While doing routing code review I noticed that commit 6df8d9aaf3af
("Drivers: hv: balloon: Correctly update onlined page count") introduced
an issue with num_pages_onlined accounting on memory offlining. Deeper look
showed that the accounting was always buggy. This is fixed in PATCH3.
PATCHes 1 and 2 are preparatory cleanups, PATCH4 adds a tracepoint to
post_status so it's now possible to see what's being sent to the host and
where the data comes from.
One more thing: currently core MM code forbids offlining blocks with
reserved pages in it, I plat to send a patch making it possible.
Tested with WS2016 only.
Vitaly Kuznetsov (4):
hv_balloon: fix printk loglevel
hv_balloon: simplify hv_online_page()/hv_page_online_one()
hv_balloon: fix bugs in num_pages_onlined accounting
hv_balloon: trace post_status
drivers/hv/Makefile | 1 +
drivers/hv/hv_balloon.c | 121 +++++++++++++++++++++++++++++-------------
drivers/hv/hv_trace_balloon.h | 48 +++++++++++++++++
3 files changed, 132 insertions(+), 38 deletions(-)
create mode 100644 drivers/hv/hv_trace_balloon.h
--
2.13.6
_______________________________________________
devel mailing list
devel@xxxxxxxxxxxxxxxxxxxxxx
http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
