On 19/10/2016 19:01, Dr. David Alan Gilbert wrote: > * Paolo Bonzini (pbonz...@redhat.com) wrote: >> >> >> On 18/10/2016 16:01, Daniel P. Berrange wrote: >>>>> >>>>> I already use error_report's in places in migration threads of various >>>>> types; I'm not sure if that's a problem. >>> Unless those places are protected by the big qemu lock, that sounds >>> not good. error_report calls into error_vprintf which checks the >>> 'cur_mon' global "Monitor" pointer. This variable is updated at >>> runtime - eg in qmp_human_monitor_command(), monitor_qmp_read(), >>> monitor_read(), etc. So if migration threads outside the BQL are >>> calling error_report() that could well cause problems. If you >>> are lucky messages will merely end up going to stderr instead of >>> the monitor, but in worst case I wouldn't be surprised if there >>> is a crash possibility in some race conditions. >> >> Writes to chardevs *are* thread-safe (assuming qio_channel_create_watch >> is thread-safe; it seems to be). > > Hmm that's useful (although it doesn't solve error_report because > error_vprintf > is racy itself).
How is it racy? Because of the case where cur_mon changes under the feet of error_vprintf? I guess that can be ignored for now (just a TODO comment will do). > How deadlock safe is it? In particular imagine I've got another > thread which is doing: > take bql > <many layers of stuff> > write to a chardev > release bql > > if that chardev is registered on the main thread, will that > deadlock? No, it won't. Since your usecase is about the monitor, note that the monitor does its own buffering; if the nonblocking write leaves stuff in the buffer, the monitor will process the watch only after the BQL is released by the thread. However, there's no deadlock. Paolo >> Only reads aren't, in the sense that they require an event loop so they >> use that event loop for serialization. > > Hmm OK. > > Dave > >> >> Paolo > -- > Dr. David Alan Gilbert / dgilb...@redhat.com / Manchester, UK > >