Unlike the SVA case where each PASID of a device has an SVA domain
attached to it, the I/O page faults are handled by the fault handler
of the SVA domain. The I/O page faults for a user page table might
be handled by the domain attached to RID or the domain attached to
the PASID, depending on whether the PASID table is managed by user
space or kernel. As a result, there is a need for the domain attach
group interfaces to have attach handle support. The attach handle
will be forwarded to the fault handler of the user domain.
Add some variants of the domain attaching group interfaces so that they
could support the attach handle and export them for use in IOMMUFD.
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
---
drivers/iommu/iommu-priv.h | 8 +++
drivers/iommu/iommu.c | 99 ++++++++++++++++++++++++++++++++++++++
2 files changed, 107 insertions(+)
diff --git a/drivers/iommu/iommu-priv.h b/drivers/iommu/iommu-priv.h
index f1536a5ebb0d..c37801c32f33 100644
--- a/drivers/iommu/iommu-priv.h
+++ b/drivers/iommu/iommu-priv.h
@@ -31,4 +31,12 @@ void iommu_device_unregister_bus(struct iommu_device *iommu,
struct iommu_attach_handle *iommu_attach_handle_get(struct iommu_group *group,
ioasid_t pasid,
unsigned int type);
+int iommu_attach_group_handle(struct iommu_domain *domain,
+ struct iommu_group *group,
+ struct iommu_attach_handle *handle);
+void iommu_detach_group_handle(struct iommu_domain *domain,
+ struct iommu_group *group);
+int iommu_replace_group_handle(struct iommu_group *group,
+ struct iommu_domain *new_domain,
+ struct iommu_attach_handle *handle);
#endif /* __LINUX_IOMMU_PRIV_H */
diff --git a/drivers/iommu/iommu.c b/drivers/iommu/iommu.c
index c506185a2fad..78e29fc97236 100644
--- a/drivers/iommu/iommu.c
+++ b/drivers/iommu/iommu.c
@@ -3475,3 +3475,102 @@ iommu_attach_handle_get(struct iommu_group *group, ioasid_t pasid, unsigned int
return handle;
}
EXPORT_SYMBOL_NS_GPL(iommu_attach_handle_get, IOMMUFD_INTERNAL);
+
+/**
+ * iommu_attach_group_handle - Attach an IOMMU domain to an IOMMU group
+ * @domain: IOMMU domain to attach
+ * @group: IOMMU group that will be attached
+ * @handle: attach handle
+ *
+ * Returns 0 on success and error code on failure.
+ *
+ * This is a variant of iommu_attach_group(). It allows the caller to provide
+ * an attach handle and use it when the domain is attached. This is currently
+ * used by IOMMUFD to deliver the I/O page faults.
+ */
+int iommu_attach_group_handle(struct iommu_domain *domain,
+ struct iommu_group *group,
+ struct iommu_attach_handle *handle)
+{
+ int ret;
+
+ mutex_lock(&group->mutex);
+ ret = xa_insert(&group->pasid_array, IOMMU_NO_PASID, handle, GFP_KERNEL);
+ if (ret)
+ goto err_unlock;
+
+ ret = __iommu_attach_group(domain, group);
+ if (ret)
+ goto err_erase;
+ mutex_unlock(&group->mutex);
+
+ return 0;
+err_erase:
+ xa_erase(&group->pasid_array, IOMMU_NO_PASID);
+err_unlock:
+ mutex_unlock(&group->mutex);
+ return ret;
+}
+EXPORT_SYMBOL_NS_GPL(iommu_attach_group_handle, IOMMUFD_INTERNAL);
+
+/**
+ * iommu_detach_group_handle - Detach an IOMMU domain from an IOMMU group
+ * @domain: IOMMU domain to attach
+ * @group: IOMMU group that will be attached
+ *
+ * Detach the specified IOMMU domain from the specified IOMMU group.
+ * It must be used in conjunction with iommu_attach_group_handle().
+ */
+void iommu_detach_group_handle(struct iommu_domain *domain,
+ struct iommu_group *group)
+{
+ mutex_lock(&group->mutex);
+ __iommu_group_set_core_domain(group);
+ xa_erase(&group->pasid_array, IOMMU_NO_PASID);
+ mutex_unlock(&group->mutex);
+}
+EXPORT_SYMBOL_NS_GPL(iommu_detach_group_handle, IOMMUFD_INTERNAL);
+
+/**
+ * iommu_replace_group_handle - replace the domain that a group is attached to
+ * @group: IOMMU group that will be attached to the new domain
+ * @new_domain: new IOMMU domain to replace with
+ * @handle: attach handle
+ *
+ * This is a variant of iommu_group_replace_domain(). It allows the caller to
+ * provide an attach handle for the new domain and use it when the domain is
+ * attached.
