MCE on shutdown when nouveau driver loaded

[Impact]
When rebooting with the focal kernel, my system always MCEs. Installing an nvidia driver - or simply blacklisting the nouveau driver - avoids the issue.

Sometimes it hard hangs the system, requiring a manual power cycle:

[ OK ] Reached target Reboot.
[ 402.489755] Disabling lock debugging due to kernel taint
[ 402.495319] mce: [Hardware Error]: CPU 24: Machine Check Exception: 5 Bank 6: bb80000000000e0b
[ 402.503924] mce: [Hardware Error]: RIP !INEXACT! 10:<ffffffff9ead91c7> {intel_idle+0x87/0x130}
[ 402.512530] mce: [Hardware Error]: TSC 29fb4740af0 MISC d7000000
[ 402.518622] mce: [Hardware Error]: PROCESSOR 0:50654 TIME 1601415822 SOCKET 1 APIC 40 microcode 2006906
[ 402.527998] mce: [Hardware Error]: Run the above through 'mcelog --ascii'

Other times it emits the MCE tombstone, but goes ahead and reboots itself:

[ OK ] Reached target Reboot.
[ 870.372933] Disabling lock debugging due to kernel taint
[ 870.378505] mce: [Hardware Error]: CPU 24: Machine Check Exception: 5 Bank 6: bb80000000000e0b
[ 870.387110] mce: [Hardware Error]: RIP !INEXACT! 10:<ffffffff8e2d4847> {intel_idle+0x87/0x130}
[ 870.395716] mce: [Hardware Error]: TSC 44e0f5e602c MISC d7000000
[ 870.401801] mce: [Hardware Error]: PROCESSOR 0:50654 TIME 1589320331 SOCKET 1 APIC 40 microcode 2000064
[ 870.411185] mce: [Hardware Error]: Run the above through 'mcelog --ascii'
[ 870.420531] mce: [Hardware Error]: Machine check: Processor context corrupt
[ 870.427488] Kernel panic - not syncing: Fatal machine check
[ 870.433108] Kernel Offset: 0xc800000 from 0xffffffff81000000 (relocation range: 0xffffffff80000000-0xffffffffbfffffff)
[ 871.054820] Rebooting in 30 seconds..
[ 900.901238] ACPI MEMORY or I/O RESET_REG.

Copyright(c) 2015 American Megatrends, Inc.
0x19 : Pre-memory SB Initialization.
Copyright(c) 2016 American Megatrends, Inc.