On Fri, Jun 24, 2016 at 03:22:19PM +0200, Marc-André Lureau wrote: > Hi > > On Thu, Jun 23, 2016 at 6:31 AM, Michael S. Tsirkin <m...@redhat.com> wrote: > > > > OK so if it's ok for read or write to fail, then I think > > the callback should just return success. > > This will make sure backends such as vhost net in kernel > > where failure indicates an internal bug are handled > > correctly, while vhost user ignores errors and attempts > > to go on. > > > > That's what we are trying to do, but since it's not yet well defined > how to disconnect/reconnect and there are many bugs around this, it > would help a lot to add error checking and error reporting for now.
Just for debugging? OK, how about a macro with #ifdef DEBUG or something like that then? > >> Since there is feature checks on reconnection, qemu should wait for > >> the initial connection feature negotiation to complete. The test added > >> demonstrates this. > > > > > > So on disconnect, we keep the guest going? But then when connect is > > attempted, then we for some reason block guest until negotiation > > is done? Sounds rather strange to me. > > When qemu is the server, it already waits for an initial connection > from vhost-user backend. Here it is just enhanced to wait for the > initial nego to complete, which is necessary for proper device > initialization and also further reconnections checks. OH, it's just for the initial connection? I didn't realize. Will have to re-read it to figure it out. Some comments in commit log and code pointing this out could also help. > -- > Marc-André Lureau