On 2013-08-13 16:26, Alex Bligh wrote: > > On 13 Aug 2013, at 15:22, Stefan Hajnoczi wrote: > >> We can change the semantics of aio_poll() so long as we don't break >> existing callers and tests. It would make sense to do that after >> merging the io_flush and AioContext timers series. > > Whilst I think we should wait until your 'drop ioflush' stuff > is merged, without doing an audit, I doubt there is any > existing code which uses an AioContext with just its > notifier fd, and is calls aio_poll with blocking=1. If there is, > I doubt that code does anything useful :-)
It would be great to have a use case for this upstream, maybe one that is less special than mine. Ping Fan was talking about that he's working on the HPET. Decoupling it from our lousy iothread should improve its timer accuracy noticeably, maybe finally making it usable for Windows guest. On the other side, it's far from easy to disentangle its logic from the rest of a PC platform devices. Jan -- Siemens AG, Corporate Technology, CT RTC ITP SES-DE Corporate Competence Center Embedded Linux
