Re: 4.15.17 regression: bisected: timeout during microcode update

From: Raj, Ashok
Date: Wed Apr 18 2018 - 09:53:38 EST

On Wed, Apr 18, 2018 at 02:22:12PM +0200, Borislav Petkov wrote:
> On Wed, Apr 18, 2018 at 02:08:40PM +0200, Vitezslav Samel wrote:
> > I switched to firmware-in-kernel early loading and that works OK.

firmware-in-kernel means you compile your microcode image in linux?

Can you tell me which distro you are using? The ones we used
doesn't do late-loading (i.e echo 1 > /sys/devices/system/cpu/microcode/reload)

I suspect that might be the problem.

- Can you remove your builtin microcode,
- rename the /lib/firmware/intel-ucode so we don't find it during late
- let the system boot completely
- then rename the intel-ucode back for this test.
- write 1 to reload and see if that update succeeds or fails?

> Ok, and keep using that from now on.
> People should all move away from that late loading dance. I'm saying
> that in case someone else reads this on lkml.
> > But still, the reported issue is regression in 4.15.17 and 4.16+.
> Oh, I know it is a regression.
> @Ashok: anything particular about his microcode revision not being able
> to stomach late loading?

nothing about the microcode itself comes to mind. I'm wondering if this
similar to the Arch linux that used late-load during booting might be an issue.