Re: [for-4.15] Re: [PATCH XENSTORE v1 00/10] Code analysis fixes

[Julien Grall <julien@xxxxxxx>]

Hi Andrew,

On 01/03/2021 19:39, Andrew Cooper wrote:
> On 01/03/2021 18:01, Julien Grall wrote:
> > Hi,
> >
> > I have tagged the e-mail with 4.15 as I think we likely want some of
> > the patches to be in the next release.
> >
> > As a minimum, we get the following:
> >    - patch #7: xenstore: handle do_mkdir and do_rm failure
> >    - patch #8: xenstore: add missing NULL check
> >    - patch #10: xs: add error handling
> >
> > The first two add missing NULL check in runtime code in XenStored. The
> > 3rd one adds a missing NULL check in xs_is_domain_introduced() in
> > libxenstore (can be used at runtime by xenpaging at least).
> >
> > In addition to that, I would like to consider patch #3: xenstore:
> > check formats of trace. It is allowing the compiler to check the
> > format printf for trace(). This should be low-risk.
> >
> > For the rest is a mix of silencing coverity and potential errors
> > either at init or in a standalone binaries.
> >
> > The init ones would be useful (patch #1, #5, #9) for Xenstored
> > LiveUpdate as they would be potential triggered when upgrading the
> > binary. But I am not sure whether we consider LiveUpdate supported.
> >
> > Any thoughts?
> Live Update is a headline feature, even if only tech preview.
I thought it was a tech preview but I couldn't find the statement in 
SUPPORT.MD. I guess we should update it before 4.15 is released.
> I'd say that all bugfixes are fair game, and low risk.  All these fixes
> (other than the evtchn one which has an outstanding question) look to be
> reasonable to take.  They're all simple and obvious fixes pointed out by
> static analysis.
That's a fair point. I wanted to set an order as I know the rules are 
getting tighter. So this gives an opportunity to Ian to have enough data 
to decide what's the best.
Cheers,

--
Julien Grall