On Wed, Sep 27, 2017 at 03:14:40PM +0200, Paolo Bonzini wrote: > On 27/09/2017 14:36, Fam Zheng wrote: > > On Wed, 09/27 13:17, Stefan Hajnoczi wrote: > >> On Tue, Sep 26, 2017 at 07:13:43PM +0800, Fam Zheng wrote: > >>> On Tue, 09/26 17:11, Peter Xu wrote: > >>> void aio_context_unref(AioContext *ctx) > >>> { > >>> + assert(ctx->refcnt > 0); > >>> + if (--ctx->refcnt == 0) { > >>> + aio_set_event_notifier(ctx, &ctx->notifier, false, NULL, NULL); > >>> + } > >> > >> This isn't a general solution because Linux AIO also has a file > >> descriptor that is removed in aio_ctx_finalize(). > > > > Right. Another option is to move everything in aio_context_finalize() into > > the > > "if (--ctx->refcnt == 0) { ... }" block, before calling g_source_unref(). > > I think I prefer Stefan's solution. If the GSource has been ref'ed > independent of the AioContext object, it may be a problem to free it early.
Stefan's patch works for me. I'll put Stefan's patch into iothread series and repost soon. Thanks Stefan (and all)! -- Peter Xu