RDSEED32 Zen5 CPU issue & November ISO

Issues & Assistance
1

Hi,

My installation using the 29th November ISO stalls after the Live ISO’s GRUB loader selection screen (i.e. it never loads up CachyOS itself). It simply stops with the error “RDSEED32 is broken. Disabling the corresponding CPUID bit.” Leaving it some minutes, it simply doesn’t move on.

This relates to an AMD issue with RDSEED in Zen5 CPUs (the PC has a Ryzen 5 9600) which is covered in the article here and includes quotes from @ptr1337.

My BIOS/UEFI is from September whilst I read somewhere the updated AMD code is due ‘sometime in November’ but should this hang the CachyOS installation i.e. is the expected behaviour?

The AMD page includes the following suggestions:

- Use the 64-bit form of RDSEED 
- Mask the CPUID Fn0000_0007_EBX[18] RDSEED from software discovery. For example, by adding clearcpuid=rdseed to the boot command line; or with the -rdseed option on the qemu command line for a VM 
- Software can treat RDSEED returning 0 equivalent to when CF=0. Retry RDSEED later until a non-zero value is returned with CF=1

As it’s the ISO rather than an existing installation I can’t (as far as I know) change the boot command line to try adding the clearcpuid=rdseed line.

Related discussion on the Arch forums here.

I’ve been waiting for the November ISO has the previous versions don’t support the HW used meaning I can’t install (due to a different error as per this thread). Is my only option to wait for updated BIOS/UEFI from the mobo manufacturer (it’s a Sapphire Nitro+ B850M) or is there anything I can do to get the installation going (as it’s a Christmas present for my son)?

TIA

2

Ill look into. It should not block the boot process tho.

FYI, the old ISO is still valid and working.

3

So, I have researched a bit - it appears the patches that enables RDSEED32 again with the new microcode are not backported by upstream LTS Kernel.

But:

I have booted now on 2 Zen5 machnes:

  1. Strix Point HX370 Framework 16 - no problems
  2. Zen5 9950X with NVIDIA GPU - no problems
4

I have the same problem on my Framework 13 (AMD Ryzen AI 7 350 w/ Radeon 860M), although post-install. If I boot the LTS kernel, I get the same warning and SDDM fails to start due to the missing rdseed.

5

So… fortified with a large glass of thinking juice, I was doing some head scratching and poking around trying to think what the root cause is and I noticed that e.g. the speed of the RAM wasn’t being reported as high as it should have been in the BIOS. Long story, short; it was a RAM issue (presumably the significant increase in RAM requirement going from boot loader/manager to OS is what caused the hang/failures). The RDSEED32 message still flashes up but CachyOS continues to load and I now write from a fully installed instance of CachyOS!

Many thanks to everyone for their ideas and info! My son will be very pleased with his new PC!

6

no fix for me however.

my memory is 4xDDR5-3000 sticks (6000 in bios)

7

Just posting my fix since it might help others:

Once at the GRUB menu:

  1. Use arrow keys to highlight your CachyOS entry

  2. Press e to enter edit mode

  3. Find the line beginning with linux (scroll down with arrow keys)

  4. Navigate to the end of that line

  5. Add a space, then type: clearcpuid=rdseed random.trust_cpu=0 systemd.unit=multi-user.target
    1.

    random.trust_cpu=0 Prevents kernel from trusting CPU RNG for entropy seeding
    1. systemd.unit=multi-user.target - Boot to text mode, bypassing graphical environment

  6. With these 3 things, it will allow you to login and insert text into the terminal directly

From here, the terminal you can, this will make it such that all next logins will have the first two options by default

  1. sudo nano /etc/default/grub
  2. GRUB_CMDLINE_LINUX_DEFAULT=“… clearcpuid=rdseed random.trust_cpu=0” (added this to the end)
  3. sudo grub-mkconfig -o /boot/grub/grub.cfg

This allows you to turn on your computer

If you want to solve it for the future, update the BIOS:

  1. Update packages - sudo pacman -Syu linux-firmware amd-ucode
  2. Regenerate initramfs to include new microcode - sudo mkinitcpio -P
  3. Reboot and check if fix was applied - sudo reboot
  4. After reboot, verify microcode version
    1. cat /proc/cpuinfo | grep microcode
      dmesg | grep microcode

In my case this update did not work, so I did a BIOS update and now everything works as normal and revert the first Steps

Edited: I use Hyprland (Wayland compositor, not a full DE),
Since Hyprland doesn’t depend on Qt like KDE does, I didn’t hit the same RDSEED issue that you’re experiencing. @D_B . With just clearcpuid=rdseed systemd.unit=multi-user.target maybe you can enter your system.

8

Hello everyone.
If add to grub “lcearcpuid=rdseed random.trust_cpu=0” , then startplasma-wayland does not work: incompatible processor. The following functions are required for this Qt build: rdseed

Which DE do you use?

9

what exactly you did after Step 5?
i am using Ubuntu on Lenovo IdeaPad Slinm 5 AMD machine.
after adding the step 5, i am booting it right away and but the problem still exist and i can’t find the command I added into linux in Grub.
Can you please help, my booting is not working.