AMD Bulldozer/Jaguar CPUs Will No Longer Advertise RdRand Support Under Linux

[pal666]

So just rely on the kernel interface.

do you just rely on kernel interface instead of using sse instructions?
if rdrand is faulty, it will be faulty in kernel. but they are just relying on kernel interface btw - kernel reports rdrand capability, and as kernel stopped reporting it, they stopped using it. exactly as you asked

[torsionbar28]

Right in the 'AMD support is excellent for Linux' meme.

It's clear you don't understand how rdrand is used. Also that you haven't read the proposed patch, which clearly states this workaround is for faulty BIOS implementations, not a flaw in the CPU. You'd think that here in 2019, the intel fanboys would have plenty of their own intel CPU hardware flaws to keep them occupied, sheesh.

[xpue]

AMD FineWine™.

[Ardje]

lol, kernel can't produce random numbers without hardware. while patch is correct, you can't use kernel instead of hardware, you have to use some other hardware

I don't know why you are LOLing. How the kernel generates random values has been described clearly. And the kernel is constantly reseeding it's generator with random measurements from the system. Patches to let the kernel's random generator solely rely on RdRand has always been rejected. There have been enough problems with it (also on intel) that nobody really wants to use it.
So just rely on the kernel interface.

nranger - I hear you, I've faced the same kinds of problems with AMD desktop cards as well. I have an R9 390 which is also GCN 1.1 so I know the problems you faced.

Luckily, I splurged and bought a Thinkpad X1 Carbon when it was on sale, so now I have a nice laptop with Intel CPU and GPU and Intel WiFi that Just Works(TM).

[nranger]

Well...THAT sux !! So, I have a Lenovo Ideapad with a Mullins APU (Puma + cores which are orocess enhanced Jaguar cores) ...

I also have a Lenovo IdeaPad G35 (one of the Mullins APU machines) and for the $230 or so I paid for it, have been pretty happy with the performance. I'm amazed how well Firefox handles 50+ tabs on a 8Gb machine with a four core CPU designed for tablets and 2in1s.

That being said, my main issues with the machine are
(1) the plastics are shite -- all the screws in the bottom cover and in the ones in bottom case near the hinges sheared their inserts after a few 20" drops off the bed onto carpet,
(2) power/battery management is pretty "meh", new I was getting 5+ hours of life, now 11 months into my third battery it's under 2,
(3) GPU drivers -- I did switch to amdgpu from radeonsi around kernel 4.18 and it more or less worked, but various things have been broken then fixed so it's never perfect. Backlight control was busted for a while, HDMI audio has never really worked, the "DC" functions of the amdgpu driver seemed to work, and then Gnome stopped loading around kernel 5.1 and disabling DC fixed it.

Kind of annoying that my kernel cmdline changes almost every release. Now it includes "modprobe.blacklist=radeon amdgpu.ppfeaturemask=0xffffffff amdgpu.cik_support=1 amdgpu.si_support=1 amdgpu.dc=0 amdgpu.dpm=1"

In hindsight I probably should have just found a used Thinkpad X220 like all the Redhat devs used.

[pal666]

Having rdrand even user visible was a mistake in the first place. Only the kernel should be handling that. User space always seems to get it wrong.

but in our case user space did get it right and nothing would change by adding kernel middle man

RDRAND literally is spec'ed to return crap in certain circumstances and you're supposed to detect that and work around it, but does anything? No, they just blindly trust that it works. Yeah, don't ever check return codes, folks. Let's see how well that works.

who are they?

[pal666]

What about Windows? Does the same issue occur there

same issue of non-random random numbers - yes, same issue of broken boot in systemd - no. i heard some game was broken

[pal666]

It does matter however if applications use it directly. So the patch of AMD is correct, it actually says: do not rely on this hardware, use the linux kernel instead.

lol, kernel can't produce random numbers without hardware. while patch is correct, you can't use kernel instead of hardware, you have to use some other hardware

[PuckPoltergeist]

Well...THAT sux !! So, I have a Lenovo Ideapad with a Mullins APU (Puma + cores which are orocess enhanced Jaguar cores) and a Lenovo Ideapad with a Bristol Ridge APU (process enhanced Carrizo APU which, in turn, is based on Excavator cores which are in turn based on Bulldozer. I've never had issues with either laptop concerning wake from sleep or suspend operations but everyone's mileage varies.

So you can just enforce the rdrand capability, if you need it. What's the problem?