(In reply to Lars Viklund from comment #572)
> (You people should be happy you're not running FreeBSD, there I can
> reasonably reliably hang Ryzens within hours by sending ZFS snapshots :D )
Looks like they've addressed Ryzen errata issues around August 2018:
Also, they've made machdep.idle=hlt and machdep.idle_mwait=0 default for Ryzen processors.
I've additionally checked the registers mentioned in FresBSD's source. Latest BIOS 4406 (w/ AGESA 0070) for my ASUS A320M-K seems to apply fixes for all affected errata.
Eventhough I can't reliably reproduce the stability issues anymore, I think it's still good practice to have Boris' patch applied, since it's addresses a known erratum. I'll test the patch and check for regressions.
(In reply to Lars Viklund from comment #572)
> (You people should be happy you're not running FreeBSD, there I can
> reasonably reliably hang Ryzens within hours by sending ZFS snapshots :D )
Looks like they've addressed Ryzen errata issues around August 2018:
https:/ /github. com/freebsd/ freebsd/ blob/75ee4f08d3 acd4bf70f24b320 3fa440255873973 /sys/amd64/ amd64/initcpu. c#L133
Also, they've made machdep.idle=hlt and machdep. idle_mwait= 0 default for Ryzen processors.
I've additionally checked the registers mentioned in FresBSD's source. Latest BIOS 4406 (w/ AGESA 0070) for my ASUS A320M-K seems to apply fixes for all affected errata.
Eventhough I can't reliably reproduce the stability issues anymore, I think it's still good practice to have Boris' patch applied, since it's addresses a known erratum. I'll test the patch and check for regressions.