* Add option max_size for regexp module to skip expensive regexp on long messages

src/plugins/regexp.c

@@ -57,12 +57,13 @@ struct autolearn_data {
 };
 
 struct regexp_ctx {
-	gint                            (*filter) (struct worker_task * task);
+	gint                          (*filter) (struct worker_task * task);
 	GHashTable                     *autolearn_symbols;
-	gchar                           *statfile_prefix;
+	gchar                          *statfile_prefix;
 
 	memory_pool_t                  *regexp_pool;
 	memory_pool_t                  *dynamic_pool;
+	gsize                           max_size;
 };
 
 struct regexp_json_buf {
@@ -444,6 +445,7 @@ regexp_module_init (struct config_file *cfg, struct module_ctx **ctx)
 
 	(void)luaopen_regexp (cfg->lua_state);
 	register_module_opt ("regexp", "dynamic_rules", MODULE_OPT_TYPE_STRING);
+	register_module_opt ("regexp", "max_size", MODULE_OPT_TYPE_SIZE);
 	register_module_opt ("regexp", "/^\\S+$/", MODULE_OPT_TYPE_STRING);
 
 	return 0;
@@ -497,6 +499,12 @@ regexp_module_config (struct config_file *cfg)
 	else {
 		regexp_module_ctx->statfile_prefix = DEFAULT_STATFILE_PREFIX;
 	}
+	if ((value = get_module_opt (cfg, "regexp", "max_size")) != NULL) {
+		regexp_module_ctx->max_size = parse_limit (value);
+	}
+	else {
+		regexp_module_ctx->max_size = 0;
+	}
 	if ((value = get_module_opt (cfg, "regexp", "dynamic_rules")) != NULL) {
 		jb = g_malloc (sizeof (struct regexp_json_buf));
 		pjb = g_malloc (sizeof (struct regexp_json_buf *));
@@ -725,6 +733,12 @@ process_regexp (struct rspamd_regexp *re, struct worker_task *task, const gchar
 				cur = g_list_next (cur);
 				continue;
 			}
+			/* Skip too large parts */
+			if (regexp_module_ctx->max_size != 0 && part->content->len > regexp_module_ctx->max_size) {
+				msg_info ("<%s> skip part of size %Hud", task->message_id, part->content->len);
+				cur = g_list_next (cur);
+				continue;
+			}
 			/* Check raw flags */
 			if (part->is_raw) {
 				regexp = re->raw_regexp;
@@ -795,6 +809,10 @@ process_regexp (struct rspamd_regexp *re, struct worker_task *task, const gchar
 		ct = task->msg->begin;
 		clen = task->msg->len;
 
+		if (regexp_module_ctx->max_size != 0 && clen > regexp_module_ctx->max_size) {
+			msg_info ("<%s> skip message of size %Hz", task->message_id, clen);
+			return 0;
+		}
 		/* If we have limit, apply regexp so much times as we can */
 		if (f != NULL && limit > 1) {
 			end = 0;

src/printf.c

@@ -25,6 +25,69 @@
 #include "fstring.h"
 #include "main.h"
 
+/**
+ * From FreeBSD libutil code
+ */
+static const int maxscale = 7;
+
+static gchar *
+humanize_number (gchar *buf, gchar *last, gint64 num, gboolean bytes)
+{
+	const gchar *prefixes;
+	int i, r, remainder, sign;
+	gint64 divisor;
+	gsize baselen, len = last - buf;
+
+	remainder = 0;
+
+	baselen = 1;
+	if (!bytes) {
+		divisor = 1000;
+		prefixes = "\0\0\0k\0\0M\0\0G\0\0T\0\0P\0\0E";
+	}
+	else {
+		divisor = 1024;
+		prefixes = "B\0\0k\0\0M\0\0G\0\0T\0\0P\0\0E";
+	}
+
+
+#define SCALE2PREFIX(scale)     (&prefixes[(scale) * 3])
+
+	if (num < 0) {
+		sign = -1;
+		num = -num;
+		baselen += 2; /* sign, digit */
+	}
+	else {
+		sign = 1;
+		baselen += 1; /* digit */
+	}
+
+	/* Check if enough room for `x y' + suffix + `\0' */
+	if (len < baselen + 1) {
+		return buf;
+	}
+
+	/*
+	 * Divide the number until it fits the given column.
+	 * If there will be an overflow by the rounding below,
+	 * divide once more.
+	 */
+	for (i = 0; i < maxscale && num > divisor; i++) {
+		remainder = num % divisor;
+		num /= divisor;
+	}
+
+	r = rspamd_snprintf (buf, len, "%L%s",
+			sign * (num + (remainder + 50) / 1000),
+			SCALE2PREFIX (i));
+
+#undef SCALE2PREFIX
+
+	return buf + r;
+}
+
+
 static gchar *
 rspamd_sprintf_num (gchar *buf, gchar *last, guint64 ui64, gchar zero,
 	guint                           hexadecimal, guint width)
@@ -214,7 +277,7 @@ rspamd_vsnprintf (gchar *buf, glong max, const gchar *fmt, va_list args)
 	size_t              len, slen;
 	gint64              i64;
 	guint64             ui64;
-	guint               width, sign, hex, max_width, frac_width, i;
+	guint               width, sign, hex, humanize, bytes, max_width, frac_width, i;
 	f_str_t			   *v;
 	GString            *gs;
 
@@ -240,6 +303,8 @@ rspamd_vsnprintf (gchar *buf, glong max, const gchar *fmt, va_list args)
 			width = 0;
 			sign = 1;
 			hex = 0;
+			bytes = 0;
+			humanize = 0;
 			max_width = 0;
 			frac_width = 0;
 			slen = (size_t) -1;
@@ -273,6 +338,17 @@ rspamd_vsnprintf (gchar *buf, glong max, const gchar *fmt, va_list args)
 					sign = 0;
 					fmt++;
 					continue;
+				case 'H':
+					humanize = 1;
+					bytes = 1;
+					sign = 0;
+					fmt ++;
+					continue;
+				case 'h':
+					humanize = 1;
+					sign = 0;
+					fmt ++;
+					continue;
 				case '.':
 					fmt++;
 
@@ -566,7 +642,12 @@ rspamd_vsnprintf (gchar *buf, glong max, const gchar *fmt, va_list args)
 				}
 			}
 
-			buf = rspamd_sprintf_num (buf, last, ui64, zero, hex, width);
+			if (!humanize) {
+				buf = rspamd_sprintf_num (buf, last, ui64, zero, hex, width);
+			}
+			else {
+				buf = humanize_number (buf, last, ui64, bytes);
+			}
 
 			fmt++;
 

src/printf.h

@@ -31,11 +31,11 @@
  * supported formats:
  *	%[0][width][x][X]O		    off_t
  *	%[0][width]T			    time_t
- *	%[0][width][u][x|X]z	    ssize_t/size_t
- *	%[0][width][u][x|X]d	    gint/guint
- *	%[0][width][u][x|X]l	    long
- *	%[0][width][u][x|X]D	    gint32/guint32
- *	%[0][width][u][x|X]L	    gint64/guint64
+ *	%[0][width][u][x|X|h|H]z	    ssize_t/size_t
+ *	%[0][width][u][x|X|h|H]d	    gint/guint
+ *	%[0][width][u][x|X|h|H]l	    long
+ *	%[0][width][u][x|X|h|H]D	    gint32/guint32
+ *	%[0][width][u][x|X|h|H]L	    gint64/guint64
  *	%[0][width][.width]f	    double
  *	%[0][width][.width]F	    long double
  *	%[0][width][.width]g	    double