From: Jon Masters <jcm@redhat.com>
To: speck@linutronix.de
Subject: [MODERATED] Re: L!TF Bulletin #3: State of the horrors
Date: Fri, 6 Jul 2018 11:32:07 -0400 [thread overview]
Message-ID: <4c7cf110-8ba5-0a6d-f88e-62474b590765@redhat.com> (raw)
In-Reply-To: <20180706003150.GE30152@char.US.ORACLE.com>
[-- Attachment #1: Type: text/plain, Size: 2280 bytes --]
On 07/05/2018 08:31 PM, speck for Konrad Rzeszutek Wilk wrote:
> On Wed, Jul 04, 2018 at 02:34:04PM -0700, speck for Linus Torvalds wrote:
>>
>>
>> On Wed, 4 Jul 2018, speck for Thomas Gleixner wrote:
>>>
>>> Aside of that the distro necrophiliacs have backported the lot already with
>>> the 'commit SHA1 upstream' references to their frankenkernels. Any opinion
>>> from the distro folks?
>>
>> Ugh, oh well. If that's the case, then perhaps leaving it alone is just
>> the right thing to do. I was thinking nobody had this yet, but I was
>> clearly wrong, so..
>
> With the FPU embargo breaking, spectre breaking, etc, we are thinking that this
> embargo may break too with the TLBleed going out early (and folks making
> the L1TF connection, not sure how) - so have been very busy backporting and
> running performance testing, various QA tests, etc.
On TLBleed. I met Ben last week and went over his current research
direction. He's onto NetSpectre and will probably independently have
that working within a few weeks. He's also exploring page table walks
BUT he's looking at them from the point of view of whether the walker
pulling in PTEs/PMDs is interesting from a leakage PoV (and isn't
looking at them speculatively from what I could tell from his summary).
I think it's very unlikely that he will find L1TF before next month at
this rate based upon everything I have seen. I also have the repro code
now (however he is not willing to share it more broadly just yet). The
one risk is that the SGX team behind Foreshadow are nearby.
[ I tossed out value prediction as an interesting area of security
research. Not entirely as a distraction, but it wouldn't hurt if folks
were looking at that right now rather than elsewhere. Value predictors
in mainstream parts that do other than guess always zero/one are still
quite novel. In theory you could combine a value predictor with a
variable latency ALU instruction or similar that has a shortcut for zero
operations to do some interesting attacks against non-hardened crypto.
It would be good to see a few papers actively hostile to value
predictors before the x86 crowd get too carried away with them. ]
Jon.
--
Computer Architect | Sent from my Fedora powered laptop
next prev parent reply other threads:[~2018-07-06 15:32 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-07-04 20:42 L!TF Bulletin #3: State of the horrors Thomas Gleixner
2018-07-04 21:05 ` [MODERATED] " Linus Torvalds
2018-07-04 21:30 ` Thomas Gleixner
2018-07-04 21:34 ` [MODERATED] " Linus Torvalds
2018-07-06 0:31 ` Konrad Rzeszutek Wilk
2018-07-06 15:32 ` Jon Masters [this message]
2018-07-05 18:56 ` Jon Masters
2018-07-05 20:07 ` Jiri Kosina
2018-07-05 20:11 ` Jiri Kosina
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4c7cf110-8ba5-0a6d-f88e-62474b590765@redhat.com \
--to=jcm@redhat.com \
--cc=speck@linutronix.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).