Re: [Xen-devel] [PATCH RFC 01/13] x86/mm: export {get, put}_pg_owner

[Wei Liu <wei.liu2@xxxxxxxxxx>]

On Wed, Mar 29, 2017 at 10:10:39AM +0100, Andrew Cooper wrote:
> On 29/03/17 10:03, Jan Beulich wrote:
> >>>> On 28.03.17 at 23:11, <andrew.cooper3@xxxxxxxxxx> wrote:
> >> On 27/03/2017 10:10, Wei Liu wrote:
> >>> Prefix them with "mm_" and add declarations to asm-x86/mm.h.
> >>>
> >>> They will be needed when we split PV specific code out of x86/mm.c.
> > Is that actually the case? They're about PV (target) domains, so
> > I'd kind of expect them to move together with the PV-only code,
> > even if the caller may not be PV.
> >
> >>> No functional change.
> >>>
> >>> Signed-off-by: Wei Liu <wei.liu2@xxxxxxxxxx>
> >> I have to admit that I don't understand why they are called
> >> {get,put}_pg_owner.  Perhaps very historical from Linux?
> > I don't think any of this code has Linux origin.
> >
> >> They are nothing to do with pages, and get a reference on the domain.
> > Depends on the perspective you take: For all of their callers,
> > they have precisely that meaning.
> >
> >> I'd recommend s/pg_owner/domain/ so the function calls actually indicate
> >> what object is having the reference taken on it.
> > Well, to make clear what uses are legitimate, perhaps
> > s/pg_owner/foreign_domain/ (if you really continue to think
> > these should be renamed in the first place)? Using just "domain"
> > pretty clearly results in too generic names. Perhaps additionally
> > they should be prefixed mm_?
> 
> Sorry - I had intended my suggestion to be in combination with the mm_
> prefixes, so mm_{get,put}_domain().

Fine by me. I've always found the original names cryptic.

Assuming exporting the functions are still necessary, I will use the
suggested names.

> 
> ~Andrew

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