On Fri, 10/16 11:32, Stefan Hajnoczi wrote: > On Tue, Oct 13, 2015 at 07:10:55PM +0800, Fam Zheng wrote: > > +static bool aio_epoll_try_enable(AioContext *ctx) > > +{ > > + AioHandler *node; > > + struct epoll_event event; > > + if (!ctx->epoll_available) { > > + return false; > > + } > > Why check this here since aio_epoll_check_poll() already checks it?
You're right, it's redundant. I will remove it. > > > +static int aio_epoll(AioContext *ctx, GPollFD *pfds, > > + unsigned npfd, int64_t timeout) > > +{ > > + AioHandler *node; > > + int i, ret = 0; > > + struct epoll_event events[128]; > > The strategy is to support up to 128 events per epoll_wait(2) call and > then wait for the next event loop iteration to harvest any remaining > events? Yes.