James Carlson wrote: > It'd be an interesting idea for testing, but I think it'd actually be > counter-productive to do this. The problem is that the actual privilege > enforcement (and thus the effects of each privilege bit) are hard-coded > into the kernel itself. There's no good way to replicate that logic out > into a user-space wrapper so that the code somehow 'knows' whether a > given system call should have succeeded give a privilege set.
Also for privilege debugging it shouldn't be necessary. This is what the "Privilege Debug Mode" is for see ppriv(1). For the cases where that isn't sufficient or accurate then the Sun Blueprint "Privilege Debugging in the Solaris 10 Operating System"[1] is useful. [1] http://www.sun.com/blueprints/0206/819-5507.pdf Not that I'm biased by being a co-author on the above blueprint, but I think that is a better way of dealing with privilege debugging that attempting to do a "fakeroot" for privileges which by its very nature of being upstream will rot and be wrong. It will also be even more of an issue if/when FMAC makes its way into OpenSolaris distributions. Having said all that I have no problem with fakeroot being delivered. I would have possible issues if I see OpenSolaris originated projects wanting to depend on fakeroot. -- Darren J Moffat
