On 03/09/2018 05:33 PM, Wei Liu wrote:
On Mon, Mar 05, 2018 at 07:41:54AM -0700, Jan Beulich wrote:
On 05.03.18 at 15:18, <julien.gr...@arm.com> wrote:
On 02/03/18 15:34, Jan Beulich wrote:
On 21.02.18 at 15:02, <julien.gr...@arm.com> wrote:
@@ -95,11 +101,18 @@ static unsigned int max_order(const struct domain *d)
       return min(order, MAX_ORDER + 0U);
+/* Helper to copy a typesafe MFN to guest */
+#define copy_mfn_to_guest(hnd, off, mfn)            \
+    ({                                              \
+        xen_pfn_t mfn_ = mfn_x(mfn);                \
+        __copy_to_guest_offset(hnd, off, &mfn_, 1); \
+    })

Hmm, not really nice, but what do you do.

I am open to better suggestion. I wanted to avoid the conversion all
over the code.

I have no better suggestion, I'm sorry, hence the "but what do
you do."

Also, do you have an opinion on Wei's suggestion:

"What I meant was to make copy_{to,from}_guest* type-safe. I just feel it
a bit strange you only created a wrapper for this file. I wonder why.

Note I'm just asking question. That's not necessarily a good idea to
turn them all in the end."

Well, I didn't really understand what he's after (in the context of
this series) - copy_{to,from}_guest() don't take or return MFNs or

Fundamentally Julien's patch is to wrap around an existing API for this
one file only. Why is this file special? Why not just make that class of
APIs do what he wants?

But that is going to be intrusive and a bit counter-intuitive.

I have quickly looked at it. The major problem I can see is it is not possible to generically define for any typesafe. Indeed, TYPE_SAFE(...) cannot define new macro and, AFAICT, it is not feasible to define static inline for copy_* helpers.

So we would need to introduce macros for each typesafe by hand. I can move copy_mfn_to_guest in xen/mm.h if people think it could be useful.


Julien Grall

Xen-devel mailing list

Reply via email to