Re: [Xen-devel] [PATCH] IOMMU/x86: per-domain control structure is not HVM-specific

[George Dunlap <george.dunlap@xxxxxxxxxx>]

On 03/05/16 11:18, Andrew Cooper wrote:
> On 03/05/16 10:59, Jan Beulich wrote:
> > > > > On 03.05.16 at 11:36, <andrew.cooper3@xxxxxxxxxx> wrote:
> > > On 02/05/16 13:55, Jan Beulich wrote:
> > > > ... and hence should not live in the HVM part of the PV/HVM union. In
> > > > fact it's not even architecture specific (there already is a per-arch
> > > > extension type to it), so it gets moved out right to common struct
> > > > domain.
> > > >
> > > > Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
> > > Yikes - this really does need fixing ASAP.
> > The more interesting question here is how to deal with the stable
> > trees: The patch, due to its size, doesn't really lend itself to
> > backporting. Yet the alternative (mentioned in a remark in the
> > original submission) would be pretty clumsy too.
> >
> > Opinions?
> Once the structure layout has moved, the compiler (and/or grep) will
> highlight all the other uses.
>
> In this instance, I would be tempted to suggest that we just do a custom
> backport for each stable tree.
>
> The only thing a BUILD_BUG_ON() will do is force someone else to
> backport this patch if it actually triggers.
It will, however, avoid the code churn to the stable trees if (as 
expected) the BUILD_BUG_ON() does *not* trigger.
I think both are reasonable approaches.  I find it hard to imagine how 
new bugs could be introduced by moving the structure; the main risk 
would be that it exposes latent bugs.  But there's an equal probability 
that there are latent bugs with the current layout.
Doing a more complete backport may make future bug fixes which touch 
this code easier to backport as well.
 -George

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