Announcement

Collapse
No announcement yet.

Linux 5.15-rc5 x86 Changes Aim To Fix "Yet Another Hardware Trainwreck"

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

  • Linux 5.15-rc5 x86 Changes Aim To Fix "Yet Another Hardware Trainwreck"

    Phoronix: Linux 5.15-rc5 x86 Changes Aim To Fix "Yet Another Hardware Trainwreck"

    Sent in this morning were an urgent set of x86 updates for the Linux 5.15-rc5 kernel due out later today...

    https://www.phoronix.com/scan.php?pa...x-5.15-rc5-x86

  • #2
    Honestly, the changes to clocksource/timers introduced with kernel 5.13 or 5.14 are a mess. I did get constant TSC errors and failed recalibrations (3 retries, then TSC gets disabled) via HPET due to the new WATCHDOG_MAX_SKEW (kernel/time/clocksource.c) setting of 50 being way too tight (the limit before was set to 100 - double of what it's now).

    R9 3900X
    X570 Aorus Pro r1.0 F34

    I manually had to patch my kernel in order to increase the WATCHDOG_MAX_SKEW to 60. Now I get 2-3 successful (at the first retry) recalibrations in 48hrs of uptime and no more warnings of TSC drift being too big. Honestly, please just leave the timers alone on X86 as in "never touch a running system". They've been working fine before afterall…

    Comment


    • #3
      That might explain why my Rocket Lake system runs in slow motion when I force HPET as clock source. Great job, Intel...
      Edit: Seems to be much worse on Windows than on Linux. Not sure if Linux even is affected in a similar way.

      kiffmet TSC gets disabled for you due to failed HPET calibrations? What happens when you boot with hpet=disable ?
      Last edited by aufkrawall; 10 October 2021, 07:32 AM.

      Comment


      • #4
        I know CPUs are complex beasts, but does anyone know why it is that hard to fix the bugs in hardware?

        Comment


        • #5
          Originally posted by ms178 View Post
          I know CPUs are complex beasts, but does anyone know why it is that hard to fix the bugs in hardware?
          My guess is backward compatibility stuff that prevents x86 architecture from moving forward.
          Not to mention that AMD/Intel often releases instructions specific to their own CPU set.

          Comment


          • #6
            Originally posted by kiffmet View Post
            Honestly, the changes to clocksource/timers introduced with kernel 5.13 or 5.14 are a mess. I did get constant TSC errors and failed recalibrations (3 retries, then TSC gets disabled) via HPET due to the new WATCHDOG_MAX_SKEW (kernel/time/clocksource.c) setting of 50 being way too tight (the limit before was set to 100 - double of what it's now).

            R9 3900X
            X570 Aorus Pro r1.0 F34

            I manually had to patch my kernel in order to increase the WATCHDOG_MAX_SKEW to 60. Now I get 2-3 successful (at the first retry) recalibrations in 48hrs of uptime and no more warnings of TSC drift being too big. Honestly, please just leave the timers alone on X86 as in "never touch a running system". They've been working fine before afterall…
            Do you rely on the watchdog timer for anything else?
            If not, simply booting your Linux kernel with the "nowatchdog" parameter might also be an option.
            Personally done so for years on multiple PCs and never had any problems.

            Comment


            • #7
              Originally posted by Weasel
              Your guess is pure and utter bullshit.

              Reading is hard I guess.
              Your posts are needlessly antagonistic and contribute nothing.

              Hard to be even a half-decent human being I guess.

              Comment


              • #8
                Agreed, that kind of stuff doesn't belong in the forum. Truly the definition of uncalled for, Weasel. Being serious.

                edit: By default, by us using Linux and caring enough to frequent this forum, we should hold ourselves to a higher standard the average population. Not lower. I think it can be done. I'll start, no more aggressive posts from me. I wish there was a block button personally. But I'm done with the aggressive replies back to trolls. I think my point has been made. Cheers.
                Last edited by perpetually high; 10 October 2021, 10:42 AM.

                Comment


                • #9
                  Originally posted by kiffmet View Post
                  R9 3900X
                  X570 Aorus Pro r1.0 F34
                  .....
                  Honestly, please just leave the timers alone on X86 as in "never touch a running system". They've been working fine before afterall…
                  honestly, please stop posting nonsense. they never been working before on your hardware because your hardware did not exist few years ago. and after they were made to work on your hardware, they have to keep up with other hardware released after your

                  Comment


                  • #10
                    Originally posted by ms178 View Post
                    why it is that hard to fix the bugs in hardware?
                    because to fix bug in hardware you have to throw it away and buy new hardware with bugfix

                    Comment

                    Working...
                    X