void rspamd_regexp_cache_insert (struct rspamd_regexp_cache* cache,
const gchar *pattern,
const gchar *flags, rspamd_regexp_t *re)
{
g_assert (re != NULL);
g_assert (pattern != NULL);
if (cache == NULL) {
rspamd_regexp_library_init ();
cache = global_re_cache;
}
g_assert (cache != NULL);
/* Generate custom id */
rspamd_regexp_generate_id (pattern, flags, re->id);
REF_RETAIN (re);
g_hash_table_insert (cache->tbl, re->id, re);
}
rspamd_regexp_t*
rspamd_regexp_cache_query (struct rspamd_regexp_cache* cache,
const gchar *pattern,
const gchar *flags)
{
rspamd_regexp_t *res = NULL;
regexp_id_t id;
if (cache == NULL) {
rspamd_regexp_library_init ();
cache = global_re_cache;
}
g_assert (cache != NULL);
rspamd_regexp_generate_id (pattern, flags, id);
res = g_hash_table_lookup (cache->tbl, id);
return res;
}
rspamd_regexp_t*
rspamd_regexp_new (const gchar *pattern, const gchar *flags,
GError **err)
{
const gchar *start = pattern, *end, *flags_str = NULL, *err_str;
rspamd_regexp_t *res;
pcre *r;
gchar sep = 0, *real_pattern;
gint regexp_flags = 0, rspamd_flags = 0, err_off, study_flags = 0, ncaptures;
gboolean strict_flags = FALSE;
rspamd_regexp_library_init ();
if (flags == NULL) {
/* We need to parse pattern and detect flags set */
if (*start == '/') {
sep = '/';
}
else if (*start == 'm') {
start ++;
sep = *start;
/* Paired braces */
if (sep == '{') {
sep = '}';
}
rspamd_flags |= RSPAMD_REGEXP_FLAG_FULL_MATCH;
}
if (sep == '\0' || g_ascii_isalnum (sep)) {
/* We have no flags, no separators and just use all line as expr */
start = pattern;
end = start + strlen (pattern);
rspamd_flags &= ~RSPAMD_REGEXP_FLAG_FULL_MATCH;
}
else {
end = strrchr (pattern, sep);
if (end == NULL || end <= start) {
g_set_error (err, rspamd_regexp_quark(), EINVAL,
"pattern is not enclosed with %c: %s",
sep, pattern);
return NULL;
}
flags_str = end + 1;
start ++;
}
}
else {
/* Strictly check all flags */
strict_flags = TRUE;
start = pattern;
end = pattern + strlen (pattern);
flags_str = flags;
}
rspamd_flags |= RSPAMD_REGEXP_FLAG_RAW;
regexp_flags &= ~PCRE_UTF8;
if (flags_str != NULL) {
while (*flags_str) {
switch (*flags_str) {
case 'i':
regexp_flags |= PCRE_CASELESS;
break;
case 'm':
regexp_flags |= PCRE_MULTILINE;
break;
case 's':
regexp_flags |= PCRE_DOTALL;
break;
case 'x':
regexp_flags |= PCRE_EXTENDED;
break;
case 'u':
rspamd_flags &= ~RSPAMD_REGEXP_FLAG_RAW;
regexp_flags |= PCRE_UTF8;
break;
case 'O':
/* We optimize all regexps by default */
rspamd_flags |= RSPAMD_REGEXP_FLAG_NOOPT;
break;
case 'r':
rspamd_flags |= RSPAMD_REGEXP_FLAG_RAW;
regexp_flags &= ~PCRE_UTF8;
break;
default:
if (strict_flags) {
g_set_error (err, rspamd_regexp_quark(), EINVAL,
"invalid regexp flag: %c in pattern %s",
*flags_str, pattern);
return NULL;
}
msg_warn ("invalid flag '%c' in pattern %s", *flags_str, pattern);
goto fin;
break;
}
flags_str++;
}
}
fin:
real_pattern = g_malloc (end - start + 1);
rspamd_strlcpy (real_pattern, start, end - start + 1);
r = pcre_compile (real_pattern, regexp_flags, &err_str, &err_off, NULL);
if (r == NULL) {
g_set_error (err, rspamd_regexp_quark(), EINVAL,
"invalid regexp pattern: '%s': %s at position %d",
pattern, err_str, err_off);
g_free (real_pattern);
return NULL;
}
/* Now allocate the target structure */
res = g_slice_alloc0 (sizeof (*res));
REF_INIT_RETAIN (res, rspamd_regexp_dtor);
