...
This is the reason why I've requested in the previous review that the pinning information would be merged into the iothread data structure. You then would not have to synchronise two data structures.
Understood ... I started down that path, but then got bogged down in cputune information inside iothreadid's and the difference with the vCPU code. So I kept it this way to reduce the number of changes.
I think it's worthy of being done, but I hope a follow-up patch will be acceptable.
ug... starting looking at it again with the "existing" view of iothreadid's fresh in mind and it seems I'll have to jump in during this series because there's a check in virDomainIOThreadPinDefParseXML for pin->id > iothreads, which won't be true once "Add" hits, so I'll work it in now <sigh> John