Response to Meltdown and Spectre

[Paul Pathiakis]

https://www.amd.com/en/corporate/speculative-execution

This is the issue at hand regarding AMD chips.

Meltdown - Not susceptible

Specter variant 1 - It looks like nothing from recent production.  They 
state early Athlon/Opteron processors (year 2000 or so)  OS patches can 
correct the issue.  (FreeBSD has always had a much more robust memory 
management system than Linux and Windows including what some of call 
protected memory and a 'virtual fence' on the memory of the kernel.

Specter variant 2 - Possibly susceptible.  Everything is being examined 
and, again, OS patches can resolve the issue.  However, in both variants 
they are working on firmware to correct 'POSSIBLE' issues.

This has been the ongoing information coming out of AMD.  What I'm more 
interested in is the 'performance hit' after all the firmware updates 
come out.  This could be something similar to the use of IIS versus 
Apache.  IIS showed how 'fast' it was in benchmarks against Apache.  
However, once it was delved into, it turned out that IIS was not doing 
proper checking on the content..... it was actually doing very little 
which made it significantly faster, but more susceptible to exploit.  
So, once we get all the firmware patches, does it turn out that if Intel 
had been doing things correctly in their design does it slow their chips 
down by 20%?  If so, given their performance would drop below most 
high-end AMD chips that, effectively, already cost 1/2 of the 
performance, I'm dying to see what market share looks like after the 
mid-point of this year.

P.

On 2/4/2018 10:36 AM, Paul Pathiakis via freebsd-questions wrote:
>
>
> On 02/03/2018 20:00, Christian Weisgerber wrote:
>> On 2018-02-03, "Valeri Galtsev" <galtsev at kicp.uchicago.edu> wrote:
>>
>>> With all due respect, one person saying, it didn't affect me, doesn't
>>> prove it is not disastrous for somebody else. Even if it is one machine
>>> out of thousand that is "bricked" for some time, it is a disaster for
>>> sysadmin who has that machine as a production server
>> Of course, but who at all is saying that Intel's microcode updates
>> have "bricked" any machines?  This appears to be an entirely spurious
>> claim, based on nothing other than grievous exaggeration that turns
>> "higher system reboots" into "bricked". You guys are talking each
>> other into a frenzy of fear over nothing.
>>
> I would say we are not panicking in any manner.  Nor are we in a 
> frenzy.  Real sysadmins are cautious on everything/anything that can 
> affect the availability of the machines.  Any machine that is 
> 'unreliable' is 'bricked' or 'near-bricked' to us.  It causes a major 
> question about availability of the machine.  If Intel's patch hadn't 
> immediately caused issues, it may have caused something that might not 
> have been caught after it was rolled out to farms (aka 1000s) of 
> machine.   What then?
>
> There was a similar issue in the UK (I have since forgotten the name 
> of the data center owner) about 8 (or more) years regarding fujitsu 
> hard drives.  The data center was doing an excellent job of tracking 
> hard drive replacement every five years.  Fujitsu gave a great bid and 
> shipped the drives..... After less than 6 months, the drives started 
> to fail.  All the drives needed to be replaced after they had just 
> replaced all the drives a few months previous.  The company almost 
> went under.
>
> It's that simple in the world of a sysadmins in charge of a large 
> number of systems.
>
> Had that patch been rolled out to thousands and failed months later.....
>
> P.
> _______________________________________________
> freebsd-questions at freebsd.org mailing list
> https://lists.freebsd.org/mailman/listinfo/freebsd-questions
> To unsubscribe, send any mail to 
> "freebsd-questions-unsubscribe at freebsd.org"


---
This email has been checked for viruses by AVG.
http://www.avg.com