Bug 30189 - Booting with >1 connected displays on various X5XX Intel chipset motherboards is not supported.
Summary: Booting with >1 connected displays on various X5XX Intel chipset motherboards...
Status: RESOLVED FIXED
Alias: None
Product: Mageia
Classification: Unclassified
Component: RPM Packages (show other bugs)
Version: 8
Hardware: x86_64 Linux
Priority: Normal major
Target Milestone: ---
Assignee: Kernel and Drivers maintainers
QA Contact:
URL: https://gitlab.freedesktop.org/drm/in...
Whiteboard:
Keywords: IN_ERRATA8
Depends on:
Blocks:
 
Reported: 2022-03-18 06:58 CET by Felix Miata
Modified: 2022-03-31 07:31 CEST (History)
1 user (show)

See Also:
Source RPM: kernel-5.15.28-1.mga8.src.rpm
CVE:
Status comment:


Attachments

Description Felix Miata 2022-03-18 06:58:36 CET
Original Summary:
Booting with >1 connected displays on various X5XX Intel chipset motherboards is not supported.

DRM problem (boot freeze) caused by defective BIOS behavior was reported and "fixed" upstream for 5.16.11+ and 5.17 kernels, so 9's 5.16.14 is OK, but 8's 5.15.28 is not:
https://gitlab.freedesktop.org/drm/intel/-/issues/4762

Whether any other distro has included this fix in their 5.15LTS kernels I haven't been able to discover.

On my Asus Prime B560-M, Mageia 8 works OK with >1 displays if additional video cables beyond the first are connected only after beginning the boot process. Connecting an HDMI cable to a powered up device is a process generally recommended against.
Comment 1 Lewis Smith 2022-03-18 09:11:30 CET
Thank you for the report.

Since this is a kernel matter, assigning the bugg accordingly.

Assignee: bugsquad => kernel

Comment 2 Thomas Backlund 2022-03-18 09:40:23 CET
the fix is in 5.15.29 in mga8 core updates testing and in backports in 5.16.14
Thomas Backlund 2022-03-18 13:13:05 CET

Component: Backports => RPM Packages

Comment 3 Felix Miata 2022-03-18 18:41:39 CET
Linux ab560 5.16.14-desktop-1.mga9 from cauldron/x86_64/media/core/release on 11 March installed on 16 March doesn't exhibit the problem.
Comment 4 Morgan Leijström 2022-03-19 08:53:18 CET
Potential for errata: is this problem in release ISOs?

CC: (none) => fri

Comment 5 Thomas Backlund 2022-03-19 09:57:03 CET
well, since the isos were out before the hw, so obviously there are non-existant hw support on them...

but technically installing with online medias added during install, it will work oob as soon as the next kernel lands in updates...
Comment 6 Thomas Backlund 2022-03-19 10:00:38 CET
and to be clear, the "fix" for this "bug" in 5.15.29 is actually a workaround for broken hw, so the vendor if they would care they could provide updated firmware
Comment 7 Felix Miata 2022-03-19 22:36:13 CET
(In reply to Morgan Leijström from comment #4)
> Potential for errata: is this problem in release ISOs?

On my Rocket Lake, the 8 GUI installer came up normally running FBDEV(0) in 1024x768 mode by loading vmlinuz and all-nonfree.rdz with Grub-efi. I didn't try to proceed from there, as I already had 8 installed via a backup image from a Haswell installation. Still, once an offline installation has been completed, the original 5.10.16 kernel will result in lockup unless: 1-displays beyond the first are not connected until after startup; or 2-KMS is somehow disabled. If booting continues one of these two ways until 5.15.29 or newer is installed, then all should be fine.

(In reply to Thomas Backlund from comment #6)
> and to be clear, the "fix" for this "bug" in 5.15.29 is actually a
> workaround for broken hw, so the vendor if they would care they could
> provide updated firmware

Both Asus and ASRock are known vendors AFAICT providing a problem BIOS, not problem hardware. As usual, the problem didn't manifest on Windows, so the vendors have little apparent motivation to provide a real fix. For my Asus B560M-A, there was a BIOS update to v1410 released last week with note "1.Function optimization." but I can't tell after installing it that it improved anything to do with booting kernels lacking the workaround fix.
Comment 8 Morgan Leijström 2022-03-20 15:08:14 CET
For problems like this, users will/should consult our Errata, so I will enter something there when new kernel is tested.
Remember, sometimes users need to install without updates possible (i.e problem with default downloader on some networks).

I am already running 5.16.14-desktop-1 OK (but am not affected by this bug)
But I dont see 5.15.29 in repos. 
I see already next kernels, 5.15.30 in updates_testing, 5.16.16-1 in backports.
Tell when they are ready for QA.

BTW, I also see since a month without accompaniyng bug(?) yet
microcode 0.20220207-2.mga8.nonfree
nvidia-current 470.103.01-1.mga8.nonfree


Suggested Errata to be entered under https://wiki.mageia.org/en/Mageia_8_Errata#Intel:

'''[[WORKAROUND]]''' in kernel update. {{Bug|30189}} '''More than one screen on late Intel X5XX chipsets.'''  Intel released new hardware but problems not yet fixed in their driver.  Workaround is implemented in Mageia kernels 5.15.30 and up in normal updates, and 5.16.14 and up in backports.  So if you are affected, boot with only one screen connected, and update your system.

Keywords: (none) => FOR_ERRATA8

Comment 9 katnatek 2022-03-21 03:58:31 CET
(In reply to Morgan Leijström from comment #8) 
> Suggested Errata to be entered under
> https://wiki.mageia.org/en/Mageia_8_Errata#Intel:
> 
> '''[[WORKAROUND]]''' in kernel update. {{Bug|30189}} '''More than one screen
> on late Intel X5XX chipsets.'''  Intel released new hardware but problems
> not yet fixed in their driver.  Workaround is implemented in Mageia kernels
> 5.15.30 and up in normal updates, and 5.16.14 and up in backports.  So if
> you are affected, boot with only one screen connected, and update your
> system.

Added to Erratas

Keywords: FOR_ERRATA8 => IN_ERRATA8

Comment 10 Morgan Leijström 2022-03-29 16:32:48 CEST
Now finally a fixed normal update (non backports) kernel is shipped, errata updated.
Bug 30199 - Update request: kernel-5.15.32-1.mga8

Resolution: (none) => FIXED
Status: NEW => RESOLVED

Comment 11 Felix Miata 2022-03-31 07:31:46 CEST
Failed with kernel-5.15.28-1.mga8, but fix confirmed with kernel-5.15.32-1.mga8. :)

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