Re: [Xen-devel] [PATCH MINI-OS] xenbus: notify the other end when necessary

[Wei Liu <wei.liu2@xxxxxxxxxx>]

On Mon, Nov 02, 2015 at 03:08:26PM +0000, Ian Campbell wrote:
> On Mon, 2015-11-02 at 15:00 +0000, Wei Liu wrote:
> > On Mon, Nov 02, 2015 at 12:14:50PM +0000, Ian Campbell wrote:
> > > On Mon, 2015-10-26 at 09:47 +0000, Wei Liu wrote:
> > > > The xenbus thread didn't send notification to other end when it
> > > > expected
> > > > more data or consumed responses, which led to stalling the ring from
> > > > time to time.
> > > > 
> > > > This is the culprit that guest was less responsive when using stubdom
> > > > because the device model was stalled.
> > > > 
> > > > Fix this by sending notification to the other end at the right
> > > > places.
> > > 
> > > Could any of this code benefit from using the various macros in ring.h
> > > which produce or consume entries on the ring and return a _notify bit?
> > > 
> > 
> > The bug is it doesn't send notification at all. The existing code in
> > Linux doesn't seem to care about mitigating notifications -- xenbus is
> > supposed to exchange small amount of information anyway so whether a few
> > more notifications are sent is not essential. I think it would be better
> > if we follow something that works at this stage.
> > 
> > The use of ring macro can be considered an improvement but not
> > essential to fix the bug. It can be considered as improvement in the
> > future.
> 
> My reason for suggesting it was more that we trust that those macros are
> already correct, compared with needing to reason from scratch about the
> open coded version used in this code which is what appeared to be going on.
> 

I don't think macros help here. Xenstore initialisation doesn't follow
conventional frontend and backend drivers, so the macros don't really
apply to it -- not without some refactoring.

Wei.

> Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel