[LTP] [PATCH 4/7] syscalls/kill05: Use any two unprivileged users

[Martin Doucha]

On 06. 09. 21 14:50, Cyril Hrubis wrote:
> Hi!
>>> I do wonder if we have to lookup the UIDs in this case. As long as I
>>> understand the test correctly we are testing that no signals are send to
>>> a process if the sender real of effective UID do not match either of
>>> real and saved set UIDs of the receiving process. If my understanding is
>>> correct we can just take two arbitrary UIDs that != 0 and the test
>>> should work just fine.
>>
>> We could just use UID 1 and 2 for this test but I'd rather not assume
>> that set*uid() functions allow setting unused UIDs. Even if it works
>> now, it could easily break in the future or on some special system auth
>> backends.
> 
> Don't we already depend on being able to use unused GID? How is this
> different?

We only set unused user/group IDs on files and directories using
chown(). That's perfectly safe because inode owner is just an arbitrary
number that doesn't need to exist as a real entry in the user/group
database. You can get the same result by creating the files under a
valid owner and then deleting the respective user/group so the system
has to be able to handle that situation.

But setting a non-existent UID/GID on a running process is different
because the kernel might try to look up some additional data in the
user/group database and block the credentials change if the lookup fails.

-- 
Martin Doucha   mdoucha@suse.cz
QA Engineer for Software Maintenance
SUSE LINUX, s.r.o.
CORSO IIa
Krizikova 148/34
186 00 Prague 8
Czech Republic