back to article Google 'cubists' fix bug in Linux network congestion control, boost performance

A bit of “quality, non-glamorous engineering” could give a bunch of Linux servers a boost by addressing an unnoticed bug in a congestion control algorithm. This little code snippet addresses the ten-year-old slip-up in the open-source kernel's net/ipv4/tcp_cubic.c code: static void bictcp_cwnd_event(struct sock *sk, enum …

  1. petur
    Pint

    Nice one

    Here's to hoping it gets backported

    1. Anonymous Coward
      Anonymous Coward

      Re: Nice one

      The patch is simple enough so it should be easily applicable to older versions of the Kernel. Whether it actually gets backported will depend on the distro manufacturers of course.

    2. Vic

      Re: Nice one

      Here's to hoping it gets backported

      If it's important enough to you to get it into a given kernel version, you can either do it yourself, or pay someone to do it for you...

      Vic.

  2. Anonymous Coward
    Anonymous Coward

    The power of open source

    This just goes to demonstrate the advantages of the open source model. A 10-year old screw-up identified and fixed, everything completely open and above board. Let's compare that to the closed-source model used by certain companies Who Will Not Be Named; in that case they would probably never have bothered to review the code and identify the bug, and even if they did they would probably have hushed it up and never admitted to the mistake.

    1. sabroni Silver badge
      WTF?

      Re: The power of open source

      And probably fucked your dog. Probably.

      But I see your point, a bug like this hanging around for 10 years definitely proves that open source is better than closed.

      1. dotdavid

        Re: The power of open source

        "a bug like this hanging around for 10 years definitely proves that open source is better than closed"

        Prove that similar bugs are not in equivalent closed-source software and I'll believe your sarcasm is warranted.

        1. TeeCee Gold badge
          Meh

          Re: The power of open source

          Oh I think Heartbleed did that for us. Every web server in teh hole wurld fucked, except the Windows/IIS ones which didn't use the open source library with the "programming for dummies" example memory usage fuckup in it.

          The sad fact is that either open or closed is equally as likely to have cockups in it and these are only usually found when somebody either spots a problem and looks for the cause or runs across it while modifying the code for another reason. The idea of thousands of highly-skilled people burning the midnight oil scrutinising every line of open-source code for bugs on the off-chance is a lovely one, but in the real world it just doesn't happen.

          1. Alan Brown Silver badge

            Re: The power of open source

            "The idea of thousands of highly-skilled people burning the midnight oil scrutinising every line of open-source code for bugs on the off-chance is a lovely one, but in the real world it just doesn't happen."

            It's far more profitable to look for them in closed source and sell to the highest bidder.

            1. Anonymous Coward
              Facepalm

              Re: The power of open source

              "The idea of thousands of highly-skilled people burning the midnight oil scrutinising every line of open-source code for bugs on the off-chance is a lovely one, but in the real world it just doesn't happen."

              RTFA

          2. e^iπ+1=0

            Re: The power of open source

            "The sad fact is that either open or closed is equally as likely to have cockups in it and these are only usually found when somebody either spots a problem and looks for the cause or runs across it while modifying the code for another reason."

            The advantage for me as a programmer using the open source code is that when my program doesn't behave as I expected I can just read the source code for the api / subroutine I was trying to use rather than just moan about the (closed source) api's behaviour.

      2. Vic

        Re: The power of open source

        a bug like this hanging around for 10 years definitely proves that open source is better than closed

        Read the description - it's a minor annoyance that will cause a bit more congestion on certain peaky networks. It's no big deal for most of us, and no show-stopper for the others. This is more upgrade than bug-fix.

        It would appear to be a good patch that will improve performance for one group of users. And that improvement is possible because someone who did not write the code was able to look at it.

        Vic.

      3. hollymcr

        Re: The power of open source

        Surely the point here about open source is that a user of the software (not its "owner") found the problem, created a fix for the problem, and released the fix making it available to other users.

        Finding a problem in (eg) Windows isn't hard, but the next two steps would be impossible as an end user.

        1. sabroni Silver badge
          Facepalm

          Re: the point here about open source is that a user of the software created a fix for the problem

          Well ok, if you explain it nice and clearly like that it does make sense.....

  3. Pascal Monett Silver badge

    This seems to be mostly a server-side issue

    I know next to nothing about network management, but it seems to me that this bug mostly impacts the server, not the client.

    As such, it will be fixed, because generally admins are vastly more consciencious of how their servers work than plain old desktop users.

    So this is good news.

    1. Anonymous Coward
      Anonymous Coward

      Re: This seems to be mostly a server-side issue

      Although server side, the client will still be impacted, albeit the impact may be less perceptible. Both ends of the conversation will benefit from a smoother data flow.

    2. Anonymous Coward
      Anonymous Coward

      Re: This seems to be mostly a server-side issue

      The "clients" running Linux are quite affected indeed, both when uploading and downloading data from a server running any OS. The bug is in the core implementation of TCP on Linux.

      "Since Linux 2.6.13, BIC had been included in the standard Linux distribution and set to the default TCP. Currently, the successor of the BIC, CUBIC, is set to default. If your kernel version is greater than 2.6.13"

    3. Pascal Monett Silver badge

      Right. Wrong wording on my part.

      I agree that the issue impacts everyone. What I meant to say is that I think this can only be _corrected_ on the server.

      Once the correction is in place (and, as I said, since it is a server-side issue, it will be), then indeed everyone will be happier.

      1. Vic

        What I meant to say is that I think this can only be _corrected_ on the server.

        No. This can be corrected everywhere. And, over time, it will be.

        The issue occurs anywhere a machine can dump peaky traffic onto a network. This might be a big server - or it might be a laptop uploading a bunch of photos. Both scenarios will be improved.

        Vic.

        1. DougS Silver badge

          Linux clients matter

          Given that there are over a billion Android phones, all of which would have this bug. Few existing ones will ever see the fix, but once the fix is applied to Android and that version starts getting into new phones and old phones get tossed there will be fewer broken clients out there over time.

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