From: "Dr. David Alan Gilbert" <dgilb...@redhat.com>
This is a small set of postcopy changes, the largest of which
is an x86 test for postcopy.
Andrea's libqtest change came about from running my test under very heavy
load.
The test includes a self contained migration workload that rapidly changes
RAM in a predictable fashion allowing us to end up in postcopy mode and
also to be able to check the contents of RAM.
Note this sometimes fails on Linux kernels 4.5 (and current 4.6) which
have a KVM+THP bug. Use this fix:
https://lists.gnu.org/archive/html/qemu-devel/2016-04/msg04028.html
v6:
Fix for postcopy test on non-linux/linux-without-userfault
Tested on freebsd (x86-64) where v5 failed
linux (aarch64) where v5 worked for me, but failed for Peter
linux (x86-64)
v5:
Swing sys/ #include's into the ifdef __linux__ on the postcopy test
v4:
Add a patch to fail migrate-set-capability when turning on postcopy if we
don't have the kernel support.
v3:
Change all the test asserts to be g_assert macros (Marcel's suggestion)
v2:
Split 'Add stats...' into two (Eric's comment)
Test:
Survive qmp events landing when we're expecting a response from the
command (qmp/libqtest doesn't help in that)
Fix a race where we'd start postcopy early
Dave
Andrea Arcangeli (1):
tests: fix libqtest socket timeouts
Dr. David Alan Gilbert (5):
Postcopy: Avoid 0 length discards
Migration: Split out ram part of qmp_query_migrate
Postcopy: Add stats on page requests
test: Postcopy
Postcopy: Check for support when setting the capability
hmp.c | 4 +
include/migration/migration.h | 2 +
migration/migration.c | 73 +++----
migration/ram.c | 5 +-
qapi-schema.json | 6 +-
tests/Makefile.include | 2 +
tests/libqtest.c | 2 +-
tests/postcopy-test.c | 455 ++++++++++++++++++++++++++++++++++++++++++
8 files changed, 511 insertions(+), 38 deletions(-)
create mode 100644 tests/postcopy-test.c
--
2.7.4
