		uri->querylen -= shift;
		remain = (uri->urllen - (uri->queryshift)) - old_shift;
		g_assert(remain >= 0);
		memmove(rspamd_url_query_unsafe(uri) + uri->querylen,
				rspamd_url_query_unsafe(uri) + old_shift,
				remain);
		uri->urllen -= shift;
		uri->flags |= RSPAMD_URL_FLAG_QUERYENCODED;
		break;
	case UF_FRAGMENT:
		if (nlen >= uri->fragmentlen) {
			return;
		}
		else {
			shift = uri->fragmentlen - nlen;
		}

		uri->fragmentlen -= shift;
		uri->urllen -= shift;
		break;
	default:
		break;
	}

	/* Now adjust lengths and offsets */
	switch (field) {
	case UF_SCHEMA:
		if (uri->userlen > 0) {
			uri->usershift -= shift;
		}
		if (uri->hostlen > 0) {
			uri->hostshift -= shift;
		}
		/* Go forward */
		/* FALLTHRU */
	case UF_HOST:
		if (uri->datalen > 0) {
			uri->datashift -= shift;
		}
		/* Go forward */
		/* FALLTHRU */
	case UF_PATH:
		if (uri->querylen > 0) {
			uri->queryshift -= shift;
		}
		/* Go forward */
		/* FALLTHRU */
	case UF_QUERY:
		if (uri->fragmentlen > 0) {
			uri->fragmentshift -= shift;
		}
		/* Go forward */
		/* FALLTHRU */
	case UF_FRAGMENT:
	default:
		break;
	}
}

static void
rspamd_telephone_normalise_inplace(struct rspamd_url *uri)
{
	char *t, *h, *end;
	int i = 0, w, orig_len;
	UChar32 uc;

	t = rspamd_url_host_unsafe(uri);
	h = t;
	end = t + uri->hostlen;
	orig_len = uri->hostlen;

	if (*h == '+') {
		h++;
		t++;
	}

	while (h < end) {
		i = 0;
		U8_NEXT(h, i, end - h, uc);

		if (u_isdigit(uc)) {
			w = 0;
			U8_APPEND_UNSAFE(t, w, uc);
			t += w;
		}

		h += i;
	}

	uri->hostlen = t - rspamd_url_host_unsafe(uri);
	uri->urllen -= (orig_len - uri->hostlen);
}

static inline bool
is_idna_label_dot(UChar ch)
{
	switch (ch) {
	case 0x3002:
	case 0xFF0E:
	case 0xFF61:
		return true;
	default:
		return false;
	}
}

/*
 * All credits for this investigation should go to
 * Dr. Hajime Shimada and Mr. Shirakura as they have revealed this case in their
 * research.
 */

/*
 * This function replaces unsafe IDNA dots in host labels. Unfortunately,
 * IDNA extends dot definition from '.' to multiple other characters that
 * should be treated equally.
 * This function replaces such dots and returns `true` if these dots are found.
 * In this case, it should be treated as obfuscation attempt.
 */
static bool
rspamd_url_remove_dots(struct rspamd_url *uri)
{
	const char *hstart = rspamd_url_host_unsafe(uri);
	char *t;
	UChar32 uc;
	int i = 0, hlen;
	bool ret = false;

	if (uri->hostlen == 0) {
		return false;
	}

	hlen = uri->hostlen;
	t = rspamd_url_host_unsafe(uri);

	while (i < hlen) {
		int prev_i = i;
		U8_NEXT(hstart, i, hlen, uc);

		if (is_idna_label_dot(uc)) {
			*t++ = '.';
			ret = true;
		}
		else {
			if (ret) {
				/* We have to shift the remaining stuff */
				while (prev_i < i) {
					*t++ = *(hstart + prev_i);
					prev_i++;
				}
			}
			else {
				t += (i - prev_i);
			}
		}
	}

	if (ret) {
		rspamd_url_shift(uri, t - hstart, UF_HOST);
	}

	return ret;
}

enum uri_errno
rspamd_url_parse(struct rspamd_url *uri,
				 char *uristring, gsize len,
				 rspamd_mempool_t *pool,
				 enum rspamd_url_parse_flags parse_flags)
{
	struct http_parser_url u;
	char *p;
	const char *end;
	unsigned int complen, ret, flags = 0;
	gsize unquoted_len = 0;

	memset(uri, 0, sizeof(*uri));
	memset(&u, 0, sizeof(u));
	uri->count = 1;
	/* Undefine order */
	uri->order = -1;
	uri->part_order = -1;

	if (*uristring == '\0') {
		return URI_ERRNO_EMPTY;
	}

	if (len >= G_MAXUINT16 / 2) {
		flags |= RSPAMD_URL_FLAG_TRUNCATED;
		len = G_MAXUINT16 / 2;
	}

	p = uristring;
	uri->protocol = PROTOCOL_UNKNOWN;

	if (len > sizeof("mailto:") - 1) {
		/* For mailto: urls we also need to add slashes to make it a valid URL */
		if (g_ascii_strncasecmp(p, "mailto:", sizeof("mailto:") - 1) == 0) {
			ret = rspamd_mailto_parse(&u, uristring, len, &end, parse_flags,
									  &flags);
		}
		else if (g_ascii_strncasecmp(p, "tel:", sizeof("tel:") - 1) == 0 ||
				 g_ascii_strncasecmp(p, "callto:", sizeof("callto:") - 1) == 0) {
			ret = rspamd_telephone_parse(&u, uristring, len, &end, parse_flags,
										 &flags);
			uri->protocol = PROTOCOL_TELEPHONE;
		}
		else {
			ret = rspamd_web_parse(&u, uristring, len, &end, parse_flags,
								   &flags);
		}
	}
	else {
		ret = rspamd_web_parse(&u, uristring, len, &end, parse_flags, &flags);
	}

	if (ret != 0) {
		return URI_ERRNO_BAD_FORMAT;
	}

	if (end > uristring && (unsigned int) (end - uristring) != len) {
		len = end - uristring;
	}

	uri->raw = p;
	uri->rawlen = len;

	if (flags & RSPAMD_URL_FLAG_MISSINGSLASHES) {
		len += 2;
		uri->string = rspamd_mempool_alloc(pool, len + 1);
		memcpy(uri->string, p, u.field_data[UF_SCHEMA].len);
		memcpy(uri->string + u.field_data[UF_SCHEMA].len, "://", 3);
		rspamd_strlcpy(uri->string + u.field_data[UF_SCHEMA].len + 3,
					   p + u.field_data[UF_SCHEMA].len + 1,
					   len - 2 - u.field_data[UF_SCHEMA].len);
		/* Compensate slashes added */
		for (int i = UF_SCHEMA + 1; i < UF_MAX; i++) {
			if (u.field_set & (1 << i)) {
				u.field_data[i].off += 2;
			}
		}
	}
	else {
		uri->string = rspamd_mempool_alloc(pool, len + 1);
		rspamd_strlcpy(uri->string, p, len + 1);
	}

	uri->urllen = len;
	uri->flags = flags;

	for (unsigned int i = 0; i < UF_MAX; i++) {
		if (u.field_set & (1 << i)) {
			unsigned int shift = u.field_data[i].off;
			complen = u.field_data[i].len;

			if (complen >= G_MAXUINT16) {
				/* Too large component length */
				return URI_ERRNO_BAD_FORMAT;
			}

			switch (i) {
			case UF_SCHEMA:
				uri->protocollen = u.field_data[i].len;
				break;
			case UF_HOST:
				uri->hostshift = shift;
				uri->hostlen = complen;
				break;
			case UF_PATH:
				uri->datashift = shift;
				uri->datalen = complen;
				break;
			case UF_QUERY:
				uri->queryshift = shift;
				uri->querylen = complen;
				break;
			case UF_FRAGMENT:
				uri->fragmentshift = shift;
				uri->fragmentlen = complen;
				break;
			case UF_USERINFO:
				uri->usershift = shift;
				uri->userlen = complen;
				break;
			default:
				break;
			}
		}
	}

	/* Port is 'special' in case of url_parser as it is not a part of UF_* macro logic */
	if (u.port != 0) {
		if (!uri->ext) {
			uri->ext = rspamd_mempool_alloc0_type(pool, struct rspamd_url_ext);
		}
		uri->flags |= RSPAMD_URL_FLAG_HAS_PORT;
		uri->ext->port = u.port;
	}

	if (!uri->hostlen) {
		return URI_ERRNO_HOST_MISSING;
	}

	/* Now decode url symbols */
	unquoted_len = rspamd_url_decode(uri->string,
									 uri->string,
									 uri->protocollen);
	rspamd_url_shift(uri, unquoted_len, UF_SCHEMA);
	unquoted_len = rspamd_url_decode(rspamd_url_host_unsafe(uri),
									 rspamd_url_host_unsafe(uri), uri->hostlen);

	rspamd_url_normalise_propagate_flags(pool, rspamd_url_host_unsafe(uri),
										 &unquoted_len, uri->flags);

	rspamd_url_shift(uri, unquoted_len, UF_HOST);

	if (rspamd_url_remove_dots(uri)) {
		uri->flags |= RSPAMD_URL_FLAG_OBSCURED;
	}

	if (uri->protocol & (PROTOCOL_HTTP | PROTOCOL_HTTPS | PROTOCOL_MAILTO | PROTOCOL_FTP | PROTOCOL_FILE)) {
		/* Ensure that hostname starts with something sane (exclude numeric urls) */
		const char *host = rspamd_url_host_unsafe(uri);

		if (!(is_domain_start(host[0]) || host[0] == ':')) {
			return URI_ERRNO_BAD_FORMAT;
		}
	}

