Bug 605038 – sys-kernel/gentoo-sources-4.4.39: early microcode update causes hang on resume from suspend

• Status: RESOLVED UPSTREAM
• Product: Gentoo Linux

[Description Jeremy Murphy 2017-01-08 10:32:47 UTC]

I enabled early microcode update for my Intel Core i5 2540M and it was fine until I tried to resume from suspend (to RAM). Black screen, LED lights in unusual states, hard power off.

Seems to work fine in 4.8.16 but I have only tried once so far.  :)

I wouldn't usually worry about reporting a bug on an old kernel version, but it's the latest stable version at the moment.

[Comment 1 Matthew Thode ( prometheanfire ) 2017-01-10 02:41:00 UTC]

I wonder if this caused the problem I had (not resuming, system said it was resumed though (bios thought it was)).

Will update if I have problems on 4.4.40/41

[Comment 2 Jeremy Murphy 2017-01-10 06:19:10 UTC]

If anyone wants me to do further debugging, I'm happy to but you'll have to remind me how.  :)

(Still no problems on 4.8.16 btw.)

[Comment 3 Alice Ferrazzi 2017-01-24 06:18:31 UTC]

can you check with 4.4.44?

[Comment 4 Jeremy Murphy 2017-01-28 01:34:15 UTC]

(In reply to Alice Ferrazzi from comment #3)
> can you check with 4.4.44?

Same problem.

[Comment 5 Thomas Deutschmann (RETIRED) 2017-01-28 13:21:24 UTC]

Can you please attach dmesg from normal boot and "lspci -vvvv" output?

Have you only set CONFIG_MICROCODE=y or are you actual using microcode updates? Than, how do you do your updates (using genkernel, own initramfs, embedded into kernel...)?


If you found a working kernel it would be nice if you could bisect the kernel which would allow kernel upstream to backport the fix. See https://wiki.gentoo.org/wiki/Kernel_git-bisect for more details.

[Comment 6 Jeremy Murphy 2017-02-03 11:10:55 UTC]

Created attachment 462342 [details]
4.8.17 dmesg, normal boot with suspends+resumes

I wasn't sure which kernel version you wanted a dmesg from so I just started with the one that I'm using, which (unsurprisingly) is from the 4.8 series, that does not exhibit the bug.
Do you want a dmesg from 4.4 too?

[Comment 7 Jeremy Murphy 2017-02-03 11:14:27 UTC]

Created attachment 462344 [details]
lspci -vvv output

[Comment 8 Jeremy Murphy 2017-02-03 11:17:44 UTC]

Oh, I missed a 'v'; is it really -vvvv? The man page only documents up to -vvv.

[Comment 9 Jeremy Murphy 2017-02-03 11:46:56 UTC]

I have enabled the early microcode update and loaded the microcode via CONFIG_EXTRA_FIRMWARE. So no initram or whatever that some how-tos describe.

[Comment 10 Alice Ferrazzi 2017-02-04 13:53:05 UTC]

4.8 is end of life and the upstream will no release bug-fixes for this kernel version.
You should consider to upgrade to a more recent kernel and check if you still have problem

[Comment 11 Alice Ferrazzi 2017-02-04 13:54:32 UTC]

so it works in 4.8.17 ?

[Comment 12 Jeremy Murphy 2017-02-04 14:03:35 UTC]

(In reply to Alice Ferrazzi from comment #11)
> so it works in 4.8.17 ?

Correct.

[Comment 13 Jeremy Murphy 2017-02-05 10:29:40 UTC]

I haven't started the bisect yet, but I'm going to guess that the fix was introduced in 4.6.0: https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=9cf8d6360c1589a97a98313729ed9e5db187f80b

[Comment 14 Robin Johnson 2018-01-10 23:09:05 UTC]

Resolved upstream it seems, closing it. Reopen if occurring in newer kernels.