Boot freeze due to intel-microcode

[Stephane]

The version "3.20180108.0+really20170707ubuntu17.10.1" of the package "intel-microcode" randomly freezes Ubuntu at boot (at a very early stage, nothing is written on the screen).
The number of hard reset of the computer required looks random but it works at third boot most of the time.

I was previously affected by the 20180808 regression :
https://bugs.launchpad.net/ubuntu/+source/intel-microcode/+bug/1742933
Symptoms are the same.

user@PC:~$ dmesg | grep microcode
[ 0.000000] microcode: microcode updated early to revision 0xba, date = 2017-04-09
[ 0.986254] microcode: sig=0x506e3, pf=0x20, revision=0xba
[ 0.986730] microcode: Microcode Update Driver: v2.2.
If I uninstall "intel-microcode" everything is fine.

user@PC:~$ lsb_release -rd
Description: Ubuntu 17.10
Release: 17.10
This install is relatively fresh as I did it when I was affected by the 20180808 regression.
Xubuntu 17.10 installed on a SSD with LUKS encryption

user@PC:~$ uname -a
Linux PC 4.13.0-32-generic #35-Ubuntu SMP Thu Jan 25 09:13:46 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux

Package Manager = synaptic 0.84.2

Computer :
Model = Asus N752VX-GC164T
Bios = v301 (13 oct 2016)
Processor = Intel Core i7-6700HQ (6th gen Skylake)
Nvidia graphic card = Grub is configured with "nouveau.modeset=0"

Sorry I can't give more accurate description of the bug.
I don't see any error message on the screen at boot.
Is there somewhere a boot log I can provide ?

[dino99 (9d9)]

Please post the output files of:
- journalctl -b | grep error > error.txt
- journalctl -b | grep ERROR > ERROR.txt
- journalctl -b [ grep intel > intel.txt

[Stephane (stephane21)]

See attached, the output of :

journalctl -b | grep error > error.txt

[Stephane (stephane21)]

See attached, the output of :

journalctl -b | grep ERROR > ERROR.txt

[Stephane (stephane21)]

See attached, the output of :

journalctl -b | grep intel > intel.txt

[dino99 (9d9)]

There is nothing into the files above that can explain your issue.
So i propose to :
- remove 'quiet splash' from /etc/deault/grub and then run 'sudo update-grub'
- reboot and enter the bios/uefi to check the activated/deactivated settings and choose the good ones. Also upgrade the bios/uefi if possible. Save the change(s) and reboot.
- then check 'journalctl -b' again to find something new.

[Stephane (stephane21)]

Thanks for the help,

I have checked the timestamps into the 3 logs : It always corresponds to the successful boot.

"journalctl -b" gives text I can't interpret, anyway it always corresponds to the last boot (successful).

The Bios version is the most up to date one : v301 (13 oct 2016)

The "quiet splash" option doesn't give any output to the screen when removed. I already checked that in the past and I re-tested today. I forgot to mention that the freeze occurs just after the Grub automatic selection. I guess the crash occurs in between Grub and kernel init.

The bios UI is really limited, and I don't remember going there since a long time (except for boot priority). Moreover that bug doesn't occur when the package "intel-microcode" is removed.

Additional hints :
- There is a big chance that I installed Xubuntu from a USB key created with UnetBootin (I saw it gives problem with Debian)
- I am using the rEFInd Boot Manager on top of Grub

[Kai-Heng Feng (kaihengfeng)]

>
> Additional hints :
> - There is a big chance that I installed Xubuntu from a USB key created with UnetBootin (I saw it gives problem with Debian)
> - I am using the rEFInd Boot Manager on top of Grub

So I guess it’s a good idea to try a “vanilla” Ubuntu live USB.

