Re: port-xen/57121: Pure NetBSD PVH guests do not appear to be able to use more than 4GB of memory

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Mon, Dec 19, 2022 at 03:20:00PM +0000, brad%anduin.eldar.org@localhost wrote:
> With a pure PVH 64bit NetBSD guest using the GENERIC kernel, the guest
> does not appear to be able to use more than 4GB of memory.  That is,
> no matter how much memory you give the guest the guest caps it at 4GB.
> This does not happen with ArchLinux guests on the same DOM0.  The same
> guest configured as a PV system with pvshim sees all of the memory.

Hello
I just commited the attached patch to HEAD; it should apply cleanly to
netbsd-10 too. Can you test ?
I can't test with much more than 4GB on my devel box (not enough RAM) ...

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--

Index: x86/x86/x86_machdep.c
===================================================================
RCS file: /cvsroot/src/sys/arch/x86/x86/x86_machdep.c,v
retrieving revision 1.152
diff -u -p -r1.152 x86_machdep.c
--- x86/x86/x86_machdep.c	20 Aug 2022 23:48:51 -0000	1.152
+++ x86/x86/x86_machdep.c	23 Dec 2022 16:02:54 -0000
@@ -870,11 +870,13 @@ x86_add_xen_clusters(void)
 			switch (map_entry[i].type) {
 			case XEN_HVM_MEMMAP_TYPE_RAM:
 				x86_add_cluster(map_entry[i].addr,
-				    map_entry[i].size, BIM_Memory);
+				    map_entry[i].addr + map_entry[i].size,
+				    BIM_Memory);
 				break;
 			case XEN_HVM_MEMMAP_TYPE_ACPI:
 				x86_add_cluster(map_entry[i].addr,
-				    map_entry[i].size, BIM_ACPI);
+				    map_entry[i].addr + map_entry[i].size,
+				    BIM_ACPI);
 				break;
 			}
 		}