res->flags = rspamd_flags;
res->pattern = real_pattern;
if (rspamd_flags & RSPAMD_REGEXP_FLAG_RAW) {
res->raw_re = r;
}
else {
res->re = r;
res->raw_re = pcre_compile (pattern, regexp_flags & ~PCRE_UTF8,
&err_str, &err_off, NULL);
if (res->raw_re == NULL) {
msg_warn ("invalid raw regexp pattern: '%s': %s at position %d",
pattern, err_str, err_off);
}
}
#ifdef HAVE_PCRE_JIT
study_flags |= PCRE_STUDY_JIT_COMPILE;
#endif
if (!(rspamd_flags & RSPAMD_REGEXP_FLAG_NOOPT)) {
/* Optimize regexp */
if (res->re) {
res->extra = pcre_study (res->re, study_flags, &err_str);
if (res->extra != NULL) {
#ifdef HAVE_PCRE_JIT
gint jit, n;
if (can_jit) {
jit = 0;
n = pcre_fullinfo (res->re, res->extra,
PCRE_INFO_JIT, &jit);
if (n != 0 || jit != 1) {
msg_debug ("jit compilation of %s is not supported", pattern);
res->jstack = NULL;
}
else {
res->jstack = pcre_jit_stack_alloc (32 * 1024, 512 * 1024);
pcre_assign_jit_stack (res->extra, NULL, res->jstack);
}
}
#endif
}
else {
msg_warn ("cannot optimize regexp pattern: '%s': %s",
pattern, err_str);
}
}
if (res->raw_re) {
if (res->raw_re != res->re) {
res->raw_extra = pcre_study (res->raw_re, study_flags, &err_str);
if (res->raw_extra != NULL) {
#ifdef HAVE_PCRE_JIT
gint jit, n;
if (can_jit) {
jit = 0;
n = pcre_fullinfo (res->raw_re, res->raw_extra,
PCRE_INFO_JIT, &jit);
if (n != 0 || jit != 1) {
msg_debug ("jit compilation of %s is not supported",
pattern);
res->raw_jstack = NULL;
}
else {
res->raw_jstack = pcre_jit_stack_alloc (32 * 1024,
512 * 1024);
pcre_assign_jit_stack (res->raw_extra, NULL,
res->raw_jstack);
}
}
#endif
}
else {
msg_warn ("cannot optimize raw regexp pattern: '%s': %s",
pattern, err_str);
}
}
else {
#ifdef HAVE_PCRE_JIT
/* Just alias pointers */
res->raw_extra = res->extra;
res->raw_jstack = res->jstack;
#endif
}
}
}
rspamd_regexp_generate_id (pattern, flags, res->id);
/* Check number of captures */
if (pcre_fullinfo (res->raw_re, res->extra, PCRE_INFO_CAPTURECOUNT,
&ncaptures) == 0) {
res->ncaptures = ncaptures;
}
return res;
}
rspamd_regexp_t*
rspamd_regexp_new (const gchar *pattern, const gchar *flags,
GError **err)
{
const gchar *start = pattern, *end, *flags_str = NULL;
gchar *err_str;
rspamd_regexp_t *res;
PCRE_T *r;
gchar sep = 0, *real_pattern;
#ifndef WITH_PCRE2
gint err_off;
#else
gsize err_off;
#endif
gint regexp_flags = 0, rspamd_flags = 0, err_code, ncaptures;
gboolean strict_flags = FALSE;
rspamd_regexp_library_init (NULL);
if (flags == NULL) {
/* We need to parse pattern and detect flags set */
if (*start == '/') {
sep = '/';
}
else if (*start == 'm') {
start ++;
sep = *start;
/* Paired braces */
if (sep == '{') {
sep = '}';
}
rspamd_flags |= RSPAMD_REGEXP_FLAG_FULL_MATCH;
}
if (sep == '\0' || g_ascii_isalnum (sep)) {
/* We have no flags, no separators and just use all line as expr */
start = pattern;
end = start + strlen (pattern);
rspamd_flags &= ~RSPAMD_REGEXP_FLAG_FULL_MATCH;
}
else {
end = strrchr (pattern, sep);
if (end == NULL || end <= start) {
g_set_error (err, rspamd_regexp_quark(), EINVAL,
"pattern is not enclosed with %c: %s",
sep, pattern);
return NULL;
}
flags_str = end + 1;
start ++;
}
}
else {
/* Strictly check all flags */
strict_flags = TRUE;
start = pattern;
end = pattern + strlen (pattern);
flags_str = flags;
}
rspamd_flags |= RSPAMD_REGEXP_FLAG_RAW;
#ifndef WITH_PCRE2
