Re: [PATCH kernel] vfio/spapr_tce: Set window when adding additional groups to container

On Tue, Feb 07, 2017 at 12:35:59PM -0700, Alex Williamson wrote:

On Tue,  7 Feb 2017 17:26:57 +1100
Alexey Kardashevskiy [off-list ref] wrote:

quoted

If a container already has a group attached, attaching a new group
should just program already created IOMMU tables to the hardware via
the iommu_table_group_ops::set_window() callback.

However 6f01cc692 "vfio/spapr: Add a helper to create default DMA window"
did not just simplify the code but also removed the set_window() calls
in the case of attaching groups to a container which already has tables
so it broke VFIO PCI hotplug.

This reverts set_window() bits in tce_iommu_take_ownership_ddw().

Fixes: 6f01cc692

See Documentation/process/submitting-patches.rst for the expected
format here, 12 char SHA-1 + --oneline summary.

quoted

Signed-off-by: Alexey Kardashevskiy <redacted>
---
 drivers/vfio/vfio_iommu_spapr_tce.c | 21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)

diff --git a/drivers/vfio/vfio_iommu_spapr_tce.c b/drivers/vfio/vfio_iommu_spapr_tce.c
index 7690e5bf3cf1..14c62a8495a2 100644
--- a/drivers/vfio/vfio_iommu_spapr_tce.c
+++ b/drivers/vfio/vfio_iommu_spapr_tce.c

@@ -1245,6 +1245,8 @@ static void tce_iommu_release_ownership_ddw(struct tce_container *container,
 static long tce_iommu_take_ownership_ddw(struct tce_container *container,
 		struct iommu_table_group *table_group)
 {
+	long i, ret = 0;
+
 	if (!table_group->ops->create_table || !table_group->ops->set_window ||
 			!table_group->ops->release_ownership) {
 		WARN_ON_ONCE(1);

@@ -1253,7 +1255,26 @@ static long tce_iommu_take_ownership_ddw(struct tce_container *container,
 
 	table_group->ops->take_ownership(table_group);
 
+	/* Set all windows to the new group */
+	for (i = 0; i < IOMMU_TABLE_GROUP_MAX_TABLES; ++i) {
+		struct iommu_table *tbl = container->tables[i];

checkpatch wants a blank line here.

quoted

+		if (!tbl)
+			continue;
+
+		ret = table_group->ops->set_window(table_group, i, tbl);
+		if (ret)
+			goto release_exit;
+	}
+
 	return 0;
+
+release_exit:
+	for (i = 0; i < IOMMU_TABLE_GROUP_MAX_TABLES; ++i)
+		table_group->ops->unset_window(table_group, i);
+
+	table_group->ops->release_ownership(table_group);
+
+	return ret;
 }
 
 static int tce_iommu_attach_group(void *iommu_data,

I've fixed these as shown below.  David, please review/ack as well and
I'll send a pull for v4.10.  Thanks,

Alex

Reviewed-by: David Gibson <redacted>

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson