Re: [tip:x86/urgent] x86/microcode: Fix double vfree() and remove redundant pointer checks before vfree()

[Henrique de Moraes Holschuh <hmh@hmh.eng.br>]

On Mon, 27 Dec 2010, tip-bot for Jesper Juhl wrote:
> In arch/x86/kernel/microcode_intel.c::generic_load_microcode()
> we have  this:
> 
> 	while (leftover) {
> 		...
> 		if (get_ucode_data(mc, ucode_ptr, mc_size) ||
> 		    microcode_sanity_check(mc) < 0) {
> 			vfree(mc);
> 			break;
> 		}
> 		...
> 	}
> 
> 	if (mc)
> 		vfree(mc);

Which is trivial to trigger from userspace, fortunately limited to root in
any sane distro (and by default).

Please send it to -stable after it gets merged in mainline and is deemed
safe...

The Intel microcode driver will also accept a bogus microcode which has a
valid header, but all zeros for the cyphertext.  This is a design defect on
the current Intel microcode format, but can be worked around by the driver
if one assumes such a cyphertext will never be valid (seems like a safe
assumption).  I sure hope the processors will reject that kind of bogosity
even if the kernel doesn't and tries to upload it, though...  I was not
crazy enough to try.

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh