Re: [PATCH 02/10] userns: Add per user namespace sysctls.

[ebiederm-aS9lmoZGLiVWk0Htik3J/w@public.gmane.org (Eric W. Biederman)]

"Serge E. Hallyn" <serge-A9i7LUbDfNHQT0dZR+AlfA@public.gmane.org> writes:

> Quoting Eric W. Biederman (ebiederm-aS9lmoZGLiVWk0Htik3J/w@public.gmane.org):
>> "Serge E. Hallyn" <serge-A9i7LUbDfNHQT0dZR+AlfA@public.gmane.org> writes:
>> 
>> > Quoting Eric W. Biederman (ebiederm-aS9lmoZGLiVWk0Htik3J/w@public.gmane.org):
>> >> +static bool setup_userns_sysctls(struct user_namespace *ns)
>> >> +{
>> >> +#ifdef CONFIG_SYSCTL
>> >> +	struct ctl_table *tbl;
>> >> +	setup_sysctl_set(&ns->set, &set_root, set_is_seen);
>> >> +	tbl = kmemdup(userns_table, sizeof(userns_table), GFP_KERNEL);
>> >> +	if (tbl) {
>> >> +		ns->sysctls = __register_sysctl_table(&ns->set, "userns", tbl);
>> >> +	}
>> >
>> > What happens if tbl is null due to oom?  Would it be better to just
>> > return false in that case here?
>> 
>> ns->sysctls is initialized to NULL and kfree(NULL) is a noop.
>> So I don't see any problems.
>> 
>> I admit it isn't a usual pattern for error handling.
>
> Right I didn't mean that - I meant, is there a way that a namespace
> could end up escaping its limits as a result?  I think not, since
> either it has hierarchical limits which will still be enforced, or
> the admin tries to set a new limit and that step will fail visibly,
> which suffices.

And what I meant is that we do actually return false in the case you are
worrying about.  And if we return false we don't even create the user
namespace.


Eric