Re: [Qemu-devel] [PATCH 3/6] Make char muxer more robust wrt small FIFOs

[Amit Shah]

On (Wed) Apr 07 2010 [16:32:04], Alexander Graf wrote:
> Amit Shah wrote:
> > On (Thu) Apr 01 2010 [18:42:38], Alexander Graf wrote:
> >   
> >> Virtio-Console can only process one character at a time.
> >>     
> >
> > The host can process as many as you give it, depending on the buffer
> > size exposed by the guest.
> >
> > On older guests (guest kernels w/o multiport support), the guest reads
> > input from host, processes it and only then opens up another buffer for
> > the host to write into.
> >
> > On newer guests (guest kernels that support multiport), the guest
> > fills the entire vq so that host can send as many buffers as possible
> > without getting throttled. I guess you're getting hit by this.
> >   
> 
> Probably, yes.
> 
> >> Using it on S390
> >> gave me strage "lags" where I got the character I pressed before when
> >> pressing one. So I typed in "abc" and only received "a", then pressed "d"
> >> but the guest received "b" and so on.
> >>     
> >
> > This might be because qemu-char would not be able to send out 'b' while
> > the guest still processes 'a' and has no free buffers to write out to.
> > On seeing 'd', it flushes its queue.
> >
> > Can you try using a 2.6.34-rc3 kernel without this patch to see if
> > things work fine?
> >   
> 
> Hrm - would it actually matter? We need to have older guest kernels
> working anyways. And my S390 LPAR doesn't exactly have a lot of disk
> space, so compiling a kernel is anything but fun :-).

Oh I just want to find out if it works for you -- you shouldn't get hit
by the 1-char limit anymore.

                Amit
-- 
http://log.amitshah.net/