[Info-vax] Intel junk...Kernel-memory-leaking Intel processor design flaw forces Linux, Windows redesign

[JF Mezei]

On 2018-01-06 10:27, Alan Browne wrote:

> The CPU memory controller is (usually) the arbiter of whether a fetch is 
> "legal" in the privilege scheme - so if something is allowed to be 
> fetched, then it is fetched.

Because the fetching is the longest one, you get it started as soon as
you get physical memory. While the fetching is happening, you do the
other checks such as access violation checks.  If you did the later
first, then it would slow down the computer because you are delaying teh
start of the longest portion of a memory access.

It seems to me the biggest vulnerability is that an unprivileged process
can access the CPU cache and bypass memnory access checks since that
isn't considered accessing memory.

More worrysome is that ARM would have similar design flaw.

I am guessing this has to do with cache coherence when multiple
processes on different cores share memory and when process 2 gets a
signal that process 1 has deposited memory, it needs to ensure that the
cache serving its core has been refreshed. (doing in in user mode
probably saves a lot of overhead of switching to kernel, doing it, and
switching back)