Linux 5.5 Begins Sanity Checking RdRand Output Due To Buggy Processor Behavior

[phoronix]

Phoronix: Linux 5.5 Begins Sanity Checking RdRand Output Due To Buggy Processor Behavior

The in-development Linux 5.5 kernel will begin sanity checking the RdRand instruction output for randomness on CPU boot/resume due to the recent spat of AMD CPUs that have yielded non-random RdRand output...

http://www.phoronix.com/scan.php?page=news_item&px=Linux-5.5-RdRand-Sanity-Check

[StandaSK]

So, could this mean that RdRand could get enabled for Family 15h/16h CPUs again?

[Guest]

So, could this mean that RdRand could get enabled for Family 15h/16h CPUs again?

No, AMD hasn't bothered with that. They don't care about fixing the issue, and OEMs don't want to issue updates for old hardware either. It's just been disabled, and will pretty much remain that way forever.

<sarcasm> Such amazing AMD support, yay! </sarcasm>

Annoying given those are old and weak CPUs, and can benefit heavily from such instructions.

[Adarion]

No, AMD hasn't bothered with that. They don't care about fixing the issue, and OEMs don't want to issue updates for old hardware either. It's just been disabled, and will pretty much remain that way forever.

<sarcasm> Such amazing AMD support, yay! </sarcasm>

Annoying given those are old and weak CPUs, and can benefit heavily from such instructions.

Not sure if that is entirely correct. As far as I remember AMD wanted to disable (possibly) faulty RdRand units via Kernel code, yes. Otherwise, as I understood, they'd have to send out BIOS/FW updates, which is well possible, but in reality we know that a lot of boards (esp. older) won't receive this updated code. And RdRand was mainly the thing with early boot, so you want that already in firmware, okay?
I've seen boards that never received a single update. Of course, that varies from vendor to vendors, but what is AMD to do here?
After all this rendered some recent distros unbootable.

On the blue side of things... ah, shall I really start?
Selling insecure CPUs knowingly?
Brian Krzanich selling much of his stocks in intel right before meltdown became public?
The countless new things that were discovered? Another one, recently, intel was aware of and they still claimed a recently launched generation was safe. And it wasn't.
We could go back in history now, or look e.g. at socket support. How often did that change? Uh-oh!
Oh wait! There's more! Very recently they announced that they'd end support to all their own motherboards. Well, one could discuss that if that is okay or not, but what was absolutely abominable is that they also announced (and did!) - with very very short notice - to remove all downloads for these boards. Now tell me: The once great market leader doesn't have freaking 512 MiB webspace for some silly BIOS images? Seriously? Just leave the latest state online so people who haven't updated or just got a used board off a popular platform can install the latest code. Now that is impossible and might even render some boards unusable.
*claps* Yeah! Go intel! *claps*

So don't be silly. It's not awesome with RdRand, that's true and I'd also rather have it fixed. But then, hoping for board vendors to update is like hoping for no taxes to pay that year.

[Danny3]

I hope this doesn't have a significant delay on boot process!

[PluMGMK]

Wouldn't it be more user-friendly to disable it and allow it to be explicitly enabled with a command-line parameter? I mean, a message containing the word "funky" in dmesg isn't exactly helpful to someone who, say, tries to boot a LiveCD on their AMD+dodgy-motherboard-powered PC…