Re: [PATCH v6 1/1] numa: add 'memmap-type' option for memory type configuration

[Huang, FangSheng (Jerry)]

On 3/6/2026 5:06 AM, Gregory Price wrote:

On Thu, Feb 26, 2026 at 06:50:23PM +0800, fanhuang wrote:

+
+            if (addr < x86ms->below_4g_mem_size) {
+                if (addr + node_size <= x86ms->below_4g_mem_size) {
+                    guest_addr = addr;
+                } else {
+                    error_report("NUMA node %d with memmap-type spans across "
+                                 "4GB boundary, not supported", i);
+                    exit(EXIT_FAILURE);
+                }
+            } else {
+                guest_addr = 0x100000000ULL +
+                            (addr - x86ms->below_4g_mem_size);
+            }
+

I missed this on my first go around

Should this be:

   guest_addr = x86ms->above_4g_mem_start +
               (addr - x86ms->below_4g_mem_size);

Or is there a reason for the hard-code?

~Gregory

Hi Gregory,

Good catch, thanks for spotting this!

You're right, there's no reason to hardcode 4 GiB here — on AMD
hosts with IOMMU, above_4g_mem_start gets relocated to above 1 TB,
so the hardcode would produce a wrong address.

You're right, it should be:

  guest_addr = x86ms->above_4g_mem_start +
              (addr - x86ms->below_4g_mem_size);

I'll send a v7 shortly with this fix.

Thanks,
Jerry Huang