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.

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

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

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