Bug 1639134 - Broken console after kexec in UEFI mode
Summary: Broken console after kexec in UEFI mode
Keywords:
Status: CLOSED INSUFFICIENT_DATA
Alias: None
Product: Red Hat Enterprise Linux 7
Classification: Red Hat
Component: kexec-tools
Version: 7.5
Hardware: Unspecified
OS: Unspecified
unspecified
unspecified
Target Milestone: rc
: ---
Assignee: Kairui Song
QA Contact: Emma Wu
URL:
Whiteboard:
Depends On:
Blocks: 1780482 1653509
TreeView+ depends on / blocked
 
Reported: 2018-10-15 06:23 UTC by Lukas Zapletal
Modified: 2023-03-24 14:17 UTC (History)
11 users (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
Environment:
Last Closed: 2020-01-06 02:15:20 UTC
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
Red Hat Knowledge Base (Solution) 5321971 0 None None None 2020-08-17 09:59:08 UTC

Description Lukas Zapletal 2018-10-15 06:23:19 UTC
Hello,

a customer is suffering from broken console on HP hardware after kexec. We use kexec in Satellite 6 in certain workflows to boot up Anaconda from a LiveCD (Discovery Node). This works nice in BIOS as long as we disable framebuffer, but in UEFI it looks like the framebuffer is the only option.

Reproducer steps:

1) Boot RHEL 7.6
2) Download Anaconda initramdisk and kernel
3) kexec --force --reset-vga --initrd=initram kernel

Comment 4 Dave Young 2018-10-15 06:53:01 UTC
In step 3, does it work without --reser-vga?

Comment 8 Kairui Song 2019-06-19 09:38:34 UTC
Hi, sorry it has been a long time, as I tried many times, still can't reproduce this issue on any machine I have.

What is the graphic driver being used in first boot and kexec kernel? If you append "nomodeset" to both first boot kernel and kexec boot kernel parameter will it still happen?

Comment 10 Lukas Matecki 2019-06-27 17:46:27 UTC
I am facing the same today while trying to deploy a gen10 HPE Server.
Any progress on this ?

Comment 11 Kairui Song 2019-07-01 06:41:36 UTC
Hi, could you help to try the workaround mentioned in comment 8? Add nomodeset kernel param to both first kernel and kexec kernel.

Comment 12 Lukas Zapletal 2019-07-15 12:15:48 UTC
Hello,

recent version of Satellite (6.6, TBA) includes a change - FDI is always booted with "nokaslr" kernel command line option which greatly improves chances of successful kexec on various hardware. We have a good experience, for older Satellite versions all you need to do is to change Global PXELinux template and put this option to the Discovery PXELinux menu item, then deploy the template to all TFTP/PXE Capsules. For PXE-less workflow this parameter must be added on every boot manually or using discovery-remaster script and then booting the modified ISO with "nokaslr" option.

I must say at this point that overall satisfaction with using kexec to boot into Anaconda in PXE-less Satellite workflows is not great and we are keeping this feature in Technology Preview state (best effort support) while we rethink different approach to the problem.

Comment 13 Suraj Patil 2019-07-15 12:30:13 UTC
Adding to Lukas's comment in PXE-less workflow. 

We should also add the "nokaslr" option in "Discovery Red Hat kexec" template, so while doing the kexec into new kernel "nokaslr" is present.

Comment 14 roarora 2019-08-15 21:45:41 UTC
Another use reported with a  HP Z6 G4 Workstation, Screen just stays where it was before kexec and doesn't go weird. nokaslr and nomodeset added both in booted and kexec kernels

Comment 15 Kairui Song 2019-09-12 05:47:15 UTC
Still, can't reproduce it, can anyone help provide a reproducer?

Comment 16 Lukas Matecki 2019-11-07 07:13:04 UTC
(In reply to Kairui Song from comment #11)
> Hi, could you help to try the workaround mentioned in comment 8? Add
> nomodeset kernel param to both first kernel and kexec kernel.

Sorry atm there are no system for installation available. As soon i will get some i will try it out.

Greetings

Lukas

Comment 17 Kairui Song 2020-01-02 17:43:11 UTC
Hi, it has been some time since last update, however there is no much progress because I can't reproduce (on multiple HPE server over ilo, it all worked well).

And, kexec'ing between different kernels is not well supported. And for the graphic as there are many different drivers and hardwares, it's hard to garenteen all of them will work, most can only be solved case by case.

For now, maybe it's worth trying to blacklist the drm drivers (if any in use) in both kernel and have "nomodeset" in cmdline, to see if it kexec can work.

Comment 18 Dave Young 2020-01-06 02:12:49 UTC
Hi,

According to the current status, we can not help.  It seems the same issue we had closed before about kexec across different kernel versions.  And using "nokaslr" can help on some cases.

I suggest to close the bug, and please open a new bug if it is not "kexec into a different kernel" problem.

Thanks
Dave


Note You need to log in before you can comment on or make changes to this bug.