Modern CPU’s are very complex—much worse than the behemoths of old. Bugs creep in.

I don’t have references at hand but I recall that Intel recently committed, for each of their products, to promptly disclose all the deviations from specs as they became aware of them. This was partly in response to the delay in Intel’s admitting to the famous Pentium divide problem. I am pleased with this new policy. As I read the description of the problems with the recent Duo line of processors, however, I am disturbed that it seems impossible to determine whether a given bug is exploitable by malware. Of course there is the eternal dilemma that if the bug is exploitable then publication in such a document may hasten exploitation. I suppose that information that allows a kernel designer to maintain control over malware also informs the design of such malware. As the document stands it provides fertile ground for experimentation by both black hats and white hats.

The format of the document seems oblivious to the issue of malware. But then again I know no format that avoids the dilemma. A common phrase is “Intel has not observed this erratum with any commercially available software, or system.”. It seems like a prescription for malware however.

Some problems sound like only a denial of service attack like the XEC * problem with CTSS. That problem restricted the applicability of the CTSS technology, but did not keep the system from being mostly useful in its actual context.

Another system state “cause[s] unpredictable processor behavior”. A secure system needs some bounds on such behavior when it can be invoked by user mode programs.

I suspect that some modern RISC processors may have rather fewer exploitable bugs but that is mainly just a hope. I have not seen commitments by other manufactures to report such “Spec Updates”.