[patches] Re: [box 1/1] box: Change checks on usage access

[Konstantin Osipov]

* imarkov <imarkov at tarantool.org> [18/01/29 15:46]:
> -	    sequence_access & ~seq->access[cr->auth_token].effective) {
> +	if (sequence_access == PRIV_U || (seq->def->uid != cr->uid &&
> +	    sequence_access & ~seq->access[cr->auth_token].effective)) {

Why == PRIV_U and not & PRIV_U? Couldn't owner access be missing other
access bits? If I am right, this needs to be covered by a separate
test case.

>  		/* Access violation, report error. */
>  		struct user *user = user_find(cr->uid);
>  		if (user != NULL) {
> diff --git a/src/box/space.c b/src/box/space.c
> index c02eb88..e9d6fdd 100644
> --- a/src/box/space.c
> +++ b/src/box/space.c
> @@ -55,8 +55,9 @@ access_check_space(struct space *space, user_access_t access)
>  	 */
>  	user_access_t space_access = access & ~cr->universal_access;
>  
> -	if (space_access && space->def->uid != cr->uid &&
> -	    space_access & ~space->access[cr->auth_token].effective) {
> +	if (space_access == PRIV_U ||
> +	    (space_access && space->def->uid != cr->uid &&
> +	    space_access & ~space->access[cr->auth_token].effective)) {

I can see you're checking spaces and sequences. What about other
objects? You need to always look at enum object_type when writing
patches. Please mention in the changeset comment why other object
types are not affected if I am wrong.


-- 
Konstantin Osipov, Moscow, Russia, +7 903 626 22 32
http://tarantool.org - www.twitter.com/kostja_osipov