Re: [Qemu-devel] [PATCH 1.1] coroutine: Avoid ucontext usage on i386 Linux host

[Anthony Liguori]

On 05/09/2012 04:27 PM, Peter Maydell wrote:
> On 9 May 2012 21:59, Anthony Liguori<address@hidden>  wrote:
> > On 05/09/2012 03:46 PM, Peter Maydell wrote:
> > > Longer term (ie post 1.1) I'm strongly in favour of kicking
> > > out coroutines, because I think there clearly is no single
> > > solid portable implementation possible. C just isn't designed
> > > to allow them; better not to try to swim against the current.
>
> > Unfortunately, voting for code to be different doesn't actually make it
> > different.
>
> Yeah, I agree with this sentiment...
>
> > If you're volunteering to rewrite the block layer to not require coroutines
> > (either by using a state machine or by using re-entrant threads and fixing
> > any locking issues associated with that) that's wonderful.
> >
> > But we decided to not do synchronous I/O years ago and still haven't removed
> > it all from the tree.  Coroutines got us much closer to getting rid of
> > synchronous I/O.
>
> ...but I would at least like us to take the position that we don't
> introduce *more* users of coroutines.

I think the long term plan has been:

1) replace synchronous I/O users with coroutines + async I/O

2) promote coroutines to threads by introducing fine grain locking.

I don't think avoiding coroutines helps us along this route nor does it help 
eliminate immediate users of coroutines.

I think our best strategy forward is to get rid of async I/O in the blocker 
layer and in devices.  Then I think we should promote coroutines as much as 
possible.

Regards,

Anthony Liguori

> -- PMM