On 05/07/19 11:58, Peter Maydell wrote: > On Fri, 5 Jul 2019 at 10:48, Paolo Bonzini <pbonz...@redhat.com> wrote: >> You're right, the compatibility causes wrong behavior for the default >> -rtc settings (the RC pauses across migration). The right thing to do >> would be to store the base rather than the offset: that is, you store >> the time at which LR was written. Then the offset is s->lr - s->base >> and it's independent of the machine on which the rtc_clock is being read. > > Right. How do we handle this for back-compat purposes? I guess > we need to have a new migration subsection, so if it's present > it has the 'base' value and we ignore the 'offset' in the > main migration data, and if it's not present we assume an > old->new migration and use the existing offset code. New->old > migration would not be possible as the new subsection is > always-present.
Yes, something like that but I would just bump the version. Version 1 has the old meaning for the first field, version 2 has the new meaning. And also, since our brains are fresh on pl031... currently s->lr is always 0; besides the bug that writing RTC_LR should update it, the datasheet says the counter counts up from 1 so perhaps at startup s->lr should be set to a nonzero value? That would be qemu_ref_timedate(QEMU_CLOCK_VIRTUAL) - 1. Paolo
