Reply to post: Puzzled

Meltdown, Spectre: The password theft bugs at the heart of Intel CPUs

Ken Hagan Gold badge


The description in the article would seem to allow a fairly simple fix in the OS.

When the original page fault occurs, control is passed from user-space (or guest space) to kernel space (or host space). The handler can determine whether the faulting address is outside user-space or not. In fact, it probably already has to do that in order to process the fault. If not, the fault is legitimate and will be related to (say) stack guard pages or virtual memory paging. We wouldn't want to penalise those, so we proceed as usual.

However, if it *is* outside user-space, I can't see any reason not to "punish" the application program (or guest kernel) by performing a full cache flush. This blocks the information disclosure. It is obviously quite costly, but as long as the bill is charged to the offending application (or guest, and in the case of cloud providers that will really mean *charged* so the provider is still happy) then it doesn't count as a DoS attack and no properly written application will ever have to pay the bill.

What have I missed?

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2019