	/* Apply nameprep algorithm */
	static UStringPrepProfile *nameprep = NULL;
	UErrorCode uc_err = U_ZERO_ERROR;

	if (nameprep == NULL) {
		/* Open and cache profile */
		nameprep = usprep_openByType(USPREP_RFC3491_NAMEPREP, &uc_err);

		g_assert(U_SUCCESS(uc_err));
	}

	UChar *utf16_hostname, *norm_utf16;
	int32_t utf16_len, norm_utf16_len, norm_utf8_len;
	UParseError parse_error;

	utf16_hostname = rspamd_mempool_alloc(pool, uri->hostlen * sizeof(UChar));
	struct UConverter *utf8_conv = rspamd_get_utf8_converter();

	utf16_len = ucnv_toUChars(utf8_conv, utf16_hostname, uri->hostlen,
							  rspamd_url_host_unsafe(uri), uri->hostlen, &uc_err);

	if (!U_SUCCESS(uc_err)) {

		return URI_ERRNO_BAD_FORMAT;
	}

	norm_utf16 = rspamd_mempool_alloc(pool, utf16_len * sizeof(UChar));
	norm_utf16_len = usprep_prepare(nameprep, utf16_hostname, utf16_len,
									norm_utf16, utf16_len, USPREP_DEFAULT, &parse_error, &uc_err);

	if (!U_SUCCESS(uc_err)) {

		return URI_ERRNO_BAD_FORMAT;
	}

	/* Convert back to utf8, sigh... */
	norm_utf8_len = ucnv_fromUChars(utf8_conv,
									rspamd_url_host_unsafe(uri), uri->hostlen,
									norm_utf16, norm_utf16_len, &uc_err);

	if (!U_SUCCESS(uc_err)) {

		return URI_ERRNO_BAD_FORMAT;
	}

	/* Final shift of lengths */
	rspamd_url_shift(uri, norm_utf8_len, UF_HOST);

	/* Process data part */
	if (uri->datalen) {
		unquoted_len = rspamd_url_decode(rspamd_url_data_unsafe(uri),
										 rspamd_url_data_unsafe(uri), uri->datalen);

		rspamd_url_normalise_propagate_flags(pool, rspamd_url_data_unsafe(uri),
											 &unquoted_len, uri->flags);

		rspamd_url_shift(uri, unquoted_len, UF_PATH);
		/* We now normalize path */
		rspamd_normalize_path_inplace(rspamd_url_data_unsafe(uri),
									  uri->datalen, &unquoted_len);
		rspamd_url_shift(uri, unquoted_len, UF_PATH);
	}

	if (uri->querylen) {
		unquoted_len = rspamd_url_decode(rspamd_url_query_unsafe(uri),
										 rspamd_url_query_unsafe(uri),
										 uri->querylen);

		rspamd_url_normalise_propagate_flags(pool, rspamd_url_query_unsafe(uri),
											 &unquoted_len, uri->flags);
		rspamd_url_shift(uri, unquoted_len, UF_QUERY);
	}

	if (uri->fragmentlen) {
		unquoted_len = rspamd_url_decode(rspamd_url_fragment_unsafe(uri),
										 rspamd_url_fragment_unsafe(uri),
										 uri->fragmentlen);

		rspamd_url_normalise_propagate_flags(pool, rspamd_url_fragment_unsafe(uri),
											 &unquoted_len, uri->flags);
		rspamd_url_shift(uri, unquoted_len, UF_FRAGMENT);
	}

	rspamd_str_lc(uri->string, uri->protocollen);
	unquoted_len = rspamd_str_lc_utf8(rspamd_url_host_unsafe(uri), uri->hostlen);
	rspamd_url_shift(uri, unquoted_len, UF_HOST);

	if (uri->protocol == PROTOCOL_UNKNOWN) {
		for (int i = 0; i < G_N_ELEMENTS(rspamd_url_protocols); i++) {
			if (uri->protocollen == rspamd_url_protocols[i].len) {
				if (memcmp(uri->string,
						   rspamd_url_protocols[i].name, uri->protocollen) == 0) {
					uri->protocol = rspamd_url_protocols[i].proto;
					break;
				}
			}
		}
	}

	if (uri->protocol & (PROTOCOL_HTTP | PROTOCOL_HTTPS | PROTOCOL_MAILTO | PROTOCOL_FTP | PROTOCOL_FILE)) {
		/* Find TLD part */
		if (url_scanner->search_trie_full) {
			rspamd_multipattern_lookup(url_scanner->search_trie_full,
									   rspamd_url_host_unsafe(uri), uri->hostlen,
									   rspamd_tld_trie_callback, uri, NULL);
		}

		if (uri->tldlen == 0) {
			/*
			 * If we have not detected eSLD, but there are no dots in the hostname,
			 * then we should treat the whole hostname as eSLD - a rule of thumb
			 *
			 * We also check that a hostname ends with a permitted character, and all characters are forming
			 * DNS label. We also need to check for a numeric IP within this check.
			 */
			const char *dot_pos = memchr(rspamd_url_host_unsafe(uri), '.', uri->hostlen);
			bool is_whole_hostname_tld = false;

			if (uri->hostlen > 0 && (dot_pos == NULL || dot_pos == rspamd_url_host_unsafe(uri) + uri->hostlen - 1)) {
				bool all_chars_domain = true;

				for (int i = 0; i < uri->hostlen; i++) {
					if (!is_domain(rspamd_url_host_unsafe(uri)[i])) {
						all_chars_domain = false;
						break;
					}
				}

				char last_c = rspamd_url_host_unsafe(uri)[uri->hostlen - 1];

				if (all_chars_domain) {
					/* Also check the last character to be either a dot or alphanumeric character */
					if (last_c != '.' && !g_ascii_isalnum(last_c)) {
						all_chars_domain = false;
					}
				}

				if (all_chars_domain) {
					/* Additionally check for a numeric IP as we can have some number here... */
					rspamd_url_maybe_regenerate_from_ip(uri, pool);

					if (last_c == '.' && uri->hostlen > 1) {
						/* Skip the last dot */
						uri->tldlen = uri->hostlen - 1;
					}
					else {
						uri->tldlen = uri->hostlen;
					}

					uri->tldshift = uri->hostshift;
					is_whole_hostname_tld = true;
				}
			}

			if (!is_whole_hostname_tld) {
				if (uri->protocol != PROTOCOL_MAILTO) {
					if (url_scanner->has_tld_file && !(parse_flags & RSPAMD_URL_PARSE_HREF)) {
						/* Ignore URL's without TLD if it is not a numeric URL */
						if (!rspamd_url_maybe_regenerate_from_ip(uri, pool)) {
							return URI_ERRNO_TLD_MISSING;
						}
					}
					else {
						if (!rspamd_url_maybe_regenerate_from_ip(uri, pool)) {
							/* Assume tld equal to host */
							uri->tldshift = uri->hostshift;
							uri->tldlen = uri->hostlen;
						}
						else if (uri->flags & RSPAMD_URL_FLAG_SCHEMALESS) {
							/* Ignore urls with both no schema and no tld */
							return URI_ERRNO_TLD_MISSING;
						}

						uri->flags |= RSPAMD_URL_FLAG_NO_TLD;
					}
				}
				else {
					/* Ignore IP like domains for mailto, as it is really never supported */
					return URI_ERRNO_TLD_MISSING;
				}
			}
		}

