Announcement

Collapse
No announcement yet.

L1 Terminal Fault - The Latest Speculative Execution Side Channel Attack

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #31
    Originally posted by oiaohm View Post

    This is causing those processing big data to be looking at risc-v and other things. Including strange ideas like having a cpu core having a thread management core. So you have a in-order core with upto 1024 threads cpu managed. So ever time it has to wait it changes to the next thread.

    In order means you don't have speculative execution issues. Yes we are use to hyperthreading being 2 thread per core. 1024 per core is a completely different beast.

    All these speculative execution issues could see the x86 loss the super computer and big server markets.
    Nobody is looking seriously at RISC-V. Some server farms have already moved to ARM.

    Comment


    • #32
      Originally posted by brrrrttttt View Post
      Nobody is looking seriously at RISC-V. Some server farms have already moved to ARM.
      https://insidehpc.com/2018/05/mateo-...architectures/
      Funny developer and maintainer of a super computer is nobody apparently.

      Sorry there are many videos covering the current RISC-V prototype chips being developed for different super computers and big data processing systems. True big data is looking at risc-v due to the fact that can invest custom silicon to in fact speed up processing. I do think it still going to be years before risc-v comes general server farm.

      Due to the silicon limit we are seeing the general processor loss its place in the top end.

      CPU managed threading is returning to the old cray barrel system.

      brrrrttttt so general server farm is having arm and power attempting to get traction. You see big data server farms and super computers are getting more custom hardware based around risc-v.

      Of course the question is will some form of new general chip appear out of big data and super computer usages of risc-v.

      Comment


      • #33
        Originally posted by oiaohm View Post
        https://insidehpc.com/2018/05/mateo-...architectures/
        Funny developer and maintainer of a super computer is nobody apparently.

        Sorry there are many videos covering the current RISC-V prototype chips being developed for different super computers and big data processing systems. True big data is looking at risc-v due to the fact that can invest custom silicon to in fact speed up processing. I do think it still going to be years before risc-v comes general server farm.

        Due to the silicon limit we are seeing the general processor loss its place in the top end.

        CPU managed threading is returning to the old cray barrel system.

        brrrrttttt so general server farm is having arm and power attempting to get traction. You see big data server farms and super computers are getting more custom hardware based around risc-v.

        Of course the question is will some form of new general chip appear out of big data and super computer usages of risc-v.
        Yeah, that seems likely in the longer term, but in the short term nobody is looking at it for general purpose computing.

        Comment


        • #34
          I am beginning to think all of the Intel-specific security issues, and I am assuming this one is "Intel-specific" until proven otherwise, are like the house guest (or in-laws, if you prefer) that simply will not leave.

          Comment


          • #35
            Originally posted by brrrrttttt View Post
            Yeah, that seems likely in the longer term, but in the short term nobody is looking at it for general purpose computing.
            There are a few different risc-v working groups attempting to get laptop and other general purpose computing worked out. This is still a few years out. Debates about how what would be called north and south bridges in intel architecture should be done for risc-v has not settled down yet there are a few competitors.
            https://genzconsortium.org/about-us/gen-z-technology/
            Its bus like gen-z that are competing.

            Basically there is a uplimit how long those using x86 platform will tolerate poor performance before risc-v, power, arm or combination of all three start eating into its market share.

            brrrrttttt if you define of short term in 5 years there is every possibility we will see the start of general purpose computing coming in risc-v. Of course standard bus will allow CPU to be made by 1 vendor and motherboard to be made by a different one.

            Do take close look at gen-z its not design to be CPU vendor particular. So we could have motherboards that don't care if you put a x86, arm, power, risc-v cpu in them. The board would be gen-z. Also gen-z you don't have to put identical cpu types in. So you could have a x86 with risc-v or any other combination like that.

            Yes the new busses for general purpose computing are not being designed with only x86 in mind. So where general purpose computing will start heading in the next few years is not locked in.

            Comment


            • #36
              Originally posted by torsionbar28 View Post
              Um, because intent requires evidence?
              I'm guessing you mean 'proving intent' not just 'intent'. Obviously you don't need evidence to have intent. But in a court room for instance, you need evidence to prove intent.

              I'm not in a court room, I suggested a plausible possibility, so I don't require evidence.

              Originally posted by torsionbar28 View Post
              Or is it just your belief?
              You repeat that back like a child in a playground.