regexp_flags &= ~PCRE_FLAG(UTF8);
regexp_flags |= PCRE_FLAG(NEWLINE_ANYCRLF);
#else
regexp_flags &= ~PCRE_FLAG(UTF);
#endif
if (flags_str != NULL) {
while (*flags_str) {
switch (*flags_str) {
case 'i':
regexp_flags |= PCRE_FLAG(CASELESS);
break;
case 'm':
regexp_flags |= PCRE_FLAG(MULTILINE);
break;
case 's':
regexp_flags |= PCRE_FLAG(DOTALL);
break;
case 'x':
regexp_flags |= PCRE_FLAG(EXTENDED);
break;
case 'u':
rspamd_flags &= ~RSPAMD_REGEXP_FLAG_RAW;
#ifndef WITH_PCRE2
regexp_flags |= PCRE_FLAG(UTF8);
#else
regexp_flags |= PCRE_FLAG(UTF);
#endif
break;
case 'O':
/* We optimize all regexps by default */
rspamd_flags |= RSPAMD_REGEXP_FLAG_NOOPT;
break;
case 'r':
rspamd_flags |= RSPAMD_REGEXP_FLAG_RAW;
#ifndef WITH_PCRE2
regexp_flags &= ~PCRE_FLAG(UTF8);
#else
regexp_flags &= ~PCRE_FLAG(UTF);
#endif
break;
default:
if (strict_flags) {
g_set_error (err, rspamd_regexp_quark(), EINVAL,
"invalid regexp flag: %c in pattern %s",
*flags_str, pattern);
return NULL;
}
msg_warn ("invalid flag '%c' in pattern %s", *flags_str, pattern);
goto fin;
break;
}
flags_str++;
}
}
fin:
real_pattern = g_malloc (end - start + 1);
rspamd_strlcpy (real_pattern, start, end - start + 1);
#ifndef WITH_PCRE2
r = pcre_compile (real_pattern, regexp_flags,
(const char **)&err_str, &err_off, NULL);
(void)err_code;
#else
r = pcre2_compile (real_pattern, PCRE2_ZERO_TERMINATED,
regexp_flags,
&err_code, &err_off, pcre2_ctx);
if (r == NULL) {
err_str = g_alloca (1024);
memset (err_str, 0, 1024);
pcre2_get_error_message (err_code, err_str, 1024);
}
#endif
if (r == NULL) {
g_set_error (err, rspamd_regexp_quark(), EINVAL,
"regexp parsing error: '%s' at position %d",
err_str, (gint)err_off);
g_free (real_pattern);
return NULL;
}
/* Now allocate the target structure */
res = g_malloc0 (sizeof (*res));
REF_INIT_RETAIN (res, rspamd_regexp_dtor);
res->flags = rspamd_flags;
res->pattern = real_pattern;
res->cache_id = RSPAMD_INVALID_ID;
res->pcre_flags = regexp_flags;
res->max_hits = 0;
res->re = r;
if (rspamd_flags & RSPAMD_REGEXP_FLAG_RAW) {
res->raw_re = r;
}
else {
#ifndef WITH_PCRE2
res->raw_re = pcre_compile (real_pattern, regexp_flags & ~PCRE_FLAG(UTF8),
(const char **)&err_str, &err_off, NULL);
(void)err_code;
#else
res->raw_re = pcre2_compile (real_pattern, PCRE2_ZERO_TERMINATED,
regexp_flags & ~PCRE_FLAG(UTF),
&err_code, &err_off, pcre2_ctx);
if (res->raw_re == NULL) {
err_str = g_alloca (1024);
memset (err_str, 0, 1024);
pcre2_get_error_message (err_code, err_str, 1024);
}
#endif
if (res->raw_re == NULL) {
msg_warn ("raw regexp parsing error: '%s': '%s' at position %d",
err_str, real_pattern, (gint)err_off);
}
}
rspamd_regexp_post_process (res);
rspamd_regexp_generate_id (pattern, flags, res->id);
#ifndef WITH_PCRE2
/* Check number of captures */
if (pcre_fullinfo (res->raw_re, res->extra, PCRE_INFO_CAPTURECOUNT,
&ncaptures) == 0) {
res->ncaptures = ncaptures;
}
/* Check number of backrefs */
if (pcre_fullinfo (res->raw_re, res->extra, PCRE_INFO_BACKREFMAX,
&ncaptures) == 0) {
res->nbackref = ncaptures;
}
#else
/* Check number of captures */
if (pcre2_pattern_info (res->raw_re, PCRE2_INFO_CAPTURECOUNT,
&ncaptures) == 0) {
res->ncaptures = ncaptures;
}
/* Check number of backrefs */
if (pcre2_pattern_info (res->raw_re, PCRE2_INFO_BACKREFMAX,
&ncaptures) == 0) {
res->nbackref = ncaptures;
}
#endif
return res;
}
