[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH] libxl: make libxl_cdrom_insert async

On Wed, 2012-07-25 at 12:14 +0100, Ian Jackson wrote:
> Ian Campbell writes ("Re: [PATCH] libxl: make libxl_cdrom_insert async"):
> > On Tue, 2012-07-24 at 17:49 +0100, Ian Jackson wrote:
> > > This doesn't look actually /incorrect/ but I'm a bit suspicious: if
> > > changing to a new non-empty disk, you first write params := "", in one
> > > xenstore transaction, and then the non-empty value in a second
> > > transaction.  Why ?
> > 
> > Mostly because xend did it that way, I agree that it seems pointless
> > without adding any additional interlock to the protocol.
> > 
> > At one point I was hoping that copying xend closely would lead to
> > working stubdom support etc, but that turned out to be a pipe dream.
> > 
> > Just writing the appropriate key once did seem to work, at least in the
> > cases where it works at all, in my experiments so I guess I might as
> > well change to that here.
> If any other xenstore client depended on seeing the intermediate
> value, the system would be buggy and it would only work by them
> winning the race.  I guess it's possible that there would be a peer
> with a strange bug which means that it misses the first watch event
> but I think we should assume that there isn't unless we discover
> otherwise.

Agreed and changed in V2.


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.