		/* Replace stupid '\' with '/' after schema */
		if (uri->protocol & (PROTOCOL_HTTP | PROTOCOL_HTTPS | PROTOCOL_FTP) &&
			uri->protocollen > 0 && uri->urllen > uri->protocollen + 2) {

			char *pos = &uri->string[uri->protocollen],
				 *host_start = rspamd_url_host_unsafe(uri);

			while (pos < host_start) {
				if (*pos == '\\') {
					*pos = '/';
					uri->flags |= RSPAMD_URL_FLAG_OBSCURED;
				}
				pos++;
			}
		}
	}
	else if (uri->protocol & PROTOCOL_TELEPHONE) {
		/* We need to normalise phone number: remove all spaces and braces */
		rspamd_telephone_normalise_inplace(uri);

		if (rspamd_url_host_unsafe(uri)[0] == '+') {
			uri->tldshift = uri->hostshift + 1;
			uri->tldlen = uri->hostlen - 1;
		}
		else {
			uri->tldshift = uri->hostshift;
			uri->tldlen = uri->hostlen;
		}
	}

	if (uri->protocol == PROTOCOL_UNKNOWN) {
		if (!(parse_flags & RSPAMD_URL_PARSE_HREF)) {
			return URI_ERRNO_INVALID_PROTOCOL;
		}
		else {
			/* Hack, hack, hack */
			uri->protocol = PROTOCOL_UNKNOWN;
		}
	}

	return URI_ERRNO_OK;
}

struct tld_trie_cbdata {
	const char *begin;
	gsize len;
	rspamd_ftok_t *out;
};

static int
rspamd_tld_trie_find_callback(struct rspamd_multipattern *mp,
							  unsigned int strnum,
							  int match_start,
							  int match_pos,
							  const char *text,
							  gsize len,
							  void *context)
{
	struct url_matcher *matcher;
	const char *start, *pos, *p;
	struct tld_trie_cbdata *cbdata = context;
	int ndots = 1;

	matcher = &g_array_index(url_scanner->matchers_full, struct url_matcher,
							 strnum);

	if (matcher->flags & URL_MATCHER_FLAG_STAR_MATCH) {
		/* Skip one more tld component */
		ndots = 2;
	}

	pos = text + match_start;
	p = pos - 1;
	start = text;

	if (*pos != '.' || match_pos != (int) cbdata->len) {
		/* Something weird has been found */
		if (match_pos != (int) cbdata->len - 1) {
			/* Search more */
			return 0;
		}
	}

	/* Now we need to find top level domain */
	pos = start;

	while (p >= start && ndots > 0) {
		if (*p == '.') {
			ndots--;
			pos = p + 1;
		}
		else {
			pos = p;
		}

		p--;
	}

	if (ndots == 0 || p == start - 1) {
		if (cbdata->begin + cbdata->len - pos > cbdata->out->len) {
			cbdata->out->begin = pos;
			cbdata->out->len = cbdata->begin + cbdata->len - pos;
		}
	}

	return 0;
}

gboolean
rspamd_url_find_tld(const char *in, gsize inlen, rspamd_ftok_t *out)
{
	struct tld_trie_cbdata cbdata;

	g_assert(in != NULL);
	g_assert(out != NULL);
	g_assert(url_scanner != NULL);

	cbdata.begin = in;
	cbdata.len = inlen;
	cbdata.out = out;
	out->len = 0;

	if (url_scanner->search_trie_full) {
		rspamd_multipattern_lookup(url_scanner->search_trie_full, in, inlen,
								   rspamd_tld_trie_find_callback, &cbdata, NULL);
	}

	if (out->len > 0) {
		return TRUE;
	}

	return FALSE;
}

static const char url_braces[] = {
	'(', ')',
	'{', '}',
	'[', ']',
	'<', '>',
	'|', '|',
	'\'', '\''};


static gboolean
url_file_start(struct url_callback_data *cb,
			   const char *pos,
			   url_match_t *match)
{
	match->m_begin = pos;

	if (pos > cb->begin) {
		match->st = *(pos - 1);
	}
	else {
		match->st = '\0';
	}

	return TRUE;
}

static gboolean
url_file_end(struct url_callback_data *cb,
			 const char *pos,
			 url_match_t *match)
{
	const char *p;
	char stop;
	unsigned int i;

	p = pos + strlen(match->pattern);
	stop = *p;
	if (*p == '/') {
		p++;
	}

	for (i = 0; i < G_N_ELEMENTS(url_braces) / 2; i += 2) {
		if (*p == url_braces[i]) {
			stop = url_braces[i + 1];
			break;
		}
	}

	while (p < cb->end && *p != stop && is_urlsafe(*p)) {
		p++;
	}

	if (p == cb->begin) {
		return FALSE;
	}
	match->m_len = p - match->m_begin;

	return TRUE;
}

static gboolean
url_tld_start(struct url_callback_data *cb,
			  const char *pos,
			  url_match_t *match)
{
	const char *p = pos;
	unsigned int processed = 0;
	static const unsigned int max_shift = 253 + sizeof("https://");

	/* Try to find the start of the url by finding any non-urlsafe character or whitespace/punctuation */
	while (p >= cb->begin) {
		if (!is_domain(*p) || g_ascii_isspace(*p) || is_url_start(*p) ||
			p == match->prev_newline_pos) {
			if (!is_url_start(*p) && !g_ascii_isspace(*p) &&
				p != match->prev_newline_pos) {
				return FALSE;
			}

			if (p != match->prev_newline_pos) {
				match->st = *p;

				p++;
			}
			else {
				match->st = '\n';
			}

			if (!g_ascii_isalnum(*p)) {
				/* Urls cannot start with strange symbols */
				return FALSE;
			}

			match->m_begin = p;
			return TRUE;
		}
		else if (p == cb->begin && p != pos) {
			match->st = '\0';
			match->m_begin = p;

			return TRUE;
		}
		else if (*p == '.') {
			if (p == cb->begin) {
				/* Urls cannot start with a dot */
				return FALSE;
			}
			if (!g_ascii_isalnum(p[1])) {
				/* Wrong we have an invalid character after dot */
				return FALSE;
			}
		}
		else if (*p == '/') {
			/* Urls cannot contain '/' in their body */
			return FALSE;
		}

		p--;
		processed++;

		if (processed > max_shift) {
			/* Too long */
			return FALSE;
		}
	}

	return FALSE;
}

static gboolean
url_tld_end(struct url_callback_data *cb,
			const char *pos,
			url_match_t *match)
{
	const char *p;
	gboolean ret = FALSE;

	p = pos + match->m_len;

	if (p == cb->end) {
		match->m_len = p - match->m_begin;
		return TRUE;
	}
	else if (*p == '/' || *p == ':' || is_url_end(*p) || is_lwsp(*p) ||
			 (match->st != '<' && p == match->newline_pos)) {
		/* Parse arguments, ports by normal way by url default function */
		p = match->m_begin;
		/* Check common prefix */
		if (g_ascii_strncasecmp(p, "http://", sizeof("http://") - 1) == 0) {
			ret = url_web_end(cb,
							  match->m_begin + sizeof("http://") - 1,
							  match);
		}
		else {
			ret = url_web_end(cb, match->m_begin, match);
		}
	}
	else if (*p == '.') {
		p++;
		if (p < cb->end) {
			if (g_ascii_isspace(*p) || *p == '/' ||
				*p == '?' || *p == ':') {
				ret = url_web_end(cb, match->m_begin, match);
			}
		}
	}

	if (ret) {
		/* Check sanity of match found */
		if (match->m_begin + match->m_len <= pos) {
			return FALSE;
		}
	}

	return ret;
}

static gboolean
url_web_start(struct url_callback_data *cb,
			  const char *pos,
			  url_match_t *match)
{
	/* Check what we have found */
	if (pos > cb->begin) {
		if (g_ascii_strncasecmp(pos, "www", 3) == 0) {

			if (!(is_url_start(*(pos - 1)) ||
				  g_ascii_isspace(*(pos - 1)) ||
				  pos - 1 == match->prev_newline_pos ||
				  (*(pos - 1) & 0x80))) { /* Chinese trick */
				return FALSE;
			}
		}
		else {
			unsigned char prev = *(pos - 1);

			if (g_ascii_isalnum(prev)) {
				/* Part of another url */
				return FALSE;
			}
		}
	}

	if (*pos == '.') {
		/* Urls cannot start with . */
		return FALSE;
	}

	if (pos > cb->begin) {
		match->st = *(pos - 1);
	}
	else {
		match->st = '\0';
	}

	match->m_begin = pos;

	return TRUE;
}

static gboolean
url_web_end(struct url_callback_data *cb,
			const char *pos,
			url_match_t *match)
{
	const char *last = NULL;
	int len = cb->end - pos;
	unsigned int flags = 0;

	if (match->newline_pos && match->st != '<') {
		/* We should also limit our match end to the newline */
		len = MIN(len, match->newline_pos - pos);
	}

	if (rspamd_web_parse(NULL, pos, len, &last,
						 RSPAMD_URL_PARSE_CHECK, &flags) != 0) {
		return FALSE;
	}

	if (last < cb->end && (*last == '>' && last != match->newline_pos)) {
		/* We need to ensure that url also starts with '>' */
		if (match->st != '<') {
			if (last + 1 < cb->end) {
				if (g_ascii_isspace(last[1])) {
					return FALSE;
				}
			}
			else {
				return FALSE;
			}
		}
	}

	match->m_len = (last - pos);
	cb->fin = last + 1;

	return TRUE;
}


static gboolean
url_email_start(struct url_callback_data *cb,
				const char *pos,
				url_match_t *match)
{
	if (!match->prefix || match->prefix[0] == '\0') {
		/* We have mailto:// at the beginning */
		match->m_begin = pos;

		if (pos >= cb->begin + 1) {
			match->st = *(pos - 1);
		}
		else {
			match->st = '\0';
		}
	}
	else {
		/* Just '@' */

