Re: [Xen-devel] [RFC PATCH 8/8]: PVH: privcmd changes

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Thu, 2012-09-13 at 19:27 +0100, Mukesh Rathor wrote:
> > I suppose it depends on whether the core "takes over" the reference
> > which you hold. I think it doesn't, so this is just a leak, rather
> > than putting a ballooned page back into the general allocation pool
> > (things would be crashing left & right if it was doing this I reckon)
> > 
> > > 
> > > I had looked for other hooks initially when I did this, but 
> > > vm_operations_struct->close was the only one to pan out.
> > > 
> > > I can't really move pvh_privcmd_resv_pfns to mmu.c because the 
> > > xen_remap_domain_mfn_range is called one page at a time, and I need
> > > to allocate the array first. I'd have to change it to linked list,
> > > worth it? Or I'd have to move and export it.
> > 
> > Another alternative would be to add the page array as a parameter to
> > the map/unmap function, rather than relying on it propagating via
> > vma_private.
> 
> I thought it was a no-no to change an exported API. Konrad, is it OK
> to change an exported API like xen_remap_domain_mfn_range, I mean, are
> there any guidelines?

You are thinking of system calls, which cannot be changed.

Kernel internal APIs, including those exported to modules are fair game
to change. See Documentation/stable_api_nonsense.txt.

Ian.

> 
> thanks,
> Mukesh
> 
> 



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