> No it hasn't - I'm still waiting for someone to test it and confirm that it
> fixes the issue on their boxes. And by issue I don't mean erratum 1033 -
> that is a red herring anyway - but the lockups people are reporting on some
> broken BIOSes.
Yes, your patch "seems" to work in my case.
Created attachment 281999
Add rifw kernel parameter to test a couple of patch to workaround ryzen freezes
Add rifw kernel parameter.
rifw=none - don't do anything
rifw=alfie - ignore BIOS
rifw=boris - use Borislav Petkov patch
I works with kernels 4.19, 4.20 and 5.0.2
--
You received this bug
(In reply to Another User from comment #564)
> Can your system reach turbo frequencies with this patch?
> Maybe I understand something wrongly...
Yes, I have a ryzen 1600 and I can see 1/2 cores running at ~3.6 GHz
under some peculiar stress case.
(In reply to Borislav Petkov from comment #565)
This is pretty strange, could someone explain it to me? Wasn't mwait
bugged in ryzen?
Tired of the random freezes, I tried this:
--- ./arch/x86/kernel/acpi/cstate.c.orig 2018-10-22 08:37:37.0 +0200
+++ ./arch/x86/kernel/acpi/cstate.c2019-03-20 19:26:45.261101857 +0100
@@
The PSU problem sounds like a weak excuse from AMD or may they are
talking about very very bad 10 dollars units.
The BIOS options "typical/low current" doesn't seems to change the c-states set
up by the kernel routines here. With any selection, I always end up with
c0 POLL
c1 ACPI HLT
c2 ACPI
kernel 4.19.27 with rcu_nocbs=0-11 idle=halt nopti, yesterday at around
11:30 p.m. I had an "usual" total lookup of the system during light
usage (I just had an ssh session opened). With idle=halt, I get no "ACPM
WAIT C-state bug" line in dmesg.
The problem with this freeze problem is that is
Here (Asus CrossHair VI, 1600, gentoo with any kernel around), no
suggested workaround works. The system freezes when in very light use
and I was never able to obtain a single line of debug via serial (I put
a small linux mini itx box near this computer just to receive the serial
log).
It seems
7 matches
Mail list logo