		/* Check if this match is a part of the previous mailto: email */
		if (cb->last_at != NULL && cb->last_at == pos) {
			cb->last_at = NULL;
			return FALSE;
		}
		else if (pos == cb->begin) {
			/* Just @ at the start of input */
			return FALSE;
		}

		match->st = '\0';
	}

	return TRUE;
}

static gboolean
url_email_end(struct url_callback_data *cb,
			  const char *pos,
			  url_match_t *match)
{
	const char *last = NULL;
	struct http_parser_url u;
	int len = cb->end - pos;
	unsigned int flags = 0;

	if (match->newline_pos && match->st != '<') {
		/* We should also limit our match end to the newline */
		len = MIN(len, match->newline_pos - pos);
	}

	if (!match->prefix || match->prefix[0] == '\0') {
		/* We have mailto:// at the beginning */
		if (rspamd_mailto_parse(&u, pos, len, &last,
								RSPAMD_URL_PARSE_CHECK, &flags) != 0) {
			return FALSE;
		}

		if (!(u.field_set & (1 << UF_USERINFO))) {
			return FALSE;
		}

		cb->last_at = match->m_begin + u.field_data[UF_USERINFO].off +
					  u.field_data[UF_USERINFO].len;

		g_assert(*cb->last_at == '@');
		match->m_len = (last - pos);

		return TRUE;
	}
	else {
		const char *c, *p;
		/*
		 * Here we have just '@', so we need to find both start and end of the
		 * pattern
		 */
		g_assert(*pos == '@');

		if (pos >= cb->end - 2 || pos < cb->begin + 1) {
			/* Boundary violation */
			return FALSE;
		}

		/* Check the next character after `@` */
		if (!g_ascii_isalnum(pos[1]) || !g_ascii_isalnum(*(pos - 1))) {
			return FALSE;
		}


		c = pos - 1;
		while (c > cb->begin) {
			if (!is_mailsafe(*c)) {
				break;
			}
			if (c == match->prev_newline_pos) {
				break;
			}

			c--;
		}
		/* Rewind to the first alphanumeric character */
		while (c < pos && !g_ascii_isalnum(*c)) {
			c++;
		}

		/* Find the end of email */
		p = pos + 1;
		while (p < cb->end && is_domain(*p)) {
			if (p == match->newline_pos) {
				break;
			}

			p++;
		}

		/* Rewind it again to avoid bad emails to be detected */
		while (p > pos && p < cb->end && !g_ascii_isalnum(*p)) {
			p--;
		}

		if (p < cb->end && g_ascii_isalnum(*p) &&
			(match->newline_pos == NULL || p < match->newline_pos)) {
			p++;
		}

		if (p > c) {
			match->m_begin = c;
			match->m_len = p - c;
			return TRUE;
		}
	}

	return FALSE;
}

static gboolean
url_tel_start(struct url_callback_data *cb,
			  const char *pos,
			  url_match_t *match)
{
	match->m_begin = pos;

	if (pos >= cb->begin + 1) {
		match->st = *(pos - 1);
	}
	else {
		match->st = '\0';
	}

	return TRUE;
}

static gboolean
url_tel_end(struct url_callback_data *cb,
			const char *pos,
			url_match_t *match)
{
	const char *last = NULL;
	struct http_parser_url u;
	int len = cb->end - pos;
	unsigned int flags = 0;

	if (match->newline_pos && match->st != '<') {
		/* We should also limit our match end to the newline */
		len = MIN(len, match->newline_pos - pos);
	}

	if (rspamd_telephone_parse(&u, pos, len, &last,
							   RSPAMD_URL_PARSE_CHECK, &flags) != 0) {
		return FALSE;
	}

	if (!(u.field_set & (1 << UF_HOST))) {
		return FALSE;
	}

	match->m_len = (last - pos);

	return TRUE;
}


static gboolean
rspamd_url_trie_is_match(struct url_matcher *matcher, const char *pos,
						 const char *end, const char *newline_pos)
{
	if (matcher->flags & URL_MATCHER_FLAG_TLD_MATCH) {
		/* Immediately check pos for valid chars */
		if (pos < end) {
			if (pos != newline_pos && !g_ascii_isspace(*pos) && *pos != '/' && *pos != '?' &&
				*pos != ':' && !is_url_end(*pos)) {
				if (*pos == '.') {
					/* We allow . at the end of the domain however */
					pos++;
					if (pos < end) {
						if (!g_ascii_isspace(*pos) && *pos != '/' &&
							*pos != '?' && *pos != ':' && !is_url_end(*pos)) {
							return FALSE;
						}
					}
				}
				else {
					return FALSE;
				}
			}
		}
	}

	return TRUE;
}

static int
rspamd_url_trie_callback(struct rspamd_multipattern *mp,
						 unsigned int strnum,
						 int match_start,
						 int match_pos,
						 const char *text,
						 gsize len,
						 void *context)
{
	struct url_matcher *matcher;
	url_match_t m;
	const char *pos, *newline_pos = NULL;
	struct url_callback_data *cb = context;

	pos = text + match_pos;

	if (cb->fin > pos) {
		/* Already seen */
		return 0;
	}

	matcher = &g_array_index(cb->matchers, struct url_matcher,
							 strnum);

	if ((matcher->flags & URL_MATCHER_FLAG_NOHTML) && cb->how == RSPAMD_URL_FIND_STRICT) {
		/* Do not try to match non-html like urls in html texts */
		return 0;
	}

	memset(&m, 0, sizeof(m));
	m.m_begin = text + match_start;
	m.m_len = match_pos - match_start;

	if (cb->newlines && cb->newlines->len > 0) {
		newline_pos = g_ptr_array_index(cb->newlines, cb->newline_idx);

		while (pos > newline_pos && cb->newline_idx < cb->newlines->len) {
			cb->newline_idx++;
			newline_pos = g_ptr_array_index(cb->newlines, cb->newline_idx);
		}

		if (pos > newline_pos) {
			newline_pos = NULL;
		}

		if (cb->newline_idx > 0) {
			m.prev_newline_pos = g_ptr_array_index(cb->newlines,
												   cb->newline_idx - 1);
		}
	}

	if (!rspamd_url_trie_is_match(matcher, pos, cb->end, newline_pos)) {
		return 0;
	}

	m.pattern = matcher->pattern;
	m.prefix = matcher->prefix;
	m.add_prefix = FALSE;
	m.newline_pos = newline_pos;
	pos = cb->begin + match_start;

	if (matcher->start(cb, pos, &m) &&
		matcher->end(cb, pos, &m)) {
		if (m.add_prefix || matcher->prefix[0] != '\0') {
			cb->len = m.m_len + strlen(matcher->prefix);
			cb->url_str = rspamd_mempool_alloc(cb->pool, cb->len + 1);
			cb->len = rspamd_snprintf(cb->url_str,
									  cb->len + 1,
									  "%s%*s",
									  m.prefix,
									  (int) m.m_len,
									  m.m_begin);
			cb->prefix_added = TRUE;
		}
		else {
			cb->url_str = rspamd_mempool_alloc(cb->pool, m.m_len + 1);
			rspamd_strlcpy(cb->url_str, m.m_begin, m.m_len + 1);
		}

		cb->start = m.m_begin;

		if (pos > cb->fin) {
			cb->fin = pos;
		}

		return 1;
	}
	else {
		cb->url_str = NULL;
	}

	/* Continue search */
	return 0;
}

gboolean
rspamd_url_find(rspamd_mempool_t *pool,
				const char *begin, gsize len,
				char **url_str,
				enum rspamd_url_find_type how,
				goffset *url_pos,
				gboolean *prefix_added)
{
	struct url_callback_data cb;
	int ret;

	memset(&cb, 0, sizeof(cb));
	cb.begin = begin;
	cb.end = begin + len;
	cb.how = how;
	cb.pool = pool;

	if (how == RSPAMD_URL_FIND_ALL) {
		if (url_scanner->search_trie_full) {
			cb.matchers = url_scanner->matchers_full;
			ret = rspamd_multipattern_lookup(url_scanner->search_trie_full,
											 begin, len,
											 rspamd_url_trie_callback, &cb, NULL);
		}
		else {
			cb.matchers = url_scanner->matchers_strict;
			ret = rspamd_multipattern_lookup(url_scanner->search_trie_strict,
											 begin, len,
											 rspamd_url_trie_callback, &cb, NULL);
		}
	}
	else {
		cb.matchers = url_scanner->matchers_strict;
		ret = rspamd_multipattern_lookup(url_scanner->search_trie_strict,
										 begin, len,
										 rspamd_url_trie_callback, &cb, NULL);
	}

	if (ret) {
		if (url_str) {
			*url_str = cb.url_str;
		}

		if (url_pos) {
			*url_pos = cb.start - begin;
		}

		if (prefix_added) {
			*prefix_added = cb.prefix_added;
		}

		return TRUE;
	}

	return FALSE;
}

static int
rspamd_url_trie_generic_callback_common(struct rspamd_multipattern *mp,
										unsigned int strnum,
										int match_start,
										int match_pos,
										const char *text,
										gsize len,
										void *context,
										gboolean multiple)
{
	struct rspamd_url *url;
	struct url_matcher *matcher;
	url_match_t m;
	const char *pos, *newline_pos = NULL;
	struct url_callback_data *cb = context;
	int rc;
	rspamd_mempool_t *pool;

	pos = text + match_pos;

	if (cb->fin > pos) {
		/* Already seen */
		return 0;
	}

	matcher = &g_array_index(cb->matchers, struct url_matcher,
							 strnum);
	pool = cb->pool;

	if ((matcher->flags & URL_MATCHER_FLAG_NOHTML) && cb->how == RSPAMD_URL_FIND_STRICT) {
		/* Do not try to match non-html like urls in html texts, continue matching */
		return 0;
	}

	memset(&m, 0, sizeof(m));


