On Sun, Feb 09, 2025 at 02:41:52PM -0400, Jason Gunthorpe wrote:

On Sat, Feb 08, 2025 at 01:02:41AM -0800, Nicolin Chen wrote:

quoted

@@ -433,8 +433,8 @@ static int iommufd_group_setup_msi(struct iommufd_group *igroup,
 	list_for_each_entry(cur, &ictx->sw_msi_list, sw_msi_item) {
 		int rc;
 
-		if (cur->sw_msi_start != igroup->sw_msi_start ||
-		    !test_bit(cur->id, igroup->required_sw_msi.bitmap))
+		if (cur->sw_msi_start != idev->sw_msi_start ||
+		    !test_bit(cur->id, idev->igroup->required_sw_msi.bitmap))
 			continue;

So we end up creating seperate sw_msi_list items with unique IDs for
every sw_msi_start?

That indeed might work well, I will try to check it and think about
this harder.

The sw_msi_list is still per-ictx, so there won't be items/ids
that overlap with their sw_msi windows, right?

Then, the per-HWPT bitmap could still protect the iommu_map(),
as the design wanted to? No?

Nicolin