HNNewShowAskJobs
Built with Tanstack Start
AMD Errata: RDSEED failure on AMD Zen 5 Processors(amd.com)
22 points by jcalvinowens a day ago | 6 comments
  • meltyness3 hours ago

    It's funny to me that the patch threads attempting to fix a concern with the CPU from the OS both immediately derail into "oh no, you broke something!!1"

    the CPU is assumed to work, despite Linux being designed with portability, theres not really a programming interface to cleanly disable arbitrary features, I guess we just screw with it until it seems to work OK and no one is yelling.

    Granted, Linux isnt a product, and a patch in a thread is unlikely to impact anyone

  • jcalvinowensa day ago

    I wrote a silly little testcase: https://gist.github.com/jcalvinowens/06530e6fc86f3b2c97a7027...

    ...but both affected znver5 machines I have "pass", so clearly there's more to it than AMD is telling us.

    You should be able to get away with "-march=znver5 -mno-rdrnd -mno-rdseed" in CFLAGS if you care.

    • jcalvinowens20 hours ago |parent

      Oh, actually the original report was very specific: https://lore.kernel.org/all/20251018024010.4112396-1-gourry@...

  • jcalvinowensa day ago

    See also: https://lore.kernel.org/lkml/20251103142800.GFaQi78AgT-qjY30...

  • rasza day ago

    again? https://arstechnica.com/gadgets/2019/10/how-a-months-old-amd... ZEN seems to have suspiciously a lot of those rdrand "bugs"

    • hulitu18 hours ago |parent

      CIA and NSA really need access to your data. /s