You are not logged in.
It seems that amd-ucode which is derived from linux-firmware does not include microcode update for mobile Ryzen APUs. This is extremely nasty with mainline kernel, currently at 5.7.x. Multi-threaded workloads (such as make -j 8) simply hard crashed without any log with fan spinning at high-speed. Older kernels do not have such problem, presumably anything between 4.19.x and 5.4.x.
The CPU microcode update version included on my Ryzen 2500U laptop is:
microcode : 0x8101007
And, mainline kernel hard crashed from time to time, especially during make jobs.
Lenovo E485/E585 has a critical BIOS/UEFI update from May 2020 and it included the new microcode. Windows 10 build 2004 mcupdate_AuthenticAMD.dll includes the same CPU microcode.
microcode : 0x810100b
One can either ping the laptop manufacturers to release a new BIOS/UEFI update or follow Gentoo wiki to extract and rebuild the image for linux early microcode update.
Offline
@liewkj have you contacted the upstream maintainers of linux-firmware or filed a bug report on the arch bug tracker?
Offline
It seems that amd-ucode which is derived from linux-firmware does not include microcode update for mobile Ryzen APUs..
Proof?
Offline
liewkj wrote:It seems that amd-ucode which is derived from linux-firmware does not include microcode update for mobile Ryzen APUs..
Proof?
I am glad to be proven wrong.
Offline
yochananmarqos wrote:liewkj wrote:It seems that amd-ucode which is derived from linux-firmware does not include microcode update for mobile Ryzen APUs..
Proof?
I am glad to be proven wrong.
I was just curious how you knew, that's all.
Offline
I was just curious how you knew, that's all.
From this thread and it always suggested that one should have installed amd-ucode while amd-ucode did not help at all.
Offline
What CPU does your system have? Did the amd-ucode update the CPU's microcode version for your system?
You extracted a microcode update from mcupdate_AuthenticAMD.dll and that did update the microcode and resolve the issue? If not where did you obtain a microcode update?
Edit:
Are you saying the PKGBUILD for amd-ucode should be updated so that amd_sev_fam17h_model0xh.sbin is included in amd-ucode.img ?
Last edited by loqs (2020-07-14 00:03:35)
Offline
I extracted the CPU microcode from Lenovo E485/E585 BIOS/UEFI update for Ryzen 2500U APU. Windows 10 mcupdate_AuthenticAMD.dll just happened to have the exact same revision. And YES, that solved the hard crash on make jobs completely.
I don't think this is a problem for package amd-ucode. Upstream git.kernel.org has never included any CPU microcode for mobile Ryzen APUs and I don't know why. If Windows 10 can have them, upstream git.kernel.org should have them, too.
Offline
I would suggest contacting John Allen who authored https://git.kernel.org/pub/scm/linux/ke … f2e79e96e6 or Sherry Hurwitz who authored https://git.kernel.org/pub/scm/linux/ke … a498e91609 to see if AMD will add the ucode to linux-firmware upstream.
Offline
$ pacman -Qo /usr/lib/firmware/amd/amd_sev_fam17h_model0xh.sbin
/usr/lib/firmware/amd/amd_sev_fam17h_model0xh.sbin is owned by linux-firmware 20200519.8ba6fa6-1
$
https://git.kernel.org/pub/scm/linux/ke … el0xh.sbin
If that's the file that solved liewkj problem, then it seems to be already presen in linux-firmware, just in an incorrect location .
Smells like a kernel bug .
Disliking systemd intensely, but not satisfied with alternatives so focusing on taming systemd.
(A works at time B) && (time C > time B ) ≠ (A works at time C)
Offline
I guess that's not the right file.
+-------------------------------------------+
| MC Extractor v1.44.0 r153 |
+-------------------------------------------+
amd_sev_fam17h_model0xh.sbin (1/1)
File does not contain CPU microcodes
Offline