	/* Find the next newline after our pos */
	if (cb->newlines && cb->newlines->len > 0) {
		newline_pos = g_ptr_array_index(cb->newlines, cb->newline_idx);

		while (pos > newline_pos && cb->newline_idx < cb->newlines->len - 1) {
			cb->newline_idx++;
			newline_pos = g_ptr_array_index(cb->newlines, cb->newline_idx);
		}

		if (pos > newline_pos) {
			newline_pos = NULL;
		}
		if (cb->newline_idx > 0) {
			m.prev_newline_pos = g_ptr_array_index(cb->newlines,
												   cb->newline_idx - 1);
		}
	}

	if (!rspamd_url_trie_is_match(matcher, pos, text + len, newline_pos)) {
		/* Mismatch, continue */
		return 0;
	}

	pos = cb->begin + match_start;
	m.pattern = matcher->pattern;
	m.prefix = matcher->prefix;
	m.add_prefix = FALSE;
	m.m_begin = text + match_start;
	m.m_len = match_pos - match_start;
	m.newline_pos = newline_pos;

	if (matcher->start(cb, pos, &m) &&
		matcher->end(cb, pos, &m)) {
		if (m.add_prefix || matcher->prefix[0] != '\0') {
			cb->len = m.m_len + strlen(matcher->prefix);
			cb->url_str = rspamd_mempool_alloc(cb->pool, cb->len + 1);
			cb->len = rspamd_snprintf(cb->url_str,
									  cb->len + 1,
									  "%s%*s",
									  m.prefix,
									  (int) m.m_len,
									  m.m_begin);
			cb->prefix_added = TRUE;
		}
		else {
			cb->url_str = rspamd_mempool_alloc(cb->pool, m.m_len + 1);
			cb->len = rspamd_strlcpy(cb->url_str, m.m_begin, m.m_len + 1);
		}

		cb->start = m.m_begin;

		if (pos > cb->fin) {
			cb->fin = pos;
		}

		url = rspamd_mempool_alloc0(pool, sizeof(struct rspamd_url));
		g_strstrip(cb->url_str);
		rc = rspamd_url_parse(url, cb->url_str,
							  strlen(cb->url_str), pool,
							  RSPAMD_URL_PARSE_TEXT);

		if (rc == URI_ERRNO_OK && url->hostlen > 0) {
			if (cb->prefix_added) {
				url->flags |= RSPAMD_URL_FLAG_SCHEMALESS;
				cb->prefix_added = FALSE;
			}

			if (cb->func) {
				if (!cb->func(url, cb->start - text, (m.m_begin + m.m_len) - text,
							  cb->funcd)) {
					/* We need to stop here in any case! */
					return -1;
				}
			}
		}
		else if (rc != URI_ERRNO_OK) {
			msg_debug_pool_check("extract of url '%s' failed: %s",
								 cb->url_str,
								 rspamd_url_strerror(rc));
		}
	}
	else {
		cb->url_str = NULL;
		/* Continue search if no pattern has been found */
		return 0;
	}

	/* Continue search if required (return 0 means continue) */
	return !multiple;
}

static int
rspamd_url_trie_generic_callback_multiple(struct rspamd_multipattern *mp,
										  unsigned int strnum,
										  int match_start,
										  int match_pos,
										  const char *text,
										  gsize len,
										  void *context)
{
	return rspamd_url_trie_generic_callback_common(mp, strnum, match_start,
												   match_pos, text, len, context, TRUE);
}

static int
rspamd_url_trie_generic_callback_single(struct rspamd_multipattern *mp,
										unsigned int strnum,
										int match_start,
										int match_pos,
										const char *text,
										gsize len,
										void *context)
{
	return rspamd_url_trie_generic_callback_common(mp, strnum, match_start,
												   match_pos, text, len, context, FALSE);
}

struct rspamd_url_mimepart_cbdata {
	struct rspamd_task *task;
	struct rspamd_mime_text_part *part;
	gsize url_len;
	uint16_t *cur_url_order; /* Global ordering */
	uint16_t cur_part_order; /* Per part ordering */
};

static gboolean
rspamd_url_query_callback(struct rspamd_url *url, gsize start_offset,
						  gsize end_offset, gpointer ud)
{
	struct rspamd_url_mimepart_cbdata *cbd =
		(struct rspamd_url_mimepart_cbdata *) ud;
	struct rspamd_task *task;

	task = cbd->task;

	if (url->protocol == PROTOCOL_MAILTO) {
		if (url->userlen == 0) {
			return FALSE;
		}
	}
	/* Also check max urls */
	if (cbd->task->cfg && cbd->task->cfg->max_urls > 0) {
		if (kh_size(MESSAGE_FIELD(task, urls)) > cbd->task->cfg->max_urls) {
			msg_err_task("part has too many URLs, we cannot process more: "
						 "%d urls extracted ",
						 (unsigned int) kh_size(MESSAGE_FIELD(task, urls)));

			return FALSE;
		}
	}

	url->flags |= RSPAMD_URL_FLAG_QUERY;


	if (rspamd_url_set_add_or_increase(MESSAGE_FIELD(task, urls), url, false)) {
		if (cbd->part && cbd->part->mime_part->urls) {
			g_ptr_array_add(cbd->part->mime_part->urls, url);
		}

		url->part_order = cbd->cur_part_order++;

		if (cbd->cur_url_order) {
			url->order = (*cbd->cur_url_order)++;
		}
	}

	return TRUE;
}

static gboolean
rspamd_url_text_part_callback(struct rspamd_url *url, gsize start_offset,
							  gsize end_offset, gpointer ud)
{
	struct rspamd_url_mimepart_cbdata *cbd =
		(struct rspamd_url_mimepart_cbdata *) ud;
	struct rspamd_process_exception *ex;
	struct rspamd_task *task;

	task = cbd->task;
	ex = rspamd_mempool_alloc0(task->task_pool, sizeof(struct rspamd_process_exception));

	ex->pos = start_offset;
	ex->len = end_offset - start_offset;
	ex->type = RSPAMD_EXCEPTION_URL;
	ex->ptr = url;

	cbd->url_len += ex->len;

	if (cbd->part->utf_stripped_content &&
		cbd->url_len > cbd->part->utf_stripped_content->len * 10) {
		/* Absurd case, stop here now */
		msg_err_task("part has too many URLs, we cannot process more: %z url len; "
					 "%d stripped content length",
					 cbd->url_len, cbd->part->utf_stripped_content->len);

		return FALSE;
	}

	if (url->protocol == PROTOCOL_MAILTO) {
		if (url->userlen == 0) {
			return FALSE;
		}
	}
	/* Also check max urls */
	if (cbd->task->cfg && cbd->task->cfg->max_urls > 0) {
		if (kh_size(MESSAGE_FIELD(task, urls)) > cbd->task->cfg->max_urls) {
			msg_err_task("part has too many URLs, we cannot process more: "
						 "%d urls extracted ",
						 (unsigned int) kh_size(MESSAGE_FIELD(task, urls)));

			return FALSE;
		}
	}

	url->flags |= RSPAMD_URL_FLAG_FROM_TEXT;

	if (rspamd_url_set_add_or_increase(MESSAGE_FIELD(task, urls), url, false) &&
		cbd->part->mime_part->urls) {
		url->part_order = cbd->cur_part_order++;

		if (cbd->cur_url_order) {
			url->order = (*cbd->cur_url_order)++;
		}
		g_ptr_array_add(cbd->part->mime_part->urls, url);
	}

	cbd->part->exceptions = g_list_prepend(
		cbd->part->exceptions,
		ex);

