On 13/03/14 10:17, Ian Campbell wrote:

Pulling out this one comment for the attention on the core Xen/Linux
maintainers.

On Thu, 2014-03-06 at 21:48 +0000, Zoltan Kiss wrote:
[...]

quoted

@@ -343,8 +347,26 @@ struct xenvif *xenvif_alloc(struct device *parent, domid_t domid,
  	vif->pending_prod = MAX_PENDING_REQS;
  	for (i = 0; i < MAX_PENDING_REQS; i++)
  		vif->pending_ring[i] = i;
-	for (i = 0; i < MAX_PENDING_REQS; i++)
-		vif->mmap_pages[i] = NULL;
+	spin_lock_init(&vif->callback_lock);
+	spin_lock_init(&vif->response_lock);
+	/* If ballooning is disabled, this will consume real memory, so you
+	 * better enable it. The long term solution would be to use just a
+	 * bunch of valid page descriptors, without dependency on ballooning
+	 */

I wonder if we ought to enforce this via Kconfig? i.e. making
CONFIG_XEN_BACKEND (or the individual backends) depend on BALLOON (or
select?) or by making CONFIG_XEN_BALLOON non-optional etc.

I support this idea, but let's have other's views.

IIRC David V was looking into a solution involving auto hotplugging a
new region to use for this case, but then I guess
CONFIG_XEN_BALLOON_MEMORY_HOTPLUG would equally need to be enabled.

Yes, if that happens in the near future, we shouldn't be bothered by the 
above right now.

quoted

+	err = alloc_xenballooned_pages(MAX_PENDING_REQS,
+				       vif->mmap_pages,
+				       false);

[...]