              You made a statement of fact:
              "when it's incompetence in the case of speculative execution vulnerabilities."

              You don't appear to have evidence to back your statement; hence I asked if it's just your belief.

              I have not made any statements. I just suggested a possibility. There are a number of possibilities and if you know how to think, you should know that someone presenting a possibility is not the same as them presenting a statement of fact (as you did).

              I don't require evidence for my suggestion because I'm not saying it's true. You do require evidence as you made a statement of fact. If you don't have evidence then how about you admit that it's just your belief, or that you misspoke or that you simply take it back. Any of those would be better than making ridiculous, strawman, statements like this:

              Originally posted by torsionbar28 View Post
              Every bug is not a conspiracy.

              Comment


              • #37
                Originally posted by oiaohm View Post
                This is causing those processing big data to be looking at risc-v and other things. Including strange ideas like having a cpu core having a thread management core. So you have a in-order core with upto 1024 threads cpu managed. So ever time it has to wait it changes to the next thread.

                In order means you don't have speculative execution issues. Yes we are use to hyperthreading being 2 thread per core. 1024 per core is a completely different beast.

                All these speculative execution issues could see the x86 loss the super computer and big server markets.
                In-order would make the CPU extremely slow. IBM tried already a similar experiment and it was total failure.

                To be clear: I'm talking about general purpose computing, not supercomputers, which nobody here really gives a crap about. You know, the kind of thing with millions of branches all over the place and very little straight parallelization potential. That's why VLIW was a failure in this context.
                Last edited by Weasel; 15 August 2018, 07:41 AM.

                Comment


                • #38
                  Originally posted by audir8 View Post
                  I also don't think Intel meant to design something like SGX and then self-disclose these vulnerabilities that affect SGX. We didn't see meltdown/L1T/spectre come up in the Snowden or other NSA/CIA hacks, so that makes me believe this was just losey-gosey CPU design. Which didn't happen at AMD and others.
                  Well, SGX is only for DRM, and not much else, and would still work like that with NSA access, but then why would they want your Netflix show?

                  Comment


                  • #39
                    Originally posted by carewolf View Post

                    Well, SGX is only for DRM, and not much else, and would still work like that with NSA access, but then why would they want your Netflix show?
                    SGX can be used for more than just DRM.

                    For example, Cornell University people proposed that you could increase the performance of blockchains by doing off-chain transactions using Intel SGX or similar "secure", trusted hardware technologies. If this Teechan technology was implemented and the promises made by Intel SGX technology were broken due to L1TF (or similar), then people using this Teechan technology could be defrauded.

                    Intel SGX (and similar) could also be used to implement anti-cheat features in games. It could be used to reduce sybil attacks (IE remote services could require that your Intel SGX chip answers a question that only it can answer to prove that it's a physical CPU. This would raise the cost of sybil attacking services, e.g. voting many times in a poll).

                    So Intel SGX exploits could potentially be used in the future to manipulate polls to manipulate public opinion, to steal from you and even to cheat in video games. I doubt the NSA want to cheat against you in video games, but I'm pretty sure they'd be interested in the former 2 items.

                    Comment


                    • #40
                      Originally posted by WesternSemiconductor View Post

                      Of course there's a reason, and a very good one: plausible deniability.

                      Only a moron would plant backdoors that can be traced back to them. The intelligence agencies are, unfortunately, not run by morons.



                      More likely because it was classified SCI, not TS. Hardware vulnerabilities take almost a decade to pervade the majority of the gear out there; if they're discovered that's a decade of work lost. It's the kind of exploit that can be used even if only ten people know about it. Sprawling behemoths like PRISM need a massive workforce to keep them going, so they really can't be kept secret from the majority of NSA's own workforce.
                      L1TF is in the same category of bugs as Meltdown and Spectre, so to believe what you're saying, this whole class of bugs had to have been planted by the NSA a decade ago with some cooperation from multiple vendors or their employees, only to be found recently and in this case be self-disclosed now. Except for certain variants of Spectre, all the other flaws only show up in Intel chips.

                      I guess you can believe what you want. I think it's much more likely that Intel did have 12 year olds designing processors.

                      carewolf SGX isn't only for DRM, it can just be considered an extension of TPM, which would just make it good security if done well.

                      Comment

                      Working...
                      X