	/* We also search the query for additional url inside */
	if (url->querylen > 0) {
		rspamd_url_find_multiple(task->task_pool,
								 rspamd_url_query_unsafe(url), url->querylen,
								 RSPAMD_URL_FIND_ALL, NULL,
								 rspamd_url_query_callback, cbd);
	}

	return TRUE;
}

void rspamd_url_text_extract(rspamd_mempool_t *pool,
							 struct rspamd_task *task,
							 struct rspamd_mime_text_part *part,
							 uint16_t *cur_url_order,
							 enum rspamd_url_find_type how)
{
	struct rspamd_url_mimepart_cbdata mcbd;

	if (part->utf_stripped_content == NULL || part->utf_stripped_content->len == 0) {
		msg_warn_task("got empty text part");
		return;
	}

	mcbd.task = task;
	mcbd.part = part;
	mcbd.url_len = 0;
	mcbd.cur_url_order = cur_url_order;
	mcbd.cur_part_order = 0;

	rspamd_url_find_multiple(task->task_pool, part->utf_stripped_content->data,
							 part->utf_stripped_content->len, how, part->newlines,
							 rspamd_url_text_part_callback, &mcbd);
}

void rspamd_url_find_multiple(rspamd_mempool_t *pool,
							  const char *in,
							  gsize inlen,
							  enum rspamd_url_find_type how,
							  GPtrArray *nlines,
							  url_insert_function func,
							  gpointer ud)
{
	struct url_callback_data cb;

	g_assert(in != NULL);

	if (inlen == 0) {
		inlen = strlen(in);
	}

	memset(&cb, 0, sizeof(cb));
	cb.begin = in;
	cb.end = in + inlen;
	cb.how = how;
	cb.pool = pool;

	cb.funcd = ud;
	cb.func = func;
	cb.newlines = nlines;

	if (how == RSPAMD_URL_FIND_ALL) {
		if (url_scanner->search_trie_full) {
			cb.matchers = url_scanner->matchers_full;
			rspamd_multipattern_lookup(url_scanner->search_trie_full,
									   in, inlen,
									   rspamd_url_trie_generic_callback_multiple, &cb, NULL);
		}
		else {
			cb.matchers = url_scanner->matchers_strict;
			rspamd_multipattern_lookup(url_scanner->search_trie_strict,
									   in, inlen,
									   rspamd_url_trie_generic_callback_multiple, &cb, NULL);
		}
	}
	else {
		cb.matchers = url_scanner->matchers_strict;
		rspamd_multipattern_lookup(url_scanner->search_trie_strict,
								   in, inlen,
								   rspamd_url_trie_generic_callback_multiple, &cb, NULL);
	}
}

void rspamd_url_find_single(rspamd_mempool_t *pool,
							const char *in,
							gsize inlen,
							enum rspamd_url_find_type how,
							url_insert_function func,
							gpointer ud)
{
	struct url_callback_data cb;

	g_assert(in != NULL);

	if (inlen == 0) {
		inlen = strlen(in);
	}

	/*
	 * We might have a situation when we need to parse URLs on config file
	 * parsing, but there is no valid url_scanner loaded. Hence, we just load
	 * some defaults and it should be fine...
	 */
	if (url_scanner == NULL) {
		rspamd_url_init(NULL);
	}

	memset(&cb, 0, sizeof(cb));
	cb.begin = in;
	cb.end = in + inlen;
	cb.how = how;
	cb.pool = pool;

	cb.funcd = ud;
	cb.func = func;

	if (how == RSPAMD_URL_FIND_ALL) {
		if (url_scanner->search_trie_full) {
			cb.matchers = url_scanner->matchers_full;
			rspamd_multipattern_lookup(url_scanner->search_trie_full,
									   in, inlen,
									   rspamd_url_trie_generic_callback_single, &cb, NULL);
		}
		else {
			cb.matchers = url_scanner->matchers_strict;
			rspamd_multipattern_lookup(url_scanner->search_trie_strict,
									   in, inlen,
									   rspamd_url_trie_generic_callback_single, &cb, NULL);
		}
	}
	else {
		cb.matchers = url_scanner->matchers_strict;
		rspamd_multipattern_lookup(url_scanner->search_trie_strict,
								   in, inlen,
								   rspamd_url_trie_generic_callback_single, &cb, NULL);
	}
}


gboolean
rspamd_url_task_subject_callback(struct rspamd_url *url, gsize start_offset,
								 gsize end_offset, gpointer ud)
{
	struct rspamd_task *task = ud;
	char *url_str = NULL;
	struct rspamd_url *query_url;
	int rc;
	gboolean prefix_added;

	/* It is just a displayed URL, we should not check it for certain things */
	url->flags |= RSPAMD_URL_FLAG_HTML_DISPLAYED | RSPAMD_URL_FLAG_SUBJECT;

	if (url->protocol == PROTOCOL_MAILTO) {
		if (url->userlen == 0) {
			return FALSE;
		}
	}

	rspamd_url_set_add_or_increase(MESSAGE_FIELD(task, urls), url, false);

	/* We also search the query for additional url inside */
	if (url->querylen > 0) {
		if (rspamd_url_find(task->task_pool, rspamd_url_query_unsafe(url), url->querylen,
							&url_str, RSPAMD_URL_FIND_ALL, NULL, &prefix_added)) {

			query_url = rspamd_mempool_alloc0(task->task_pool,
											  sizeof(struct rspamd_url));
			rc = rspamd_url_parse(query_url,
								  url_str,
								  strlen(url_str),
								  task->task_pool,
								  RSPAMD_URL_PARSE_TEXT);

			if (rc == URI_ERRNO_OK &&
				url->hostlen > 0) {
				msg_debug_task("found url %s in query of url"
							   " %*s",
							   url_str, url->querylen, rspamd_url_query_unsafe(url));

				if (prefix_added) {
					query_url->flags |= RSPAMD_URL_FLAG_SCHEMALESS;
				}

				if (query_url->protocol == PROTOCOL_MAILTO) {
					if (query_url->userlen == 0) {
						return TRUE;
					}
				}

				rspamd_url_set_add_or_increase(MESSAGE_FIELD(task, urls),
											   query_url, false);
			}
		}
	}

	return TRUE;
}

static inline khint_t
rspamd_url_hash(struct rspamd_url *url)
{
	if (url->urllen > 0) {
		return (khint_t) rspamd_cryptobox_fast_hash(url->string, url->urllen,
													rspamd_hash_seed());
	}

	return 0;
}

static inline khint_t
rspamd_url_host_hash(struct rspamd_url *url)
{
	if (url->hostlen > 0) {
		return (khint_t) rspamd_cryptobox_fast_hash(rspamd_url_host_unsafe(url),
													url->hostlen,
													rspamd_hash_seed());
	}

	return 0;
}

/* Compare two emails for building emails tree */
static inline bool
rspamd_emails_cmp(struct rspamd_url *u1, struct rspamd_url *u2)
{
	int r;

	if (u1->hostlen != u2->hostlen || u1->hostlen == 0) {
		return FALSE;
	}
	else {
		if ((r = rspamd_lc_cmp(rspamd_url_host_unsafe(u1),
							   rspamd_url_host_unsafe(u2), u1->hostlen)) == 0) {
			if (u1->userlen != u2->userlen || u1->userlen == 0) {
				return FALSE;
			}
			else {
				return (rspamd_lc_cmp(rspamd_url_user_unsafe(u1),
									  rspamd_url_user_unsafe(u2),
									  u1->userlen) == 0);
			}
		}
		else {
			return r == 0;
		}
	}

	return FALSE;
}

static inline bool
rspamd_urls_cmp(struct rspamd_url *u1, struct rspamd_url *u2)
{
	int r = 0;

	if (u1->protocol != u2->protocol || u1->urllen != u2->urllen) {
		return false;
	}
	else {
		if (u1->protocol & PROTOCOL_MAILTO) {
			return rspamd_emails_cmp(u1, u2);
		}

		r = memcmp(u1->string, u2->string, u1->urllen);
	}

	return r == 0;
}

static inline bool
rspamd_urls_host_cmp(struct rspamd_url *u1, struct rspamd_url *u2)
{
	int r = 0;

	if (u1->hostlen != u2->hostlen) {
		return false;
	}
	else {
		r = memcmp(rspamd_url_host_unsafe(u1), rspamd_url_host_unsafe(u2),
				   u1->hostlen);
	}

	return r == 0;
}

gsize rspamd_url_decode(char *dst, const char *src, gsize size)
{
	char *d, ch, c, decoded;
	const char *s;
	enum {
		sw_usual = 0,
		sw_quoted,
		sw_quoted_second
	} state;

	d = dst;
	s = src;

	state = 0;
	decoded = 0;

	while (size--) {

		ch = *s++;

		switch (state) {
		case sw_usual:

			if (ch == '%') {
				state = sw_quoted;
				break;
			}
			else if (ch == '+') {
				*d++ = ' ';
			}
			else {
				*d++ = ch;
			}
			break;

		case sw_quoted:

			if (ch >= '0' && ch <= '9') {
				decoded = (ch - '0');
				state = sw_quoted_second;
				break;
			}

			c = (ch | 0x20);
			if (c >= 'a' && c <= 'f') {
				decoded = (c - 'a' + 10);
				state = sw_quoted_second;
				break;
			}

			/* the invalid quoted character */

			state = sw_usual;

			*d++ = ch;

			break;

		case sw_quoted_second:

			state = sw_usual;

			if (ch >= '0' && ch <= '9') {
				ch = ((decoded << 4) + ch - '0');
				*d++ = ch;

				break;
			}

			c = (u_char) (ch | 0x20);
			if (c >= 'a' && c <= 'f') {
				ch = ((decoded << 4) + c - 'a' + 10);

