Pavel Emelyanov wrote:
> Subrata Modak wrote:
>> Hi Pavel,
>>
>> We have PID, SYSVIPC and UTSNAMESPACE test cases written recently in LTP
>> (Linux Test Project) and used by people for their namespace testing. Would
>> you like to use them and review them for us, telling about itś effectiveness
>> and other stuff:
Well, current tests look sane, but there's something that caught my
eyes immediately.
1) The very important thing that has to be checked, but is missed is
that parent from init_pid_ns can waitpid (not just wait) a child
from a sub namespace by its pid returned from clone. It will be *not*
1, but some number above it.
I.e. test like this:
int child(void *x)
{
return 0;
}
int main(void)
{
int pid;
pid = clone(child, CLONE_NEWPID, stack, NULL);
if (waitpid(pid) == -1 && errno == ESRCH)
/* FAIL - pid must exist */
else
/* Either we're OK, or some other bad
thing (but not pid-ns-related) happened */
}
2) The other thing to check (also very important) is that child from
newly created namespace cannot see its parent from the init one by this
parent's pid.
I.e. smth like this
int child(void *x)
{
int ppid = (int)x;
if (kill(ppid, 0) == -1 && errno == ESRCH)
/* PASS - parent is invisible */
else
/* FAIL - parent is visible (but maybe
just not enough perms to kill one */
return 0;
}
int main(void)
{
int pid;
pid = clone(child, CLONE_NEWPID, stack, (void *)getpid());
wait();
}
I will try to produce more feedback in the future :)
>> http://ltp.cvs.sourceforge.net/ltp/ltp/testcases/kernel/containers/
>
> I will try to take a look early.
>
> Thanks,
> Pavel
>
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list
