Messages in this thread | | | From | Vlastimil Babka <> | Subject | [PATCH 0/5] page_owner improvements for debugging | Date | Wed, 4 Nov 2015 16:00:56 +0100 |
| |
Hi,
I know it's merge window, but this might potentially help us with some outstanding bugs if page_owner was enabled e.g. for trinity runs, so here you go.
Patch 1 is a bug fix, patch 2 reduces page_owner overhead when compiled in but not enabled on boot. Patch 3 is something I suggested before [1] and it was deemed a good idea, that the page_owner info should follow the page during migration. Patch 4 allows us again to know that a migration happened and for which reason.
Patch 5 will hopefully help us when debugging, as it makes all the info be printed as part of e.g. VM_BUG_ON_PAGE(). Until now it was only accessible via /sys file.
Patches are based on today's -next. Hugh's migration patches caused conflicts for patches 3 and 4 when rebasing from 4.3.
[1] https://lkml.org/lkml/2015/7/23/47
Vlastimil Babka (5): mm, page_owner: print migratetype of a page, not pageblock mm, page_owner: convert page_owner_inited to static key mm, page_owner: copy page owner info during migration mm, page_owner: track last migrate reason mm, page_owner: dump page owner info from dump_page()
Documentation/vm/page_owner.txt | 9 +++--- include/linux/page_ext.h | 1 + include/linux/page_owner.h | 50 ++++++++++++++++++++++++--------- mm/debug.c | 2 ++ mm/migrate.c | 11 ++++++-- mm/page_owner.c | 61 +++++++++++++++++++++++++++++++++++++---- mm/vmstat.c | 2 +- 7 files changed, 110 insertions(+), 26 deletions(-)
-- 2.6.2
| |