MT#63826 db_redis: revert support of NULL values

db_redis module implicitly supports conversion
of <null> key values into empty "" strings.

Remove empty-string/zero-length guards introduced
previously, because they break this behavior.

E.g.: dialog and usrloc modules can in fact
handle <null> key values.

Additionally: introduce the memcpy() guard.

Change-Id: I96515b8a4dfea6eef16a05b5d9f970946a9fc915
(cherry picked from commit 743265a8c4)

debian/patches/series

@@ -46,6 +46,8 @@ sipwise/presence_offline_cleanup.patch
 sipwise/permissions-don-t-allow-reloads-in-the-middle-of-ong.patch
 sipwise/dlg_get_var_error_more_verbose_on_dlg
 sipwise/db_redis_protect_length_overflow.patch
+sipwise/db_redis_support_null_key_values.patch
+
 ### Don't just put stuff in any order
 ### use gbp pq import/export tooling to help maintain patches
 ###

debian/patches/sipwise/db_redis_support_null_key_values.patch

@@ -0,0 +1,31 @@
+--- a/src/modules/db_redis/redis_table.c
++++ b/src/modules/db_redis/redis_table.c
+@@ -32,11 +32,6 @@ int db_redis_key_add_string(redis_key_t
+ {
+ 	redis_key_t *k;
+ 
+-	if (!entry || !len) {
+-		LM_ERR("Empty entry or zero length\n");
+-		return -1;
+-	}
+-
+ 	if (db_redis_max_key_len > 0 && len > db_redis_max_key_len) {
+ 		LM_ERR("Too big length for key being added: allowed '%u' / given '%zu'\n",
+ 				db_redis_max_key_len, len);
+@@ -56,8 +51,14 @@ int db_redis_key_add_string(redis_key_t
+ 		goto err;
+ 	}
+ 
+-	memcpy(k->key.s, entry, len);
+-	k->key.s[len] = '\0';
++	/* run memcpy only on non-NULL pointer, because in fact it may happen
++	 * it comes here empty and with len = 0, this is then an implicit
++	 * conversion of <null> redis key value into the empty "" string.
++	 * see `db_redis_val2str()`
++	 * This is the allowed behavior, but avoid then running memcpy() on it. */
++	if (entry && len > 0)
++		memcpy(k->key.s, entry, len);
++	k->key.s[len] = '\0'; /* at least 1 byte is already pre-allocated before */
+ 	k->key.len = len;
+ 
+ 	if(!*list) {