5.X kernels Flag Tainted Kernels during boot tests.

[Sean Feole]

5.4 & 5.8 Focal & Groovy Generic/Lowlatency Kernel Tainted Flag

5.4 Kernels all flag ERRORs of tainted kernels, all of which are flag != 0: 2048

Linux grumman 5.4.0-66-generic #74-Ubuntu SMP Wed Jan 27 22:54:38 UTC 2021 x86_64 x86_64 x86_64 GNU/Linux

This does not affect fips or kvm

01/28 10:59:32 INFO |ubuntu_boo:0042| Checking kernel tainted flags in /proc/sys/kernel/tainted
01/28 10:59:32 INFO |ubuntu_boo:0046| ERROR: kernel tainted flag != 0: 2048
01/28 10:59:32 INFO |ubuntu_boo:0046|

See Attached dmesg.gz and syslog

-------------------------------

[Sean Feole (sfeole)]

Attached dmesg is from Focal - AMD64

[Tim Gardner (timg-tpi)]

Nothing in the attached dmesg indicates a taint.

[Thadeu Lima de Souza Cascardo (cascardo)]

2048 is TAINT_FIRMWARE_WORKAROUND. That seems to be set when linux thinks there is something wrong with ACPI tables:

arch/x86/pci/fixup.c: add_taint(TAINT_FIRMWARE_WORKAROUND, LOCKDEP_STILL_OK);
drivers/acpi/arm64/iort.c: if (WARN_TAINT(iort_node >= iort_end, TAINT_FIRMWARE_WORKAROUND,
drivers/dma/ioat/dca.c: add_taint(TAINT_FIRMWARE_WORKAROUND, LOCKDEP_STILL_OK);
drivers/firmware/efi/runtime-wrappers.c: add_taint(TAINT_FIRMWARE_WORKAROUND, LOCKDEP_NOW_UNRELIABLE);
drivers/iommu/intel/dmar.c: add_taint(TAINT_FIRMWARE_WORKAROUND, LOCKDEP_STILL_OK);

That would make sense to be happening on some systems where the tests were not running before.

Cascardo.

[Kelsey Steele (kelsey-steele)]

Seeing this on Bionic 4.15.0-136.140 lowlatency/generic as well

[Po-Hsu Lin (cypressyew)]

[ 5.511280] systemd[1]: Set hostname to <grumman>.

This is a dup with bug 1912316

[Po-Hsu Lin (cypressyew)]

BTW this issue is limited to the node in Intel cloud, I didn't see this on other bare-metals in our pool.