On Sat, 29 Aug 2020 15:07:40 +0100, druck wrote:

> On 28/08/2020 21:53, Pancho wrote:
>> On 28/08/2020 20:19, Dennis Lee Bieber wrote:
>>>     If ARM is considered RISC, it is still pipelined:
>>> https://www.nccgroup.com/us/about-us/newsroom-and-events/blog/2011/
september/arm-pipeline-and-gdb-oh-my/
>>>
>>> https://en.wikipedia.org/wiki/
ARM_architecture#Pipelines_and_other_implementation_issues
>>>
>>>
>>>
>> ARM is pipelined, I think all chips are now.
>
> Even the ARM2 was pipelined, you'll probably have to go back to the 8
> bit ear such as the 6502, to find one which wasn't.
>
> Pipelining isn't the problem though, its the vastly more complex aspects
> of modern processors such as branch prediction, speculative execution
> and caching which are targetted by these exploits.
>
> No protected information is directly leaked by the bugs, but
> manipulations of these features can allow protected information to be
> deduced from timing differences of cached or uncached memory accesses.
>
Fair comment - the 6809, which it the 8/16 bit chip I know best from
writing a fair amount assembler on one back on the day, of isn't
pipelined either.

I also have a lot of time on a 68020, running OS-9/68000 but only writing
C and a 4GL on it, so I am relatively unfamiliar with its execution
modes.

The 68020 has a 3-stage pipeline and a 256 byte instruction cache, but no
protected address space or protection rings. I'd guess from this that it
is impervious to Spectre or similar attacks - not that you'd ever need
them because all RAM is accessible to any process. BTW, under OS-9 all
processes are written in position-independent code, so processes can be
loaded anywhere in RAM and the OS can move any user process in RAM,
provided its not running - something that is guaranteed while the OS is
rearranging processes in memory.


--
Martin    | martin at
Gregorie  | gregorie dot org

--- SoupGate-Win32 v1.05