BUG: non-zero pgtables_bytes on freeing mm: -16384

[Dan Streetman]

[impact]

This message is printed repeatedly in the logs:
BUG: non-zero pgtables_bytes on freeing mm: -16384

[test case]

boot the 4.15.0-58 kernel on s390x

[regression potential]

this affects task pud accounting; regressions may be around cleaning up task memory.

[Dan Streetman (ddstreet)]

patches to fix this already pulled in via stable update bug 1837664

[Nivedita Singhvi (niveditasinghvi)]

I'm not sure this bug should be DUP'd to the stable-release
bug. Might confuse the verification and handling triggers,
perhaps?

Will need to make sure the fix is tested once the fix is
uploaded.

[Nivedita Singhvi (niveditasinghvi)]

I'll unDUP it unless the kernel team says otherwise in IRC.

[Nivedita Singhvi (niveditasinghvi)]

I unduped it for test process clarity.

Trying to get the relevant people to test the fix.

[Nivedita Singhvi (niveditasinghvi)]

I'll update here once kernel is uploaded.

[Nivedita Singhvi (niveditasinghvi)]

This issue has been tested and successfully verified:

Verification successful !

"...test appliance built with 4.15.0-58 was unusable ... hundreds of "BUG: non-zero pgtables_bytes on freeing mm: -16384" in syslog, RestAPI interface timeouts, failed to produce FFDC data using sosreport.

Build with 4.15.0-60.67 displays none of these behaviors ... smoke test completed successfully."