> >> > At least the floppy DMA engine is fine with it, it uses idle bottom > >> > halves (which are a hack and could be replaced by timers, but that's > >> > not relevant now). > > > > I thought idle bottom halves were one of the things that made this timout > > necessary. How else are they going to get run? > > The timeout is reduced to 10 ms when an idle bottom half is scheduled. > See qemu_bh_update_timeout in async.c.
Ah, I see. Idle BH are indeed a nasty hack that should be removed, but not directly relevant to this 1s timeout. I don't think this changes my overall conlusion: Either we need this timeout to poll below the user-thinks-qemu-died threshold, or we should be blocking indefinitely. Paul
