Hello,
On (06/16/16 19:12), Minchan Kim wrote:
[..]
> > I'll copy-paste one more backtrace I swa today [originally was posted to
> > another
> > mail thread].
>
> Please, look at http://lkml.kernel.org/r/20160616100932.GS17127@bbox
I don't have a solid/stable reproducer for this one, but after some
mixed workloads beating (mempressure + zsmalloc + compiler workload)
with reverted b3ceb05f4bae844f67ce I haven't seen any problems.
So I think you nailed it Minchan!
reverted the entire patch set (for simplicity):
Revert "mm/compaction: split freepages without holding the zone lock"
Revert "mm/page_owner: initialize page owner without holding the zone lock"
Revert "mm/page_owner: copy last_migrate_reason in copy_page_owner()"
Revert "mm/page_owner: introduce split_page_owner and replace manual
handling"
Revert "tools/vm/page_owner: increase temporary buffer size"
Revert "mm/page_owner: use stackdepot to store stacktrace"
Revert "mm/page_owner: avoid null pointer dereference"
Revert "mm/page_alloc: introduce post allocation processing on page
allocator"
adding "mm/compaction: split freepages without holding the zone lock"
back seem to introduce the page->map_count bug after some time.
-ss
