[Resending with the mailing lists CCed - sorry for spamming]
Hi Andrew,
Oscar has reported two issue when playing with the memory hotplug
[1][2]. The first one seems more serious and patch 1 should address it.
In short we are overly optimistic about zone movable not containing any
non-movable pages and after 72b39cfc4d75 ("mm, memory_hotplug: do not
fail offlining too early") this can lead to a seemingly stuck (still
interruptible by a signal) memory offline.
Patch 2 fixes an over-eager warning which is not harmful but surely
annoying.
I know we are late in the release cycle but I guess both would be
candidates for rc7. They are simple enough and they should be
"obviously" correct. If you would like more time for them for testing
then I am perfectly fine postponing to the next merge window of course.
[1] http://lkml.kernel.org/r/20180523073547.GA29266@xxxxxxxxxxxxxxxxxx
[2] http://lkml.kernel.org/r/20180523080108.GA30350@xxxxxxxxxxxxxxxxxx
Michal Hocko (2):
mm, memory_hotplug: make has_unmovable_pages more robust
mm: do not warn on offline nodes unless the specific node is explicitly requested
Diffstat
include/linux/gfp.h | 2 +-
mm/page_alloc.c | 16 ++++++++++------
2 files changed, 11 insertions(+), 7 deletions(-)