+ */
+int iommu_replace_group_handle(struct iommu_group *group,
+ struct iommu_domain *new_domain,
+ struct iommu_attach_handle *handle)
+{
+ struct iommu_domain *old_domain = group->domain;
+ void *curr;
+ int ret;
+
+ if (!new_domain)
+ return -EINVAL;
+
+ mutex_lock(&group->mutex);
+ ret = __iommu_group_set_domain(group, new_domain);
+ if (ret)
+ goto err_unlock;
+ xa_erase(&group->pasid_array, IOMMU_NO_PASID);
+ if (handle) {
+ curr = xa_store(&group->pasid_array, IOMMU_NO_PASID, handle, GFP_KERNEL);
+ if (xa_err(curr)) {
+ ret = xa_err(curr);
+ goto err_restore;
+ }
+ }
+ mutex_unlock(&group->mutex);
+
+ return 0;
+err_restore:
+ __iommu_group_set_domain(group, old_domain);
+err_unlock:
+ mutex_unlock(&group->mutex);
+ return ret;
+}
+EXPORT_SYMBOL_NS_GPL(iommu_replace_group_handle, IOMMUFD_INTERNAL);
--
2.34.1On Mon, May 27, 2024 at 12:05:11PM +0800, Lu Baolu wrote: > Unlike the SVA case where each PASID of a device has an SVA domain > attached to it, the I/O page faults are handled by the fault handler > of the SVA domain. The I/O page faults for a user page table might > be handled by the domain attached to RID or the domain attached to > the PASID, depending on whether the PASID table is managed by user > space or kernel. As a result, there is a need for the domain attach > group interfaces to have attach handle support. The attach handle > will be forwarded to the fault handler of the user domain. > > Add some variants of the domain attaching group interfaces so that they > could support the attach handle and export them for use in IOMMUFD. > > Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com> > --- > drivers/iommu/iommu-priv.h | 8 +++ > drivers/iommu/iommu.c | 99 ++++++++++++++++++++++++++++++++++++++ > 2 files changed, 107 insertions(+) I don't have an objection to it like this, but I wonder if we could be smaller to teach iommu_attach_device_pasid to use IOMMU_NO_PASID to attach the handle to the rid? It would have an if there to call the __iommu_attach_group() instead of the pasid attach ? Jason
On 6/12/24 9:41 PM, Jason Gunthorpe wrote: > On Mon, May 27, 2024 at 12:05:11PM +0800, Lu Baolu wrote: >> Unlike the SVA case where each PASID of a device has an SVA domain >> attached to it, the I/O page faults are handled by the fault handler >> of the SVA domain. The I/O page faults for a user page table might >> be handled by the domain attached to RID or the domain attached to >> the PASID, depending on whether the PASID table is managed by user >> space or kernel. As a result, there is a need for the domain attach >> group interfaces to have attach handle support. The attach handle >> will be forwarded to the fault handler of the user domain. >> >> Add some variants of the domain attaching group interfaces so that they >> could support the attach handle and export them for use in IOMMUFD. >> >> Signed-off-by: Lu Baolu<baolu.lu@linux.intel.com> >> --- >> drivers/iommu/iommu-priv.h | 8 +++ >> drivers/iommu/iommu.c | 99 ++++++++++++++++++++++++++++++++++++++ >> 2 files changed, 107 insertions(+) > I don't have an objection to it like this, but I wonder if we could be > smaller to teach iommu_attach_device_pasid to use IOMMU_NO_PASID to > attach the handle to the rid? > > It would have an if there to call the __iommu_attach_group() instead > of the pasid attach ? This is a good idea. I guess that in the future, we will have iommu_attach_device_pasid(IOMMU_NO_PASID) to replace iommu_attach_group(). The group->domain and domain_ops::attach_dev will also be removed. I'd suggest making such refactoring in a separate series with wider discussions. For now, let's make this special interface for iommufd. Best regards, baolu
© 2016 - 2026 Red Hat, Inc.