On Tue, 11 Oct 2022 17:58:29 +0200 Christian Borntraeger <borntraeger@xxxxxxxxxxxxx> wrote: > Am 11.10.22 um 17:14 schrieb Nico Boehr: > > v1->v2: > > --- > > - remove unneeded include > > - advance clock by 10 minutes instead of 1 minute (thanks Claudio) > > - express get_clock_us() using stck() (thanks Claudio) > > > > The guest TOD clock should be preserved on migration. Add a test to > > verify that. > > I do not fully agree with this assumption. Its the way it curently is, but we might want to have a configurable or different behaviour in the future. > > For example if the difference is smaller than time x it could be allowed to move the time forward to get the guest synced to the new host (never go backward though). the test is actually testing that the clock does not go backwards, rather than staying the same > Or to preserve the time but then slowly step towards the target system clock etc (or for this testcase step the epoch difference towards the original difference). > > So we maybe want to have a comment in here somehow that this is the as-is behaviour. comments never hurt :) > > > To reduce code duplication, move some of the time-related defined > > from the sck test to the library. > > > > Nico Boehr (2): > > lib/s390x: move TOD clock related functions to library > > s390x: add migration TOD clock test > > > > lib/s390x/asm/time.h | 50 ++++++++++++++++++++++++++++++++++++++++++- > > s390x/Makefile | 1 + > > s390x/migration-sck.c | 44 +++++++++++++++++++++++++++++++++++++ > > s390x/sck.c | 32 --------------------------- > > s390x/unittests.cfg | 4 ++++ > > 5 files changed, 98 insertions(+), 33 deletions(-) > > create mode 100644 s390x/migration-sck.c > >
