search

intel / Intel-Linux-Processor-Microcode-Data-Files

Other
668 stars 71 forks source link

On 06-5e-03 20200609 was good, but 20200616 removes SRBDS mitigation. #36

Open zarniwhoop73 opened 4 years ago

zarniwhoop73 commented 4 years ago

I have an i3 which was working ok with the early-loaded 20200616 microcode (0xdc), so I assumed that the issues with 06-4e-03 for that version were specific to -4e-. With linux-5.7.2 I have Microcode listed as the mitigation for srbds.

But the 20200616 release downgrades both skylake variants to the microcode from last year and therefore loses the mitigation.

esyr-rh commented 4 years ago

On Tue, Jun 16, 2020 at 06:20:35PM -0700, Ken Moffat wrote:

I have an i3 which was working ok with the early-loaded 20200616 microcode (0xdc), so I assumed that the issues with 06-4e-03 for that version were specific to -4e-. With linux-5.7.2 I have Microcode listed as the mitigation for srbds.

But the 20200616 release downgrades both skylake variants to the microcode from last year and therefore loses the mitigation.

See [1].

[1] https://github.com/intel/Intel-Linux-Processor-Microcode-Data-Files/issues/31#issuecomment-644885826

hmh commented 4 years ago

Yes, whatever issue can happen with 0x506e3 does not happen to everyone. We also had at least one good report (as in: it works) in Debian for 0x506e3.

eworm-de commented 4 years ago

I had a report for Arch Linux as well....

hmh commented 4 years ago

@eworm-de: was the report good (20200609 works on 0x506e3) or bad (20200609 causes issues on 0x506e3) ?

hmh commented 4 years ago

Anyway, it clearly can cause trouble on some systems: someone just posted a confirmation of a hang on their 0x506e3 (Xeon E3) on another issue thread [1]

[1] https://github.com/intel/Intel-Linux-Processor-Microcode-Data-Files/issues/31#issuecomment-645181792

eworm-de commented 4 years ago

20200609 caused issues: https://bugs.archlinux.org/task/67011

esyr-rh commented 4 years ago

06-5e-03 microcode is updated to revision 0xe2 (that contains SRBDS mitigations) with microcode-20201110 release.

hmh commented 4 years ago

Since 0x406e3 boothang is reportedly fixed -- but not everywhere, there are now also reports of it causing hangs on warm boots again -- the one with 0x506e3 might have been fixed and it contains the SRBDS mitigations.

It would be nice to get direct confirmation that the new microcode works on 0x506e3, from someone that had issues with the pre-revert version, since it did work for many people.

zarniwhoop73 commented 4 years ago

0xe2 works on my 06-5e-03 (i3 skylake), but then the old pre-revert version did too.

hmh commented 4 years ago

Yeah, we'd need testing from someone who had issues with the pre-revert microcode update :-( And now some complains of failure-to-reboot for 0x406e3 with the newest update surfaced, so it might not have been fixed (or some bad cases still remain)...

hmh commented 3 years ago

This is related to #31

Systems with 06-4e-03 and 06-5e-03 processors where the firmware has a microcode revision older than 0x80 cannot be safely updated to revisions 0xd8 and above, and may hang.

As usual "update your firmware" if such an update is available. If the microcode revision in the firmware is newer than 0x80, it should be possible to use the microcode updates in the latest Intel release (20201118 as I write this).

esyr-rh commented 3 years ago

New revision 0xea of 06-5e-03 microcode file has been published as part of microcode-20210608 release, it may be worth to try it out.