				*d++ = ch;
				break;
			}

			/* the invalid quoted character */
			break;
		}
	}

	return (d - dst);
}

enum rspamd_url_char_class {
	RSPAMD_URL_UNRESERVED = (1 << 0),
	RSPAMD_URL_SUBDELIM = (1 << 1),
	RSPAMD_URL_PATHSAFE = (1 << 2),
	RSPAMD_URL_QUERYSAFE = (1 << 3),
	RSPAMD_URL_FRAGMENTSAFE = (1 << 4),
	RSPAMD_URL_HOSTSAFE = (1 << 5),
	RSPAMD_URL_USERSAFE = (1 << 6),
};

#define RSPAMD_URL_FLAGS_HOSTSAFE (RSPAMD_URL_UNRESERVED | RSPAMD_URL_HOSTSAFE | RSPAMD_URL_SUBDELIM)
#define RSPAMD_URL_FLAGS_USERSAFE (RSPAMD_URL_UNRESERVED | RSPAMD_URL_USERSAFE | RSPAMD_URL_SUBDELIM)
#define RSPAMD_URL_FLAGS_PATHSAFE (RSPAMD_URL_UNRESERVED | RSPAMD_URL_PATHSAFE | RSPAMD_URL_SUBDELIM)
#define RSPAMD_URL_FLAGS_QUERYSAFE (RSPAMD_URL_UNRESERVED | RSPAMD_URL_QUERYSAFE | RSPAMD_URL_SUBDELIM)
#define RSPAMD_URL_FLAGS_FRAGMENTSAFE (RSPAMD_URL_UNRESERVED | RSPAMD_URL_FRAGMENTSAFE | RSPAMD_URL_SUBDELIM)

static const unsigned char rspamd_url_encoding_classes[256] = {
	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
	0, 0, 0, 0, 0, 0, 0 /*   */, RSPAMD_URL_SUBDELIM /* ! */, 0 /* " */, 0 /* # */,
	RSPAMD_URL_SUBDELIM /* $ */, 0 /* % */, RSPAMD_URL_SUBDELIM /* & */,
	RSPAMD_URL_SUBDELIM /* ' */, RSPAMD_URL_SUBDELIM /* ( */,
	RSPAMD_URL_SUBDELIM /* ) */, RSPAMD_URL_SUBDELIM /* * */,
	RSPAMD_URL_SUBDELIM /* + */, RSPAMD_URL_SUBDELIM /* , */,
	RSPAMD_URL_UNRESERVED /* - */, RSPAMD_URL_UNRESERVED /* . */,
	RSPAMD_URL_PATHSAFE | RSPAMD_URL_QUERYSAFE | RSPAMD_URL_FRAGMENTSAFE /* / */,
	RSPAMD_URL_UNRESERVED /* 0 */, RSPAMD_URL_UNRESERVED /* 1 */,
	RSPAMD_URL_UNRESERVED /* 2 */, RSPAMD_URL_UNRESERVED /* 3 */,
	RSPAMD_URL_UNRESERVED /* 4 */, RSPAMD_URL_UNRESERVED /* 5 */,
	RSPAMD_URL_UNRESERVED /* 6 */, RSPAMD_URL_UNRESERVED /* 7 */,
	RSPAMD_URL_UNRESERVED /* 8 */, RSPAMD_URL_UNRESERVED /* 9 */,
	RSPAMD_URL_USERSAFE | RSPAMD_URL_HOSTSAFE | RSPAMD_URL_PATHSAFE | RSPAMD_URL_QUERYSAFE | RSPAMD_URL_FRAGMENTSAFE /* : */,
	RSPAMD_URL_SUBDELIM /* ; */, 0 /* < */, RSPAMD_URL_SUBDELIM /* = */, 0 /* > */,
	RSPAMD_URL_QUERYSAFE | RSPAMD_URL_FRAGMENTSAFE /* ? */,
	RSPAMD_URL_PATHSAFE | RSPAMD_URL_QUERYSAFE | RSPAMD_URL_FRAGMENTSAFE /* @ */,
	RSPAMD_URL_UNRESERVED /* A */, RSPAMD_URL_UNRESERVED /* B */,
	RSPAMD_URL_UNRESERVED /* C */, RSPAMD_URL_UNRESERVED /* D */,
	RSPAMD_URL_UNRESERVED /* E */, RSPAMD_URL_UNRESERVED /* F */,
	RSPAMD_URL_UNRESERVED /* G */, RSPAMD_URL_UNRESERVED /* H */,
	RSPAMD_URL_UNRESERVED /* I */, RSPAMD_URL_UNRESERVED /* J */,
	RSPAMD_URL_UNRESERVED /* K */, RSPAMD_URL_UNRESERVED /* L */,
	RSPAMD_URL_UNRESERVED /* M */, RSPAMD_URL_UNRESERVED /* N */,
	RSPAMD_URL_UNRESERVED /* O */, RSPAMD_URL_UNRESERVED /* P */,
	RSPAMD_URL_UNRESERVED /* Q */, RSPAMD_URL_UNRESERVED /* R */,
	RSPAMD_URL_UNRESERVED /* S */, RSPAMD_URL_UNRESERVED /* T */,
	RSPAMD_URL_UNRESERVED /* U */, RSPAMD_URL_UNRESERVED /* V */,
	RSPAMD_URL_UNRESERVED /* W */, RSPAMD_URL_UNRESERVED /* X */,
	RSPAMD_URL_UNRESERVED /* Y */, RSPAMD_URL_UNRESERVED /* Z */,
	RSPAMD_URL_HOSTSAFE /* [ */, 0 /* \ */, RSPAMD_URL_HOSTSAFE /* ] */, 0 /* ^ */,
	RSPAMD_URL_UNRESERVED /* _ */, 0 /* ` */, RSPAMD_URL_UNRESERVED /* a */,
	RSPAMD_URL_UNRESERVED /* b */, RSPAMD_URL_UNRESERVED /* c */,
	RSPAMD_URL_UNRESERVED /* d */, RSPAMD_URL_UNRESERVED /* e */,
	RSPAMD_URL_UNRESERVED /* f */, RSPAMD_URL_UNRESERVED /* g */,
	RSPAMD_URL_UNRESERVED /* h */, RSPAMD_URL_UNRESERVED /* i */,
	RSPAMD_URL_UNRESERVED /* j */, RSPAMD_URL_UNRESERVED /* k */,
	RSPAMD_URL_UNRESERVED /* l */, RSPAMD_URL_UNRESERVED /* m */,
	RSPAMD_URL_UNRESERVED /* n */, RSPAMD_URL_UNRESERVED /* o */,
	RSPAMD_URL_UNRESERVED /* p */, RSPAMD_URL_UNRESERVED /* q */,
	RSPAMD_URL_UNRESERVED /* r */, RSPAMD_URL_UNRESERVED /* s */,
	RSPAMD_URL_UNRESERVED /* t */, RSPAMD_URL_UNRESERVED /* u */,
	RSPAMD_URL_UNRESERVED /* v */, RSPAMD_URL_UNRESERVED /* w */,
	RSPAMD_URL_UNRESERVED /* x */, RSPAMD_URL_UNRESERVED /* y */,
	RSPAMD_URL_UNRESERVED /* z */, 0 /* { */, 0 /* | */, 0 /* } */,
	RSPAMD_URL_UNRESERVED /* ~ */, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
	0, 0, 0, 0, 0, 0, 0, 0, 0};

#define CHECK_URL_COMPONENT(beg, len, flags)                                              \
	do {                                                                                  \
		for (i = 0; i < (len); i++) {                                                     \
			if ((rspamd_url_encoding_classes[(unsigned char) (beg)[i]] & (flags)) == 0) { \
				dlen += 2;                                                                \
			}                                                                             \
		}                                                                                 \
	} while (0)

#define ENCODE_URL_COMPONENT(beg, len, flags)                                             \
	do {                                                                                  \
		for (i = 0; i < (len) && dend > d; i++) {                                         \
			if ((rspamd_url_encoding_classes[(unsigned char) (beg)[i]] & (flags)) == 0) { \
				*d++ = '%';                                                               \
				*d++ = hexdigests[(unsigned char) ((beg)[i] >> 4) & 0xf];                 \
				*d++ = hexdigests[(unsigned char) (beg)[i] & 0xf];                        \
			}                                                                             \
			else {                                                                        \
				*d++ = (beg)[i];                                                          \
			}                                                                             \
		}                                                                                 \
	} while (0)

const char *
rspamd_url_encode(struct rspamd_url *url, gsize *pdlen,
				  rspamd_mempool_t *pool)
{
	unsigned char *dest, *d, *dend;
	static const char hexdigests[16] = "0123456789ABCDEF";
	unsigned int i;
	gsize dlen = 0;

	g_assert(pdlen != NULL && url != NULL && pool != NULL);