>
> --
> You received this bug notification because you are a member of Intel
> Team, which is subscribed to intel-microcode in Ubuntu.
> Matching subscriptions: intel-microcode
> https://bugs.launchpad.net/bugs/1746200
>
> Title:
> Regression in 2017-07-07 updates
>
> Status in intel-microcode package in Ubuntu:
> Incomplete
>
> Bug description:
> The version "3.20180108.0+really20170707ubuntu17.10.1" of the package "intel-microcode" randomly freezes Ubuntu at boot (at a very early stage, nothing is written on the screen).
> The number of hard reset of the computer required looks random but it works at third boot most of the time.
>
> I was previously affected by the 20180808 regression :
> https://bugs.launchpad.net/ubuntu/+source/intel-microcode/+bug/1742933
> Symptoms are the same.
>
> user@PC:~$ dmesg | grep microcode
> [ 0.000000] microcode: microcode updated early to revision 0xba, date = 2017-04-09
> [ 0.986254] microcode: sig=0x506e3, pf=0x20, revision=0xba
> [ 0.986730] microcode: Microcode Update Driver: v2.2.
> If I uninstall "intel-microcode" everything is fine.
>
> user@PC:~$ lsb_release -rd
> Description: Ubuntu 17.10
> Release: 17.10
> This install is relatively fresh as I did it when I was affected by the 20180808 regression.
> Xubuntu 17.10 installed on a SSD with LUKS encryption
>
> user@PC:~$ uname -a
> Linux PC 4.13.0-32-generic #35-Ubuntu SMP Thu Jan 25 09:13:46 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux
>
> Package Manager = synaptic 0.84.2
>
> Computer :
> Model = Asus N752VX-GC164T
> Bios = v301 (13 oct 2016)
> Processor = Intel Core i7-6700HQ (6th gen Skylake)
> Nvidia graphic card = Grub is configured with "nouveau.modeset=0"
>
> Sorry I can't give more accurate description of the bug.
> I don't see any error message on the screen at boot.
> Is there somewhere a boot log I can provide ?
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/ubuntu/+source/intel-microcode/+bug/1746200/+subscriptions

[Stephane (stephane21)]

If I boot a “vanilla” Ubuntu live USB, Does the "last CPU firmware" will be applied at reboot or will it disappear ?

[dino99 (9d9)]

rEFInd is said 'beta quality' by its own dev http://www.rodsbooks.com/refind/

and itself rely on rEFIt which has its bugs list : https://sourceforge.net/p/refit/bugs/ but looks abandonned

http://www.rodsbooks.com/refind/todo.html

To conclude:

 intel-microcode can't be blamed at first glance. But refind have trouble to work with the microcode script. Please check: 'journalctl | grep refind > refind.txt' (after intel-microcode removal then reinstall).

Running refind over grub2 results having 2 EFI boot loaders at once !!!
http://www.rodsbooks.com/efi-bootloaders/index.html
... and that should be the main trouble source.

So test booting only with refind (purging grub2) and report back; but that one is clearly 'invalid' regarding the links above.

[Stephane (stephane21)]

Some tests I just did :

- "journalctl | grep refind > refind.txt" > Empty file

- Currently, I am booting like that : UEFI > Refind > Grub > Linux kernel

- Chainloading the UEFI directly to Grub (no use of Refind) : gives the same random bug

- Chainloading the Refind directly to Linux kernel (no use of Grub) : gives the same random bug and when it passes, it bugs later at a busybox prompt because I guess that Refind is not able to boot alone an OS installed on a LUKS partition.

- Install of Debian 9.3 Stable : Boot always works (Linux 4.9 kernel)

- UEFI > Refind > Grub (advanced menu) > Ubuntu with Linux 4.13.0-32 > Gives the same random bug but then I have a display of 2 lines before the freeze.
Those 2 lines are :
Loading Linux 4.13.0-32-generic ...
Loading initial ramdisk ...

I am wondering if I could circumvent the bug with the use of an old linux kernel on Xubuntu.
Is is possible to install the 4.9 linux version on Xubuntu 17.10 without breaking the OS ?

[dino99 (9d9)]

Grepping journalctl is case sensitive; so 'refind' could be logged with some uppercases, does not know, so watch 'journalctl -b' to find the exact spelling.

Mixing different DEs will only add confusion/conflict/headhache ...
Fightting about trouble works the other way: test step by step, the simpler the easiest.

Your report is about 'regression' pointing to a freeze on S3 recovery. Is it really the case ?
Then you said 'randomly', which is not the easiest case. About 'intel-microcode' that is used very early at boot time, i doubt it can be blamed (maybe glance at the mobo forum to know if other users are affected).

