]> git.karo-electronics.de Git - karo-tx-linux.git/commit
mm, memcg: sync allocation and memcg charge gfp flags for THP
authorMichal Hocko <mhocko@suse.cz>
Tue, 7 Apr 2015 23:44:27 +0000 (09:44 +1000)
committerStephen Rothwell <sfr@canb.auug.org.au>
Tue, 7 Apr 2015 23:44:27 +0000 (09:44 +1000)
commita8aa6cc302d53787ec26154699cb4332a46f02ba
tree391cd951da6d64407221c8c71a3ba138827f7c4b
parent8118fba02c8cd6ee17bf19d2f56dab0d6bef5d37
mm, memcg: sync allocation and memcg charge gfp flags for THP

memcg currently uses hardcoded GFP_TRANSHUGE gfp flags for all THP
charges.  THP allocations, however, might be using different flags
depending on /sys/kernel/mm/transparent_hugepage/{,khugepaged/}defrag and
the current allocation context.

The primary difference is that defrag configured to "madvise" value will
clear __GFP_WAIT flag from the core gfp mask to make the allocation
lighter for all mappings which are not backed by VM_HUGEPAGE vmas.  If
memcg charge path ignores this fact we will get light allocation but the a
potential memcg reclaim would kill the whole point of the configuration.

Fix the mismatch by providing the same gfp mask used for the allocation to
the charge functions.  This is quite easy for all paths except for
hugepaged kernel thread with !CONFIG_NUMA which is doing a pre-allocation
long before the allocated page is used in collapse_huge_page via
khugepaged_alloc_page.  To prevent from cluttering the whole code path
from khugepaged_do_scan we simply return the current flags as per
khugepaged_defrag() value which might have changed since the
preallocation.  If somebody changed the value of the knob we would charge
differently but this shouldn't happen often and it is definitely not
critical because it would only lead to a reduced success rate of one-off
THP promotion.

Signed-off-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
mm/huge_memory.c