We are seeing 

Jun 11 13:09:51 ip-172-19-14-69 kernel: BUG: Bad rss-counter state 
mm:ffff926ac6571000 idx:1 val:512
Jun 11 13:09:51 ip-172-19-14-69 kernel: BUG: non-zero nr_ptes on freeing 
mm: 1

and

./source/mm/pgtable-generic.c:33 
<http://elixir.free-electrons.com/linux/v4.9.24/source/mm/pgtable-generic.c#L33>:
 
bad pmd ffff9072a7590d10(00000169582001e2)

on 1353.8.0 on large instances (x1.32xlarge from aws) 

As per http://www.spinics.net/lists/linux-mm/msg109370.html and just 
guessing at where the stack trace is coming from, this might be huge pages 
related, so we are trying disabling huge pages to see if we can keep from 
encountering this bug.

Reply via email to