	CHECK_URL_COMPONENT(rspamd_url_host_unsafe(url), url->hostlen,
						RSPAMD_URL_FLAGS_HOSTSAFE);
	CHECK_URL_COMPONENT(rspamd_url_user_unsafe(url), url->userlen,
						RSPAMD_URL_FLAGS_USERSAFE);
	CHECK_URL_COMPONENT(rspamd_url_data_unsafe(url), url->datalen,
						RSPAMD_URL_FLAGS_PATHSAFE);
	CHECK_URL_COMPONENT(rspamd_url_query_unsafe(url), url->querylen,
						RSPAMD_URL_FLAGS_QUERYSAFE);
	CHECK_URL_COMPONENT(rspamd_url_fragment_unsafe(url), url->fragmentlen,
						RSPAMD_URL_FLAGS_FRAGMENTSAFE);

	if (dlen == 0) {
		*pdlen = url->urllen;

		return url->string;
	}

	/* Need to encode */
	dlen += url->urllen + sizeof("telephone://"); /* Protocol hack */
	dest = rspamd_mempool_alloc(pool, dlen + 1);
	d = dest;
	dend = d + dlen;

	if (url->protocollen > 0) {
		if (!(url->protocol & PROTOCOL_UNKNOWN)) {
			const char *known_proto = rspamd_url_protocol_name(url->protocol);
			d += rspamd_snprintf((char *) d, dend - d,
								 "%s://",
								 known_proto);
		}
		else {
			d += rspamd_snprintf((char *) d, dend - d,
								 "%*s://",
								 (int) url->protocollen, url->string);
		}
	}
	else {
		d += rspamd_snprintf((char *) d, dend - d, "http://");
	}

	if (url->userlen > 0) {
		ENCODE_URL_COMPONENT(rspamd_url_user_unsafe(url), url->userlen,
							 RSPAMD_URL_FLAGS_USERSAFE);
		*d++ = '@';
	}

	ENCODE_URL_COMPONENT(rspamd_url_host_unsafe(url), url->hostlen,
						 RSPAMD_URL_FLAGS_HOSTSAFE);

	if (url->datalen > 0) {
		*d++ = '/';
		ENCODE_URL_COMPONENT(rspamd_url_data_unsafe(url), url->datalen,
							 RSPAMD_URL_FLAGS_PATHSAFE);
	}

	if (url->querylen > 0) {
		*d++ = '?';
		ENCODE_URL_COMPONENT(rspamd_url_query_unsafe(url), url->querylen,
							 RSPAMD_URL_FLAGS_QUERYSAFE);
	}

	if (url->fragmentlen > 0) {
		*d++ = '#';
		ENCODE_URL_COMPONENT(rspamd_url_fragment_unsafe(url), url->fragmentlen,
							 RSPAMD_URL_FLAGS_FRAGMENTSAFE);
	}

	*pdlen = (d - dest);

	return (const char *) dest;
}

gboolean
rspamd_url_is_domain(int c)
{
	return is_domain((unsigned char) c);
}

const char *
rspamd_url_protocol_name(enum rspamd_url_protocol proto)
{
	const char *ret = "unknown";

	switch (proto) {
	case PROTOCOL_HTTP:
		ret = "http";
		break;
	case PROTOCOL_HTTPS:
		ret = "https";
		break;
	case PROTOCOL_FTP:
		ret = "ftp";
		break;
	case PROTOCOL_FILE:
		ret = "file";
		break;
	case PROTOCOL_MAILTO:
		ret = "mailto";
		break;
	case PROTOCOL_TELEPHONE:
		ret = "telephone";
		break;
	default:
		break;
	}

	return ret;
}

enum rspamd_url_protocol
rspamd_url_protocol_from_string(const char *str)
{
	enum rspamd_url_protocol ret = PROTOCOL_UNKNOWN;

	if (strcmp(str, "http") == 0) {
		ret = PROTOCOL_HTTP;
	}
	else if (strcmp(str, "https") == 0) {
		ret = PROTOCOL_HTTPS;
	}
	else if (strcmp(str, "mailto") == 0) {
		ret = PROTOCOL_MAILTO;
	}
	else if (strcmp(str, "ftp") == 0) {
		ret = PROTOCOL_FTP;
	}
	else if (strcmp(str, "file") == 0) {
		ret = PROTOCOL_FILE;
	}
	else if (strcmp(str, "telephone") == 0) {
		ret = PROTOCOL_TELEPHONE;
	}

	return ret;
}


bool rspamd_url_set_add_or_increase(khash_t(rspamd_url_hash) * set,
									struct rspamd_url *u,
									bool enforce_replace)
{
	khiter_t k;
	int r;

	k = kh_get(rspamd_url_hash, set, u);

	if (k != kh_end(set)) {
		/* Existing url */
		struct rspamd_url *ex = kh_key(set, k);
#define SUSPICIOUS_URL_FLAGS (RSPAMD_URL_FLAG_PHISHED | RSPAMD_URL_FLAG_OBSCURED | RSPAMD_URL_FLAG_ZW_SPACES)
		if (enforce_replace) {
			kh_key(set, k) = u;
			u->count++;
		}
		else {
			if (u->flags & SUSPICIOUS_URL_FLAGS) {
				if (!(ex->flags & SUSPICIOUS_URL_FLAGS)) {
					/* Propagate new url to an old one */
					kh_key(set, k) = u;
					u->count++;
				}
				else {
					ex->count++;
				}
			}
			else {
				ex->count++;
			}
		}

		return false;
	}
	else {
		k = kh_put(rspamd_url_hash, set, u, &r);
	}

	return true;
}

struct rspamd_url *
rspamd_url_set_add_or_return(khash_t(rspamd_url_hash) * set,
							 struct rspamd_url *u)
{
	khiter_t k;
	int r;

	if (set) {
		k = kh_get(rspamd_url_hash, set, u);

		if (k != kh_end(set)) {
			return kh_key(set, k);
		}
		else {
			k = kh_put(rspamd_url_hash, set, u, &r);

			return kh_key(set, k);
		}
	}

	return NULL;
}

bool rspamd_url_host_set_add(khash_t(rspamd_url_host_hash) * set,
							 struct rspamd_url *u)
{
	int r;

	if (set) {
		kh_put(rspamd_url_host_hash, set, u, &r);

		if (r == 0) {
			return false;
		}

		return true;
	}

	return false;
}

bool rspamd_url_set_has(khash_t(rspamd_url_hash) * set, struct rspamd_url *u)
{
	khiter_t k;

	if (set) {
		k = kh_get(rspamd_url_hash, set, u);

		if (k == kh_end(set)) {
			return false;
		}

		return true;
	}

	return false;
}

bool rspamd_url_host_set_has(khash_t(rspamd_url_host_hash) * set, struct rspamd_url *u)
{
	khiter_t k;

	if (set) {
		k = kh_get(rspamd_url_host_hash, set, u);

		if (k == kh_end(set)) {
			return false;
		}

		return true;
	}

	return false;
}

bool rspamd_url_flag_from_string(const char *str, int *flag)
{
	int h = rspamd_cryptobox_fast_hash_specific(RSPAMD_CRYPTOBOX_HASHFAST_INDEPENDENT,
												str, strlen(str), 0);

	for (int i = 0; i < G_N_ELEMENTS(url_flag_names); i++) {
		if (url_flag_names[i].hash == h) {
			*flag |= url_flag_names[i].flag;

			return true;
		}
	}

	return false;
}


const char *
rspamd_url_flag_to_string(int flag)
{
	for (int i = 0; i < G_N_ELEMENTS(url_flag_names); i++) {
		if (url_flag_names[i].flag & flag) {
			return url_flag_names[i].name;
		}
	}

	return NULL;
}

inline int
rspamd_url_cmp(const struct rspamd_url *u1, const struct rspamd_url *u2)
{
	int min_len = MIN(u1->urllen, u2->urllen);
	int r;

	if (u1->protocol != u2->protocol) {
		return u1->protocol - u2->protocol;
	}

	if (u1->protocol & PROTOCOL_MAILTO) {
		/* Emails specialisation (hosts must be compared in a case insensitive matter */
		min_len = MIN(u1->hostlen, u2->hostlen);

		if ((r = rspamd_lc_cmp(rspamd_url_host_unsafe(u1),
							   rspamd_url_host_unsafe(u2), min_len)) == 0) {
			if (u1->hostlen == u2->hostlen) {
				if (u1->userlen != u2->userlen || u1->userlen == 0) {
					r = (int) u1->userlen - (int) u2->userlen;
				}
				else {
					r = memcmp(rspamd_url_user_unsafe(u1),
							   rspamd_url_user_unsafe(u2),
							   u1->userlen);
				}
			}
			else {
				r = u1->hostlen - u2->hostlen;
			}
		}
	}
	else {
		if (u1->urllen != u2->urllen) {
			/* Different length, compare common part and then compare length */
			r = memcmp(u1->string, u2->string, min_len);

			if (r == 0) {
				r = u1->urllen - u2->urllen;
			}
		}
		else {
			/* Equal length */
			r = memcmp(u1->string, u2->string, u1->urllen);
		}
	}

	return r;
}

int rspamd_url_cmp_qsort(const void *_u1, const void *_u2)
{
	const struct rspamd_url *u1 = *(struct rspamd_url **) _u1,
							*u2 = *(struct rspamd_url **) _u2;

	return rspamd_url_cmp(u1, u2);
}