What can you do more: clean the system with gtkorphan & bleachbit (as root carefully), test ram/swap stress.

[Stephane (stephane21)]

- There is nothing about Refind into "journalctl -b".

- I am not mixing DEs, I use only Xfce.

- What does "S3 recovery" mean ? Samsung Galaxy S3 recovery ?

- The bug occurs since the first boot of that Xubuntu install (at that time it was the 20180108 Intel CPU microcode), so I won't try to clean the system. Moreover the freeze occurs before that I have the opportunity to enter the LUKS passphrase, at that time only the /boot partition is accessible. So, in my opinion, the only candidates for the bug are : Hardware - UEFI Bios - EFI partition - Refind - Grub - Something inside the /boot folder.

Thank you for your help and for all that time you spent to help me.
But I have to admit that I currently need to take some rest away from that bug.
I will uninstall "intel-microcode" for the moment, as anyway the current version doesn't provide any defence against Spectre/Meltdown. I will see what the next intel update gives. Also ASUS is supposed to provide "soon" a Bios update for my computer.

Have a good day,

[Stephane (stephane21)]

Some news :

- Install of Debian 9.3 Stable 31 Jan 2018 (Linux kernel 4.9) : Boot always works
- Install of Debian Testing 31 Jan 2018 (Linux kernel 4.14) : Random freeze at boot

So considering my investigations until here, I consider that "intel-microcode v3.20180108.0+really20170707ubuntu17.10.1" doesn't work on my hardware/software config with a too recent Linux kernel, but I haven't any clue or detail on what is happening.

[Stephane (stephane21)]

Some news :

Last week I saw a "grub" package update so I reinstalled the "intel-microcode" package, the freezing bug is still there on my configuration but there is one difference.

With the "intel-microcode" package, at boot there is always a message that appears during approximately 1 second :

    error: invalid video mode specification `text'.
    Booting in blind mode

It could be related to the fact that I am booting with the "nouveau.modeset=0" option.
But without the "intel-microcode" package, I can't see (or don't have the time to) see that message.

Sorry if this message is not related to the bug. I post it in case of.

[dino99 (9d9)]

'text' could be related to 'plymouth'. Maybe a theme is faulty. Please do check:

journalctl -b | grep ply

[Stephane (stephane21)]

With or without intel-microcode installed : "journalctl -b | grep ply" gives no result.

[Stephane (stephane21)]

Today, I installed manually :
sudo dpkg -i intel-microcode_3.20180312.0~ubuntu18.04.1_amd64.deb

No change, the freeze still occurs.

[camypaj (mantonijevic)]

Hi all, is it just me, or this is similar, or perhaps even a duplicate of #1759920?
I have a Skylake machine which behaves the same with this latest intel-microcode.

[Stephane (stephane21)]

In my case, the freeze occurs very early in the boot process.
In the case of #1759920, the freeze occurs at login.

Maybe the 2 bugs are related, but the behaviour is not the same.

[Carl van Schaik (navlrac)]

Seems to be affecting me on HP DL20 G9. Some boots stay alive for up to 10 minutes, other boots I don't make it to the login prompt.

Did a system resume, set dis_ucode_ldr kernel cmdline. I've uninstalled intel-microcode for now.

I agree this seems to me a duplicate of #1759920.

My affected CPU:
Intel(R) Xeon(R) CPU E3-1240 v5 @ 3.50GHz (family: 0x6, model: 0x5e, stepping: 0x3) (Skylake)

[Stephane (stephane21)]

Today, I retried "intel-microcode" on Xubuntu 18.04 fresh install and I was able to reboot 6 times in a row without any freeze.

~$ dmesg | grep microcode
[ 0.000000] microcode: microcode updated early to revision 0xc2, date = 2017-11-16
[ 0.987572] microcode: sig=0x506e3, pf=0x20, revision=0xc2
[ 0.987975] microcode: Microcode Update Driver: v2.2.

Linux kernel version 4.15.0-20

So, it looks that the bug has been repaired by that linux v4.15 branch.