Re: [Xen-devel] [PATCH] vunmap: let vunmap align virtual address by itself

[Julien Grall <julien.grall@xxxxxxx>]

Hi Jan,

On 23/07/2019 10:12, Jan Beulich wrote:
> On 23.07.2019 10:48, Andrii Anisov wrote:
> > Julien, Jan, Andrew,
> >
> > The problem addressed by [1] causes random ARM64 boot fails dependent on 
> > hypervisor code changes.
> > Yet more generic solution was requested by Andrew and supported by Julien [2].
> >
> > How to proceed with this particular patch?
> > As I understand, Jan doubts we should move page alignment to vunmap(), while 
> > Julien and Andrew wanted the commit message clarification.
> > Can we have an agreement on approach here?
> >
> > [1] https://lists.xenproject.org/archives/html/xen-devel/2019-07/msg01167.html
> > [2] https://lists.xenproject.org/archives/html/xen-devel/2019-07/msg01129.html
> First of all, let me quote Linux'es code:
>
> static void __vunmap(const void *addr, int deallocate_pages)
> {
>         struct vm_struct *area;
>
>         if (!addr)
>                 return;
>
>         if (WARN(!PAGE_ALIGNED(addr), "Trying to vfree() bad address (%p)\n",
>                         addr))
>                 return;
>
> As long as we aim to have a reasonable level of compatibility of
> similar interfaces, we should not go the suggested route.
Well, it is more likely to have Linux code moving to Xen compare to the 
opposite. So the change suggested is still compatible as we don't restrict 
anything but just open more flexibility.
> Beyond that I continue to be of the opinion that it should be
> all-or-nothing: Any pointer pointing anywhere at or inside the
> region should be accepted, or just the one pointing precisely at
> the start.
That's fine, but we can still achieve this step by step... Handling unaligned 
address is quite easy.
Cheers,

--
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel