[redhat-lspp] security context in audit records (audit.39 kernel)

[Stephen Smalley]

On Fri, 2005-05-20 at 08:49 -0400, Stephen Smalley wrote:
> Any reason you can't use your selinux_getsecurity() helper for
> selinux_getprocattr() now that you are having it also return len for a
> NULL value?  However, a note of caution:  while proc_pid_attr_read will
> never pass a NULL value (and thus you can use that as your check for
> returning the actual length), it could possibly pass a 0 size if
> userspace passed a 0 size to the read(2) call, in which case it would
> _not_ expect getprocattr to be returning a non-zero length.  So you
> might need getprocattr to still reject a zero size with -ERANGE
> (restoring the original check).  This would still allow you to use
> selinux_getsecurity(), as long as you check for the 0 size first so it
> never reaches it in that case.

Actually, that is unpleasant because then you can't pass a 0 size along
with your NULL value for your call.  Ok, what if we instead disallow a 0
size with a non-NULL buffer, but allow a 0 size with a NULL buffer?
That will let your code call with a NULL buffer and 0 size to obtain the
length, while ensuring that proc_pid_attr_read never gets confused since
it always passes a non-NULL page.

-- 
Stephen Smalley
National Security Agency