back to article Linus Torvalds opts for the scream test: Linux kernel syscall tweaked to shut data-leak hole – anyone upset, yell now

The Linux kernel will be tweaked to mitigate data-stealing attacks that exploit system page caches. As we revealed first over the weekend, a group of experts – including some of the researchers who discovered the Spectre family of chip flaws – worked out how to get operating system page caches to leak information from one …

  1. Sir Runcible Spoon Silver badge
    Mushroom

    The Scream Test

    Is a proper engineering solution to get you to your goal as quickly as possible.

    Often the threat of a scream test is enough to get people thinking about the issue at hand which they would otherwise ignore.

    Glad to hear someone still employs them when required (I don't think of them as the first tool in the box to use, but you should certainly have it in there somewhere to pull out when all else fails - mainly because it works when other methods are hamstrung by politics).

    Oh, and he didn't swear? Are you sure this was Linus?

    1. Wellyboot Silver badge

      Re: The Scream Test

      given the age of someof this software lets hope >>>shouldn't have any downstream effects.<<< hasn't just jinxed it!

      1. A.P. Veening

        Re: The Scream Test

        There is no hope in there, Murphy just was invoked with that statement.

        1. }{amis}{ Silver badge
          Meh

          Re: The Scream Test

          Meh as with any scream test the results only seem to count if the screamer is far enough up the greasy pole to be heard......

    2. Doctor Syntax Silver badge

      Re: The Scream Test

      Yes, use it sparingly. If there's too much screaming going on already another one isn't going to get noticed.

    3. JohnFen Silver badge

      Re: The Scream Test

      "Glad to hear someone still employs them when required"

      ...but too many people employ them when they aren't required. Windows 10, anyone?

  2. Anonymous Coward
    Anonymous Coward

    Surely everyone knows...

    That a 0.0001% chance of something bad means it's sure to happen?

    Here's hoping that's not the case with this though ;-)

    1. imanidiot Silver badge

      Re: Surely everyone knows...

      “Scientists have calculated that the chances of something so patently absurd actually existing are millions to one. But magicians have calculated that million-to-one chances crop up nine times out of ten.”

  3. Herby Silver badge

    At least...

    Linux admits to it, and publicly at that. I suspect that there are MANY cases where you can hear screams all the way from Redmond Washington, and with the amount of "wonderful" releases, it appears that the scream might even be continuous.

    1. Anonymous Coward
      Anonymous Coward

      Re: At least...

      Herby,

      The scream *is* continuous BUT changes pitch and modulation as new people join and others run out of breath ..... only to rejoin again when the 'horror' hits again. !!! :)

      Why does this make me think of 'Clarice' & 'Silence of the Lambs' ???!!!

  4. jch
    Linux

    Hands up everyone that’s heard of mincore(2)

    OK. Anyone who hasn’t used it in anger, please put your hands down.

    Who’s left?

    The authors of the paper noted that in their admittedly limited test that mincore wasn’t even called once.

    I’m sure there is a good use for it, but I’m unsure what.

  5. StuntMisanthrope Bronze badge

    gitbot

    That’s why he’s been quiet. #trythememcache

  6. PyLETS

    Probably an access control issue

    With anything as complex as a multi CPU chip and OS kernels capable of using such for highly concurrent loadings efficiently, there will inevitably be performance versus security trade-offs with this class of bug (including Specter, Meltdown and similar). That means that not all programs running on a system (particularly a multi-tenanted data center server) should have access to certain kernel data structures or the ability to thrash the CPU to the extent it gives up predictive execution exposed secrets.

    So it seems to me the developers of these systems are either going to end up compromising the ability of a system running trusted processes to operate at full performance when they patch these bugs, or they're going to compromise the performance of untrusted processes and have to let the system know which processes are trusted and which are not.

    1. Mike 16 Silver badge

      Re: Probably an access control issue

      ---

      have to let the system know which processes are trusted and which are not.

      ---

      I thought one of the key principles in the Evil Overlord's Guide to World Domination was to _never_ let any of your "trusted" henchmen know which of them is actually _trusted_ (for now).

  7. Maelstorm Bronze badge
    Boffin

    The manual page for mincore for FreeBSD just shows if a page is allocated and if it has been modified by the calling process or otherwise. I've never really found a use for it as paging is handled by the operating system. What I have used those is madvise, mlock, munlock, mmap, and munmap when I wrote a pooled memory manager some years ago.

    Perhaps some libraries use it.

    FreeBSD Manual: mincore

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2019